d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
53436ace02
freebsd-dev/sys/dev/isp/isp.c

8305 lines
222 KiB
C
Raw Normal View History

Start each of the license/copyright comments with /*-, minor shuffle of lines
2005-01-06 01:43:34 +00:00
/*-
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* Copyright (c) 1997-2009 by Matthew Jacob
Fix some stupid copyright mistakes that have been there for quite some time.
2007-03-10 02:39:54 +00:00
* All rights reserved.
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
*
Fix some stupid copyright mistakes that have been there for quite some time.
2007-03-10 02:39:54 +00:00
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
*
Fix some stupid copyright mistakes that have been there for quite some time.
2007-03-10 02:39:54 +00:00
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
*
Fix some stupid copyright mistakes that have been there for quite some time.
2007-03-10 02:39:54 +00:00
* THIS SOFTWARE IS PROVIDED BY AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
*
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
Some rearrangement of headers to minimize diffs with outside of FreeBSD repository and to clean up the license header so as to not pollute the license with file function. Zero all mailbox structures prior to use (just in case). Change the outgoing mailbox count for INIT_FIRMWARE to be correct.
2006-07-16 20:11:50 +00:00
/*
* Machine and OS Independent (well, as best as possible)
minor change to reduce some diff noise
2006-11-16 00:31:46 +00:00
* code for the Qlogic ISP SCSI and FC-SCSI adapters.
Some rearrangement of headers to minimize diffs with outside of FreeBSD repository and to clean up the license header so as to not pollute the license with file function. Zero all mailbox structures prior to use (just in case). Change the outgoing mailbox count for INIT_FIRMWARE to be correct.
2006-07-16 20:11:50 +00:00
*/
minor change to reduce some diff noise
2006-11-16 00:31:46 +00:00
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
* Inspiration and ideas about this driver are from Erik Moe's Linux driver
* (qlogicisp.c) and Dave Miller's SBus version of same (qlogicisp.c). Some
* ideas dredged from the Solaris driver.
*/
/*
* Include header file appropriate for platform we're building on.
*/
#ifdef __NetBSD__
Fix some stupid copyright mistakes that have been there for quite some time.
2007-03-10 02:39:54 +00:00
#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD$");
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
#include <dev/ic/isp_netbsd.h>
#endif
#ifdef __FreeBSD__
Some rearrangement of headers to minimize diffs with outside of FreeBSD repository and to clean up the license header so as to not pollute the license with file function. Zero all mailbox structures prior to use (just in case). Change the outgoing mailbox count for INIT_FIRMWARE to be correct.
2006-07-16 20:11:50 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
#include <dev/isp/isp_freebsd.h>
#endif
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
#ifdef __OpenBSD__
#include <dev/ic/isp_openbsd.h>
#endif
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
#ifdef __linux__
clarify headers;move uninit to outer layer;remove watchdog
1998-12-28 19:22:27 +00:00
#include "isp_linux.h"
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
#endif
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
#ifdef __svr4__
#include "isp_solaris.h"
#endif
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
* General defines
*/
#define MBOX_DELAY_COUNT 1000000 / 100
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
#define ISP_MARK_PORTDB(a, b, c) \
isp_prt(isp, ISP_LOGSANCFG, \
"Chan %d ISP_MARK_PORTDB@LINE %d", b, __LINE__); \
isp_mark_portdb(a, b, c)
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
* Local static data
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
static const char fconf[] = "Chan %d PortDB[%d] changed:\n current =(0x%x@0x%06x 0x%08x%08x 0x%08x%08x)\n database=(0x%x@0x%06x 0x%08x%08x 0x%08x%08x)";
static const char notresp[] = "Not RESPONSE in RESPONSE Queue (type 0x%x) @ idx %d (next %d) nlooked %d";
static const char topology[] = "Chan %d WWPN 0x%08x%08x PortID 0x%06x N-Port Handle %d, Connection '%s'";
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
static const char sc4[] = "NVRAM";
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
static const char bun[] = "bad underrun (count %d, resid %d, status %s)";
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
static const char lipd[] = "Chan %d LIP destroyed %d active commands";
static const char sacq[] = "unable to acquire scratch area";
static const uint8_t alpa_map[] = {
0xef, 0xe8, 0xe4, 0xe2, 0xe1, 0xe0, 0xdc, 0xda,
0xd9, 0xd6, 0xd5, 0xd4, 0xd3, 0xd2, 0xd1, 0xce,
0xcd, 0xcc, 0xcb, 0xca, 0xc9, 0xc7, 0xc6, 0xc5,
0xc3, 0xbc, 0xba, 0xb9, 0xb6, 0xb5, 0xb4, 0xb3,
0xb2, 0xb1, 0xae, 0xad, 0xac, 0xab, 0xaa, 0xa9,
0xa7, 0xa6, 0xa5, 0xa3, 0x9f, 0x9e, 0x9d, 0x9b,
0x98, 0x97, 0x90, 0x8f, 0x88, 0x84, 0x82, 0x81,
0x80, 0x7c, 0x7a, 0x79, 0x76, 0x75, 0x74, 0x73,
0x72, 0x71, 0x6e, 0x6d, 0x6c, 0x6b, 0x6a, 0x69,
0x67, 0x66, 0x65, 0x63, 0x5c, 0x5a, 0x59, 0x56,
0x55, 0x54, 0x53, 0x52, 0x51, 0x4e, 0x4d, 0x4c,
0x4b, 0x4a, 0x49, 0x47, 0x46, 0x45, 0x43, 0x3c,
0x3a, 0x39, 0x36, 0x35, 0x34, 0x33, 0x32, 0x31,
0x2e, 0x2d, 0x2c, 0x2b, 0x2a, 0x29, 0x27, 0x26,
0x25, 0x23, 0x1f, 0x1e, 0x1d, 0x1b, 0x18, 0x17,
0x10, 0x0f, 0x08, 0x04, 0x02, 0x01, 0x00
};
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
/*
* Local function prototypes.
*/
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
static void isp_prt_endcmd(ispsoftc_t *, XS_T *);
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
static int isp_parse_async(ispsoftc_t *, uint16_t);
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
static int isp_parse_async_fc(ispsoftc_t *, uint16_t);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
static int isp_handle_other_response(ispsoftc_t *, int, isphdr_t *, uint32_t *);
static void isp_parse_status(ispsoftc_t *, ispstatusreq_t *, XS_T *, long *); static void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_parse_status_24xx(ispsoftc_t *, isp24xx_statusreq_t *, XS_T *, long *);
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
static void isp_fastpost_complete(ispsoftc_t *, uint32_t);
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
static int isp_mbox_continue(ispsoftc_t *);
static void isp_scsi_init(ispsoftc_t *);
static void isp_scsi_channel_init(ispsoftc_t *, int);
static void isp_fibre_init(ispsoftc_t *);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
static void isp_fibre_init_2400(ispsoftc_t *);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
static void isp_mark_portdb(ispsoftc_t *, int, int);
static int isp_plogx(ispsoftc_t *, int, uint16_t, uint32_t, int, int);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
static int isp_port_login(ispsoftc_t *, uint16_t, uint32_t);
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
static int isp_port_logout(ispsoftc_t *, uint16_t, uint32_t);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
static int isp_getpdb(ispsoftc_t *, int, uint16_t, isp_pdb_t *, int);
static void isp_dump_chip_portdb(ispsoftc_t *, int, int);
static uint64_t isp_get_wwn(ispsoftc_t *, int, int, int);
static int isp_fclink_test(ispsoftc_t *, int, int);
static int isp_pdb_sync(ispsoftc_t *, int);
static int isp_scan_loop(ispsoftc_t *, int);
static int isp_gid_ft_sns(ispsoftc_t *, int);
static int isp_gid_ft_ct_passthru(ispsoftc_t *, int);
static int isp_scan_fabric(ispsoftc_t *, int);
static int isp_login_device(ispsoftc_t *, int, uint32_t, isp_pdb_t *, uint16_t *);
static int isp_register_fc4_type(ispsoftc_t *, int);
static int isp_register_fc4_type_24xx(ispsoftc_t *, int);
static uint16_t isp_nxt_handle(ispsoftc_t *, int, uint16_t);
static void isp_fw_state(ispsoftc_t *, int);
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
static void isp_mboxcmd_qnw(ispsoftc_t *, mbreg_t *, int);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
static void isp_mboxcmd(ispsoftc_t *, mbreg_t *);
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
static void isp_spi_update(ispsoftc_t *, int);
static void isp_setdfltsdparm(ispsoftc_t *);
static void isp_setdfltfcparm(ispsoftc_t *, int);
static int isp_read_nvram(ispsoftc_t *, int);
static int isp_read_nvram_2400(ispsoftc_t *, uint8_t *);
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
static void isp_rdnvram_word(ispsoftc_t *, int, uint16_t *);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
static void isp_rd_2400_nvram(ispsoftc_t *, uint32_t, uint32_t *);
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
static void isp_parse_nvram_1020(ispsoftc_t *, uint8_t *);
static void isp_parse_nvram_1080(ispsoftc_t *, int, uint8_t *);
static void isp_parse_nvram_12160(ispsoftc_t *, int, uint8_t *);
static void isp_parse_nvram_2100(ispsoftc_t *, uint8_t *);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
static void isp_parse_nvram_2400(ispsoftc_t *, uint8_t *);
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
* Reset Hardware.
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
*
Use pointer to f/w in md structure as to whether f/w exists or not. If firmware length isn't specified, extract from the 4th short into the firmware.
1999-10-30 19:32:44 +00:00
* Hit the chip over the head, download new f/w if available and set it running.
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
*
* Locking done elsewhere.
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_reset(ispsoftc_t *isp, int do_load_defaults)
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
{
mbreg_t mbs;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
uint32_t code_org, val;
If the HBA is already 'touched', still set maxluns. Othewise for CAM_QUIRK_HILUN devices we loop thru 32bits of lun. Oops. Switch to using USEC_DELAY rather than USEC_SLEEP at isp_reset time. Try to paper around a defect in clients that don't correctly registers themeselves with the fabric nameserver. Minor updates for Mirapoint support- they still use code that is not HANDLE_LOOPSTATE_IN_OUTER_LAYERS, and, surprise surprise, this old stuff had some bugs in it. Clean up some target mode stuff. MFC after: 1 week
2002-06-16 05:18:22 +00:00
int loops, i, dodnld = 1;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
const char *btype = "????";
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
static const char dcrc[] = "Downloaded RISC Code Checksum Failure";
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp->isp_state = ISP_NILSTATE;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (isp->isp_dead) {
isp_shutdown(isp);
ISP_DISABLE_INTS(isp);
return;
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
* Basic types (SCSI, FibreChannel and PCI or SBus)
* have been set in the MD code. We figure out more
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
* here. Possibly more refined types based upon PCI
* identification. Chip revision has been gathered.
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
*
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 01:07:06 +00:00
* After we've fired this chip up, zero out the conf1 register
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
* for SCSI adapters and do other settings for the 2100.
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 01:07:06 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_DISABLE_INTS(isp);
/*
* Pick an initial maxcmds value which will be used
* to allocate xflist pointer space. It may be changed
* later by the firmware.
*/
if (IS_24XX(isp)) {
isp->isp_maxcmds = 4096;
} else if (IS_2322(isp)) {
isp->isp_maxcmds = 2048;
} else if (IS_23XX(isp) || IS_2200(isp)) {
isp->isp_maxcmds = 1024;
} else {
isp->isp_maxcmds = 512;
}
/*
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* Set up DMA for the request and response queues.
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*
* We do this now so we can use the request queue
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* for dma to load firmware from.
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*/
if (ISP_MBOXDMASETUP(isp) != 0) {
isp_prt(isp, ISP_LOGERR, "Cannot setup DMA");
return;
}
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
/*
* Set up default request/response queue in-pointer/out-pointer
* register indices.
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
isp->isp_rqstinrp = BIU2400_REQINP;
isp->isp_rqstoutrp = BIU2400_REQOUTP;
isp->isp_respinrp = BIU2400_RSPINP;
isp->isp_respoutrp = BIU2400_RSPOUTP;
} else if (IS_23XX(isp)) {
When calling isp_reset, set the request/response in/out pointers all at once so there isn't a window with the ones for the 23XX cards being wrong. When being verbose, print out some more FC NVRAM values (like framesize). MFC after: 1 week
2001-09-29 19:37:49 +00:00
isp->isp_rqstinrp = BIU_REQINP;
isp->isp_rqstoutrp = BIU_REQOUTP;
isp->isp_respinrp = BIU_RSPINP;
isp->isp_respoutrp = BIU_RSPOUTP;
} else {
isp->isp_rqstinrp = INMAILBOX4;
isp->isp_rqstoutrp = OUTMAILBOX4;
isp->isp_respinrp = OUTMAILBOX5;
isp->isp_respoutrp = INMAILBOX5;
}
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 02:28:29 +00:00
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
/*
When resetting the Qlogic 2X00 units, reset the FPM (Fibre Protocol Module) and FBM (Fibre Buffer Modules). Also remember to clear the semaphore registers. Tell the RISC processor to not halt on FPM parity errors. Throw out the ISP_CFG_NOINIT silliness and instead go to the use of adapter 'roles' to see whether one completes initialization or not (mostly for Fibre Channel). The ultimate intent, btw, of all of this is to have a warm standby adapter for failover reasons. Because we do roles now, setting of Target Capable Class 3 service parameters in the ICB for the 2x00 cards reflects from role. Also, in isp_start, if we're not supporting an initiator role, we bounce outgoing commands with a Selection Timeout error. Also clean out the TOGGLE_TMODE goop for FC- there is no toggling of target mode like there is for parallel SCSI cards. Do more cleanup with respect to using target ids 0..125 in F-port topologies. Also keep track of things which *were* fabric devices so that when you rescan the fabric you can notify the outer layers when fabric devices go away. Only force a LOGOUT for fabric devices if they're still logged in (i.e., you cat their Port Database entry. Clean up the Get All Next scanning. Finally, use a new tag in the softc to store the opcode for the last mailbox command used so we can report which opcode timed out.
2001-01-15 18:33:08 +00:00
* Put the board into PAUSE mode (so we can read the SXP registers
* or write FPM/FBM registers).
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
ISP_WRITE(isp, BIU2400_HCCR, HCCR_2400_CMD_CLEAR_HOST_INT);
ISP_WRITE(isp, BIU2400_HCCR, HCCR_2400_CMD_CLEAR_RISC_INT);
ISP_WRITE(isp, BIU2400_HCCR, HCCR_2400_CMD_PAUSE);
} else {
ISP_WRITE(isp, HCCR, HCCR_CMD_PAUSE);
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
if (IS_FC(isp)) {
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
switch (isp->isp_type) {
case ISP_HA_FC_2100:
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
btype = "2100";
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
break;
case ISP_HA_FC_2200:
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
btype = "2200";
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
break;
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
case ISP_HA_FC_2300:
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
btype = "2300";
Spring MegaChange #1. ---- Make a device for each ISP- really usable only with devfs and add an ioctl entry point (this can be used to (re)set debug levels, reset the HBA, rescan the fabric, issue lips, etc). ---- Add in a kernel thread for Fibre Channel cards. The purpose of this thread is to be woken up to clean up after Fibre Channel events block things. Basically, any FC event that casts doubt on the location or identify of FC devices blocks the queues. When, and if, we get the PORT DATABASE CHANGED or NAME SERVER DATABASE CHANGED async event, we activate the kthread which will then, in full thread context, re-evaluate the local loop and/or the fabric. When it's satisfied that things are stable, it can then release the blocked queues and let commands flow again. The prior mechanism was a lazy evaluation. That is, the next command to come down the pipe after change events would pay the full price for re-evaluation. And if this was done off of a softcall, it really could hang up the system. These changes brings the FreeBSD port more in line with the Solaris, Linux and NetBSD ports. It also, more importantly, gets us being more proactive about topology changes which could then be reflected upwards to CAM so that the periph driver can be informed sooner rather than later when things arrive or depart. --- Add in the (correct) usage of locking macros- we now have lock transition macros which allow us to transition from holding the CAM lock (Giant) and grabbing the softc lock and vice versa. Switch over to having this HBA do real locking. Some folks claim this won't be a win. They're right. But you have to start somewhere, and this will begin to teach us how to DTRT for HBAs, etc. -- Start putting in prototype 2300 support. Add back in LIP and Loop Reset as async events that each platform will handle. Add in another int_bogus instrumentation point. Do some more substantial target mode cleanups. MFC after: 8 weeks
2001-05-28 21:20:43 +00:00
break;
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
case ISP_HA_FC_2312:
btype = "2312";
break;
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
case ISP_HA_FC_2322:
btype = "2322";
break;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
case ISP_HA_FC_2400:
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
btype = "2422";
break;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
case ISP_HA_FC_2500:
btype = "2532";
break;
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
default:
break;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (!IS_24XX(isp)) {
/*
* While we're paused, reset the FPM module and FBM
* fifos.
*/
ISP_WRITE(isp, BIU2100_CSR, BIU2100_FPM0_REGS);
ISP_WRITE(isp, FPM_DIAG_CONFIG, FPM_SOFT_RESET);
ISP_WRITE(isp, BIU2100_CSR, BIU2100_FB_REGS);
ISP_WRITE(isp, FBM_CMD, FBMCMD_FIFO_RESET_ALL);
ISP_WRITE(isp, BIU2100_CSR, BIU2100_RISC_REGS);
}
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
} else if (IS_1240(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdparam *sdp;
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
btype = "1240";
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
isp->isp_clock = 60;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdp = SDPARAM(isp, 0);
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
sdp->isp_ultramode = 1;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdp = SDPARAM(isp, 1);
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
sdp->isp_ultramode = 1;
/*
* XXX: Should probably do some bus sensing.
*/
Don't try reading the SXP_PINS_DIFF on the 10160 and 12160 SCSI controllers. Reading this register, for which there are indications that it doesn't really exist, returns 0 on at least some 12160 and doing so on Sun Fire V880 causes a data access error exception. Reported and tested by: Beat Gaetzi Approved by: mjacob Obtained from: OpenBSD (modulo setting isp_lvdmode)
2008-12-15 21:42:38 +00:00
} else if (IS_ULTRA3(isp)) {
sdparam *sdp = isp->isp_param;
isp->isp_clock = 100;
if (IS_10160(isp))
btype = "10160";
else if (IS_12160(isp))
btype = "12160";
else
btype = "<UNKLVD>";
sdp->isp_lvdmode = 1;
if (IS_DUALBUS(isp)) {
sdp++;
sdp->isp_lvdmode = 1;
}
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
} else if (IS_ULTRA2(isp)) {
Add in Bill Sommerfelds -Wformat changes. Set up default node && port WWNs correctly (Again!) - this time for the case that we're not going to fully init the adapter if isp_init is called (with ISP_CFG_NOINIT set in options). The pupose for this is to bring the adapter up to almost ready to go, get info out of NVRAM, but to not start it up- leaving it until later to actually start things up if wanted (and possibly with different roles selected).
2000-12-29 19:12:44 +00:00
static const char m[] = "bus %d is in %s Mode";
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
uint16_t l;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdparam *sdp = SDPARAM(isp, 0);
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
isp->isp_clock = 100;
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
if (IS_1280(isp))
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
btype = "1280";
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
else if (IS_1080(isp))
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
btype = "1080";
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
else
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
btype = "<UNKLVD>";
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
l = ISP_READ(isp, SXP_PINS_DIFF) & ISP1080_MODE_MASK;
switch (l) {
case ISP1080_LVD_MODE:
sdp->isp_lvdmode = 1;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGCONFIG, m, 0, "LVD");
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
break;
case ISP1080_HVD_MODE:
sdp->isp_diffmode = 1;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGCONFIG, m, 0, "Differential");
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
break;
case ISP1080_SE_MODE:
sdp->isp_ultramode = 1;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGCONFIG, m, 0, "Single-Ended");
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
break;
default:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGERR,
"unknown mode on bus %d (0x%x)", 0, l);
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
break;
}
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
if (IS_DUALBUS(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdp = SDPARAM(isp, 1);
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
l = ISP_READ(isp, SXP_PINS_DIFF|SXP_BANK1_SELECT);
l &= ISP1080_MODE_MASK;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
switch (l) {
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
case ISP1080_LVD_MODE:
sdp->isp_lvdmode = 1;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGCONFIG, m, 1, "LVD");
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
break;
case ISP1080_HVD_MODE:
sdp->isp_diffmode = 1;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGCONFIG,
m, 1, "Differential");
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
break;
case ISP1080_SE_MODE:
sdp->isp_ultramode = 1;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGCONFIG,
m, 1, "Single-Ended");
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
break;
default:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGERR,
"unknown mode on bus %d (0x%x)", 1, l);
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
break;
}
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdparam *sdp = SDPARAM(isp, 0);
Add some prototype deadchip detection. Set FIFO bursting (1XX0 only- it's already on for the 2XX0) and detect the broken 1040A FIFO. Change bzero to MEMZERO (portability with **nux). Use memcpy for same reason. Finally detect QUEUE FULL conditions and return this as an error that will get cam_periph_error to do it's 'tagged openings now XXX' dance.
1999-01-10 02:55:10 +00:00
i = ISP_READ(isp, BIU_CONF0) & BIU_CONF0_HW_MASK;
switch (i) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
default:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGALL, "Unknown Chip Type 0x%x", i);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
/* FALLTHROUGH */
case 1:
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
btype = "1020";
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp->isp_type = ISP_HA_SCSI_1020;
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
isp->isp_clock = 40;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
case 2:
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
/*
* Some 1020A chips are Ultra Capable, but don't
* run the clock rate up for that unless told to
* do so by the Ultra Capable bits being set.
*/
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
btype = "1020A";
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
isp->isp_type = ISP_HA_SCSI_1020A;
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
isp->isp_clock = 40;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
case 3:
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
btype = "1040";
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
isp->isp_type = ISP_HA_SCSI_1040;
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
isp->isp_clock = 60;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case 4:
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
btype = "1040A";
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp->isp_type = ISP_HA_SCSI_1040A;
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
isp->isp_clock = 60;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
case 5:
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
btype = "1040B";
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp->isp_type = ISP_HA_SCSI_1040B;
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
isp->isp_clock = 60;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
More 2300 support prep- the Request/Response in/out pointers are part of the PCI block for the 2300- not software convention usage of the mailbox registers- so we macrosize in/out pointer usage. Only report that a LIP destroyed commands if it actually destroyed commands. Get the chan/tgt/lun order correct. Fix a longstanding stupid bug that caused us to try and issue a command with a tag on Channel B because we were checking the tagged capability for the target against Channel A. A firmware crash is now vectored out to platform specific code as an async event. Some minor formatting tweaks.
2001-07-04 18:42:41 +00:00
case 6:
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
btype = "1040C";
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 02:28:29 +00:00
isp->isp_type = ISP_HA_SCSI_1040C;
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
isp->isp_clock = 60;
More 2300 support prep- the Request/Response in/out pointers are part of the PCI block for the 2300- not software convention usage of the mailbox registers- so we macrosize in/out pointer usage. Only report that a LIP destroyed commands if it actually destroyed commands. Get the chan/tgt/lun order correct. Fix a longstanding stupid bug that caused us to try and issue a command with a tag on Channel B because we were checking the tagged capability for the target against Channel A. A firmware crash is now vectored out to platform specific code as an async event. Some minor formatting tweaks.
2001-07-04 18:42:41 +00:00
break;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
/*
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
* Now, while we're at it, gather info about ultra
* and/or differential mode.
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
*/
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
if (ISP_READ(isp, SXP_PINS_DIFF) & SXP_PINS_DIFF_MODE) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGCONFIG, "Differential Mode");
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
sdp->isp_diffmode = 1;
Add some prototype deadchip detection. Set FIFO bursting (1XX0 only- it's already on for the 2XX0) and detect the broken 1040A FIFO. Change bzero to MEMZERO (portability with **nux). Use memcpy for same reason. Finally detect QUEUE FULL conditions and return this as an error that will get cam_periph_error to do it's 'tagged openings now XXX' dance.
1999-01-10 02:55:10 +00:00
} else {
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
sdp->isp_diffmode = 0;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
i = ISP_READ(isp, RISC_PSR);
if (isp->isp_bustype == ISP_BT_SBUS) {
i &= RISC_PSR_SBUS_ULTRA;
} else {
i &= RISC_PSR_PCI_ULTRA;
}
if (i != 0) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGCONFIG, "Ultra Mode Capable");
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
sdp->isp_ultramode = 1;
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
/*
Revert previous commit. Violates Maintainer (O'Brien knows how to reach me directly), but more importantly, breaks compiles on non-FreeBSD platforms.
2003-08-25 17:58:23 +00:00
* If we're in Ultra Mode, we have to be 60MHz clock-
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
* even for the SBus version.
*/
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
isp->isp_clock = 60;
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
} else {
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
sdp->isp_ultramode = 0;
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
/*
* Clock is known. Gronk.
*/
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
/*
* Machine dependent clock (if set) overrides
* our generic determinations.
*/
if (isp->isp_mdvec->dv_clock) {
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
if (isp->isp_mdvec->dv_clock < isp->isp_clock) {
isp->isp_clock = isp->isp_mdvec->dv_clock;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Make the Not RESPONSE in RESPONSE QUEUE message have a bit more info (specifically, how many entries we've looked at so far). Maintain interrupt instrumentation. Use USEC_SLEEP instead of USEC_DELAY in a number of places (this allows us to drop locks and sleep instead of spin). Track changes to configuration options for topology preference. Fix botched order of printout for Channel, Target, Lun.
2000-12-02 18:08:35 +00:00
/*
* Clear instrumentation
*/
isp->isp_intcnt = isp->isp_intbogus = 0;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
* Do MD specific pre initialization
*/
ISP_RESET0(isp);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
* Hit the chip over the head with hammer,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* and give it a chance to recover.
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
if (IS_SCSI(isp)) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
ISP_WRITE(isp, BIU_ICR, BIU_ICR_SOFT_RESET);
/*
* A slight delay...
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(100);
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
* Clear data && control DMA engines.
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_WRITE(isp, CDMA_CONTROL, DMA_CNTRL_CLEAR_CHAN | DMA_CNTRL_RESET_INT);
ISP_WRITE(isp, DDMA_CONTROL, DMA_CNTRL_CLEAR_CHAN | DMA_CNTRL_RESET_INT);
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else if (IS_24XX(isp)) {
/*
* Stop DMA and wait for it to stop.
*/
ISP_WRITE(isp, BIU2400_CSR, BIU2400_DMA_STOP|(3 << 4));
for (val = loops = 0; loops < 30000; loops++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(10);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
val = ISP_READ(isp, BIU2400_CSR);
if ((val & BIU2400_DMA_ACTIVE) == 0) {
break;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (val & BIU2400_DMA_ACTIVE) {
Implement ISP_RESET0 for PCI and SBUS attachments- isp_reset has been modified to call ISP_RESET0 if it fails to do a reset. This gives us a chance to disable interrupts.
2006-12-16 05:54:29 +00:00
ISP_RESET0(isp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_prt(isp, ISP_LOGERR, "DMA Failed to Stop on Reset");
return;
}
/*
* Hold it in SOFT_RESET and STOP state for 100us.
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_WRITE(isp, BIU2400_CSR, BIU2400_SOFT_RESET|BIU2400_DMA_STOP|(3 << 4));
ISP_DELAY(100);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
for (loops = 0; loops < 10000; loops++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(5);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
val = ISP_READ(isp, OUTMAILBOX0);
}
for (val = loops = 0; loops < 500000; loops ++) {
val = ISP_READ(isp, BIU2400_CSR);
if ((val & BIU2400_SOFT_RESET) == 0) {
break;
}
}
if (val & BIU2400_SOFT_RESET) {
Implement ISP_RESET0 for PCI and SBUS attachments- isp_reset has been modified to call ISP_RESET0 if it fails to do a reset. This gives us a chance to disable interrupts.
2006-12-16 05:54:29 +00:00
ISP_RESET0(isp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_prt(isp, ISP_LOGERR, "Failed to come out of reset");
return;
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
} else {
ISP_WRITE(isp, BIU2100_CSR, BIU2100_SOFT_RESET);
/*
* A slight delay...
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(100);
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
/*
* Clear data && control DMA engines.
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_WRITE(isp, CDMA2100_CONTROL, DMA_CNTRL2100_CLEAR_CHAN | DMA_CNTRL2100_RESET_INT);
ISP_WRITE(isp, TDMA2100_CONTROL, DMA_CNTRL2100_CLEAR_CHAN | DMA_CNTRL2100_RESET_INT);
ISP_WRITE(isp, RDMA2100_CONTROL, DMA_CNTRL2100_CLEAR_CHAN | DMA_CNTRL2100_RESET_INT);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
/*
* Wait for ISP to be ready to go...
*/
loops = MBOX_DELAY_COUNT;
for (;;) {
More code cleanup. Go back to using FULL_LOGIN Fibre Chan if f/w is less than 1.17.0 level. Change where we do the loop database init. Add in the CMD_RQLATER return. Add some register debounce.
1999-08-16 19:59:55 +00:00
if (IS_SCSI(isp)) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (!(ISP_READ(isp, BIU_ICR) & BIU_ICR_SOFT_RESET)) {
break;
}
} else if (IS_24XX(isp)) {
if (ISP_READ(isp, OUTMAILBOX0) == 0) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
} else {
if (!(ISP_READ(isp, BIU2100_CSR) & BIU2100_SOFT_RESET))
break;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(100);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (--loops < 0) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
ISP_DUMPREGS(isp, "chip reset timed out");
Implement ISP_RESET0 for PCI and SBUS attachments- isp_reset has been modified to call ISP_RESET0 if it fails to do a reset. This gives us a chance to disable interrupts.
2006-12-16 05:54:29 +00:00
ISP_RESET0(isp);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
return;
}
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
* After we've fired this chip up, zero out the conf1 register
* for SCSI adapters and other settings for the 2100.
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
if (IS_SCSI(isp)) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
ISP_WRITE(isp, BIU_CONF1, 0);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else if (!IS_24XX(isp)) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
ISP_WRITE(isp, BIU2100_CSR, 0);
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
/*
* Reset RISC Processor
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
ISP_WRITE(isp, BIU2400_HCCR, HCCR_2400_CMD_RESET);
ISP_WRITE(isp, BIU2400_HCCR, HCCR_2400_CMD_RELEASE);
ISP_WRITE(isp, BIU2400_HCCR, HCCR_2400_CMD_CLEAR_RESET);
} else {
ISP_WRITE(isp, HCCR, HCCR_CMD_RESET);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(100);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_WRITE(isp, BIU_SEMA, 0);
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
Add some prototype deadchip detection. Set FIFO bursting (1XX0 only- it's already on for the 2XX0) and detect the broken 1040A FIFO. Change bzero to MEMZERO (portability with **nux). Use memcpy for same reason. Finally detect QUEUE FULL conditions and return this as an error that will get cam_periph_error to do it's 'tagged openings now XXX' dance.
1999-01-10 02:55:10 +00:00
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* Post-RISC Reset stuff.
Add some prototype deadchip detection. Set FIFO bursting (1XX0 only- it's already on for the 2XX0) and detect the broken 1040A FIFO. Change bzero to MEMZERO (portability with **nux). Use memcpy for same reason. Finally detect QUEUE FULL conditions and return this as an error that will get cam_periph_error to do it's 'tagged openings now XXX' dance.
1999-01-10 02:55:10 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
for (val = loops = 0; loops < 5000000; loops++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(5);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
val = ISP_READ(isp, OUTMAILBOX0);
if (val == 0) {
break;
}
}
if (val != 0) {
Implement ISP_RESET0 for PCI and SBUS attachments- isp_reset has been modified to call ISP_RESET0 if it fails to do a reset. This gives us a chance to disable interrupts.
2006-12-16 05:54:29 +00:00
ISP_RESET0(isp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_prt(isp, ISP_LOGERR, "reset didn't clear");
return;
}
} else if (IS_SCSI(isp)) {
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
uint16_t tmp = isp->isp_mdvec->dv_conf1;
Add some prototype deadchip detection. Set FIFO bursting (1XX0 only- it's already on for the 2XX0) and detect the broken 1040A FIFO. Change bzero to MEMZERO (portability with **nux). Use memcpy for same reason. Finally detect QUEUE FULL conditions and return this as an error that will get cam_periph_error to do it's 'tagged openings now XXX' dance.
1999-01-10 02:55:10 +00:00
/*
* Busted FIFO. Turn off all but burst enables.
*/
if (isp->isp_type == ISP_HA_SCSI_1040A) {
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
tmp &= BIU_BURST_ENABLE;
Add some prototype deadchip detection. Set FIFO bursting (1XX0 only- it's already on for the 2XX0) and detect the broken 1040A FIFO. Change bzero to MEMZERO (portability with **nux). Use memcpy for same reason. Finally detect QUEUE FULL conditions and return this as an error that will get cam_periph_error to do it's 'tagged openings now XXX' dance.
1999-01-10 02:55:10 +00:00
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
ISP_SETBITS(isp, BIU_CONF1, tmp);
if (tmp & BIU_BURST_ENABLE) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
ISP_SETBITS(isp, CDMA_CONF, DMA_ENABLE_BURST);
ISP_SETBITS(isp, DDMA_CONF, DMA_ENABLE_BURST);
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (SDPARAM(isp, 0)->isp_ptisp) {
if (SDPARAM(isp, 0)->isp_ultramode) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
while (ISP_READ(isp, RISC_MTR) != 0x1313) {
ISP_WRITE(isp, RISC_MTR, 0x1313);
ISP_WRITE(isp, HCCR, HCCR_CMD_STEP);
}
} else {
ISP_WRITE(isp, RISC_MTR, 0x1212);
}
/*
* PTI specific register
*/
ISP_WRITE(isp, RISC_EMB, DUAL_BANK);
} else {
ISP_WRITE(isp, RISC_MTR, 0x1212);
}
ISP_WRITE(isp, HCCR, HCCR_CMD_RELEASE);
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
} else {
ISP_WRITE(isp, RISC_MTR2100, 0x1212);
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
if (IS_2200(isp) || IS_23XX(isp)) {
When resetting the Qlogic 2X00 units, reset the FPM (Fibre Protocol Module) and FBM (Fibre Buffer Modules). Also remember to clear the semaphore registers. Tell the RISC processor to not halt on FPM parity errors. Throw out the ISP_CFG_NOINIT silliness and instead go to the use of adapter 'roles' to see whether one completes initialization or not (mostly for Fibre Channel). The ultimate intent, btw, of all of this is to have a warm standby adapter for failover reasons. Because we do roles now, setting of Target Capable Class 3 service parameters in the ICB for the 2x00 cards reflects from role. Also, in isp_start, if we're not supporting an initiator role, we bounce outgoing commands with a Selection Timeout error. Also clean out the TOGGLE_TMODE goop for FC- there is no toggling of target mode like there is for parallel SCSI cards. Do more cleanup with respect to using target ids 0..125 in F-port topologies. Also keep track of things which *were* fabric devices so that when you rescan the fabric you can notify the outer layers when fabric devices go away. Only force a LOGOUT for fabric devices if they're still logged in (i.e., you cat their Port Database entry. Clean up the Get All Next scanning. Finally, use a new tag in the softc to store the opcode for the last mailbox command used so we can report which opcode timed out.
2001-01-15 18:33:08 +00:00
ISP_WRITE(isp, HCCR, HCCR_2X00_DISABLE_PARITY_PAUSE);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_WRITE(isp, HCCR, HCCR_CMD_RELEASE);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_WRITE(isp, isp->isp_rqstinrp, 0);
ISP_WRITE(isp, isp->isp_rqstoutrp, 0);
ISP_WRITE(isp, isp->isp_respinrp, 0);
ISP_WRITE(isp, isp->isp_respoutrp, 0);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (IS_24XX(isp)) {
ISP_WRITE(isp, BIU2400_PRI_REQINP, 0);
ISP_WRITE(isp, BIU2400_PRI_REQOUTP, 0);
ISP_WRITE(isp, BIU2400_ATIO_RSPINP, 0);
ISP_WRITE(isp, BIU2400_ATIO_RSPOUTP, 0);
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
* Do MD specific post initialization
*/
ISP_RESET1(isp);
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
/*
Misunderstanding documentation caused me to try and set 1Gbps/2Gps/Auto connection speed for the 2300 in the wrong offset in the ICB. Oops. Respect some QLogic errat wrt PCI errors on certain shared host/RISC registers.
2001-10-06 20:41:18 +00:00
* Wait for everything to finish firing up.
*
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* Avoid doing this on early 2312s because you can generate a PCI
Misunderstanding documentation caused me to try and set 1Gbps/2Gps/Auto connection speed for the 2300 in the wrong offset in the ICB. Oops. Respect some QLogic errat wrt PCI errors on certain shared host/RISC registers.
2001-10-06 20:41:18 +00:00
* parity error (chip breakage).
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (IS_2312(isp) && isp->isp_revision < 2) {
ISP_DELAY(100);
Misunderstanding documentation caused me to try and set 1Gbps/2Gps/Auto connection speed for the 2300 in the wrong offset in the ICB. Oops. Respect some QLogic errat wrt PCI errors on certain shared host/RISC registers.
2001-10-06 20:41:18 +00:00
} else {
loops = MBOX_DELAY_COUNT;
while (ISP_READ(isp, OUTMAILBOX0) == MBOX_BUSY) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(100);
Misunderstanding documentation caused me to try and set 1Gbps/2Gps/Auto connection speed for the 2300 in the wrong offset in the ICB. Oops. Respect some QLogic errat wrt PCI errors on certain shared host/RISC registers.
2001-10-06 20:41:18 +00:00
if (--loops < 0) {
Implement ISP_RESET0 for PCI and SBUS attachments- isp_reset has been modified to call ISP_RESET0 if it fails to do a reset. This gives us a chance to disable interrupts.
2006-12-16 05:54:29 +00:00
ISP_RESET0(isp);
Misunderstanding documentation caused me to try and set 1Gbps/2Gps/Auto connection speed for the 2300 in the wrong offset in the ICB. Oops. Respect some QLogic errat wrt PCI errors on certain shared host/RISC registers.
2001-10-06 20:41:18 +00:00
isp_prt(isp, ISP_LOGERR,
"MBOX_BUSY never cleared on reset");
return;
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
}
}
/*
* Up until this point we've done everything by just reading or
* setting registers. From this point on we rely on at least *some*
* kind of firmware running in the card.
*/
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* Do some sanity checking by running a NOP command.
* If it succeeds, the ROM firmware is now running.
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(&mbs, sizeof (mbs));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
mbs.param[0] = MBOX_NO_OP;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.logval = MBLOGALL;
isp_mboxcmd(isp, &mbs);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
isp_prt(isp, ISP_LOGERR, "NOP command failed (%x)", mbs.param[0]);
Implement ISP_RESET0 for PCI and SBUS attachments- isp_reset has been modified to call ISP_RESET0 if it fails to do a reset. This gives us a chance to disable interrupts.
2006-12-16 05:54:29 +00:00
ISP_RESET0(isp);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
return;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* Do some operational tests
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_SCSI(isp) || IS_24XX(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(&mbs, sizeof (mbs));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
mbs.param[0] = MBOX_MAILBOX_REG_TEST;
mbs.param[1] = 0xdead;
mbs.param[2] = 0xbeef;
mbs.param[3] = 0xffff;
mbs.param[4] = 0x1111;
mbs.param[5] = 0xa5a5;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.param[6] = 0x0000;
mbs.param[7] = 0x0000;
mbs.logval = MBLOGALL;
isp_mboxcmd(isp, &mbs);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Implement ISP_RESET0 for PCI and SBUS attachments- isp_reset has been modified to call ISP_RESET0 if it fails to do a reset. This gives us a chance to disable interrupts.
2006-12-16 05:54:29 +00:00
ISP_RESET0(isp);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
return;
}
if (mbs.param[1] != 0xdead || mbs.param[2] != 0xbeef ||
mbs.param[3] != 0xffff || mbs.param[4] != 0x1111 ||
mbs.param[5] != 0xa5a5) {
Implement ISP_RESET0 for PCI and SBUS attachments- isp_reset has been modified to call ISP_RESET0 if it fails to do a reset. This gives us a chance to disable interrupts.
2006-12-16 05:54:29 +00:00
ISP_RESET0(isp);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGERR, "Register Test Failed (0x%x 0x%x 0x%x 0x%x 0x%x)", mbs.param[1], mbs.param[2], mbs.param[3], mbs.param[4], mbs.param[5]);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
return;
}
}
/*
* Download new Firmware, unless requested not to do so.
* This is made slightly trickier in some cases where the
* firmware of the ROM revision is newer than the revision
* compiled into the driver. So, where we used to compare
* versions of our f/w and the ROM f/w, now we just see
* whether we have f/w at all and whether a config flag
* has disabled our download.
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if ((isp->isp_mdvec->dv_ispfw == NULL) || (isp->isp_confopts & ISP_CFG_NORELOAD)) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
dodnld = 0;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
code_org = ISP_CODE_ORG_2400;
} else if (IS_23XX(isp)) {
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
code_org = ISP_CODE_ORG_2300;
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
} else {
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
code_org = ISP_CODE_ORG;
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (dodnld && IS_24XX(isp)) {
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
const uint32_t *ptr = isp->isp_mdvec->dv_ispfw;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
* Keep loading until we run out of f/w.
*/
code_org = ptr[2]; /* 1st load address is our start addr */
for (;;) {
uint32_t la, wi, wl;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGDEBUG0, "load 0x%x words of code at load address 0x%x", ptr[3], ptr[2]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
wi = 0;
la = ptr[2];
wl = ptr[3];
while (wi < ptr[3]) {
uint32_t *cp;
uint32_t nw;
nw = ISP_QUEUE_SIZE(RQUEST_QUEUE_LEN(isp)) >> 2;
if (nw > wl) {
nw = wl;
}
cp = isp->isp_rquest;
for (i = 0; i < nw; i++) {
If we're going to (for 23XX and 24XX cards) DMA firmware from the request queues rather than shove it down a word at a time, we have to remember to put it into little endian format. Use the macros ISP_IOXPUT_{16,32} for this purpose. Otherwise, on sparc the firmware is loaded garbled and we get a (not surprisingly) firmware checksum failure and the card won't start and we don't attach it. Approved by: re (bruce) MFC after: 3 days
2007-06-24 01:41:16 +00:00
ISP_IOXPUT_32(isp, ptr[wi++], &cp[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
wl--;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MEMORYBARRIER(isp, SYNC_REQUEST, 0, ISP_QUEUE_SIZE(RQUEST_QUEUE_LEN(isp)));
ISP_MEMZERO(&mbs, sizeof (mbs));
if (la < 0x10000 && nw < 0x10000) {
mbs.param[0] = MBOX_LOAD_RISC_RAM_2100;
mbs.param[1] = la;
mbs.param[2] = DMA_WD1(isp->isp_rquest_dma);
mbs.param[3] = DMA_WD0(isp->isp_rquest_dma);
mbs.param[4] = nw;
mbs.param[6] = DMA_WD3(isp->isp_rquest_dma);
mbs.param[7] = DMA_WD2(isp->isp_rquest_dma);
} else {
mbs.param[0] = MBOX_LOAD_RISC_RAM;
mbs.param[1] = la;
mbs.param[2] = DMA_WD1(isp->isp_rquest_dma);
mbs.param[3] = DMA_WD0(isp->isp_rquest_dma);
mbs.param[4] = nw >> 16;
mbs.param[5] = nw;
mbs.param[6] = DMA_WD3(isp->isp_rquest_dma);
mbs.param[7] = DMA_WD2(isp->isp_rquest_dma);
mbs.param[8] = la >> 16;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.logval = MBLOGALL;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGERR, "F/W Risc Ram Load Failed");
Implement ISP_RESET0 for PCI and SBUS attachments- isp_reset has been modified to call ISP_RESET0 if it fails to do a reset. This gives us a chance to disable interrupts.
2006-12-16 05:54:29 +00:00
ISP_RESET0(isp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return;
}
la += nw;
}
if (ptr[1] == 0) {
break;
}
ptr += ptr[3];
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp->isp_loaded_fw = 1;
} else if (dodnld && IS_23XX(isp)) {
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
const uint16_t *ptr = isp->isp_mdvec->dv_ispfw;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
uint16_t wi, wl, segno;
uint32_t la;
la = code_org;
segno = 0;
for (;;) {
uint32_t nxtaddr;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGDEBUG0, "load 0x%x words of code at load address 0x%x", ptr[3], la);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
wi = 0;
wl = ptr[3];
while (wi < ptr[3]) {
uint16_t *cp;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
uint16_t nw;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
nw = ISP_QUEUE_SIZE(RQUEST_QUEUE_LEN(isp)) >> 1;
if (nw > wl) {
nw = wl;
}
if (nw > (1 << 15)) {
nw = 1 << 15;
}
cp = isp->isp_rquest;
for (i = 0; i < nw; i++) {
If we're going to (for 23XX and 24XX cards) DMA firmware from the request queues rather than shove it down a word at a time, we have to remember to put it into little endian format. Use the macros ISP_IOXPUT_{16,32} for this purpose. Otherwise, on sparc the firmware is loaded garbled and we get a (not surprisingly) firmware checksum failure and the card won't start and we don't attach it. Approved by: re (bruce) MFC after: 3 days
2007-06-24 01:41:16 +00:00
ISP_IOXPUT_16(isp, ptr[wi++], &cp[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
wl--;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MEMORYBARRIER(isp, SYNC_REQUEST, 0, ISP_QUEUE_SIZE(RQUEST_QUEUE_LEN(isp)));
ISP_MEMZERO(&mbs, sizeof (mbs));
if (la < 0x10000) {
mbs.param[0] = MBOX_LOAD_RISC_RAM_2100;
mbs.param[1] = la;
mbs.param[2] = DMA_WD1(isp->isp_rquest_dma);
mbs.param[3] = DMA_WD0(isp->isp_rquest_dma);
mbs.param[4] = nw;
mbs.param[6] = DMA_WD3(isp->isp_rquest_dma);
mbs.param[7] = DMA_WD2(isp->isp_rquest_dma);
} else {
mbs.param[0] = MBOX_LOAD_RISC_RAM;
mbs.param[1] = la;
mbs.param[2] = DMA_WD1(isp->isp_rquest_dma);
mbs.param[3] = DMA_WD0(isp->isp_rquest_dma);
mbs.param[4] = nw;
mbs.param[6] = DMA_WD3(isp->isp_rquest_dma);
mbs.param[7] = DMA_WD2(isp->isp_rquest_dma);
mbs.param[8] = la >> 16;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.logval = MBLOGALL;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGERR, "F/W Risc Ram Load Failed");
Implement ISP_RESET0 for PCI and SBUS attachments- isp_reset has been modified to call ISP_RESET0 if it fails to do a reset. This gives us a chance to disable interrupts.
2006-12-16 05:54:29 +00:00
ISP_RESET0(isp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return;
}
la += nw;
}
if (!IS_2322(isp)) {
break;
}
if (++segno == 3) {
break;
}
/*
* If we're a 2322, the firmware actually comes in
* three chunks. We loaded the first at the code_org
* address. The other two chunks, which follow right
* after each other in memory here, get loaded at
* addresses specfied at offset 0x9..0xB.
*/
nxtaddr = ptr[3];
ptr = &ptr[nxtaddr];
la = ptr[5] | ((ptr[4] & 0x3f) << 16);
}
isp->isp_loaded_fw = 1;
} else if (dodnld) {
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
union {
const uint16_t *cp;
uint16_t *np;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
} ucd;
ucd.cp = isp->isp_mdvec->dv_ispfw;
isp->isp_mbxworkp = &ucd.np[1];
isp->isp_mbxwrk0 = ucd.np[3] - 1;
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
isp->isp_mbxwrk1 = code_org + 1;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(&mbs, sizeof (mbs));
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
mbs.param[0] = MBOX_WRITE_RAM_WORD;
mbs.param[1] = code_org;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
mbs.param[2] = ucd.np[0];
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.logval = MBLOGNONE;
isp_mboxcmd(isp, &mbs);
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGERR, "F/W download failed at word %d", isp->isp_mbxwrk1 - code_org);
Implement ISP_RESET0 for PCI and SBUS attachments- isp_reset has been modified to call ISP_RESET0 if it fails to do a reset. This gives us a chance to disable interrupts.
2006-12-16 05:54:29 +00:00
ISP_RESET0(isp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
} else {
isp->isp_loaded_fw = 0;
isp_prt(isp, ISP_LOGDEBUG2, "skipping f/w download");
}
/*
* If we loaded firmware, verify its checksum
*/
if (isp->isp_loaded_fw) {
ISP_MEMZERO(&mbs, sizeof (mbs));
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.param[0] = MBOX_VERIFY_CHECKSUM;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (IS_24XX(isp)) {
mbs.param[1] = code_org >> 16;
mbs.param[2] = code_org;
} else {
mbs.param[1] = code_org;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
isp_prt(isp, ISP_LOGERR, dcrc);
Implement ISP_RESET0 for PCI and SBUS attachments- isp_reset has been modified to call ISP_RESET0 if it fails to do a reset. This gives us a chance to disable interrupts.
2006-12-16 05:54:29 +00:00
ISP_RESET0(isp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return;
Fix 2KLOGIN code to specify *ibits* (not *obits*) so that the options field in register 10 will be deterministic, not random. Correct the number of input bits for EXECUTE_FIRMWARE 0..1 to 0..2- the 2322 and 24XX cards use mailbox register 2 to specify whether the f/w being executed is freshly loaded or not. Correct the number of input bits for {READ,WRITE}_RAM_WORD_EXTENDED so that register 8 gets picked up. Fix the indexing and offset for the 2322 f/w download so that it correctly puts the different code segments where they belong. Move VERIFY_CHECKSUM to be the 'else' clause to 2322 f/w downloads- the EXECUTE_FIRMWARE command for 2322 and 24XX cards will tell you if the f/w checksum is incorrect and VERIFY_CHECKSUM only works for RISC SRAM address < 64K so you can only do a VERIFY_CHECKSUM on the first of the 3 f/w segments for the 2322. Shorten the delay for the continuation mailbox commands- 1ms is ridiculous (100us is more likely). All of the more or less is really only for the 2322/6322 cards.
2006-08-14 05:42:46 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
/*
* Now start it rolling.
*
* If we didn't actually download f/w,
* we still need to (re)start it.
*/
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_EXEC_FIRMWARE, MBLOGALL, 1000000);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
mbs.param[1] = code_org >> 16;
mbs.param[2] = code_org;
if (isp->isp_loaded_fw) {
mbs.param[3] = 0;
} else {
mbs.param[3] = 1;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (IS_25XX(isp)) {
mbs.ibits |= 0x10;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else if (IS_2322(isp)) {
mbs.param[1] = code_org;
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
if (isp->isp_loaded_fw) {
mbs.param[2] = 0;
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
} else {
mbs.param[2] = 1;
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else {
mbs.param[1] = code_org;
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Fix 2KLOGIN code to specify *ibits* (not *obits*) so that the options field in register 10 will be deterministic, not random. Correct the number of input bits for EXECUTE_FIRMWARE 0..1 to 0..2- the 2322 and 24XX cards use mailbox register 2 to specify whether the f/w being executed is freshly loaded or not. Correct the number of input bits for {READ,WRITE}_RAM_WORD_EXTENDED so that register 8 gets picked up. Fix the indexing and offset for the 2322 f/w download so that it correctly puts the different code segments where they belong. Move VERIFY_CHECKSUM to be the 'else' clause to 2322 f/w downloads- the EXECUTE_FIRMWARE command for 2322 and 24XX cards will tell you if the f/w checksum is incorrect and VERIFY_CHECKSUM only works for RISC SRAM address < 64K so you can only do a VERIFY_CHECKSUM on the first of the 3 f/w segments for the 2322. Shorten the delay for the continuation mailbox commands- 1ms is ridiculous (100us is more likely). All of the more or less is really only for the 2322/6322 cards.
2006-08-14 05:42:46 +00:00
if (IS_2322(isp) || IS_24XX(isp)) {
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Implement ISP_RESET0 for PCI and SBUS attachments- isp_reset has been modified to call ISP_RESET0 if it fails to do a reset. This gives us a chance to disable interrupts.
2006-12-16 05:54:29 +00:00
ISP_RESET0(isp);
Fix 2KLOGIN code to specify *ibits* (not *obits*) so that the options field in register 10 will be deterministic, not random. Correct the number of input bits for EXECUTE_FIRMWARE 0..1 to 0..2- the 2322 and 24XX cards use mailbox register 2 to specify whether the f/w being executed is freshly loaded or not. Correct the number of input bits for {READ,WRITE}_RAM_WORD_EXTENDED so that register 8 gets picked up. Fix the indexing and offset for the 2322 f/w download so that it correctly puts the different code segments where they belong. Move VERIFY_CHECKSUM to be the 'else' clause to 2322 f/w downloads- the EXECUTE_FIRMWARE command for 2322 and 24XX cards will tell you if the f/w checksum is incorrect and VERIFY_CHECKSUM only works for RISC SRAM address < 64K so you can only do a VERIFY_CHECKSUM on the first of the 3 f/w segments for the 2322. Shorten the delay for the continuation mailbox commands- 1ms is ridiculous (100us is more likely). All of the more or less is really only for the 2322/6322 cards.
2006-08-14 05:42:46 +00:00
return;
}
}
Some rearrangement of headers to minimize diffs with outside of FreeBSD repository and to clean up the license header so as to not pollute the license with file function. Zero all mailbox structures prior to use (just in case). Change the outgoing mailbox count for INIT_FIRMWARE to be correct.
2006-07-16 20:11:50 +00:00
If the HBA is already 'touched', still set maxluns. Othewise for CAM_QUIRK_HILUN devices we loop thru 32bits of lun. Oops. Switch to using USEC_DELAY rather than USEC_SLEEP at isp_reset time. Try to paper around a defect in clients that don't correctly registers themeselves with the fabric nameserver. Minor updates for Mirapoint support- they still use code that is not HANDLE_LOOPSTATE_IN_OUTER_LAYERS, and, surprise surprise, this old stuff had some bugs in it. Clean up some target mode stuff. MFC after: 1 week
2002-06-16 05:18:22 +00:00
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* Give it a chance to finish starting up.
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* Give the 24XX more time.
If the HBA is already 'touched', still set maxluns. Othewise for CAM_QUIRK_HILUN devices we loop thru 32bits of lun. Oops. Switch to using USEC_DELAY rather than USEC_SLEEP at isp_reset time. Try to paper around a defect in clients that don't correctly registers themeselves with the fabric nameserver. Minor updates for Mirapoint support- they still use code that is not HANDLE_LOOPSTATE_IN_OUTER_LAYERS, and, surprise surprise, this old stuff had some bugs in it. Clean up some target mode stuff. MFC after: 1 week
2002-06-16 05:18:22 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (IS_24XX(isp)) {
ISP_DELAY(500000);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* Check to see if the 24XX firmware really started.
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (mbs.param[1] == 0xdead) {
isp_prt(isp, ISP_LOGERR, "f/w didn't *really* start");
ISP_RESET0(isp);
return;
}
} else {
ISP_DELAY(250000);
if (IS_SCSI(isp)) {
/*
* Set CLOCK RATE, but only if asked to.
*/
if (isp->isp_clock) {
mbs.param[0] = MBOX_SET_CLOCK_RATE;
mbs.param[1] = isp->isp_clock;
mbs.logval = MBLOGNONE;
isp_mboxcmd(isp, &mbs);
/* we will try not to care if this fails */
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
}
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* Ask the chip for the current firmware version.
* This should prove that the new firmware is working.
*/
MBSINIT(&mbs, MBOX_ABOUT_FIRMWARE, MBLOGALL, 0);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Implement ISP_RESET0 for PCI and SBUS attachments- isp_reset has been modified to call ISP_RESET0 if it fails to do a reset. This gives us a chance to disable interrupts.
2006-12-16 05:54:29 +00:00
ISP_RESET0(isp);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
return;
}
Clarify issues about whether we have SCCLUN (65535 luns) or non-SCCLUN (16 luns) firmware for the Fibre Channel cards. We used to assume that if we didn't download firmware, we couldn't know what the firmware capability with respect to SCCLUNs is- and it's important because the lun field changes in the request queue entry based upon which firmware it is. At any rate, we *do* get back firmware attributes in mailbox register 6 when we do ABOUT FIRMWARE for all 2200/2300 cards- and for 2100 cards with at least 1.17.0 firmware. So- we now assume non-SCCLUN behaviour for 2100 cards with firmware < 1.17.0- and we check the firmware attributes for other cards (loaded firmware or not). This also allows us to get rid of the crappy test of isp_maxluns > 16- we simply can check firmware attributes for SCCLUN behaviour. This required an 'oops' fix to the outgoing mailbox count field for ABOUT FIRMWARE for FC cards. Also- while here, hardwire firmware revisions for loaded code for SBus cards. Apparently the 1.35 or 1.37 f/w we've been loading into isp1000 just doesn't report firmware revisions out to mailbox regs 1, 2 and 3 like everyone else. Grumble. Not that this fix hardly matters for FreeBSD. MFC after: 4 weeks
2001-09-03 03:09:48 +00:00
/*
* The SBus firmware that we are using apparently does not return
* major, minor, micro revisions in the mailbox registers, which
* is really, really, annoying.
*/
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
if (ISP_SBUS_SUPPORTED && isp->isp_bustype == ISP_BT_SBUS) {
Clarify issues about whether we have SCCLUN (65535 luns) or non-SCCLUN (16 luns) firmware for the Fibre Channel cards. We used to assume that if we didn't download firmware, we couldn't know what the firmware capability with respect to SCCLUNs is- and it's important because the lun field changes in the request queue entry based upon which firmware it is. At any rate, we *do* get back firmware attributes in mailbox register 6 when we do ABOUT FIRMWARE for all 2200/2300 cards- and for 2100 cards with at least 1.17.0 firmware. So- we now assume non-SCCLUN behaviour for 2100 cards with firmware < 1.17.0- and we check the firmware attributes for other cards (loaded firmware or not). This also allows us to get rid of the crappy test of isp_maxluns > 16- we simply can check firmware attributes for SCCLUN behaviour. This required an 'oops' fix to the outgoing mailbox count field for ABOUT FIRMWARE for FC cards. Also- while here, hardwire firmware revisions for loaded code for SBus cards. Apparently the 1.35 or 1.37 f/w we've been loading into isp1000 just doesn't report firmware revisions out to mailbox regs 1, 2 and 3 like everyone else. Grumble. Not that this fix hardly matters for FreeBSD. MFC after: 4 weeks
2001-09-03 03:09:48 +00:00
if (dodnld) {
#ifdef ISP_TARGET_MODE
isp->isp_fwrev[0] = 7;
isp->isp_fwrev[1] = 55;
#else
isp->isp_fwrev[0] = 1;
isp->isp_fwrev[1] = 37;
#endif
isp->isp_fwrev[2] = 0;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Clarify issues about whether we have SCCLUN (65535 luns) or non-SCCLUN (16 luns) firmware for the Fibre Channel cards. We used to assume that if we didn't download firmware, we couldn't know what the firmware capability with respect to SCCLUNs is- and it's important because the lun field changes in the request queue entry based upon which firmware it is. At any rate, we *do* get back firmware attributes in mailbox register 6 when we do ABOUT FIRMWARE for all 2200/2300 cards- and for 2100 cards with at least 1.17.0 firmware. So- we now assume non-SCCLUN behaviour for 2100 cards with firmware < 1.17.0- and we check the firmware attributes for other cards (loaded firmware or not). This also allows us to get rid of the crappy test of isp_maxluns > 16- we simply can check firmware attributes for SCCLUN behaviour. This required an 'oops' fix to the outgoing mailbox count field for ABOUT FIRMWARE for FC cards. Also- while here, hardwire firmware revisions for loaded code for SBus cards. Apparently the 1.35 or 1.37 f/w we've been loading into isp1000 just doesn't report firmware revisions out to mailbox regs 1, 2 and 3 like everyone else. Grumble. Not that this fix hardly matters for FreeBSD. MFC after: 4 weeks
2001-09-03 03:09:48 +00:00
} else {
isp->isp_fwrev[0] = mbs.param[1];
isp->isp_fwrev[1] = mbs.param[2];
isp->isp_fwrev[2] = mbs.param[3];
}
Some rearrangement of headers to minimize diffs with outside of FreeBSD repository and to clean up the license header so as to not pollute the license with file function. Zero all mailbox structures prior to use (just in case). Change the outgoing mailbox count for INIT_FIRMWARE to be correct.
2006-07-16 20:11:50 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGCONFIG, "Board Type %s, Chip Revision 0x%x, %s F/W Revision %d.%d.%d",
btype, isp->isp_revision, dodnld? "loaded" : "resident", isp->isp_fwrev[0], isp->isp_fwrev[1], isp->isp_fwrev[2]);
Clarify issues about whether we have SCCLUN (65535 luns) or non-SCCLUN (16 luns) firmware for the Fibre Channel cards. We used to assume that if we didn't download firmware, we couldn't know what the firmware capability with respect to SCCLUNs is- and it's important because the lun field changes in the request queue entry based upon which firmware it is. At any rate, we *do* get back firmware attributes in mailbox register 6 when we do ABOUT FIRMWARE for all 2200/2300 cards- and for 2100 cards with at least 1.17.0 firmware. So- we now assume non-SCCLUN behaviour for 2100 cards with firmware < 1.17.0- and we check the firmware attributes for other cards (loaded firmware or not). This also allows us to get rid of the crappy test of isp_maxluns > 16- we simply can check firmware attributes for SCCLUN behaviour. This required an 'oops' fix to the outgoing mailbox count field for ABOUT FIRMWARE for FC cards. Also- while here, hardwire firmware revisions for loaded code for SBus cards. Apparently the 1.35 or 1.37 f/w we've been loading into isp1000 just doesn't report firmware revisions out to mailbox regs 1, 2 and 3 like everyone else. Grumble. Not that this fix hardly matters for FreeBSD. MFC after: 4 weeks
2001-09-03 03:09:48 +00:00
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
if (IS_FC(isp)) {
Clarify issues about whether we have SCCLUN (65535 luns) or non-SCCLUN (16 luns) firmware for the Fibre Channel cards. We used to assume that if we didn't download firmware, we couldn't know what the firmware capability with respect to SCCLUNs is- and it's important because the lun field changes in the request queue entry based upon which firmware it is. At any rate, we *do* get back firmware attributes in mailbox register 6 when we do ABOUT FIRMWARE for all 2200/2300 cards- and for 2100 cards with at least 1.17.0 firmware. So- we now assume non-SCCLUN behaviour for 2100 cards with firmware < 1.17.0- and we check the firmware attributes for other cards (loaded firmware or not). This also allows us to get rid of the crappy test of isp_maxluns > 16- we simply can check firmware attributes for SCCLUN behaviour. This required an 'oops' fix to the outgoing mailbox count field for ABOUT FIRMWARE for FC cards. Also- while here, hardwire firmware revisions for loaded code for SBus cards. Apparently the 1.35 or 1.37 f/w we've been loading into isp1000 just doesn't report firmware revisions out to mailbox regs 1, 2 and 3 like everyone else. Grumble. Not that this fix hardly matters for FreeBSD. MFC after: 4 weeks
2001-09-03 03:09:48 +00:00
/*
* We do not believe firmware attributes for 2100 code less
If we're using ancient (pre 1.17.0) 2100 f/w (for the cards that cannot load f/w images > 0x7fff words), set ISP_FW_ATTR_SCCLUN. We explicitly don't believe we can find attributes if f/w is < 1.17.0, so we have to set SCCLUN for the 1.15.37 f/w we're using manually- otherwise every target will replicate itself across all 16 supported luns for non-SCCLUN f/w. Correctly set things up for 23XX and either fast posting or ZIO. The 23XX, it turns out, does not support RIO. If you put a non-zero value in xfwoptions, this will disable fast posting. If you put ICBXOPT_ZIO in xfwoptions, then the 23XX will do interrupt delays but post to the response queue- apparently QLogic *now* believes that reading multiple handles from registers is less of a win than writing (and delaying) multiple 64 byte responses to the response queue. At the end of taking a a good f/w crash dump, send the ISPASYNC_FW_DUMPED event to the outer layers (who can then do things like wake a user daemon to *fetch* the crash image, etc.).
2002-08-17 17:29:15 +00:00
* than 1.17.0, unless it's the firmware we specifically
* are loading.
*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* Note that all 22XX and later f/w is greater than 1.X.0.
Clarify issues about whether we have SCCLUN (65535 luns) or non-SCCLUN (16 luns) firmware for the Fibre Channel cards. We used to assume that if we didn't download firmware, we couldn't know what the firmware capability with respect to SCCLUNs is- and it's important because the lun field changes in the request queue entry based upon which firmware it is. At any rate, we *do* get back firmware attributes in mailbox register 6 when we do ABOUT FIRMWARE for all 2200/2300 cards- and for 2100 cards with at least 1.17.0 firmware. So- we now assume non-SCCLUN behaviour for 2100 cards with firmware < 1.17.0- and we check the firmware attributes for other cards (loaded firmware or not). This also allows us to get rid of the crappy test of isp_maxluns > 16- we simply can check firmware attributes for SCCLUN behaviour. This required an 'oops' fix to the outgoing mailbox count field for ABOUT FIRMWARE for FC cards. Also- while here, hardwire firmware revisions for loaded code for SBus cards. Apparently the 1.35 or 1.37 f/w we've been loading into isp1000 just doesn't report firmware revisions out to mailbox regs 1, 2 and 3 like everyone else. Grumble. Not that this fix hardly matters for FreeBSD. MFC after: 4 weeks
2001-09-03 03:09:48 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if ((ISP_FW_OLDER_THAN(isp, 1, 17, 1))) {
If we're using ancient (pre 1.17.0) 2100 f/w (for the cards that cannot load f/w images > 0x7fff words), set ISP_FW_ATTR_SCCLUN. We explicitly don't believe we can find attributes if f/w is < 1.17.0, so we have to set SCCLUN for the 1.15.37 f/w we're using manually- otherwise every target will replicate itself across all 16 supported luns for non-SCCLUN f/w. Correctly set things up for 23XX and either fast posting or ZIO. The 23XX, it turns out, does not support RIO. If you put a non-zero value in xfwoptions, this will disable fast posting. If you put ICBXOPT_ZIO in xfwoptions, then the 23XX will do interrupt delays but post to the response queue- apparently QLogic *now* believes that reading multiple handles from registers is less of a win than writing (and delaying) multiple 64 byte responses to the response queue. At the end of taking a a good f/w crash dump, send the ISPASYNC_FW_DUMPED event to the outer layers (who can then do things like wake a user daemon to *fetch* the crash image, etc.).
2002-08-17 17:29:15 +00:00
#ifdef USE_SMALLER_2100_FIRMWARE
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp->isp_fwattr = ISP_FW_ATTR_SCCLUN;
If we're using ancient (pre 1.17.0) 2100 f/w (for the cards that cannot load f/w images > 0x7fff words), set ISP_FW_ATTR_SCCLUN. We explicitly don't believe we can find attributes if f/w is < 1.17.0, so we have to set SCCLUN for the 1.15.37 f/w we're using manually- otherwise every target will replicate itself across all 16 supported luns for non-SCCLUN f/w. Correctly set things up for 23XX and either fast posting or ZIO. The 23XX, it turns out, does not support RIO. If you put a non-zero value in xfwoptions, this will disable fast posting. If you put ICBXOPT_ZIO in xfwoptions, then the 23XX will do interrupt delays but post to the response queue- apparently QLogic *now* believes that reading multiple handles from registers is less of a win than writing (and delaying) multiple 64 byte responses to the response queue. At the end of taking a a good f/w crash dump, send the ISPASYNC_FW_DUMPED event to the outer layers (who can then do things like wake a user daemon to *fetch* the crash image, etc.).
2002-08-17 17:29:15 +00:00
#else
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp->isp_fwattr = 0;
If we're using ancient (pre 1.17.0) 2100 f/w (for the cards that cannot load f/w images > 0x7fff words), set ISP_FW_ATTR_SCCLUN. We explicitly don't believe we can find attributes if f/w is < 1.17.0, so we have to set SCCLUN for the 1.15.37 f/w we're using manually- otherwise every target will replicate itself across all 16 supported luns for non-SCCLUN f/w. Correctly set things up for 23XX and either fast posting or ZIO. The 23XX, it turns out, does not support RIO. If you put a non-zero value in xfwoptions, this will disable fast posting. If you put ICBXOPT_ZIO in xfwoptions, then the 23XX will do interrupt delays but post to the response queue- apparently QLogic *now* believes that reading multiple handles from registers is less of a win than writing (and delaying) multiple 64 byte responses to the response queue. At the end of taking a a good f/w crash dump, send the ISPASYNC_FW_DUMPED event to the outer layers (who can then do things like wake a user daemon to *fetch* the crash image, etc.).
2002-08-17 17:29:15 +00:00
#endif
Clarify issues about whether we have SCCLUN (65535 luns) or non-SCCLUN (16 luns) firmware for the Fibre Channel cards. We used to assume that if we didn't download firmware, we couldn't know what the firmware capability with respect to SCCLUNs is- and it's important because the lun field changes in the request queue entry based upon which firmware it is. At any rate, we *do* get back firmware attributes in mailbox register 6 when we do ABOUT FIRMWARE for all 2200/2300 cards- and for 2100 cards with at least 1.17.0 firmware. So- we now assume non-SCCLUN behaviour for 2100 cards with firmware < 1.17.0- and we check the firmware attributes for other cards (loaded firmware or not). This also allows us to get rid of the crappy test of isp_maxluns > 16- we simply can check firmware attributes for SCCLUN behaviour. This required an 'oops' fix to the outgoing mailbox count field for ABOUT FIRMWARE for FC cards. Also- while here, hardwire firmware revisions for loaded code for SBus cards. Apparently the 1.35 or 1.37 f/w we've been loading into isp1000 just doesn't report firmware revisions out to mailbox regs 1, 2 and 3 like everyone else. Grumble. Not that this fix hardly matters for FreeBSD. MFC after: 4 weeks
2001-09-03 03:09:48 +00:00
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp->isp_fwattr = mbs.param[6];
isp_prt(isp, ISP_LOGDEBUG0, "Firmware Attributes = 0x%x", mbs.param[6]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
} else {
#ifndef ISP_TARGET_MODE
isp->isp_fwattr = ISP_FW_ATTR_TMODE;
#else
isp->isp_fwattr = 0;
#endif
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (!IS_24XX(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_GET_FIRMWARE_STATUS, MBLOGALL, 0);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Implement ISP_RESET0 for PCI and SBUS attachments- isp_reset has been modified to call ISP_RESET0 if it fails to do a reset. This gives us a chance to disable interrupts.
2006-12-16 05:54:29 +00:00
ISP_RESET0(isp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return;
}
if (isp->isp_maxcmds >= mbs.param[2]) {
isp->isp_maxcmds = mbs.param[2];
}
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGCONFIG, "%d max I/O command limit set", isp->isp_maxcmds);
/*
* If we don't have Multi-ID f/w loaded, we need to restrict channels to one.
* Only make this check for non-SCSI cards (I'm not sure firmware attributes
* work for them).
*/
if (IS_FC(isp) && ISP_CAP_MULTI_ID(isp) == 0 && isp->isp_nchan > 1) {
isp_prt(isp, ISP_LOGWARN, "non-MULTIID f/w loaded, only can enable 1 of %d channels", isp->isp_nchan);
isp->isp_nchan = 1;
}
for (i = 0; i < isp->isp_nchan; i++) {
isp_fw_state(isp, i);
}
if (isp->isp_dead) {
isp_shutdown(isp);
ISP_DISABLE_INTS(isp);
return;
}
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp->isp_state = ISP_RESETSTATE;
Once we have firmware running (if isp_reset) and this is the first time through, establish what our LUN width is. Unfortunately, we can't ask the f/w. If we loaded the f/w, we'll now assume we have expanded LUNs (SCCLUN for fibre channel, just plain 32 LUN for SCSI). If we didn't load firmware, assume 8 LUNs for SCSI and 1 LUN for Fibre Channel. We have to assume only one LUN for Fibre Channel because the LUN setting in Request Queue entries is in different places whether we have SCCLUN firmware or not, so the only LUN guaranteed to work for both is LUN 0. Clean up the rest of isp.c so that ISP2100_SCCLUN defines aren't used- instead use run time determinants based upon isp->isp_maxluns. After starting firmware, delay 500us to give it a chance to get rolling. Fix the interrupt service routine to check for both isr && sema being zero before thinking this was a spurious interrupt. Following the manuals, allow for both Mailbox as well as Queue Reponse type interrupts for regular SCSI.
2000-06-18 04:56:17 +00:00
/*
* Okay- now that we have new firmware running, we now (re)set our
* notion of how many luns we support. This is somewhat tricky because
Clarify issues about whether we have SCCLUN (65535 luns) or non-SCCLUN (16 luns) firmware for the Fibre Channel cards. We used to assume that if we didn't download firmware, we couldn't know what the firmware capability with respect to SCCLUNs is- and it's important because the lun field changes in the request queue entry based upon which firmware it is. At any rate, we *do* get back firmware attributes in mailbox register 6 when we do ABOUT FIRMWARE for all 2200/2300 cards- and for 2100 cards with at least 1.17.0 firmware. So- we now assume non-SCCLUN behaviour for 2100 cards with firmware < 1.17.0- and we check the firmware attributes for other cards (loaded firmware or not). This also allows us to get rid of the crappy test of isp_maxluns > 16- we simply can check firmware attributes for SCCLUN behaviour. This required an 'oops' fix to the outgoing mailbox count field for ABOUT FIRMWARE for FC cards. Also- while here, hardwire firmware revisions for loaded code for SBus cards. Apparently the 1.35 or 1.37 f/w we've been loading into isp1000 just doesn't report firmware revisions out to mailbox regs 1, 2 and 3 like everyone else. Grumble. Not that this fix hardly matters for FreeBSD. MFC after: 4 weeks
2001-09-03 03:09:48 +00:00
* if we haven't loaded firmware, we sometimes do not have an easy way
* of knowing how many luns we support.
Once we have firmware running (if isp_reset) and this is the first time through, establish what our LUN width is. Unfortunately, we can't ask the f/w. If we loaded the f/w, we'll now assume we have expanded LUNs (SCCLUN for fibre channel, just plain 32 LUN for SCSI). If we didn't load firmware, assume 8 LUNs for SCSI and 1 LUN for Fibre Channel. We have to assume only one LUN for Fibre Channel because the LUN setting in Request Queue entries is in different places whether we have SCCLUN firmware or not, so the only LUN guaranteed to work for both is LUN 0. Clean up the rest of isp.c so that ISP2100_SCCLUN defines aren't used- instead use run time determinants based upon isp->isp_maxluns. After starting firmware, delay 500us to give it a chance to get rolling. Fix the interrupt service routine to check for both isr && sema being zero before thinking this was a spurious interrupt. Following the manuals, allow for both Mailbox as well as Queue Reponse type interrupts for regular SCSI.
2000-06-18 04:56:17 +00:00
*
* Expanded lun firmware gives you 32 luns for SCSI cards and
Scale back # of luns supported for SCC to 16384- oops- top 3 bits are a lun address modifier of sorts. Only an HP XP-512 seems to have cared. Fix a few misplaced pointers for the new fabric goop, which has been demonstrated to work on newer Brocades and McData switches now. Put in commented out code which would run GFF_ID if the QLogic f/w allowed it. Don't whine about not being able to find a handle for a command if it was a command aborted (by us).
2002-04-16 19:55:35 +00:00
* 16384 luns for Fibre Channel cards.
Once we have firmware running (if isp_reset) and this is the first time through, establish what our LUN width is. Unfortunately, we can't ask the f/w. If we loaded the f/w, we'll now assume we have expanded LUNs (SCCLUN for fibre channel, just plain 32 LUN for SCSI). If we didn't load firmware, assume 8 LUNs for SCSI and 1 LUN for Fibre Channel. We have to assume only one LUN for Fibre Channel because the LUN setting in Request Queue entries is in different places whether we have SCCLUN firmware or not, so the only LUN guaranteed to work for both is LUN 0. Clean up the rest of isp.c so that ISP2100_SCCLUN defines aren't used- instead use run time determinants based upon isp->isp_maxluns. After starting firmware, delay 500us to give it a chance to get rolling. Fix the interrupt service routine to check for both isr && sema being zero before thinking this was a spurious interrupt. Following the manuals, allow for both Mailbox as well as Queue Reponse type interrupts for regular SCSI.
2000-06-18 04:56:17 +00:00
*
Clarify issues about whether we have SCCLUN (65535 luns) or non-SCCLUN (16 luns) firmware for the Fibre Channel cards. We used to assume that if we didn't download firmware, we couldn't know what the firmware capability with respect to SCCLUNs is- and it's important because the lun field changes in the request queue entry based upon which firmware it is. At any rate, we *do* get back firmware attributes in mailbox register 6 when we do ABOUT FIRMWARE for all 2200/2300 cards- and for 2100 cards with at least 1.17.0 firmware. So- we now assume non-SCCLUN behaviour for 2100 cards with firmware < 1.17.0- and we check the firmware attributes for other cards (loaded firmware or not). This also allows us to get rid of the crappy test of isp_maxluns > 16- we simply can check firmware attributes for SCCLUN behaviour. This required an 'oops' fix to the outgoing mailbox count field for ABOUT FIRMWARE for FC cards. Also- while here, hardwire firmware revisions for loaded code for SBus cards. Apparently the 1.35 or 1.37 f/w we've been loading into isp1000 just doesn't report firmware revisions out to mailbox regs 1, 2 and 3 like everyone else. Grumble. Not that this fix hardly matters for FreeBSD. MFC after: 4 weeks
2001-09-03 03:09:48 +00:00
* It turns out that even for QLogic 2100s with ROM 1.10 and above
* we do get a firmware attributes word returned in mailbox register 6.
*
Correct typos, mostly s/ a / an / where appropriate. Some whitespace cleanup, especially in troff files.
2003-01-01 18:49:04 +00:00
* Because the lun is in a different position in the Request Queue
Once we have firmware running (if isp_reset) and this is the first time through, establish what our LUN width is. Unfortunately, we can't ask the f/w. If we loaded the f/w, we'll now assume we have expanded LUNs (SCCLUN for fibre channel, just plain 32 LUN for SCSI). If we didn't load firmware, assume 8 LUNs for SCSI and 1 LUN for Fibre Channel. We have to assume only one LUN for Fibre Channel because the LUN setting in Request Queue entries is in different places whether we have SCCLUN firmware or not, so the only LUN guaranteed to work for both is LUN 0. Clean up the rest of isp.c so that ISP2100_SCCLUN defines aren't used- instead use run time determinants based upon isp->isp_maxluns. After starting firmware, delay 500us to give it a chance to get rolling. Fix the interrupt service routine to check for both isr && sema being zero before thinking this was a spurious interrupt. Following the manuals, allow for both Mailbox as well as Queue Reponse type interrupts for regular SCSI.
2000-06-18 04:56:17 +00:00
* Entry structure for Fibre Channel with expanded lun firmware, we
* can only support one lun (lun zero) when we don't know what kind
* of firmware we're running.
*/
If the HBA is already 'touched', still set maxluns. Othewise for CAM_QUIRK_HILUN devices we loop thru 32bits of lun. Oops. Switch to using USEC_DELAY rather than USEC_SLEEP at isp_reset time. Try to paper around a defect in clients that don't correctly registers themeselves with the fabric nameserver. Minor updates for Mirapoint support- they still use code that is not HANDLE_LOOPSTATE_IN_OUTER_LAYERS, and, surprise surprise, this old stuff had some bugs in it. Clean up some target mode stuff. MFC after: 1 week
2002-06-16 05:18:22 +00:00
if (IS_SCSI(isp)) {
if (dodnld) {
if (IS_ULTRA2(isp) || IS_ULTRA3(isp)) {
Once we have firmware running (if isp_reset) and this is the first time through, establish what our LUN width is. Unfortunately, we can't ask the f/w. If we loaded the f/w, we'll now assume we have expanded LUNs (SCCLUN for fibre channel, just plain 32 LUN for SCSI). If we didn't load firmware, assume 8 LUNs for SCSI and 1 LUN for Fibre Channel. We have to assume only one LUN for Fibre Channel because the LUN setting in Request Queue entries is in different places whether we have SCCLUN firmware or not, so the only LUN guaranteed to work for both is LUN 0. Clean up the rest of isp.c so that ISP2100_SCCLUN defines aren't used- instead use run time determinants based upon isp->isp_maxluns. After starting firmware, delay 500us to give it a chance to get rolling. Fix the interrupt service routine to check for both isr && sema being zero before thinking this was a spurious interrupt. Following the manuals, allow for both Mailbox as well as Queue Reponse type interrupts for regular SCSI.
2000-06-18 04:56:17 +00:00
isp->isp_maxluns = 32;
} else {
Clarify issues about whether we have SCCLUN (65535 luns) or non-SCCLUN (16 luns) firmware for the Fibre Channel cards. We used to assume that if we didn't download firmware, we couldn't know what the firmware capability with respect to SCCLUNs is- and it's important because the lun field changes in the request queue entry based upon which firmware it is. At any rate, we *do* get back firmware attributes in mailbox register 6 when we do ABOUT FIRMWARE for all 2200/2300 cards- and for 2100 cards with at least 1.17.0 firmware. So- we now assume non-SCCLUN behaviour for 2100 cards with firmware < 1.17.0- and we check the firmware attributes for other cards (loaded firmware or not). This also allows us to get rid of the crappy test of isp_maxluns > 16- we simply can check firmware attributes for SCCLUN behaviour. This required an 'oops' fix to the outgoing mailbox count field for ABOUT FIRMWARE for FC cards. Also- while here, hardwire firmware revisions for loaded code for SBus cards. Apparently the 1.35 or 1.37 f/w we've been loading into isp1000 just doesn't report firmware revisions out to mailbox regs 1, 2 and 3 like everyone else. Grumble. Not that this fix hardly matters for FreeBSD. MFC after: 4 weeks
2001-09-03 03:09:48 +00:00
isp->isp_maxluns = 8;
Once we have firmware running (if isp_reset) and this is the first time through, establish what our LUN width is. Unfortunately, we can't ask the f/w. If we loaded the f/w, we'll now assume we have expanded LUNs (SCCLUN for fibre channel, just plain 32 LUN for SCSI). If we didn't load firmware, assume 8 LUNs for SCSI and 1 LUN for Fibre Channel. We have to assume only one LUN for Fibre Channel because the LUN setting in Request Queue entries is in different places whether we have SCCLUN firmware or not, so the only LUN guaranteed to work for both is LUN 0. Clean up the rest of isp.c so that ISP2100_SCCLUN defines aren't used- instead use run time determinants based upon isp->isp_maxluns. After starting firmware, delay 500us to give it a chance to get rolling. Fix the interrupt service routine to check for both isr && sema being zero before thinking this was a spurious interrupt. Following the manuals, allow for both Mailbox as well as Queue Reponse type interrupts for regular SCSI.
2000-06-18 04:56:17 +00:00
}
} else {
If the HBA is already 'touched', still set maxluns. Othewise for CAM_QUIRK_HILUN devices we loop thru 32bits of lun. Oops. Switch to using USEC_DELAY rather than USEC_SLEEP at isp_reset time. Try to paper around a defect in clients that don't correctly registers themeselves with the fabric nameserver. Minor updates for Mirapoint support- they still use code that is not HANDLE_LOOPSTATE_IN_OUTER_LAYERS, and, surprise surprise, this old stuff had some bugs in it. Clean up some target mode stuff. MFC after: 1 week
2002-06-16 05:18:22 +00:00
isp->isp_maxluns = 8;
}
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (ISP_CAP_SCCFW(isp)) {
If the HBA is already 'touched', still set maxluns. Othewise for CAM_QUIRK_HILUN devices we loop thru 32bits of lun. Oops. Switch to using USEC_DELAY rather than USEC_SLEEP at isp_reset time. Try to paper around a defect in clients that don't correctly registers themeselves with the fabric nameserver. Minor updates for Mirapoint support- they still use code that is not HANDLE_LOOPSTATE_IN_OUTER_LAYERS, and, surprise surprise, this old stuff had some bugs in it. Clean up some target mode stuff. MFC after: 1 week
2002-06-16 05:18:22 +00:00
isp->isp_maxluns = 16384;
} else {
isp->isp_maxluns = 16;
Once we have firmware running (if isp_reset) and this is the first time through, establish what our LUN width is. Unfortunately, we can't ask the f/w. If we loaded the f/w, we'll now assume we have expanded LUNs (SCCLUN for fibre channel, just plain 32 LUN for SCSI). If we didn't load firmware, assume 8 LUNs for SCSI and 1 LUN for Fibre Channel. We have to assume only one LUN for Fibre Channel because the LUN setting in Request Queue entries is in different places whether we have SCCLUN firmware or not, so the only LUN guaranteed to work for both is LUN 0. Clean up the rest of isp.c so that ISP2100_SCCLUN defines aren't used- instead use run time determinants based upon isp->isp_maxluns. After starting firmware, delay 500us to give it a chance to get rolling. Fix the interrupt service routine to check for both isr && sema being zero before thinking this was a spurious interrupt. Following the manuals, allow for both Mailbox as well as Queue Reponse type interrupts for regular SCSI.
2000-06-18 04:56:17 +00:00
}
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MFP4: Move default setting to the end of isp_reset instead of the front of isp_init so we can read NVRAM even if we're role ISP_NONE. Prepare for reintroduction of channels (for FC) for N-Port Virtualization. Fix a botch in handle assignment that caused us to nuke one device when a new one arrives and end up with two devices with the same identity in the virtual target mapping table.
2007-01-20 04:00:21 +00:00
/*
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* We get some default values established. As a side
* effect, NVRAM is read here (unless overriden by
* a configuration flag).
MFP4: Move default setting to the end of isp_reset instead of the front of isp_init so we can read NVRAM even if we're role ISP_NONE. Prepare for reintroduction of channels (for FC) for N-Port Virtualization. Fix a botch in handle assignment that caused us to nuke one device when a new one arrives and end up with two devices with the same identity in the virtual target mapping table.
2007-01-20 04:00:21 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (do_load_defaults) {
if (IS_SCSI(isp)) {
isp_setdfltsdparm(isp);
} else {
for (i = 0; i < isp->isp_nchan; i++) {
isp_setdfltfcparm(isp, i);
}
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
}
MFP4: Move default setting to the end of isp_reset instead of the front of isp_init so we can read NVRAM even if we're role ISP_NONE. Prepare for reintroduction of channels (for FC) for N-Port Virtualization. Fix a botch in handle assignment that caused us to nuke one device when a new one arrives and end up with two devices with the same identity in the virtual target mapping table.
2007-01-20 04:00:21 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
/*
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
* Initialize Parameters of Hardware to a known state.
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
*
* Locks are held before coming here.
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
void
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
isp_init(ispsoftc_t *isp)
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
{
When resetting the Qlogic 2X00 units, reset the FPM (Fibre Protocol Module) and FBM (Fibre Buffer Modules). Also remember to clear the semaphore registers. Tell the RISC processor to not halt on FPM parity errors. Throw out the ISP_CFG_NOINIT silliness and instead go to the use of adapter 'roles' to see whether one completes initialization or not (mostly for Fibre Channel). The ultimate intent, btw, of all of this is to have a warm standby adapter for failover reasons. Because we do roles now, setting of Target Capable Class 3 service parameters in the ICB for the 2x00 cards reflects from role. Also, in isp_start, if we're not supporting an initiator role, we bounce outgoing commands with a Selection Timeout error. Also clean out the TOGGLE_TMODE goop for FC- there is no toggling of target mode like there is for parallel SCSI cards. Do more cleanup with respect to using target ids 0..125 in F-port topologies. Also keep track of things which *were* fabric devices so that when you rescan the fabric you can notify the outer layers when fabric devices go away. Only force a LOGOUT for fabric devices if they're still logged in (i.e., you cat their Port Database entry. Clean up the Get All Next scanning. Finally, use a new tag in the softc to store the opcode for the last mailbox command used so we can report which opcode timed out.
2001-01-15 18:33:08 +00:00
if (IS_FC(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (IS_24XX(isp)) {
isp_fibre_init_2400(isp);
} else {
isp_fibre_init(isp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
When resetting the Qlogic 2X00 units, reset the FPM (Fibre Protocol Module) and FBM (Fibre Buffer Modules). Also remember to clear the semaphore registers. Tell the RISC processor to not halt on FPM parity errors. Throw out the ISP_CFG_NOINIT silliness and instead go to the use of adapter 'roles' to see whether one completes initialization or not (mostly for Fibre Channel). The ultimate intent, btw, of all of this is to have a warm standby adapter for failover reasons. Because we do roles now, setting of Target Capable Class 3 service parameters in the ICB for the 2x00 cards reflects from role. Also, in isp_start, if we're not supporting an initiator role, we bounce outgoing commands with a Selection Timeout error. Also clean out the TOGGLE_TMODE goop for FC- there is no toggling of target mode like there is for parallel SCSI cards. Do more cleanup with respect to using target ids 0..125 in F-port topologies. Also keep track of things which *were* fabric devices so that when you rescan the fabric you can notify the outer layers when fabric devices go away. Only force a LOGOUT for fabric devices if they're still logged in (i.e., you cat their Port Database entry. Clean up the Get All Next scanning. Finally, use a new tag in the softc to store the opcode for the last mailbox command used so we can report which opcode timed out.
2001-01-15 18:33:08 +00:00
} else {
isp_scsi_init(isp);
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
GET_NANOTIME(&isp->isp_init_time);
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
}
static void
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
isp_scsi_init(ispsoftc_t *isp)
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
{
sdparam *sdp_chan0, *sdp_chan1;
mbreg_t mbs;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdp_chan0 = SDPARAM(isp, 0);
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
sdp_chan1 = sdp_chan0;
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
if (IS_DUALBUS(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdp_chan1 = SDPARAM(isp, 1);
When resetting the Qlogic 2X00 units, reset the FPM (Fibre Protocol Module) and FBM (Fibre Buffer Modules). Also remember to clear the semaphore registers. Tell the RISC processor to not halt on FPM parity errors. Throw out the ISP_CFG_NOINIT silliness and instead go to the use of adapter 'roles' to see whether one completes initialization or not (mostly for Fibre Channel). The ultimate intent, btw, of all of this is to have a warm standby adapter for failover reasons. Because we do roles now, setting of Target Capable Class 3 service parameters in the ICB for the 2x00 cards reflects from role. Also, in isp_start, if we're not supporting an initiator role, we bounce outgoing commands with a Selection Timeout error. Also clean out the TOGGLE_TMODE goop for FC- there is no toggling of target mode like there is for parallel SCSI cards. Do more cleanup with respect to using target ids 0..125 in F-port topologies. Also keep track of things which *were* fabric devices so that when you rescan the fabric you can notify the outer layers when fabric devices go away. Only force a LOGOUT for fabric devices if they're still logged in (i.e., you cat their Port Database entry. Clean up the Get All Next scanning. Finally, use a new tag in the softc to store the opcode for the last mailbox command used so we can report which opcode timed out.
2001-01-15 18:33:08 +00:00
}
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
/* First do overall per-card settings. */
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 01:07:06 +00:00
/*
* If we have fast memory timing enabled, turn it on.
*/
various fixes
2000-08-27 23:38:44 +00:00
if (sdp_chan0->isp_fast_mttr) {
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 01:07:06 +00:00
ISP_WRITE(isp, RISC_MTR, 0x1313);
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
* Set Retry Delay and Count.
* You set both channels at the same time.
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_SET_RETRY_COUNT, MBLOGALL, 0);
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
mbs.param[1] = sdp_chan0->isp_retry_count;
mbs.param[2] = sdp_chan0->isp_retry_delay;
mbs.param[6] = sdp_chan1->isp_retry_count;
mbs.param[7] = sdp_chan1->isp_retry_delay;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
return;
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
/*
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
* Set ASYNC DATA SETUP time. This is very important.
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_SET_ASYNC_DATA_SETUP_TIME, MBLOGALL, 0);
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
mbs.param[1] = sdp_chan0->isp_async_data_setup;
mbs.param[2] = sdp_chan1->isp_async_data_setup;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
return;
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
/*
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
* Set ACTIVE Negation State.
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_SET_ACT_NEG_STATE, MBLOGNONE, 0);
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
mbs.param[1] =
(sdp_chan0->isp_req_ack_active_neg << 4) |
(sdp_chan0->isp_data_line_active_neg << 5);
mbs.param[2] =
(sdp_chan1->isp_req_ack_active_neg << 4) |
(sdp_chan1->isp_data_line_active_neg << 5);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGERR,
"failed to set active negation state (%d,%d), (%d,%d)",
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
sdp_chan0->isp_req_ack_active_neg,
sdp_chan0->isp_data_line_active_neg,
sdp_chan1->isp_req_ack_active_neg,
sdp_chan1->isp_data_line_active_neg);
/*
* But don't return.
*/
}
/*
* Set the Tag Aging limit
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_SET_TAG_AGE_LIMIT, MBLOGALL, 0);
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
mbs.param[1] = sdp_chan0->isp_tag_aging;
mbs.param[2] = sdp_chan1->isp_tag_aging;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGERR, "failed to set tag age limit (%d,%d)",
sdp_chan0->isp_tag_aging, sdp_chan1->isp_tag_aging);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
return;
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
/*
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
* Set selection timeout.
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_SET_SELECT_TIMEOUT, MBLOGALL, 0);
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
mbs.param[1] = sdp_chan0->isp_selection_timeout;
mbs.param[2] = sdp_chan1->isp_selection_timeout;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
return;
}
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
/* now do per-channel settings */
isp_scsi_channel_init(isp, 0);
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
if (IS_DUALBUS(isp))
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
isp_scsi_channel_init(isp, 1);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
/*
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
* Now enable request/response queues
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
*/
If we have a 1240 or an ULTRA2 or better card, use MBOX_INIT_RES_QUEUE_A64 (preparation for DAC/A64 support)
2002-09-23 04:59:42 +00:00
if (IS_ULTRA2(isp) || IS_1240(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_INIT_RES_QUEUE_A64, MBLOGALL, 0);
If we have a 1240 or an ULTRA2 or better card, use MBOX_INIT_RES_QUEUE_A64 (preparation for DAC/A64 support)
2002-09-23 04:59:42 +00:00
mbs.param[1] = RESULT_QUEUE_LEN(isp);
mbs.param[2] = DMA_WD1(isp->isp_result_dma);
mbs.param[3] = DMA_WD0(isp->isp_result_dma);
mbs.param[4] = 0;
mbs.param[6] = DMA_WD3(isp->isp_result_dma);
mbs.param[7] = DMA_WD2(isp->isp_result_dma);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
If we have a 1240 or an ULTRA2 or better card, use MBOX_INIT_RES_QUEUE_A64 (preparation for DAC/A64 support)
2002-09-23 04:59:42 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
return;
}
isp->isp_residx = mbs.param[5];
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_INIT_REQ_QUEUE_A64, MBLOGALL, 0);
If we have a 1240 or an ULTRA2 or better card, use MBOX_INIT_RES_QUEUE_A64 (preparation for DAC/A64 support)
2002-09-23 04:59:42 +00:00
mbs.param[1] = RQUEST_QUEUE_LEN(isp);
mbs.param[2] = DMA_WD1(isp->isp_rquest_dma);
mbs.param[3] = DMA_WD0(isp->isp_rquest_dma);
mbs.param[5] = 0;
mbs.param[6] = DMA_WD3(isp->isp_result_dma);
mbs.param[7] = DMA_WD2(isp->isp_result_dma);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
If we have a 1240 or an ULTRA2 or better card, use MBOX_INIT_RES_QUEUE_A64 (preparation for DAC/A64 support)
2002-09-23 04:59:42 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
return;
}
isp->isp_reqidx = isp->isp_reqodx = mbs.param[4];
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_INIT_RES_QUEUE, MBLOGALL, 0);
If we have a 1240 or an ULTRA2 or better card, use MBOX_INIT_RES_QUEUE_A64 (preparation for DAC/A64 support)
2002-09-23 04:59:42 +00:00
mbs.param[1] = RESULT_QUEUE_LEN(isp);
mbs.param[2] = DMA_WD1(isp->isp_result_dma);
mbs.param[3] = DMA_WD0(isp->isp_result_dma);
mbs.param[4] = 0;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
If we have a 1240 or an ULTRA2 or better card, use MBOX_INIT_RES_QUEUE_A64 (preparation for DAC/A64 support)
2002-09-23 04:59:42 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
return;
}
isp->isp_residx = mbs.param[5];
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_INIT_REQ_QUEUE, MBLOGALL, 0);
If we have a 1240 or an ULTRA2 or better card, use MBOX_INIT_RES_QUEUE_A64 (preparation for DAC/A64 support)
2002-09-23 04:59:42 +00:00
mbs.param[1] = RQUEST_QUEUE_LEN(isp);
mbs.param[2] = DMA_WD1(isp->isp_rquest_dma);
mbs.param[3] = DMA_WD0(isp->isp_rquest_dma);
mbs.param[5] = 0;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
If we have a 1240 or an ULTRA2 or better card, use MBOX_INIT_RES_QUEUE_A64 (preparation for DAC/A64 support)
2002-09-23 04:59:42 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
return;
}
isp->isp_reqidx = isp->isp_reqodx = mbs.param[4];
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
/*
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
* Turn on LVD transitions for ULTRA2 or better and other features
Clean up some f/w revision checking wrt enabling fast posting. Make sure we set defaults sanely for dual-bus adapters.
1999-12-20 01:34:01 +00:00
*
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
* Now that we have 32 bit handles, don't do any fast posting
* any more. For Ultra2/Ultra3 cards, we can turn on 32 bit RIO
* operation or use fast posting. To be conservative, we'll only
* do this for Ultra3 cards now because the other cards are so
* rare for this author to find and test with.
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_SET_FW_FEATURES, MBLOGALL, 0);
Clean up some f/w revision checking wrt enabling fast posting. Make sure we set defaults sanely for dual-bus adapters.
1999-12-20 01:34:01 +00:00
if (IS_ULTRA2(isp))
mbs.param[1] |= FW_FEATURE_LVD_NOTIFY;
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
#ifdef ISP_NO_RIO
if (IS_ULTRA3(isp))
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
mbs.param[1] |= FW_FEATURE_FAST_POST;
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
#else
if (IS_ULTRA3(isp))
mbs.param[1] |= FW_FEATURE_RIO_32BIT;
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
#endif
Clean up some f/w revision checking wrt enabling fast posting. Make sure we set defaults sanely for dual-bus adapters.
1999-12-20 01:34:01 +00:00
if (mbs.param[1] != 0) {
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
uint16_t sfeat = mbs.param[1];
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (mbs.param[0] == MBOX_COMMAND_COMPLETE) {
isp_prt(isp, ISP_LOGINFO,
"Enabled FW features (0x%x)", sfeat);
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
}
}
isp->isp_state = ISP_INITSTATE;
}
static void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_scsi_channel_init(ispsoftc_t *isp, int chan)
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
{
sdparam *sdp;
mbreg_t mbs;
int tgt;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdp = SDPARAM(isp, chan);
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
/*
* Set (possibly new) Initiator ID.
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_SET_INIT_SCSI_ID, MBLOGALL, 0);
mbs.param[1] = (chan << 7) | sdp->isp_initiator_id;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
return;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGINFO, "Chan %d Initiator ID is %d",
chan, sdp->isp_initiator_id);
Inintialize the queue index stuff from what the f/w sends back- just in case it's insane enough to not do what you tell it to. Print out (LOGINFO level) initiator ID.
2000-09-21 17:06:45 +00:00
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
/*
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records either what's in NVRAM or what the safe defaults would be if we lack NVRAM. Then we rename cur_XXXX to actv_XXXX (these are the currently active settings) and the dev_XXX settings to goal_XXXX (these are the settings which we want cur_XXXX to converge to).
2001-07-30 00:59:06 +00:00
* Set current per-target parameters to an initial safe minimum.
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
*/
for (tgt = 0; tgt < MAX_TARGETS; tgt++) {
Once we have firmware running (if isp_reset) and this is the first time through, establish what our LUN width is. Unfortunately, we can't ask the f/w. If we loaded the f/w, we'll now assume we have expanded LUNs (SCCLUN for fibre channel, just plain 32 LUN for SCSI). If we didn't load firmware, assume 8 LUNs for SCSI and 1 LUN for Fibre Channel. We have to assume only one LUN for Fibre Channel because the LUN setting in Request Queue entries is in different places whether we have SCCLUN firmware or not, so the only LUN guaranteed to work for both is LUN 0. Clean up the rest of isp.c so that ISP2100_SCCLUN defines aren't used- instead use run time determinants based upon isp->isp_maxluns. After starting firmware, delay 500us to give it a chance to get rolling. Fix the interrupt service routine to check for both isr && sema being zero before thinking this was a spurious interrupt. Following the manuals, allow for both Mailbox as well as Queue Reponse type interrupts for regular SCSI.
2000-06-18 04:56:17 +00:00
int lun;
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
uint16_t sdf;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
if (sdp->isp_devparam[tgt].dev_enable == 0) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
continue;
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
}
After loading f/w, for FC cards print out Firmware Attributes. Redo establishment of default SCSI parameters whether or not we've been compiled for target mode. Unfortunately, the Qlogic f/w is confused so that if we set all targets to be 'safe' (i.e., narrow/async), it will also then report narrow, async if we're contacted in target mode from that target (acting in initiator role). D'oh! Fix ISPCTL_TOGGLE_TMODE to correctly enable the right channel for dual channel cards. Add some more opcodes. Fix a stupid NULL pointer bug.
2001-04-04 21:42:59 +00:00
#ifndef ISP_TARGET_MODE
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records either what's in NVRAM or what the safe defaults would be if we lack NVRAM. Then we rename cur_XXXX to actv_XXXX (these are the currently active settings) and the dev_XXX settings to goal_XXXX (these are the settings which we want cur_XXXX to converge to).
2001-07-30 00:59:06 +00:00
sdf = sdp->isp_devparam[tgt].goal_flags;
sdf &= DPARM_SAFE_DFLT;
/*
* It is not quite clear when this changed over so that
* we could force narrow and async for 1000/1020 cards,
* but assume that this is only the case for loaded
* firmware.
*/
if (isp->isp_loaded_fw) {
sdf |= DPARM_NARROW | DPARM_ASYNC;
After loading f/w, for FC cards print out Firmware Attributes. Redo establishment of default SCSI parameters whether or not we've been compiled for target mode. Unfortunately, the Qlogic f/w is confused so that if we set all targets to be 'safe' (i.e., narrow/async), it will also then report narrow, async if we're contacted in target mode from that target (acting in initiator role). D'oh! Fix ISPCTL_TOGGLE_TMODE to correctly enable the right channel for dual channel cards. Add some more opcodes. Fix a stupid NULL pointer bug.
2001-04-04 21:42:59 +00:00
}
#else
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
/*
After loading f/w, for FC cards print out Firmware Attributes. Redo establishment of default SCSI parameters whether or not we've been compiled for target mode. Unfortunately, the Qlogic f/w is confused so that if we set all targets to be 'safe' (i.e., narrow/async), it will also then report narrow, async if we're contacted in target mode from that target (acting in initiator role). D'oh! Fix ISPCTL_TOGGLE_TMODE to correctly enable the right channel for dual channel cards. Add some more opcodes. Fix a stupid NULL pointer bug.
2001-04-04 21:42:59 +00:00
* The !$*!)$!$)* f/w uses the same index into some
* internal table to decide how to respond to negotiations,
* so if we've said "let's be safe" for ID X, and ID X
* selects *us*, the negotiations will back to 'safe'
* (as in narrow/async). What the f/w *should* do is
* use the initiator id settings to decide how to respond.
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
*/
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records either what's in NVRAM or what the safe defaults would be if we lack NVRAM. Then we rename cur_XXXX to actv_XXXX (these are the currently active settings) and the dev_XXX settings to goal_XXXX (these are the settings which we want cur_XXXX to converge to).
2001-07-30 00:59:06 +00:00
sdp->isp_devparam[tgt].goal_flags = sdf = DPARM_DEFAULT;
After loading f/w, for FC cards print out Firmware Attributes. Redo establishment of default SCSI parameters whether or not we've been compiled for target mode. Unfortunately, the Qlogic f/w is confused so that if we set all targets to be 'safe' (i.e., narrow/async), it will also then report narrow, async if we're contacted in target mode from that target (acting in initiator role). D'oh! Fix ISPCTL_TOGGLE_TMODE to correctly enable the right channel for dual channel cards. Add some more opcodes. Fix a stupid NULL pointer bug.
2001-04-04 21:42:59 +00:00
#endif
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_SET_TARGET_PARAMS, MBLOGNONE, 0);
mbs.param[1] = (chan << 15) | (tgt << 8);
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 01:07:06 +00:00
mbs.param[2] = sdf;
various fixes
2000-08-27 23:38:44 +00:00
if ((sdf & DPARM_SYNC) == 0) {
mbs.param[3] = 0;
} else {
mbs.param[3] =
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records either what's in NVRAM or what the safe defaults would be if we lack NVRAM. Then we rename cur_XXXX to actv_XXXX (these are the currently active settings) and the dev_XXX settings to goal_XXXX (these are the settings which we want cur_XXXX to converge to).
2001-07-30 00:59:06 +00:00
(sdp->isp_devparam[tgt].goal_offset << 8) |
(sdp->isp_devparam[tgt].goal_period);
various fixes
2000-08-27 23:38:44 +00:00
}
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_prt(isp, ISP_LOGDEBUG0, "Initial Settings bus%d tgt%d flags 0x%x off 0x%x per 0x%x",
chan, tgt, mbs.param[2], mbs.param[3] >> 8, mbs.param[3] & 0xff);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 01:07:06 +00:00
sdf = DPARM_SAFE_DFLT;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_SET_TARGET_PARAMS, MBLOGALL, 0);
mbs.param[1] = (tgt << 8) | (chan << 15);
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 01:07:06 +00:00
mbs.param[2] = sdf;
various fixes
2000-08-27 23:38:44 +00:00
mbs.param[3] = 0;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
continue;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
}
Clean up some f/w revision checking wrt enabling fast posting. Make sure we set defaults sanely for dual-bus adapters.
1999-12-20 01:34:01 +00:00
Annoying little nigglet- apparently *some* Qlogic temporarily ignore settings you've just sent them and return random values if you follow the set by a get. This causes problems when you latter run a Tag-enabled command when you've command tagged mode off.
1999-03-26 00:33:13 +00:00
/*
various fixes
2000-08-27 23:38:44 +00:00
* We don't update any information directly from the f/w
* because we need to run at least one command to cause a
* new state to be latched up. So, we just assume that we
* converge to the values we just had set.
*
Annoying little nigglet- apparently *some* Qlogic temporarily ignore settings you've just sent them and return random values if you follow the set by a get. This causes problems when you latter run a Tag-enabled command when you've command tagged mode off.
1999-03-26 00:33:13 +00:00
* Ensure that we don't believe tagged queuing is enabled yet.
* It turns out that sometimes the ISP just ignores our
* attempts to set parameters for devices that it hasn't
* seen yet.
*/
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records either what's in NVRAM or what the safe defaults would be if we lack NVRAM. Then we rename cur_XXXX to actv_XXXX (these are the currently active settings) and the dev_XXX settings to goal_XXXX (these are the settings which we want cur_XXXX to converge to).
2001-07-30 00:59:06 +00:00
sdp->isp_devparam[tgt].actv_flags = sdf & ~DPARM_TQING;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
for (lun = 0; lun < (int) isp->isp_maxluns; lun++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_SET_DEV_QUEUE_PARAMS, MBLOGALL, 0);
mbs.param[1] = (chan << 15) | (tgt << 8) | lun;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
mbs.param[2] = sdp->isp_max_queue_depth;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
mbs.param[3] = sdp->isp_devparam[tgt].exc_throttle;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
}
various fixes
2000-08-27 23:38:44 +00:00
for (tgt = 0; tgt < MAX_TARGETS; tgt++) {
if (sdp->isp_devparam[tgt].dev_refresh) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdp->sendmarker = 1;
sdp->update = 1;
various fixes
2000-08-27 23:38:44 +00:00
break;
}
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
/*
* Fibre Channel specific initialization.
*/
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
static void
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
isp_fibre_init(ispsoftc_t *isp)
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
{
fcparam *fcp;
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
isp_icb_t local, *icbp = &local;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
mbreg_t mbs;
Don't attempt to load illegal hard loop addresses into an ICB. This shows up on card restarts, and usually for 2200-2300 cards. What happens is that we start up, attempting to acquire a hard address. We end up instead being an F-port topology, which reports out a loop id of 0xff (or 0xffff for 2K Login f/w). Then, if we restart, we end up telling the card to go off an acquire this loop address, which the card then rejects. Bah. Compilation fixes from Solaris port.
2007-02-23 21:59:21 +00:00
int ownloopid;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* We only support one channel on non-24XX cards
*/
fcp = FCPARAM(isp, 0);
if (fcp->role == ISP_ROLE_NONE) {
isp->isp_state = ISP_INITSTATE;
return;
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(icbp, sizeof (*icbp));
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
icbp->icb_version = ICB_VERSION1;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
icbp->icb_fwoptions = fcp->isp_fwoptions;
various fixes
2000-08-27 23:38:44 +00:00
/*
* Firmware Options are either retrieved from NVRAM or
* are patched elsewhere. We check them for sanity here
* and make changes based on board revision, but otherwise
* let others decide policy.
*/
Clean up stupidity in the isp_handle_other_response function- indexes of queue entries have to be at least 16 bits now! If we're running a 2100 less than rev 5, turn off loop fairness (per Qlogic errata). Fix typo in checking against 2200 F/W revision. Slightly fix/reorder fabric login stuff. Change to usage of isp_getrqentry for code clarity. Add some defensive dual bus assumptions. Various cleanups, etc...
1999-11-21 03:18:22 +00:00
/*
* If this is a 2100 < revision 5, we have to turn off FAIRNESS.
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_2100(isp) && isp->isp_revision < 5) {
icbp->icb_fwoptions &= ~ICBOPT_FAIRNESS;
Clean up stupidity in the isp_handle_other_response function- indexes of queue entries have to be at least 16 bits now! If we're running a 2100 less than rev 5, turn off loop fairness (per Qlogic errata). Fix typo in checking against 2200 F/W revision. Slightly fix/reorder fabric login stuff. Change to usage of isp_getrqentry for code clarity. Add some defensive dual bus assumptions. Various cleanups, etc...
1999-11-21 03:18:22 +00:00
}
various fixes
2000-08-27 23:38:44 +00:00
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
/*
More code cleanup. Go back to using FULL_LOGIN Fibre Chan if f/w is less than 1.17.0 level. Change where we do the loop database init. Add in the CMD_RQLATER return. Add some register debounce.
1999-08-16 19:59:55 +00:00
* We have to use FULL LOGIN even though it resets the loop too much
* because otherwise port database entries don't get updated after
Fix some breakage about how we build WWNs. Do some other fabric related changes: consider a new PDB entry different if Class 3 service parameter roles change (!!!). Do some checking as we're getting a port database that traps whether things change while we're doing so. Handle N-port and F-ports correctly. Fix the fabric login loop to retain a login/binding if things haven't changed (I mean, why logout a device only to log it back in). No longer accept, after fabric logins, garbage if we can't get a PDB entry that matches the device we've just logged into- if it doesn't, log it out as it is very unlikely to still be what we thought it was. Get rid of some of the debounce loops because we could get stuck there.
2000-05-09 01:14:43 +00:00
* a LIP- this is a known f/w bug for 2100 f/w less than 1.17.0.
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
*/
If we're using ancient (pre 1.17.0) 2100 f/w (for the cards that cannot load f/w images > 0x7fff words), set ISP_FW_ATTR_SCCLUN. We explicitly don't believe we can find attributes if f/w is < 1.17.0, so we have to set SCCLUN for the 1.15.37 f/w we're using manually- otherwise every target will replicate itself across all 16 supported luns for non-SCCLUN f/w. Correctly set things up for 23XX and either fast posting or ZIO. The 23XX, it turns out, does not support RIO. If you put a non-zero value in xfwoptions, this will disable fast posting. If you put ICBXOPT_ZIO in xfwoptions, then the 23XX will do interrupt delays but post to the response queue- apparently QLogic *now* believes that reading multiple handles from registers is less of a win than writing (and delaying) multiple 64 byte responses to the response queue. At the end of taking a a good f/w crash dump, send the ISPASYNC_FW_DUMPED event to the outer layers (who can then do things like wake a user daemon to *fetch* the crash image, etc.).
2002-08-17 17:29:15 +00:00
if (!ISP_FW_NEWER_THAN(isp, 1, 17, 0)) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
icbp->icb_fwoptions |= ICBOPT_FULL_LOGIN;
More code cleanup. Go back to using FULL_LOGIN Fibre Chan if f/w is less than 1.17.0 level. Change where we do the loop database init. Add in the CMD_RQLATER return. Add some register debounce.
1999-08-16 19:59:55 +00:00
}
various fixes
2000-08-27 23:38:44 +00:00
/*
* Insist on Port Database Update Async notifications
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
icbp->icb_fwoptions |= ICBOPT_PDBCHANGE_AE;
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
When resetting the Qlogic 2X00 units, reset the FPM (Fibre Protocol Module) and FBM (Fibre Buffer Modules). Also remember to clear the semaphore registers. Tell the RISC processor to not halt on FPM parity errors. Throw out the ISP_CFG_NOINIT silliness and instead go to the use of adapter 'roles' to see whether one completes initialization or not (mostly for Fibre Channel). The ultimate intent, btw, of all of this is to have a warm standby adapter for failover reasons. Because we do roles now, setting of Target Capable Class 3 service parameters in the ICB for the 2x00 cards reflects from role. Also, in isp_start, if we're not supporting an initiator role, we bounce outgoing commands with a Selection Timeout error. Also clean out the TOGGLE_TMODE goop for FC- there is no toggling of target mode like there is for parallel SCSI cards. Do more cleanup with respect to using target ids 0..125 in F-port topologies. Also keep track of things which *were* fabric devices so that when you rescan the fabric you can notify the outer layers when fabric devices go away. Only force a LOGOUT for fabric devices if they're still logged in (i.e., you cat their Port Database entry. Clean up the Get All Next scanning. Finally, use a new tag in the softc to store the opcode for the last mailbox command used so we can report which opcode timed out.
2001-01-15 18:33:08 +00:00
/*
* Make sure that target role reflects into fwoptions.
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (fcp->role & ISP_ROLE_TARGET) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
icbp->icb_fwoptions |= ICBOPT_TGT_ENABLE;
When resetting the Qlogic 2X00 units, reset the FPM (Fibre Protocol Module) and FBM (Fibre Buffer Modules). Also remember to clear the semaphore registers. Tell the RISC processor to not halt on FPM parity errors. Throw out the ISP_CFG_NOINIT silliness and instead go to the use of adapter 'roles' to see whether one completes initialization or not (mostly for Fibre Channel). The ultimate intent, btw, of all of this is to have a warm standby adapter for failover reasons. Because we do roles now, setting of Target Capable Class 3 service parameters in the ICB for the 2x00 cards reflects from role. Also, in isp_start, if we're not supporting an initiator role, we bounce outgoing commands with a Selection Timeout error. Also clean out the TOGGLE_TMODE goop for FC- there is no toggling of target mode like there is for parallel SCSI cards. Do more cleanup with respect to using target ids 0..125 in F-port topologies. Also keep track of things which *were* fabric devices so that when you rescan the fabric you can notify the outer layers when fabric devices go away. Only force a LOGOUT for fabric devices if they're still logged in (i.e., you cat their Port Database entry. Clean up the Get All Next scanning. Finally, use a new tag in the softc to store the opcode for the last mailbox command used so we can report which opcode timed out.
2001-01-15 18:33:08 +00:00
} else {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
icbp->icb_fwoptions &= ~ICBOPT_TGT_ENABLE;
When resetting the Qlogic 2X00 units, reset the FPM (Fibre Protocol Module) and FBM (Fibre Buffer Modules). Also remember to clear the semaphore registers. Tell the RISC processor to not halt on FPM parity errors. Throw out the ISP_CFG_NOINIT silliness and instead go to the use of adapter 'roles' to see whether one completes initialization or not (mostly for Fibre Channel). The ultimate intent, btw, of all of this is to have a warm standby adapter for failover reasons. Because we do roles now, setting of Target Capable Class 3 service parameters in the ICB for the 2x00 cards reflects from role. Also, in isp_start, if we're not supporting an initiator role, we bounce outgoing commands with a Selection Timeout error. Also clean out the TOGGLE_TMODE goop for FC- there is no toggling of target mode like there is for parallel SCSI cards. Do more cleanup with respect to using target ids 0..125 in F-port topologies. Also keep track of things which *were* fabric devices so that when you rescan the fabric you can notify the outer layers when fabric devices go away. Only force a LOGOUT for fabric devices if they're still logged in (i.e., you cat their Port Database entry. Clean up the Get All Next scanning. Finally, use a new tag in the softc to store the opcode for the last mailbox command used so we can report which opcode timed out.
2001-01-15 18:33:08 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (fcp->role & ISP_ROLE_INITIATOR) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
icbp->icb_fwoptions &= ~ICBOPT_INI_DISABLE;
If we have ISP_ROLE_INITIATOR set, make sure that we clear ICBOPT_INI_DISABLE from the fwoptions. Likewise, we *set* ICBOPT_INI_DISABLE if we don't have initiator role.
2004-01-23 23:23:31 +00:00
} else {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
icbp->icb_fwoptions |= ICBOPT_INI_DISABLE;
If we have ISP_ROLE_INITIATOR set, make sure that we clear ICBOPT_INI_DISABLE from the fwoptions. Likewise, we *set* ICBOPT_INI_DISABLE if we don't have initiator role.
2004-01-23 23:23:31 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
icbp->icb_maxfrmlen = DEFAULT_FRAMESIZE(isp);
Remove file unused in freebsd.
2009-09-21 01:38:22 +00:00
if (icbp->icb_maxfrmlen < ICB_MIN_FRMLEN || icbp->icb_maxfrmlen > ICB_MAX_FRMLEN) {
isp_prt(isp, ISP_LOGERR, "bad frame length (%d) from NVRAM- using %d", DEFAULT_FRAMESIZE(isp), ICB_DFLT_FRMLEN);
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 02:28:29 +00:00
icbp->icb_maxfrmlen = ICB_DFLT_FRMLEN;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
icbp->icb_maxalloc = fcp->isp_maxalloc;
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
if (icbp->icb_maxalloc < 1) {
Remove file unused in freebsd.
2009-09-21 01:38:22 +00:00
isp_prt(isp, ISP_LOGERR, "bad maximum allocation (%d)- using 16", fcp->isp_maxalloc);
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 02:28:29 +00:00
icbp->icb_maxalloc = 16;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
icbp->icb_execthrottle = DEFAULT_EXEC_THROTTLE(isp);
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 02:28:29 +00:00
if (icbp->icb_execthrottle < 1) {
Remove file unused in freebsd.
2009-09-21 01:38:22 +00:00
isp_prt(isp, ISP_LOGERR, "bad execution throttle of %d- using %d", DEFAULT_EXEC_THROTTLE(isp), ICB_DFLT_THROTTLE);
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
icbp->icb_execthrottle = ICB_DFLT_THROTTLE;
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 02:28:29 +00:00
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
icbp->icb_retry_delay = fcp->isp_retry_delay;
icbp->icb_retry_count = fcp->isp_retry_count;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
icbp->icb_hardaddr = fcp->isp_loopid;
Don't attempt to load illegal hard loop addresses into an ICB. This shows up on card restarts, and usually for 2200-2300 cards. What happens is that we start up, attempting to acquire a hard address. We end up instead being an F-port topology, which reports out a loop id of 0xff (or 0xffff for 2K Login f/w). Then, if we restart, we end up telling the card to go off an acquire this loop address, which the card then rejects. Bah. Compilation fixes from Solaris port.
2007-02-23 21:59:21 +00:00
ownloopid = (isp->isp_confopts & ISP_CFG_OWNLOOPID) != 0;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (icbp->icb_hardaddr >= LOCAL_LOOP_LIM) {
Don't attempt to load illegal hard loop addresses into an ICB. This shows up on card restarts, and usually for 2200-2300 cards. What happens is that we start up, attempting to acquire a hard address. We end up instead being an F-port topology, which reports out a loop id of 0xff (or 0xffff for 2K Login f/w). Then, if we restart, we end up telling the card to go off an acquire this loop address, which the card then rejects. Bah. Compilation fixes from Solaris port.
2007-02-23 21:59:21 +00:00
icbp->icb_hardaddr = 0;
ownloopid = 0;
On reset, make sure that we have some parameters set correctly. This fixes a longstanding issue WRT resetting the chip after startup- it would fail if we were connected as an F-port to a switch. If we were connected as an F-port, we got assigned a hard loop ID of 255, which is really a bogus loop id. Then when we turned around to reset ourselves, the firmware would reject the ICB_INIT request because the loop id was bogus. *sputter* Minor fixlet from somebody in NetBSD with too much time on their hands (dma -> DMA).
2003-09-13 01:55:44 +00:00
}
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
What the heck - make the last (most recent) 2200 f/w also do Hard Loop acquisition.
2006-07-03 20:56:48 +00:00
/*
* Our life seems so much better with 2200s and later with
* the latest f/w if we set Hard Address.
*/
Don't attempt to load illegal hard loop addresses into an ICB. This shows up on card restarts, and usually for 2200-2300 cards. What happens is that we start up, attempting to acquire a hard address. We end up instead being an F-port topology, which reports out a loop id of 0xff (or 0xffff for 2K Login f/w). Then, if we restart, we end up telling the card to go off an acquire this loop address, which the card then rejects. Bah. Compilation fixes from Solaris port.
2007-02-23 21:59:21 +00:00
if (ownloopid || ISP_FW_NEWER_THAN(isp, 2, 2, 5)) {
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
icbp->icb_fwoptions |= ICBOPT_HARD_ADDRESS;
}
various fixes
2000-08-27 23:38:44 +00:00
/*
* Right now we just set extended options to prefer point-to-point
* over loop based upon some soft config options.
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
*
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
* NB: for the 2300, ICBOPT_EXTENDED is required.
various fixes
2000-08-27 23:38:44 +00:00
*/
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
if (IS_2200(isp) || IS_23XX(isp)) {
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
icbp->icb_fwoptions |= ICBOPT_EXTENDED;
/*
* Prefer or force Point-To-Point instead Loop?
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
switch (isp->isp_confopts & ISP_CFG_PORT_PREF) {
Make the Not RESPONSE in RESPONSE QUEUE message have a bit more info (specifically, how many entries we've looked at so far). Maintain interrupt instrumentation. Use USEC_SLEEP instead of USEC_DELAY in a number of places (this allows us to drop locks and sleep instead of spin). Track changes to configuration options for topology preference. Fix botched order of printout for Channel, Target, Lun.
2000-12-02 18:08:35 +00:00
case ISP_CFG_NPORT:
Implement a call to get the actual link data rate (if 23XX) so we can set whether it's a 2Gps or 1Gps link. MFC after: 1 week
2001-10-01 03:45:54 +00:00
icbp->icb_xfwoptions |= ICBXOPT_PTP_2_LOOP;
Make the Not RESPONSE in RESPONSE QUEUE message have a bit more info (specifically, how many entries we've looked at so far). Maintain interrupt instrumentation. Use USEC_SLEEP instead of USEC_DELAY in a number of places (this allows us to drop locks and sleep instead of spin). Track changes to configuration options for topology preference. Fix botched order of printout for Channel, Target, Lun.
2000-12-02 18:08:35 +00:00
break;
case ISP_CFG_NPORT_ONLY:
Implement a call to get the actual link data rate (if 23XX) so we can set whether it's a 2Gps or 1Gps link. MFC after: 1 week
2001-10-01 03:45:54 +00:00
icbp->icb_xfwoptions |= ICBXOPT_PTP_ONLY;
Make the Not RESPONSE in RESPONSE QUEUE message have a bit more info (specifically, how many entries we've looked at so far). Maintain interrupt instrumentation. Use USEC_SLEEP instead of USEC_DELAY in a number of places (this allows us to drop locks and sleep instead of spin). Track changes to configuration options for topology preference. Fix botched order of printout for Channel, Target, Lun.
2000-12-02 18:08:35 +00:00
break;
case ISP_CFG_LPORT_ONLY:
Implement a call to get the actual link data rate (if 23XX) so we can set whether it's a 2Gps or 1Gps link. MFC after: 1 week
2001-10-01 03:45:54 +00:00
icbp->icb_xfwoptions |= ICBXOPT_LOOP_ONLY;
Make the Not RESPONSE in RESPONSE QUEUE message have a bit more info (specifically, how many entries we've looked at so far). Maintain interrupt instrumentation. Use USEC_SLEEP instead of USEC_DELAY in a number of places (this allows us to drop locks and sleep instead of spin). Track changes to configuration options for topology preference. Fix botched order of printout for Channel, Target, Lun.
2000-12-02 18:08:35 +00:00
break;
default:
Implement a call to get the actual link data rate (if 23XX) so we can set whether it's a 2Gps or 1Gps link. MFC after: 1 week
2001-10-01 03:45:54 +00:00
icbp->icb_xfwoptions |= ICBXOPT_LOOP_2_PTP;
Make the Not RESPONSE in RESPONSE QUEUE message have a bit more info (specifically, how many entries we've looked at so far). Maintain interrupt instrumentation. Use USEC_SLEEP instead of USEC_DELAY in a number of places (this allows us to drop locks and sleep instead of spin). Track changes to configuration options for topology preference. Fix botched order of printout for Channel, Target, Lun.
2000-12-02 18:08:35 +00:00
break;
}
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
if (IS_2200(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
* We can't have Fast Posting any more- we now
* have 32 bit handles.
*
* RIO seemed to have to much breakage.
*
* Just opt for safety.
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
*/
icbp->icb_xfwoptions &= ~ICBXOPT_RIO_16BIT;
Be more conservative- turn off fast posting and RIO for 22XX cards. Approved by: re (ken) MFC after: 3 days
2007-07-10 07:55:04 +00:00
icbp->icb_fwoptions &= ~ICBOPT_FAST_POST;
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
} else {
If we're using ancient (pre 1.17.0) 2100 f/w (for the cards that cannot load f/w images > 0x7fff words), set ISP_FW_ATTR_SCCLUN. We explicitly don't believe we can find attributes if f/w is < 1.17.0, so we have to set SCCLUN for the 1.15.37 f/w we're using manually- otherwise every target will replicate itself across all 16 supported luns for non-SCCLUN f/w. Correctly set things up for 23XX and either fast posting or ZIO. The 23XX, it turns out, does not support RIO. If you put a non-zero value in xfwoptions, this will disable fast posting. If you put ICBXOPT_ZIO in xfwoptions, then the 23XX will do interrupt delays but post to the response queue- apparently QLogic *now* believes that reading multiple handles from registers is less of a win than writing (and delaying) multiple 64 byte responses to the response queue. At the end of taking a a good f/w crash dump, send the ISPASYNC_FW_DUMPED event to the outer layers (who can then do things like wake a user daemon to *fetch* the crash image, etc.).
2002-08-17 17:29:15 +00:00
/*
* QLogic recommends that FAST Posting be turned
* off for 23XX cards and instead allow the HBA
* to write response queue entries and interrupt
* after a delay (ZIO).
*/
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
icbp->icb_fwoptions &= ~ICBOPT_FAST_POST;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if ((fcp->isp_xfwoptions & ICBXOPT_TIMER_MASK) == ICBXOPT_ZIO) {
Don't set ZIO for 23XX for target mode (use fast posting instead). Use the correct number of handles for multihandle returns. Very, very, rarely on some SMP systems we've seen an 'unstable' type in the response queue. I dunno whether or not it's a bug in our handling, or whether there's a cache incoherency issue, but try to guard against it. MFC after: 2 weeks
2005-01-23 06:26:45 +00:00
icbp->icb_xfwoptions |= ICBXOPT_ZIO;
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
icbp->icb_idelaytimer = 10;
Don't set ZIO for 23XX for target mode (use fast posting instead). Use the correct number of handles for multihandle returns. Very, very, rarely on some SMP systems we've seen an 'unstable' type in the response queue. I dunno whether or not it's a bug in our handling, or whether there's a cache incoherency issue, but try to guard against it. MFC after: 2 weeks
2005-01-23 06:26:45 +00:00
}
Implement a call to get the actual link data rate (if 23XX) so we can set whether it's a 2Gps or 1Gps link. MFC after: 1 week
2001-10-01 03:45:54 +00:00
if (isp->isp_confopts & ISP_CFG_ONEGB) {
Misunderstanding documentation caused me to try and set 1Gbps/2Gps/Auto connection speed for the 2300 in the wrong offset in the ICB. Oops. Respect some QLogic errat wrt PCI errors on certain shared host/RISC registers.
2001-10-06 20:41:18 +00:00
icbp->icb_zfwoptions |= ICBZOPT_RATE_ONEGB;
Implement a call to get the actual link data rate (if 23XX) so we can set whether it's a 2Gps or 1Gps link. MFC after: 1 week
2001-10-01 03:45:54 +00:00
} else if (isp->isp_confopts & ISP_CFG_TWOGB) {
Misunderstanding documentation caused me to try and set 1Gbps/2Gps/Auto connection speed for the 2300 in the wrong offset in the ICB. Oops. Respect some QLogic errat wrt PCI errors on certain shared host/RISC registers.
2001-10-06 20:41:18 +00:00
icbp->icb_zfwoptions |= ICBZOPT_RATE_TWOGB;
Implement a call to get the actual link data rate (if 23XX) so we can set whether it's a 2Gps or 1Gps link. MFC after: 1 week
2001-10-01 03:45:54 +00:00
} else {
Misunderstanding documentation caused me to try and set 1Gbps/2Gps/Auto connection speed for the 2300 in the wrong offset in the ICB. Oops. Respect some QLogic errat wrt PCI errors on certain shared host/RISC registers.
2001-10-06 20:41:18 +00:00
icbp->icb_zfwoptions |= ICBZOPT_RATE_AUTO;
Implement a call to get the actual link data rate (if 23XX) so we can set whether it's a 2Gps or 1Gps link. MFC after: 1 week
2001-10-01 03:45:54 +00:00
}
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
if (fcp->isp_zfwoptions & ICBZOPT_50_OHM) {
icbp->icb_zfwoptions |= ICBZOPT_50_OHM;
}
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
}
Support for f/w crash dumps (2200 && 23XX). If you want QLogic to look at a potential f/w problem for FC cards, you really have to provide them info in the format they expect. This involves dumping a lot of hardware registers (> 300 16 bit registers) and a lot of SRAM (> 128KB minimum). Thus all of this code is #ifdef protected which will become an option so that the memory allocation of where to dump the crash image is pretty expensive. It's worth it if you have a reproducible problem because they have some tools that can tell them, given the f/w version, the precise state of everything. MFC after: 1 week
2002-02-17 06:38:22 +00:00
}
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
If we're using ancient (pre 1.17.0) 2100 f/w (for the cards that cannot load f/w images > 0x7fff words), set ISP_FW_ATTR_SCCLUN. We explicitly don't believe we can find attributes if f/w is < 1.17.0, so we have to set SCCLUN for the 1.15.37 f/w we're using manually- otherwise every target will replicate itself across all 16 supported luns for non-SCCLUN f/w. Correctly set things up for 23XX and either fast posting or ZIO. The 23XX, it turns out, does not support RIO. If you put a non-zero value in xfwoptions, this will disable fast posting. If you put ICBXOPT_ZIO in xfwoptions, then the 23XX will do interrupt delays but post to the response queue- apparently QLogic *now* believes that reading multiple handles from registers is less of a win than writing (and delaying) multiple 64 byte responses to the response queue. At the end of taking a a good f/w crash dump, send the ISPASYNC_FW_DUMPED event to the outer layers (who can then do things like wake a user daemon to *fetch* the crash image, etc.).
2002-08-17 17:29:15 +00:00
/*
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
* For 22XX > 2.1.26 && 23XX, set some options.
If we're using ancient (pre 1.17.0) 2100 f/w (for the cards that cannot load f/w images > 0x7fff words), set ISP_FW_ATTR_SCCLUN. We explicitly don't believe we can find attributes if f/w is < 1.17.0, so we have to set SCCLUN for the 1.15.37 f/w we're using manually- otherwise every target will replicate itself across all 16 supported luns for non-SCCLUN f/w. Correctly set things up for 23XX and either fast posting or ZIO. The 23XX, it turns out, does not support RIO. If you put a non-zero value in xfwoptions, this will disable fast posting. If you put ICBXOPT_ZIO in xfwoptions, then the 23XX will do interrupt delays but post to the response queue- apparently QLogic *now* believes that reading multiple handles from registers is less of a win than writing (and delaying) multiple 64 byte responses to the response queue. At the end of taking a a good f/w crash dump, send the ISPASYNC_FW_DUMPED event to the outer layers (who can then do things like wake a user daemon to *fetch* the crash image, etc.).
2002-08-17 17:29:15 +00:00
*/
if (ISP_FW_NEWER_THAN(isp, 2, 26, 0)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_SET_FIRMWARE_OPTIONS, MBLOGALL, 0);
Remove file unused in freebsd.
2009-09-21 01:38:22 +00:00
mbs.param[1] = IFCOPT1_DISF7SWTCH|IFCOPT1_LIPASYNC|IFCOPT1_LIPF8;
Enable LIP F8, LIP Reset async events. Be more chatty about SNS failures. Fix typo for skipped phase mesage. Correct MBOX_GET_PORT_QUEUE_PARAMS options in table. MFC after: 2 weeks
2001-08-16 17:25:08 +00:00
mbs.param[2] = 0;
mbs.param[3] = 0;
Remove file unused in freebsd.
2009-09-21 01:38:22 +00:00
if (ISP_FW_NEWER_THAN(isp, 3, 16, 0)) {
mbs.param[1] |= IFCOPT1_EQFQASYNC|IFCOPT1_CTIO_RETRY;
if (fcp->role & ISP_ROLE_TARGET) {
mbs.param[3] = IFCOPT3_NOPRLI;
}
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
return;
}
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
}
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
icbp->icb_logintime = ICB_LOGIN_TOV;
icbp->icb_lunetimeout = ICB_LUN_ENABLE_TOV;
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (fcp->isp_wwnn && fcp->isp_wwpn && (fcp->isp_wwnn >> 60) != 2) {
Fix a longstanding bug- we had the sense of what bit 14 for the ICB firmware options meant- *I* had taken it to mean that if you set it, Node Name would be ignored and derived from Port Name. Actually, it meant the opposite. As a consequence- change ICBOPT_USE_PORTNAME to the define ICBOPT_BOTH_WWNS- makes more sense. Fix wrong input bitmap for MBOX_DUMP_RAM command. Call ISP_DUMPREGS if we get a f/w crash. Add ISPCTL_RUN_MBOXCMD control command (so outer layers can run a mailbox command directly) and add a ISPASYNC_UNHANDLED_RESPONSE hook so outer layers can understand response queue entries we might not know about.
2001-02-23 05:35:50 +00:00
icbp->icb_fwoptions |= ICBOPT_BOTH_WWNS;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MAKE_NODE_NAME_FROM_WWN(icbp->icb_nodename, fcp->isp_wwnn);
MAKE_NODE_NAME_FROM_WWN(icbp->icb_portname, fcp->isp_wwpn);
Redo how default Node and Port WWNs are determined (again!). This is so we don't stomp on the differences between ports for a Qlogic 2202.
2000-10-12 23:49:09 +00:00
isp_prt(isp, ISP_LOGDEBUG1,
"Setting ICB Node 0x%08x%08x Port 0x%08x%08x",
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
((uint32_t) (fcp->isp_wwnn >> 32)),
((uint32_t) (fcp->isp_wwnn)),
((uint32_t) (fcp->isp_wwpn >> 32)),
((uint32_t) (fcp->isp_wwpn)));
} else if (fcp->isp_wwpn) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
icbp->icb_fwoptions &= ~ICBOPT_BOTH_WWNS;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MAKE_NODE_NAME_FROM_WWN(icbp->icb_portname, fcp->isp_wwpn);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_prt(isp, ISP_LOGDEBUG1,
"Setting ICB Port 0x%08x%08x",
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
((uint32_t) (fcp->isp_wwpn >> 32)),
((uint32_t) (fcp->isp_wwpn)));
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 02:28:29 +00:00
} else {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_prt(isp, ISP_LOGERR, "No valid WWNs to use");
return;
Remove the Target mode functions until they're in better shape. Implement some suggested compilation cleanups from Eklund. Wire down a hard loop id if we are not on a platform that has the ability to get to a PCI BIOS (it still will float to the ID it gets after a LIP but at least we can try). Clarify that the expanded lun is based upon SCCLUN defines (in f/w).
1998-12-05 01:33:57 +00:00
}
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
icbp->icb_rqstqlen = RQUEST_QUEUE_LEN(isp);
On reset, make sure that we have some parameters set correctly. This fixes a longstanding issue WRT resetting the chip after startup- it would fail if we were connected as an F-port to a switch. If we were connected as an F-port, we got assigned a hard loop ID of 255, which is really a bogus loop id. Then when we turned around to reset ourselves, the firmware would reject the ICB_INIT request because the loop id was bogus. *sputter* Minor fixlet from somebody in NetBSD with too much time on their hands (dma -> DMA).
2003-09-13 01:55:44 +00:00
if (icbp->icb_rqstqlen < 1) {
isp_prt(isp, ISP_LOGERR, "bad request queue length");
}
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
icbp->icb_rsltqlen = RESULT_QUEUE_LEN(isp);
On reset, make sure that we have some parameters set correctly. This fixes a longstanding issue WRT resetting the chip after startup- it would fail if we were connected as an F-port to a switch. If we were connected as an F-port, we got assigned a hard loop ID of 255, which is really a bogus loop id. Then when we turned around to reset ourselves, the firmware would reject the ICB_INIT request because the loop id was bogus. *sputter* Minor fixlet from somebody in NetBSD with too much time on their hands (dma -> DMA).
2003-09-13 01:55:44 +00:00
if (icbp->icb_rsltqlen < 1) {
isp_prt(isp, ISP_LOGERR, "bad result queue length");
}
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
icbp->icb_rqstaddr[RQRSP_ADDR0015] = DMA_WD0(isp->isp_rquest_dma);
icbp->icb_rqstaddr[RQRSP_ADDR1631] = DMA_WD1(isp->isp_rquest_dma);
icbp->icb_rqstaddr[RQRSP_ADDR3247] = DMA_WD2(isp->isp_rquest_dma);
icbp->icb_rqstaddr[RQRSP_ADDR4863] = DMA_WD3(isp->isp_rquest_dma);
icbp->icb_respaddr[RQRSP_ADDR0015] = DMA_WD0(isp->isp_result_dma);
icbp->icb_respaddr[RQRSP_ADDR1631] = DMA_WD1(isp->isp_result_dma);
icbp->icb_respaddr[RQRSP_ADDR3247] = DMA_WD2(isp->isp_result_dma);
icbp->icb_respaddr[RQRSP_ADDR4863] = DMA_WD3(isp->isp_result_dma);
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (FC_SCRATCH_ACQUIRE(isp, 0)) {
isp_prt(isp, ISP_LOGERR, sacq);
return;
}
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_prt(isp, ISP_LOGDEBUG0, "isp_fibre_init: fwopt 0x%x xfwopt 0x%x zfwopt 0x%x",
If we're using ancient (pre 1.17.0) 2100 f/w (for the cards that cannot load f/w images > 0x7fff words), set ISP_FW_ATTR_SCCLUN. We explicitly don't believe we can find attributes if f/w is < 1.17.0, so we have to set SCCLUN for the 1.15.37 f/w we're using manually- otherwise every target will replicate itself across all 16 supported luns for non-SCCLUN f/w. Correctly set things up for 23XX and either fast posting or ZIO. The 23XX, it turns out, does not support RIO. If you put a non-zero value in xfwoptions, this will disable fast posting. If you put ICBXOPT_ZIO in xfwoptions, then the 23XX will do interrupt delays but post to the response queue- apparently QLogic *now* believes that reading multiple handles from registers is less of a win than writing (and delaying) multiple 64 byte responses to the response queue. At the end of taking a a good f/w crash dump, send the ISPASYNC_FW_DUMPED event to the outer layers (who can then do things like wake a user daemon to *fetch* the crash image, etc.).
2002-08-17 17:29:15 +00:00
icbp->icb_fwoptions, icbp->icb_xfwoptions, icbp->icb_zfwoptions);
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
isp_put_icb(isp, icbp, (isp_icb_t *)fcp->isp_scratch);
When resetting the Qlogic 2X00 units, reset the FPM (Fibre Protocol Module) and FBM (Fibre Buffer Modules). Also remember to clear the semaphore registers. Tell the RISC processor to not halt on FPM parity errors. Throw out the ISP_CFG_NOINIT silliness and instead go to the use of adapter 'roles' to see whether one completes initialization or not (mostly for Fibre Channel). The ultimate intent, btw, of all of this is to have a warm standby adapter for failover reasons. Because we do roles now, setting of Target Capable Class 3 service parameters in the ICB for the 2x00 cards reflects from role. Also, in isp_start, if we're not supporting an initiator role, we bounce outgoing commands with a Selection Timeout error. Also clean out the TOGGLE_TMODE goop for FC- there is no toggling of target mode like there is for parallel SCSI cards. Do more cleanup with respect to using target ids 0..125 in F-port topologies. Also keep track of things which *were* fabric devices so that when you rescan the fabric you can notify the outer layers when fabric devices go away. Only force a LOGOUT for fabric devices if they're still logged in (i.e., you cat their Port Database entry. Clean up the Get All Next scanning. Finally, use a new tag in the softc to store the opcode for the last mailbox command used so we can report which opcode timed out.
2001-01-15 18:33:08 +00:00
More code cleanup. Go back to using FULL_LOGIN Fibre Chan if f/w is less than 1.17.0 level. Change where we do the loop database init. Add in the CMD_RQLATER return. Add some register debounce.
1999-08-16 19:59:55 +00:00
/*
When resetting the Qlogic 2X00 units, reset the FPM (Fibre Protocol Module) and FBM (Fibre Buffer Modules). Also remember to clear the semaphore registers. Tell the RISC processor to not halt on FPM parity errors. Throw out the ISP_CFG_NOINIT silliness and instead go to the use of adapter 'roles' to see whether one completes initialization or not (mostly for Fibre Channel). The ultimate intent, btw, of all of this is to have a warm standby adapter for failover reasons. Because we do roles now, setting of Target Capable Class 3 service parameters in the ICB for the 2x00 cards reflects from role. Also, in isp_start, if we're not supporting an initiator role, we bounce outgoing commands with a Selection Timeout error. Also clean out the TOGGLE_TMODE goop for FC- there is no toggling of target mode like there is for parallel SCSI cards. Do more cleanup with respect to using target ids 0..125 in F-port topologies. Also keep track of things which *were* fabric devices so that when you rescan the fabric you can notify the outer layers when fabric devices go away. Only force a LOGOUT for fabric devices if they're still logged in (i.e., you cat their Port Database entry. Clean up the Get All Next scanning. Finally, use a new tag in the softc to store the opcode for the last mailbox command used so we can report which opcode timed out.
2001-01-15 18:33:08 +00:00
* Init the firmware
More code cleanup. Go back to using FULL_LOGIN Fibre Chan if f/w is less than 1.17.0 level. Change where we do the loop database init. Add in the CMD_RQLATER return. Add some register debounce.
1999-08-16 19:59:55 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_INIT_FIRMWARE, MBLOGALL, 30000000);
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
mbs.param[2] = DMA_WD1(fcp->isp_scdma);
mbs.param[3] = DMA_WD0(fcp->isp_scdma);
mbs.param[6] = DMA_WD3(fcp->isp_scdma);
mbs.param[7] = DMA_WD2(fcp->isp_scdma);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.logval = MBLOGALL;
Some rearrangement of headers to minimize diffs with outside of FreeBSD repository and to clean up the license header so as to not pollute the license with file function. Zero all mailbox structures prior to use (just in case). Change the outgoing mailbox count for INIT_FIRMWARE to be correct.
2006-07-16 20:11:50 +00:00
isp_prt(isp, ISP_LOGDEBUG0, "INIT F/W from %p (%08x%08x)",
fcp->isp_scratch, (uint32_t) ((uint64_t)fcp->isp_scdma >> 32),
(uint32_t) fcp->isp_scdma);
Add some missing braces. Add MEMORY_BARRIER for the few scratch dma ops that were missing them plus add a couple of hi 32 bit dma ops (we could probably allow 64 bit scratch and request/response queue dma now).
2006-07-14 05:14:48 +00:00
MEMORYBARRIER(isp, SYNC_SFORDEV, 0, sizeof (*icbp));
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FC_SCRATCH_RELEASE(isp, 0);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Don't attempt to load illegal hard loop addresses into an ICB. This shows up on card restarts, and usually for 2200-2300 cards. What happens is that we start up, attempting to acquire a hard address. We end up instead being an F-port topology, which reports out a loop id of 0xff (or 0xffff for 2K Login f/w). Then, if we restart, we end up telling the card to go off an acquire this loop address, which the card then rejects. Bah. Compilation fixes from Solaris port.
2007-02-23 21:59:21 +00:00
isp_print_bytes(isp, "isp_fibre_init", sizeof (*icbp), icbp);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
return;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp->isp_reqidx = 0;
isp->isp_reqodx = 0;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp->isp_residx = 0;
/*
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
* Whatever happens, we're now committed to being here.
*/
isp->isp_state = ISP_INITSTATE;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
static void
isp_fibre_init_2400(ispsoftc_t *isp)
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
{
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
fcparam *fcp;
isp_icb_2400_t local, *icbp = &local;
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
mbreg_t mbs;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
int chan;
/*
* Check to see whether all channels have *some* kind of role
*/
for (chan = 0; chan < isp->isp_nchan; chan++) {
fcp = FCPARAM(isp, chan);
if (fcp->role != ISP_ROLE_NONE) {
break;
}
}
if (chan == isp->isp_nchan) {
isp_prt(isp, ISP_LOGDEBUG0, "all %d channels with role 'none'", chan);
isp->isp_state = ISP_INITSTATE;
return;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* Start with channel 0.
*/
fcp = FCPARAM(isp, 0);
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
* Turn on LIP F8 async event (1)
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_SET_FIRMWARE_OPTIONS, MBLOGALL, 0);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.param[1] = 1;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
return;
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(icbp, sizeof (*icbp));
icbp->icb_fwoptions1 = fcp->isp_fwoptions;
if (fcp->role & ISP_ROLE_TARGET) {
icbp->icb_fwoptions1 |= ICB2400_OPT1_TGT_ENABLE;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
icbp->icb_fwoptions1 &= ~ICB2400_OPT1_TGT_ENABLE;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (fcp->role & ISP_ROLE_INITIATOR) {
icbp->icb_fwoptions1 &= ~ICB2400_OPT1_INI_DISABLE;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
icbp->icb_fwoptions1 |= ICB2400_OPT1_INI_DISABLE;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
icbp->icb_version = ICB_VERSION1;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
icbp->icb_maxfrmlen = DEFAULT_FRAMESIZE(isp);
if (icbp->icb_maxfrmlen < ICB_MIN_FRMLEN || icbp->icb_maxfrmlen > ICB_MAX_FRMLEN) {
isp_prt(isp, ISP_LOGERR, "bad frame length (%d) from NVRAM- using %d", DEFAULT_FRAMESIZE(isp), ICB_DFLT_FRMLEN);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
icbp->icb_maxfrmlen = ICB_DFLT_FRMLEN;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
icbp->icb_execthrottle = DEFAULT_EXEC_THROTTLE(isp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (icbp->icb_execthrottle < 1) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGERR, "bad execution throttle of %d- using %d", DEFAULT_EXEC_THROTTLE(isp), ICB_DFLT_THROTTLE);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
icbp->icb_execthrottle = ICB_DFLT_THROTTLE;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (icbp->icb_fwoptions1 & ICB2400_OPT1_TGT_ENABLE) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
* Get current resource count
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_GET_RESOURCE_COUNT, MBLOGALL, 0);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.obits = 0x4cf;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
return;
}
icbp->icb_xchgcnt = mbs.param[3];
}
icbp->icb_hardaddr = fcp->isp_loopid;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (icbp->icb_hardaddr >= LOCAL_LOOP_LIM) {
Don't attempt to load illegal hard loop addresses into an ICB. This shows up on card restarts, and usually for 2200-2300 cards. What happens is that we start up, attempting to acquire a hard address. We end up instead being an F-port topology, which reports out a loop id of 0xff (or 0xffff for 2K Login f/w). Then, if we restart, we end up telling the card to go off an acquire this loop address, which the card then rejects. Bah. Compilation fixes from Solaris port.
2007-02-23 21:59:21 +00:00
icbp->icb_hardaddr = 0;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* Force this on.
*/
icbp->icb_fwoptions1 |= ICB2400_OPT1_HARD_ADDRESS;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
icbp->icb_fwoptions2 = fcp->isp_xfwoptions;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
switch (isp->isp_confopts & ISP_CFG_PORT_PREF) {
#if 0
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
case ISP_CFG_NPORT:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* XXX: This causes the f/w to crash.
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
icbp->icb_fwoptions2 &= ~ICB2400_OPT2_TOPO_MASK;
icbp->icb_fwoptions2 |= ICB2400_OPT2_PTP_2_LOOP;
break;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
#endif
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
case ISP_CFG_NPORT_ONLY:
icbp->icb_fwoptions2 &= ~ICB2400_OPT2_TOPO_MASK;
icbp->icb_fwoptions2 |= ICB2400_OPT2_PTP_ONLY;
break;
case ISP_CFG_LPORT_ONLY:
icbp->icb_fwoptions2 &= ~ICB2400_OPT2_TOPO_MASK;
icbp->icb_fwoptions2 |= ICB2400_OPT2_LOOP_ONLY;
break;
default:
icbp->icb_fwoptions2 &= ~ICB2400_OPT2_TOPO_MASK;
icbp->icb_fwoptions2 |= ICB2400_OPT2_LOOP_2_PTP;
break;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/* force this on for now */
icbp->icb_fwoptions2 |= ICB2400_OPT2_ZIO;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
switch (icbp->icb_fwoptions2 & ICB2400_OPT2_TIMER_MASK) {
case ICB2400_OPT2_ZIO:
case ICB2400_OPT2_ZIO1:
icbp->icb_idelaytimer = 0;
break;
case 0:
break;
default:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGWARN, "bad value %x in fwopt2 timer field", icbp->icb_fwoptions2 & ICB2400_OPT2_TIMER_MASK);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
icbp->icb_fwoptions2 &= ~ICB2400_OPT2_TIMER_MASK;
break;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* We don't support FCTAPE, so clear it.
*/
icbp->icb_fwoptions2 &= ~ICB2400_OPT2_FCTAPE;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
icbp->icb_fwoptions3 = fcp->isp_zfwoptions;
icbp->icb_fwoptions3 &= ~ICB2400_OPT3_RATE_AUTO;
if (isp->isp_confopts & ISP_CFG_ONEGB) {
icbp->icb_fwoptions3 |= ICB2400_OPT3_RATE_ONEGB;
} else if (isp->isp_confopts & ISP_CFG_TWOGB) {
icbp->icb_fwoptions3 |= ICB2400_OPT3_RATE_TWOGB;
} else if (isp->isp_confopts & ISP_CFG_FOURGB) {
icbp->icb_fwoptions3 |= ICB2400_OPT3_RATE_FOURGB;
} else {
icbp->icb_fwoptions3 |= ICB2400_OPT3_RATE_AUTO;
}
if ((isp->isp_confopts & ISP_CFG_OWNLOOPID) == 0) {
icbp->icb_fwoptions3 |= ICB2400_OPT3_SOFTID;
}
icbp->icb_logintime = ICB_LOGIN_TOV;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (fcp->isp_wwnn && fcp->isp_wwpn && (fcp->isp_wwnn >> 60) != 2) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
icbp->icb_fwoptions1 |= ICB2400_OPT1_BOTH_WWNS;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MAKE_NODE_NAME_FROM_WWN(icbp->icb_portname, fcp->isp_wwpn);
MAKE_NODE_NAME_FROM_WWN(icbp->icb_nodename, fcp->isp_wwnn);
isp_prt(isp, ISP_LOGDEBUG1, "Setting ICB Node 0x%08x%08x Port 0x%08x%08x", ((uint32_t) (fcp->isp_wwnn >> 32)), ((uint32_t) (fcp->isp_wwnn)),
((uint32_t) (fcp->isp_wwpn >> 32)), ((uint32_t) (fcp->isp_wwpn)));
} else if (fcp->isp_wwpn) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
icbp->icb_fwoptions1 &= ~ICB2400_OPT1_BOTH_WWNS;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MAKE_NODE_NAME_FROM_WWN(icbp->icb_portname, fcp->isp_wwpn);
isp_prt(isp, ISP_LOGDEBUG1, "Setting ICB Node to be same as Port 0x%08x%08x", ((uint32_t) (fcp->isp_wwpn >> 32)), ((uint32_t) (fcp->isp_wwpn)));
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else {
isp_prt(isp, ISP_LOGERR, "No valid WWNs to use");
return;
}
icbp->icb_retry_count = fcp->isp_retry_count;
icbp->icb_rqstqlen = RQUEST_QUEUE_LEN(isp);
if (icbp->icb_rqstqlen < 8) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGERR, "bad request queue length %d", icbp->icb_rqstqlen);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return;
}
icbp->icb_rsltqlen = RESULT_QUEUE_LEN(isp);
if (icbp->icb_rsltqlen < 8) {
isp_prt(isp, ISP_LOGERR, "bad result queue length %d",
icbp->icb_rsltqlen);
return;
}
icbp->icb_rqstaddr[RQRSP_ADDR0015] = DMA_WD0(isp->isp_rquest_dma);
icbp->icb_rqstaddr[RQRSP_ADDR1631] = DMA_WD1(isp->isp_rquest_dma);
icbp->icb_rqstaddr[RQRSP_ADDR3247] = DMA_WD2(isp->isp_rquest_dma);
icbp->icb_rqstaddr[RQRSP_ADDR4863] = DMA_WD3(isp->isp_rquest_dma);
icbp->icb_respaddr[RQRSP_ADDR0015] = DMA_WD0(isp->isp_result_dma);
icbp->icb_respaddr[RQRSP_ADDR1631] = DMA_WD1(isp->isp_result_dma);
icbp->icb_respaddr[RQRSP_ADDR3247] = DMA_WD2(isp->isp_result_dma);
icbp->icb_respaddr[RQRSP_ADDR4863] = DMA_WD3(isp->isp_result_dma);
#ifdef ISP_TARGET_MODE
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/* unconditionally set up the ATIO queue if we support target mode */
icbp->icb_atioqlen = RESULT_QUEUE_LEN(isp);
if (icbp->icb_atioqlen < 8) {
isp_prt(isp, ISP_LOGERR, "bad ATIO queue length %d", icbp->icb_atioqlen);
return;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
icbp->icb_atioqaddr[RQRSP_ADDR0015] = DMA_WD0(isp->isp_atioq_dma);
icbp->icb_atioqaddr[RQRSP_ADDR1631] = DMA_WD1(isp->isp_atioq_dma);
icbp->icb_atioqaddr[RQRSP_ADDR3247] = DMA_WD2(isp->isp_atioq_dma);
icbp->icb_atioqaddr[RQRSP_ADDR4863] = DMA_WD3(isp->isp_atioq_dma);
isp_prt(isp, ISP_LOGDEBUG0, "isp_fibre_init_2400: atioq %04x%04x%04x%04x", DMA_WD3(isp->isp_atioq_dma), DMA_WD2(isp->isp_atioq_dma),
DMA_WD1(isp->isp_atioq_dma), DMA_WD0(isp->isp_atioq_dma));
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
#endif
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGDEBUG0, "isp_fibre_init_2400: fwopt1 0x%x fwopt2 0x%x fwopt3 0x%x", icbp->icb_fwoptions1, icbp->icb_fwoptions2, icbp->icb_fwoptions3);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGDEBUG0, "isp_fibre_init_2400: rqst %04x%04x%04x%04x rsp %04x%04x%04x%04x", DMA_WD3(isp->isp_rquest_dma), DMA_WD2(isp->isp_rquest_dma),
DMA_WD1(isp->isp_rquest_dma), DMA_WD0(isp->isp_rquest_dma), DMA_WD3(isp->isp_result_dma), DMA_WD2(isp->isp_result_dma),
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
DMA_WD1(isp->isp_result_dma), DMA_WD0(isp->isp_result_dma));
if (isp->isp_dblev & ISP_LOGDEBUG1) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_print_bytes(isp, "isp_fibre_init_2400", sizeof (*icbp), icbp);
}
if (FC_SCRATCH_ACQUIRE(isp, 0)) {
isp_prt(isp, ISP_LOGERR, sacq);
return;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(fcp->isp_scratch, ISP_FC_SCRLEN);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_icb_2400(isp, icbp, fcp->isp_scratch);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* Now fill in information about any additional channels
*/
if (isp->isp_nchan > 1) {
isp_icb_2400_vpinfo_t vpinfo, *vdst;
vp_port_info_t pi, *pdst;
size_t amt = 0;
uint8_t *off;
vpinfo.vp_count = isp->isp_nchan - 1;
vpinfo.vp_global_options = 0;
off = fcp->isp_scratch;
off += ICB2400_VPINFO_OFF;
vdst = (isp_icb_2400_vpinfo_t *) off;
isp_put_icb_2400_vpinfo(isp, &vpinfo, vdst);
amt = ICB2400_VPINFO_OFF + sizeof (isp_icb_2400_vpinfo_t);
for (chan = 1; chan < isp->isp_nchan; chan++) {
fcparam *fcp2;
ISP_MEMZERO(&pi, sizeof (pi));
fcp2 = FCPARAM(isp, chan);
if (fcp2->role != ISP_ROLE_NONE) {
pi.vp_port_options = ICB2400_VPOPT_ENABLED;
if (fcp2->role & ISP_ROLE_INITIATOR) {
pi.vp_port_options |= ICB2400_VPOPT_INI_ENABLE;
}
if ((fcp2->role & ISP_ROLE_TARGET) == 0) {
pi.vp_port_options |= ICB2400_VPOPT_TGT_DISABLE;
}
MAKE_NODE_NAME_FROM_WWN(pi.vp_port_portname, fcp2->isp_wwpn);
MAKE_NODE_NAME_FROM_WWN(pi.vp_port_nodename, fcp2->isp_wwnn);
}
off = fcp->isp_scratch;
off += ICB2400_VPINFO_PORT_OFF(chan);
pdst = (vp_port_info_t *) off;
isp_put_vp_port_info(isp, &pi, pdst);
amt += ICB2400_VPOPT_WRITE_SIZE;
}
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
* Init the firmware
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, 0, MBLOGALL, 30000000);
if (isp->isp_nchan > 1) {
mbs.param[0] = MBOX_INIT_FIRMWARE_MULTI_ID;
} else {
mbs.param[0] = MBOX_INIT_FIRMWARE;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.param[2] = DMA_WD1(fcp->isp_scdma);
mbs.param[3] = DMA_WD0(fcp->isp_scdma);
mbs.param[6] = DMA_WD3(fcp->isp_scdma);
mbs.param[7] = DMA_WD2(fcp->isp_scdma);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGDEBUG0, "INIT F/W from %04x%04x%04x%04x", DMA_WD3(fcp->isp_scdma), DMA_WD2(fcp->isp_scdma), DMA_WD1(fcp->isp_scdma), DMA_WD0(fcp->isp_scdma));
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
MEMORYBARRIER(isp, SYNC_SFORDEV, 0, sizeof (*icbp));
isp_mboxcmd(isp, &mbs);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FC_SCRATCH_RELEASE(isp, 0);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
return;
}
isp->isp_reqidx = 0;
isp->isp_reqodx = 0;
isp->isp_residx = 0;
/*
* Whatever happens, we're now committed to being here.
*/
isp->isp_state = ISP_INITSTATE;
}
static void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_mark_portdb(ispsoftc_t *isp, int chan, int disposition)
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
{
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcparam *fcp = FCPARAM(isp, chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
int i;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (chan < 0 || chan >= isp->isp_nchan) {
isp_prt(isp, ISP_LOGWARN, "isp_mark_portdb: bad channel %d", chan);
return;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
for (i = 0; i < MAX_FC_TARG; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (fcp->portdb[i].target_mode) {
if (disposition < 0) {
isp_prt(isp, ISP_LOGTINFO, "isp_mark_portdb: Chan %d zeroing handle 0x" "%04x port 0x%06x", chan,
fcp->portdb[i].handle, fcp->portdb[i].portid);
ISP_MEMZERO(&fcp->portdb[i], sizeof (fcportdb_t));
}
continue;
}
if (disposition == 0) {
ISP_MEMZERO(&fcp->portdb[i], sizeof (fcportdb_t));
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else {
switch (fcp->portdb[i].state) {
case FC_PORTDB_STATE_CHANGED:
case FC_PORTDB_STATE_PENDING_VALID:
case FC_PORTDB_STATE_VALID:
case FC_PORTDB_STATE_PROBATIONAL:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcp->portdb[i].state = FC_PORTDB_STATE_PROBATIONAL;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
case FC_PORTDB_STATE_ZOMBIE:
break;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
case FC_PORTDB_STATE_NIL:
default:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(&fcp->portdb[i], sizeof (fcportdb_t));
fcp->portdb[i].state = FC_PORTDB_STATE_NIL;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
}
}
}
}
/*
* Perform an IOCB PLOGI or LOGO via EXECUTE IOCB A64 for 24XX cards
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
* or via FABRIC LOGIN/FABRIC LOGOUT for other cards.
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*/
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
static int
Remove file unused in freebsd.
2009-09-21 01:38:22 +00:00
isp_plogx(ispsoftc_t *isp, int chan, uint16_t handle, uint32_t portid, int flags, int gs)
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
{
mbreg_t mbs;
uint8_t q[QENTRY_LEN];
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
isp_plogx_t *plp;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcparam *fcp;
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
uint8_t *scp;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
uint32_t sst, parm1;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
int rval, lev;
const char *msg;
char buf[64];
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
if (!IS_24XX(isp)) {
int action = flags & PLOGX_FLG_CMD_MASK;
if (action == PLOGX_FLG_CMD_PLOGI) {
return (isp_port_login(isp, handle, portid));
} else if (action == PLOGX_FLG_CMD_LOGO) {
return (isp_port_logout(isp, handle, portid));
} else {
return (MBOX_INVALID_COMMAND);
}
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(q, QENTRY_LEN);
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
plp = (isp_plogx_t *) q;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
plp->plogx_header.rqs_entry_count = 1;
plp->plogx_header.rqs_entry_type = RQSTYPE_LOGIN;
plp->plogx_handle = 0xffffffff;
plp->plogx_nphdl = handle;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
plp->plogx_vphdl = chan;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
plp->plogx_portlo = portid;
plp->plogx_rspsz_porthi = (portid >> 16) & 0xff;
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
plp->plogx_flags = flags;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (isp->isp_dblev & ISP_LOGDEBUG1) {
isp_print_bytes(isp, "IOCB LOGX", QENTRY_LEN, plp);
}
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
if (gs == 0) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (FC_SCRATCH_ACQUIRE(isp, chan)) {
isp_prt(isp, ISP_LOGERR, sacq);
return (-1);
}
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcp = FCPARAM(isp, chan);
scp = fcp->isp_scratch;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_plogx(isp, plp, (isp_plogx_t *) scp);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_EXEC_COMMAND_IOCB_A64, MBLOGALL, 500000);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.param[1] = QENTRY_LEN;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
mbs.param[2] = DMA_WD1(fcp->isp_scdma);
mbs.param[3] = DMA_WD0(fcp->isp_scdma);
mbs.param[6] = DMA_WD3(fcp->isp_scdma);
mbs.param[7] = DMA_WD2(fcp->isp_scdma);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
MEMORYBARRIER(isp, SYNC_SFORDEV, 0, QENTRY_LEN);
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
rval = mbs.param[0];
goto out;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
MEMORYBARRIER(isp, SYNC_SFORCPU, QENTRY_LEN, QENTRY_LEN);
scp += QENTRY_LEN;
isp_get_plogx(isp, (isp_plogx_t *) scp, plp);
if (isp->isp_dblev & ISP_LOGDEBUG1) {
isp_print_bytes(isp, "IOCB LOGX response", QENTRY_LEN, plp);
}
if (plp->plogx_status == PLOGX_STATUS_OK) {
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
rval = 0;
goto out;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else if (plp->plogx_status != PLOGX_STATUS_IOCBERR) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGWARN,
"status 0x%x on port login IOCB chanel %d",
plp->plogx_status, chan);
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
rval = -1;
goto out;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
sst = plp->plogx_ioparm[0].lo16 | (plp->plogx_ioparm[0].hi16 << 16);
parm1 = plp->plogx_ioparm[1].lo16 | (plp->plogx_ioparm[1].hi16 << 16);
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
rval = -1;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
lev = ISP_LOGERR;
msg = NULL;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
switch (sst) {
case PLOGX_IOCBERR_NOLINK:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
msg = "no link";
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case PLOGX_IOCBERR_NOIOCB:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
msg = "no IOCB buffer";
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case PLOGX_IOCBERR_NOXGHG:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
msg = "no Exchange Control Block";
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case PLOGX_IOCBERR_FAILED:
Don't try and re-use a handle, even if the firmware tells you that's what is logged in. PR: kern/144026 MFC after: 1 week
2010-02-18 18:35:09 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "reason 0x%x (last LOGIN state 0x%x)", parm1 & 0xff, (parm1 >> 8) & 0xff);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
msg = buf;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case PLOGX_IOCBERR_NOFABRIC:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
msg = "no fabric";
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case PLOGX_IOCBERR_NOTREADY:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
msg = "firmware not ready";
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case PLOGX_IOCBERR_NOLOGIN:
Don't try and re-use a handle, even if the firmware tells you that's what is logged in. PR: kern/144026 MFC after: 1 week
2010-02-18 18:35:09 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "not logged in (last state 0x%x)", parm1);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
msg = buf;
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
rval = MBOX_NOT_LOGGED_IN;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case PLOGX_IOCBERR_REJECT:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "LS_RJT = 0x%x", parm1);
msg = buf;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case PLOGX_IOCBERR_NOPCB:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
msg = "no PCB allocated";
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case PLOGX_IOCBERR_EINVAL:
Don't try and re-use a handle, even if the firmware tells you that's what is logged in. PR: kern/144026 MFC after: 1 week
2010-02-18 18:35:09 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "invalid parameter at offset 0x%x", parm1);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
msg = buf;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case PLOGX_IOCBERR_PORTUSED:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
lev = ISP_LOGSANCFG|ISP_LOGDEBUG0;
Don't try and re-use a handle, even if the firmware tells you that's what is logged in. PR: kern/144026 MFC after: 1 week
2010-02-18 18:35:09 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "already logged in with N-Port handle 0x%x", parm1);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
msg = buf;
rval = MBOX_PORT_ID_USED | (parm1 << 16);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case PLOGX_IOCBERR_HNDLUSED:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
lev = ISP_LOGSANCFG|ISP_LOGDEBUG0;
Don't try and re-use a handle, even if the firmware tells you that's what is logged in. PR: kern/144026 MFC after: 1 week
2010-02-18 18:35:09 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "handle already used for PortID 0x%06x", parm1);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
msg = buf;
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
rval = MBOX_LOOP_ID_USED;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case PLOGX_IOCBERR_NOHANDLE:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
msg = "no handle allocated";
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case PLOGX_IOCBERR_NOFLOGI:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
msg = "no FLOGI_ACC";
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
default:
Don't try and re-use a handle, even if the firmware tells you that's what is logged in. PR: kern/144026 MFC after: 1 week
2010-02-18 18:35:09 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "status %x from %x", plp->plogx_status, flags);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
msg = buf;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (msg) {
Don't try and re-use a handle, even if the firmware tells you that's what is logged in. PR: kern/144026 MFC after: 1 week
2010-02-18 18:35:09 +00:00
isp_prt(isp, ISP_LOGERR, "Chan %d PLOGX PortID 0x%06x to N-Port handle 0x%x: %s", chan, portid, handle, msg);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
out:
if (gs == 0) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FC_SCRATCH_RELEASE(isp, chan);
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
}
return (rval);
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
}
static int
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_port_login(ispsoftc_t *isp, uint16_t handle, uint32_t portid)
{
mbreg_t mbs;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_FABRIC_LOGIN, MBLOGNONE, 500000);
if (ISP_CAP_2KLOGIN(isp)) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.param[1] = handle;
mbs.ibits = (1 << 10);
} else {
mbs.param[1] = handle << 8;
}
mbs.param[2] = portid >> 16;
mbs.param[3] = portid;
mbs.logval = MBLOGNONE;
Increase the timeout for some SAN commands. Only complain about FC Reponse errors if they're nonzero. Shorten some PortID printouts for local loop. Add an internal isp_xcmd_t data structure which we'll use for some CT-Passthru support as part of adding SMI-S.
2006-11-16 00:39:56 +00:00
mbs.timeout = 500000;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
switch (mbs.param[0]) {
case MBOX_PORT_ID_USED:
isp_prt(isp, ISP_LOGDEBUG0,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"isp_port_login: portid 0x%06x already logged in as %u",
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
portid, mbs.param[1]);
return (MBOX_PORT_ID_USED | (mbs.param[1] << 16));
case MBOX_LOOP_ID_USED:
isp_prt(isp, ISP_LOGDEBUG0,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"isp_port_login: handle 0x%04x in use for port id 0x%02xXXXX",
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
handle, mbs.param[1] & 0xff);
return (MBOX_LOOP_ID_USED);
case MBOX_COMMAND_COMPLETE:
return (0);
case MBOX_COMMAND_ERROR:
isp_prt(isp, ISP_LOGINFO,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"isp_port_login: error 0x%x in PLOGI to port 0x%06x",
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.param[1], portid);
return (MBOX_COMMAND_ERROR);
case MBOX_ALL_IDS_USED:
isp_prt(isp, ISP_LOGINFO,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"isp_port_login: all IDs used for fabric login");
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (MBOX_ALL_IDS_USED);
default:
isp_prt(isp, ISP_LOGINFO,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"isp_port_login: error 0x%x on port login of 0x%06x@0x%0x",
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.param[0], portid, handle);
return (mbs.param[0]);
}
}
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
static int
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_port_logout(ispsoftc_t *isp, uint16_t handle, uint32_t portid)
{
mbreg_t mbs;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_FABRIC_LOGOUT, MBLOGNONE, 500000);
if (ISP_CAP_2KLOGIN(isp)) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.param[1] = handle;
mbs.ibits = (1 << 10);
} else {
mbs.param[1] = handle << 8;
}
isp_mboxcmd(isp, &mbs);
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
return (mbs.param[0] == MBOX_COMMAND_COMPLETE? 0 : mbs.param[0]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
static int
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_getpdb(ispsoftc_t *isp, int chan, uint16_t id, isp_pdb_t *pdb, int dolock)
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
{
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcparam *fcp = FCPARAM(isp, chan);
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
mbreg_t mbs;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
union {
isp_pdb_21xx_t fred;
isp_pdb_24xx_t bill;
} un;
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_GET_PORT_DB, MBLOGALL & ~MBOX_COMMAND_PARAM_ERROR, 250000);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
mbs.ibits = (1 << 9)|(1 << 10);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
mbs.param[1] = id;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
mbs.param[9] = chan;
} else if (ISP_CAP_2KLOGIN(isp)) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.param[1] = id;
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
} else {
mbs.param[1] = id << 8;
}
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
mbs.param[2] = DMA_WD1(fcp->isp_scdma);
mbs.param[3] = DMA_WD0(fcp->isp_scdma);
mbs.param[6] = DMA_WD3(fcp->isp_scdma);
mbs.param[7] = DMA_WD2(fcp->isp_scdma);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (dolock) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (FC_SCRATCH_ACQUIRE(isp, chan)) {
isp_prt(isp, ISP_LOGERR, sacq);
return (-1);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
MEMORYBARRIER(isp, SYNC_SFORDEV, 0, sizeof (un));
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
if (dolock) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FC_SCRATCH_RELEASE(isp, chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
return (mbs.param[0]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
if (IS_24XX(isp)) {
isp_get_pdb_24xx(isp, fcp->isp_scratch, &un.bill);
pdb->handle = un.bill.pdb_handle;
pdb->s3_role = un.bill.pdb_prli_svc3;
pdb->portid = BITS2WORD_24XX(un.bill.pdb_portid_bits);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMCPY(pdb->portname, un.bill.pdb_portname, 8);
ISP_MEMCPY(pdb->nodename, un.bill.pdb_nodename, 8);
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
"Chan %d Port 0x%06x flags 0x%x curstate %x",
chan, pdb->portid, un.bill.pdb_flags,
un.bill.pdb_curstate);
if (un.bill.pdb_curstate < PDB2400_STATE_PLOGI_DONE ||
un.bill.pdb_curstate > PDB2400_STATE_LOGGED_IN) {
mbs.param[0] = MBOX_NOT_LOGGED_IN;
if (dolock) {
FC_SCRATCH_RELEASE(isp, chan);
}
return (mbs.param[0]);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else {
isp_get_pdb_21xx(isp, fcp->isp_scratch, &un.fred);
pdb->handle = un.fred.pdb_loopid;
pdb->s3_role = un.fred.pdb_prli_svc3;
pdb->portid = BITS2WORD(un.fred.pdb_portid_bits);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMCPY(pdb->portname, un.fred.pdb_portname, 8);
ISP_MEMCPY(pdb->nodename, un.fred.pdb_nodename, 8);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
if (dolock) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FC_SCRATCH_RELEASE(isp, chan);
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (0);
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
static void
isp_dump_chip_portdb(ispsoftc_t *isp, int chan, int dolock)
{
isp_pdb_t pdb;
int lim, loopid;
if (ISP_CAP_2KLOGIN(isp)) {
lim = NPH_MAX_2K;
} else {
lim = NPH_MAX;
}
for (loopid = 0; loopid != lim; loopid++) {
if (isp_getpdb(isp, chan, loopid, &pdb, dolock)) {
continue;
}
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGINFO, "Chan %d Loopid 0x%04x "
"PortID 0x%06x WWPN 0x%02x%02x%02x%02x%02x%02x%02x%02x",
chan, loopid, pdb.portid, pdb.portname[0], pdb.portname[1],
pdb.portname[2], pdb.portname[3], pdb.portname[4],
pdb.portname[5], pdb.portname[6], pdb.portname[7]);
}
}
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
static uint64_t
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_get_wwn(ispsoftc_t *isp, int chan, int loopid, int nodename)
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
{
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
uint64_t wwn = INI_NONE;
fcparam *fcp = FCPARAM(isp, chan);
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
mbreg_t mbs;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (fcp->isp_fwstate < FW_READY ||
fcp->isp_loopstate < LOOP_PDB_RCVD) {
return (wwn);
}
MBSINIT(&mbs, MBOX_GET_PORT_NAME, MBLOGALL & ~MBOX_COMMAND_PARAM_ERROR, 500000);
if (ISP_CAP_2KLOGIN(isp)) {
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
mbs.param[1] = loopid;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.ibits = (1 << 10);
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
if (nodename) {
mbs.param[10] = 1;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (ISP_CAP_MULTI_ID(isp)) {
mbs.ibits |= (1 << 9);
mbs.param[9] = chan;
}
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
} else {
mbs.param[1] = loopid << 8;
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
if (nodename) {
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
mbs.param[1] |= 1;
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
}
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
return (wwn);
}
if (IS_24XX(isp)) {
wwn =
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
(((uint64_t)(mbs.param[2] >> 8)) << 56) |
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
(((uint64_t)(mbs.param[2] & 0xff)) << 48) |
(((uint64_t)(mbs.param[3] >> 8)) << 40) |
(((uint64_t)(mbs.param[3] & 0xff)) << 32) |
(((uint64_t)(mbs.param[6] >> 8)) << 24) |
(((uint64_t)(mbs.param[6] & 0xff)) << 16) |
(((uint64_t)(mbs.param[7] >> 8)) << 8) |
(((uint64_t)(mbs.param[7] & 0xff)));
} else {
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
wwn =
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
(((uint64_t)(mbs.param[2] & 0xff)) << 56) |
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
(((uint64_t)(mbs.param[2] >> 8)) << 48) |
(((uint64_t)(mbs.param[3] & 0xff)) << 40) |
(((uint64_t)(mbs.param[3] >> 8)) << 32) |
(((uint64_t)(mbs.param[6] & 0xff)) << 24) |
(((uint64_t)(mbs.param[6] >> 8)) << 16) |
(((uint64_t)(mbs.param[7] & 0xff)) << 8) |
(((uint64_t)(mbs.param[7] >> 8)));
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
}
return (wwn);
}
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* Make sure we have good FC link.
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
*/
static int
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_fclink_test(ispsoftc_t *isp, int chan, int usdelay)
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
{
mbreg_t mbs;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
int count, check_for_fabric, r;
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
uint8_t lwfs;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
int loopid;
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
fcparam *fcp;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
fcportdb_t *lp;
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
isp_pdb_t pdb;
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcp = FCPARAM(isp, chan);
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0, "Chan %d FC Link Test Entry", chan);
ISP_MARK_PORTDB(isp, chan, 1);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
/*
* Wait up to N microseconds for F/W to go to a ready state.
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
lwfs = FW_CONFIG_WAIT;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
count = 0;
while (count < usdelay) {
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
uint64_t enano;
uint32_t wrk;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
NANOTIME_T hra, hrb;
GET_NANOTIME(&hra);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_fw_state(isp, chan);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (lwfs != fcp->isp_fwstate) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGCONFIG|ISP_LOGSANCFG, "Chan %d Firmware State <%s->%s>", chan, isp_fc_fw_statename((int)lwfs), isp_fc_fw_statename((int)fcp->isp_fwstate));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
lwfs = fcp->isp_fwstate;
}
if (fcp->isp_fwstate == FW_READY) {
break;
}
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
GET_NANOTIME(&hrb);
/*
* Get the elapsed time in nanoseconds.
* Always guaranteed to be non-zero.
*/
enano = NANOTIME_SUB(&hrb, &hra);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGDEBUG1, "usec%d: 0x%lx->0x%lx enano 0x%x%08x", count, (long) GET_NANOSEC(&hra), (long) GET_NANOSEC(&hrb), (uint32_t)(enano >> 32), (uint32_t)(enano));
Add in Bill Sommerfelds -Wformat changes. Set up default node && port WWNs correctly (Again!) - this time for the case that we're not going to fully init the adapter if isp_init is called (with ISP_CFG_NOINIT set in options). The pupose for this is to bring the adapter up to almost ready to go, get info out of NVRAM, but to not start it up- leaving it until later to actually start things up if wanted (and possibly with different roles selected).
2000-12-29 19:12:44 +00:00
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
/*
* If the elapsed time is less than 1 millisecond,
* delay a period of time up to that millisecond of
* waiting.
Add in Bill Sommerfelds -Wformat changes. Set up default node && port WWNs correctly (Again!) - this time for the case that we're not going to fully init the adapter if isp_init is called (with ISP_CFG_NOINIT set in options). The pupose for this is to bring the adapter up to almost ready to go, get info out of NVRAM, but to not start it up- leaving it until later to actually start things up if wanted (and possibly with different roles selected).
2000-12-29 19:12:44 +00:00
*
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
* This peculiar code is an attempt to try and avoid
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
* invoking uint64_t math support functions for some
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
* platforms where linkage is a problem.
*/
if (enano < (1000 * 1000)) {
count += 1000;
enano = (1000 * 1000) - enano;
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
while (enano > (uint64_t) 4000000000U) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SLEEP(isp, 4000000);
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
enano -= (uint64_t) 4000000000U;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
wrk = enano;
In order to save ourselves grief with the SUNPRO compiler under Solaris (which, for reasons unknown to me, chokes on u_int16_t as a typedef of unsigned short if used in a transitional (mixed K&R and ANSI) way), we'll go the extra mile and fully ANSIfy things.
2001-03-14 04:11:56 +00:00
wrk /= 1000;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SLEEP(isp, wrk);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
} else {
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
while (enano > (uint64_t) 4000000000U) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
count += 4000000;
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
enano -= (uint64_t) 4000000000U;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
wrk = enano;
count += (wrk / 1000);
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
* If we haven't gone to 'ready' state, return.
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
if (fcp->isp_fwstate != FW_READY) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGSANCFG, "%s: chan %d not at FW_READY state", __func__, chan);
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
return (-1);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* Get our Loop ID and Port ID.
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_GET_LOOP_ID, MBLOGALL, 0);
if (ISP_CAP_MULTI_ID(isp)) {
mbs.param[9] = chan;
mbs.ibits = (1 << 9);
mbs.obits = (1 << 7);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
return (-1);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (ISP_CAP_2KLOGIN(isp)) {
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
fcp->isp_loopid = mbs.param[1];
} else {
fcp->isp_loopid = mbs.param[1] & 0xff;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_2100(isp)) {
fcp->isp_topo = TOPO_NL_PORT;
} else {
Fix some breakage about how we build WWNs. Do some other fabric related changes: consider a new PDB entry different if Class 3 service parameter roles change (!!!). Do some checking as we're getting a port database that traps whether things change while we're doing so. Handle N-port and F-ports correctly. Fix the fabric login loop to retain a login/binding if things haven't changed (I mean, why logout a device only to log it back in). No longer accept, after fabric logins, garbage if we can't get a PDB entry that matches the device we've just logged into- if it doesn't, log it out as it is very unlikely to still be what we thought it was. Get rid of some of the debounce loops because we could get stuck there.
2000-05-09 01:14:43 +00:00
int topo = (int) mbs.param[6];
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (topo < TOPO_NL_PORT || topo > TOPO_PTP_STUB) {
Fix some breakage about how we build WWNs. Do some other fabric related changes: consider a new PDB entry different if Class 3 service parameter roles change (!!!). Do some checking as we're getting a port database that traps whether things change while we're doing so. Handle N-port and F-ports correctly. Fix the fabric login loop to retain a login/binding if things haven't changed (I mean, why logout a device only to log it back in). No longer accept, after fabric logins, garbage if we can't get a PDB entry that matches the device we've just logged into- if it doesn't, log it out as it is very unlikely to still be what we thought it was. Get rid of some of the debounce loops because we could get stuck there.
2000-05-09 01:14:43 +00:00
topo = TOPO_PTP_STUB;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Fix some breakage about how we build WWNs. Do some other fabric related changes: consider a new PDB entry different if Class 3 service parameter roles change (!!!). Do some checking as we're getting a port database that traps whether things change while we're doing so. Handle N-port and F-ports correctly. Fix the fabric login loop to retain a login/binding if things haven't changed (I mean, why logout a device only to log it back in). No longer accept, after fabric logins, garbage if we can't get a PDB entry that matches the device we've just logged into- if it doesn't, log it out as it is very unlikely to still be what we thought it was. Get rid of some of the debounce loops because we could get stuck there.
2000-05-09 01:14:43 +00:00
fcp->isp_topo = topo;
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
}
Make sure we don't pick up a loopid that's larger than our current portdb max (MAX_FC_TARG == 256) now that we support 2K Login f/w. MFC after: 3 days
2006-02-02 09:02:16 +00:00
fcp->isp_portid = mbs.param[2] | (mbs.param[3] << 16);
Fix some breakage about how we build WWNs. Do some other fabric related changes: consider a new PDB entry different if Class 3 service parameter roles change (!!!). Do some checking as we're getting a port database that traps whether things change while we're doing so. Handle N-port and F-ports correctly. Fix the fabric login loop to retain a login/binding if things haven't changed (I mean, why logout a device only to log it back in). No longer accept, after fabric logins, garbage if we can't get a PDB entry that matches the device we've just logged into- if it doesn't, log it out as it is very unlikely to still be what we thought it was. Get rid of some of the debounce loops because we could get stuck there.
2000-05-09 01:14:43 +00:00
If we're using ancient (pre 1.17.0) 2100 f/w (for the cards that cannot load f/w images > 0x7fff words), set ISP_FW_ATTR_SCCLUN. We explicitly don't believe we can find attributes if f/w is < 1.17.0, so we have to set SCCLUN for the 1.15.37 f/w we're using manually- otherwise every target will replicate itself across all 16 supported luns for non-SCCLUN f/w. Correctly set things up for 23XX and either fast posting or ZIO. The 23XX, it turns out, does not support RIO. If you put a non-zero value in xfwoptions, this will disable fast posting. If you put ICBXOPT_ZIO in xfwoptions, then the 23XX will do interrupt delays but post to the response queue- apparently QLogic *now* believes that reading multiple handles from registers is less of a win than writing (and delaying) multiple 64 byte responses to the response queue. At the end of taking a a good f/w crash dump, send the ISPASYNC_FW_DUMPED event to the outer layers (who can then do things like wake a user daemon to *fetch* the crash image, etc.).
2002-08-17 17:29:15 +00:00
if (IS_2100(isp)) {
/*
* Don't bother with fabric if we are using really old
* 2100 firmware. It's just not worth it.
*/
if (ISP_FW_NEWER_THAN(isp, 1, 15, 37)) {
check_for_fabric = 1;
} else {
check_for_fabric = 0;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
} else if (fcp->isp_topo == TOPO_FL_PORT || fcp->isp_topo == TOPO_F_PORT) {
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
check_for_fabric = 1;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else {
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
check_for_fabric = 0;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* Check to make sure we got a valid loopid
* The 24XX seems to mess this up for multiple channels.
*/
if (fcp->isp_topo == TOPO_FL_PORT || fcp->isp_topo == TOPO_NL_PORT) {
uint8_t alpa = fcp->isp_portid;
if (alpa == 0) {
/* "Cannot Happen" */
isp_prt(isp, ISP_LOGWARN, "Zero AL_PA for Loop Topology?");
} else {
int i;
for (i = 0; alpa_map[i]; i++) {
if (alpa_map[i] == alpa) {
break;
}
}
if (alpa_map[i] && fcp->isp_loopid != i) {
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0, "Chan %d deriving loopid %d from AL_PA map (AL_PA 0x%x) and ignoring returned value %d (AL_PA 0x%x)", chan, i, alpa_map[i], fcp->isp_loopid, alpa);
fcp->isp_loopid = i;
}
}
}
if (IS_24XX(isp)) { /* XXX SHOULDN'T THIS BE FOR 2K F/W? XXX */
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
loopid = NPH_FL_ID;
} else {
loopid = FL_ID;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (check_for_fabric) {
r = isp_getpdb(isp, chan, loopid, &pdb, 1);
if (r && (fcp->isp_topo == TOPO_F_PORT || fcp->isp_topo == TOPO_FL_PORT)) {
isp_prt(isp, ISP_LOGWARN, "fabric topology but cannot get info about fabric controller (0x%x)", r);
fcp->isp_topo = TOPO_PTP_STUB;
}
} else {
r = -1;
}
if (r == 0) {
Fix some breakage about how we build WWNs. Do some other fabric related changes: consider a new PDB entry different if Class 3 service parameter roles change (!!!). Do some checking as we're getting a port database that traps whether things change while we're doing so. Handle N-port and F-ports correctly. Fix the fabric login loop to retain a login/binding if things haven't changed (I mean, why logout a device only to log it back in). No longer accept, after fabric logins, garbage if we can't get a PDB entry that matches the device we've just logged into- if it doesn't, log it out as it is very unlikely to still be what we thought it was. Get rid of some of the debounce loops because we could get stuck there.
2000-05-09 01:14:43 +00:00
if (IS_2100(isp)) {
Pick up topology more sanely at f/w startup. Change the restrictions of where we can have targets (based on topology). Much more importantly, make sure all mods to isp_sendmarker or |= so we don't lose the marking of a bus that needs to have a marker sent for it.
2000-04-21 02:04:34 +00:00
fcp->isp_topo = TOPO_FL_PORT;
Fix some breakage about how we build WWNs. Do some other fabric related changes: consider a new PDB entry different if Class 3 service parameter roles change (!!!). Do some checking as we're getting a port database that traps whether things change while we're doing so. Handle N-port and F-ports correctly. Fix the fabric login loop to retain a login/binding if things haven't changed (I mean, why logout a device only to log it back in). No longer accept, after fabric logins, garbage if we can't get a PDB entry that matches the device we've just logged into- if it doesn't, log it out as it is very unlikely to still be what we thought it was. Get rid of some of the debounce loops because we could get stuck there.
2000-05-09 01:14:43 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (pdb.portid == 0) {
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
/*
* Crock.
*/
fcp->isp_topo = TOPO_NL_PORT;
goto not_on_fabric;
}
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
/*
Fix some breakage about how we build WWNs. Do some other fabric related changes: consider a new PDB entry different if Class 3 service parameter roles change (!!!). Do some checking as we're getting a port database that traps whether things change while we're doing so. Handle N-port and F-ports correctly. Fix the fabric login loop to retain a login/binding if things haven't changed (I mean, why logout a device only to log it back in). No longer accept, after fabric logins, garbage if we can't get a PDB entry that matches the device we've just logged into- if it doesn't, log it out as it is very unlikely to still be what we thought it was. Get rid of some of the debounce loops because we could get stuck there.
2000-05-09 01:14:43 +00:00
* Save the Fabric controller's port database entry.
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
lp = &fcp->portdb[FL_ID];
lp->state = FC_PORTDB_STATE_PENDING_VALID;
MAKE_WWN_FROM_NODE_NAME(lp->node_wwn, pdb.nodename);
MAKE_WWN_FROM_NODE_NAME(lp->port_wwn, pdb.portname);
lp->roles = (pdb.s3_role & SVC3_ROLE_MASK) >> SVC3_ROLE_SHIFT;
lp->portid = pdb.portid;
lp->handle = pdb.handle;
lp->new_portid = lp->portid;
lp->new_roles = lp->roles;
if (IS_24XX(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcp->inorder = (mbs.param[7] & ISP24XX_INORDER) != 0;
if (ISP_FW_NEWER_THAN(isp, 4, 0, 27)) {
fcp->npiv_fabric = (mbs.param[7] & ISP24XX_NPIV_SAN) != 0;
if (fcp->npiv_fabric) {
isp_prt(isp, ISP_LOGCONFIG, "fabric supports NP-IV");
}
}
if (chan) {
fcp->isp_sns_hdl = NPH_SNS_HDLBASE + chan;
r = isp_plogx(isp, chan, fcp->isp_sns_hdl, SNS_PORT_ID, PLOGX_FLG_CMD_PLOGI | PLOGX_FLG_COND_PLOGI | PLOGX_FLG_SKIP_PRLI, 0);
if (r) {
isp_prt(isp, ISP_LOGWARN, "%s: Chan %d cannot log into SNS", __func__, chan);
return (-1);
}
} else {
fcp->isp_sns_hdl = NPH_SNS_ID;
}
r = isp_register_fc4_type_24xx(isp, chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcp->isp_sns_hdl = SNS_ID;
r = isp_register_fc4_type(isp, chan);
Check the return from registering FC4 types with the fabric name server. Don't complain about a hard loop id of 0xffff- we get this in point-to-point topologies with the 2300 and 2K Login firmware. Up the timeout on register FC4 types commands.
2007-01-05 22:59:26 +00:00
}
if (r) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGWARN|ISP_LOGSANCFG, "%s: register fc4 type failed", __func__);
Check the return from registering FC4 types with the fabric name server. Don't complain about a hard loop id of 0xffff- we get this in point-to-point topologies with the 2300 and 2K Login firmware. Up the timeout on register FC4 types commands.
2007-01-05 22:59:26 +00:00
return (-1);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
} else {
not_on_fabric:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
fcp->portdb[FL_ID].state = FC_PORTDB_STATE_NIL;
Fix some breakage about how we build WWNs. Do some other fabric related changes: consider a new PDB entry different if Class 3 service parameter roles change (!!!). Do some checking as we're getting a port database that traps whether things change while we're doing so. Handle N-port and F-ports correctly. Fix the fabric login loop to retain a login/binding if things haven't changed (I mean, why logout a device only to log it back in). No longer accept, after fabric logins, garbage if we can't get a PDB entry that matches the device we've just logged into- if it doesn't, log it out as it is very unlikely to still be what we thought it was. Get rid of some of the debounce loops because we could get stuck there.
2000-05-09 01:14:43 +00:00
}
Implement a call to get the actual link data rate (if 23XX) so we can set whether it's a 2Gps or 1Gps link. MFC after: 1 week
2001-10-01 03:45:54 +00:00
fcp->isp_gbspeed = 1;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_23XX(isp) || IS_24XX(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_GET_SET_DATA_RATE, MBLOGALL, 3000000);
Implement a call to get the actual link data rate (if 23XX) so we can set whether it's a 2Gps or 1Gps link. MFC after: 1 week
2001-10-01 03:45:54 +00:00
mbs.param[1] = MBGSD_GET_RATE;
/* mbs.param[2] undefined if we're just getting rate */
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Implement a call to get the actual link data rate (if 23XX) so we can set whether it's a 2Gps or 1Gps link. MFC after: 1 week
2001-10-01 03:45:54 +00:00
if (mbs.param[0] == MBOX_COMMAND_COMPLETE) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (mbs.param[1] == MBGSD_EIGHTGB) {
isp_prt(isp, ISP_LOGINFO, "Chan %d 8Gb link speed", chan);
fcp->isp_gbspeed = 8;
} else if (mbs.param[1] == MBGSD_FOURGB) {
isp_prt(isp, ISP_LOGINFO, "Chan %d 4Gb link speed", chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
fcp->isp_gbspeed = 4;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
} else if (mbs.param[1] == MBGSD_TWOGB) {
isp_prt(isp, ISP_LOGINFO, "Chan %d 2Gb link speed", chan);
Implement a call to get the actual link data rate (if 23XX) so we can set whether it's a 2Gps or 1Gps link. MFC after: 1 week
2001-10-01 03:45:54 +00:00
fcp->isp_gbspeed = 2;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
} else if (mbs.param[1] == MBGSD_ONEGB) {
isp_prt(isp, ISP_LOGINFO, "Chan %d 1Gb link speed", chan);
fcp->isp_gbspeed = 1;
Implement a call to get the actual link data rate (if 23XX) so we can set whether it's a 2Gps or 1Gps link. MFC after: 1 week
2001-10-01 03:45:54 +00:00
}
}
}
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* Announce ourselves, too.
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGCONFIG, topology, chan, (uint32_t) (fcp->isp_wwpn >> 32), (uint32_t) fcp->isp_wwpn, fcp->isp_portid, fcp->isp_loopid, isp_fc_toponame(fcp));
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0, "Chan %d FC Link Test Complete", chan);
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
return (0);
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
}
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* Complete the synchronization of our Port Database.
*
* At this point, we've scanned the local loop (if any) and the fabric
* and performed fabric logins on all new devices.
*
* Our task here is to go through our port database and remove any entities
* that are still marked probational (issuing PLOGO for ones which we had
* PLOGI'd into) or are dead.
*
* Our task here is to also check policy to decide whether devices which
* have *changed* in some way should still be kept active. For example,
* if a device has just changed PortID, we can either elect to treat it
* as an old device or as a newly arrived device (and notify the outer
* layer appropriately).
*
* We also do initiator map target id assignment here for new initiator
* devices and refresh old ones ot make sure that they point to the corret
* entities.
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
*/
static int
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_pdb_sync(ispsoftc_t *isp, int chan)
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
{
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcparam *fcp = FCPARAM(isp, chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
fcportdb_t *lp;
uint16_t dbidx;
if (fcp->isp_loopstate == LOOP_READY) {
return (0);
}
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
/*
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
* Make sure we're okay for doing this right now.
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
*/
Remove file unused in freebsd.
2009-09-21 01:38:22 +00:00
if (fcp->isp_loopstate != LOOP_PDB_RCVD &&
fcp->isp_loopstate != LOOP_FSCAN_DONE &&
fcp->isp_loopstate != LOOP_LSCAN_DONE) {
isp_prt(isp, ISP_LOGWARN, "isp_pdb_sync: bad loopstate %d",
fcp->isp_loopstate);
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
return (-1);
}
Remove file unused in freebsd.
2009-09-21 01:38:22 +00:00
if (fcp->isp_topo == TOPO_FL_PORT ||
fcp->isp_topo == TOPO_NL_PORT ||
fcp->isp_topo == TOPO_N_PORT) {
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
if (fcp->isp_loopstate < LOOP_LSCAN_DONE) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (isp_scan_loop(isp, chan) != 0) {
Remove file unused in freebsd.
2009-09-21 01:38:22 +00:00
isp_prt(isp, ISP_LOGWARN,
"isp_pdb_sync: isp_scan_loop failed");
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
return (-1);
}
}
}
Fix some breakage about how we build WWNs. Do some other fabric related changes: consider a new PDB entry different if Class 3 service parameter roles change (!!!). Do some checking as we're getting a port database that traps whether things change while we're doing so. Handle N-port and F-ports correctly. Fix the fabric login loop to retain a login/binding if things haven't changed (I mean, why logout a device only to log it back in). No longer accept, after fabric logins, garbage if we can't get a PDB entry that matches the device we've just logged into- if it doesn't, log it out as it is very unlikely to still be what we thought it was. Get rid of some of the debounce loops because we could get stuck there.
2000-05-09 01:14:43 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (fcp->isp_topo == TOPO_F_PORT || fcp->isp_topo == TOPO_FL_PORT) {
if (fcp->isp_loopstate < LOOP_FSCAN_DONE) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (isp_scan_fabric(isp, chan) != 0) {
Remove file unused in freebsd.
2009-09-21 01:38:22 +00:00
isp_prt(isp, ISP_LOGWARN,
"isp_pdb_sync: isp_scan_fabric failed");
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (-1);
}
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
}
}
Remove file unused in freebsd.
2009-09-21 01:38:22 +00:00
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
"Chan %d Synchronizing PDBs", chan);
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
fcp->isp_loopstate = LOOP_SYNCING_PDB;
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
for (dbidx = 0; dbidx < MAX_FC_TARG; dbidx++) {
lp = &fcp->portdb[dbidx];
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (lp->state == FC_PORTDB_STATE_NIL || lp->target_mode) {
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
continue;
}
More code cleanup. Go back to using FULL_LOGIN Fibre Chan if f/w is less than 1.17.0 level. Change where we do the loop database init. Add in the CMD_RQLATER return. Add some register debounce.
1999-08-16 19:59:55 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (lp->state == FC_PORTDB_STATE_VALID) {
if (dbidx != FL_ID) {
isp_prt(isp,
ISP_LOGERR, "portdb idx %d already valid",
dbidx);
}
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
continue;
}
More code cleanup. Go back to using FULL_LOGIN Fibre Chan if f/w is less than 1.17.0 level. Change where we do the loop database init. Add in the CMD_RQLATER return. Add some register debounce.
1999-08-16 19:59:55 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
switch (lp->state) {
case FC_PORTDB_STATE_PROBATIONAL:
case FC_PORTDB_STATE_DEAD:
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
/*
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* It's up to the outer layers to clear isp_dev_map.
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
lp->state = FC_PORTDB_STATE_NIL;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_async(isp, ISPASYNC_DEV_GONE, chan, lp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (lp->autologin == 0) {
Remove file unused in freebsd.
2009-09-21 01:38:22 +00:00
(void) isp_plogx(isp, chan, lp->handle,
lp->portid,
PLOGX_FLG_CMD_LOGO |
PLOGX_FLG_IMPLICIT |
PLOGX_FLG_FREE_NPHDL, 0);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
} else {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
lp->autologin = 0;
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
lp->new_roles = 0;
lp->new_portid = 0;
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
/*
* Note that we might come out of this with our state
* set to FC_PORTDB_STATE_ZOMBIE.
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case FC_PORTDB_STATE_NEW:
/*
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
* It's up to the outer layers to assign a virtual
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* target id in isp_dev_map (if any).
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*/
lp->portid = lp->new_portid;
lp->roles = lp->new_roles;
lp->state = FC_PORTDB_STATE_VALID;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_async(isp, ISPASYNC_DEV_ARRIVED, chan, lp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
lp->new_roles = 0;
lp->new_portid = 0;
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
lp->reserved = 0;
lp->new_reserved = 0;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case FC_PORTDB_STATE_CHANGED:
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
/*
* XXXX FIX THIS
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
lp->state = FC_PORTDB_STATE_VALID;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_async(isp, ISPASYNC_DEV_CHANGED, chan, lp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
lp->new_roles = 0;
lp->new_portid = 0;
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
lp->reserved = 0;
lp->new_reserved = 0;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case FC_PORTDB_STATE_PENDING_VALID:
lp->portid = lp->new_portid;
lp->roles = lp->new_roles;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (lp->dev_map_idx) {
int t = lp->dev_map_idx - 1;
fcp->isp_dev_map[t] = dbidx + 1;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
lp->state = FC_PORTDB_STATE_VALID;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_async(isp, ISPASYNC_DEV_STAYED, chan, lp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (dbidx != FL_ID) {
lp->new_roles = 0;
lp->new_portid = 0;
}
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
lp->reserved = 0;
lp->new_reserved = 0;
break;
case FC_PORTDB_STATE_ZOMBIE:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
default:
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
isp_prt(isp, ISP_LOGWARN,
"isp_scan_loop: state %d for idx %d",
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
lp->state, dbidx);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_dump_portdb(isp, chan);
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
}
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
/*
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
* If we get here, we've for sure seen not only a valid loop
* but know what is or isn't on it, so mark this for usage
* in isp_start.
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
*/
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
fcp->loop_seen_once = 1;
fcp->isp_loopstate = LOOP_READY;
return (0);
}
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
* Scan local loop for devices.
*/
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
static int
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_scan_loop(ispsoftc_t *isp, int chan)
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
{
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
fcportdb_t *lp, tmp;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcparam *fcp = FCPARAM(isp, chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
int i;
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
isp_pdb_t pdb;
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
uint16_t handle, lim = 0;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (fcp->isp_fwstate < FW_READY ||
fcp->isp_loopstate < LOOP_PDB_RCVD) {
return (-1);
}
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (fcp->isp_loopstate > LOOP_SCANNING_LOOP) {
return (0);
}
/*
* Check our connection topology.
*
* If we're a public or private loop, we scan 0..125 as handle values.
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* The firmware has (typically) peformed a PLOGI for us. We skip this
* step if we're a ISP_24XX in NP-IV mode.
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*
* If we're a N-port connection, we treat this is a short loop (0..1).
*/
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
switch (fcp->isp_topo) {
case TOPO_NL_PORT:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
lim = LOCAL_LOOP_LIM;
break;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
case TOPO_FL_PORT:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (IS_24XX(isp) && isp->isp_nchan > 1) {
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
"Chan %d Skipping Local Loop Scan", chan);
fcp->isp_loopstate = LOOP_LSCAN_DONE;
return (0);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
lim = LOCAL_LOOP_LIM;
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
break;
case TOPO_N_PORT:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
lim = 2;
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
break;
default:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
"Chan %d no loop topology to scan", chan);
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
fcp->isp_loopstate = LOOP_LSCAN_DONE;
return (0);
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
}
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
fcp->isp_loopstate = LOOP_SCANNING_LOOP;
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
"Chan %d FC scan loop 0..%d", chan, lim-1);
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
When resetting the Qlogic 2X00 units, reset the FPM (Fibre Protocol Module) and FBM (Fibre Buffer Modules). Also remember to clear the semaphore registers. Tell the RISC processor to not halt on FPM parity errors. Throw out the ISP_CFG_NOINIT silliness and instead go to the use of adapter 'roles' to see whether one completes initialization or not (mostly for Fibre Channel). The ultimate intent, btw, of all of this is to have a warm standby adapter for failover reasons. Because we do roles now, setting of Target Capable Class 3 service parameters in the ICB for the 2x00 cards reflects from role. Also, in isp_start, if we're not supporting an initiator role, we bounce outgoing commands with a Selection Timeout error. Also clean out the TOGGLE_TMODE goop for FC- there is no toggling of target mode like there is for parallel SCSI cards. Do more cleanup with respect to using target ids 0..125 in F-port topologies. Also keep track of things which *were* fabric devices so that when you rescan the fabric you can notify the outer layers when fabric devices go away. Only force a LOGOUT for fabric devices if they're still logged in (i.e., you cat their Port Database entry. Clean up the Get All Next scanning. Finally, use a new tag in the softc to store the opcode for the last mailbox command used so we can report which opcode timed out.
2001-01-15 18:33:08 +00:00
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* Run through the list and get the port database info for each one.
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
*/
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
for (handle = 0; handle < lim; handle++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
int r;
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
/*
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* Don't scan "special" ids.
*/
if (handle >= FL_ID && handle <= SNS_ID) {
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
continue;
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (ISP_CAP_2KLOGIN(isp)) {
if (handle >= NPH_RESERVED && handle <= NPH_FL_ID) {
continue;
}
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
* In older cards with older f/w GET_PORT_DATABASE has been
* known to hang. This trick gets around that problem.
*/
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
if (IS_2100(isp) || IS_2200(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
uint64_t node_wwn = isp_get_wwn(isp, chan, handle, 1);
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
if (fcp->isp_loopstate < LOOP_SCANNING_LOOP) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
"Chan %d FC scan loop DONE (bad)", chan);
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
return (-1);
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (node_wwn == INI_NONE) {
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
continue;
}
Add a isp_register_fc4_type function so that we work with McData switches that require us to register our FC4 types of interest. Allow ourselves, in F-port topologies, to start logging in fabric devices in the target 0..125 range. Change ISPASYNC_PDB_CHANGED (misnamed) to ISPASYNC_LOGGED_INOUT. Fix (*SMACK*) again some default WWN stuff. This is *really* hard to get right across all the range of platforms.
2001-01-09 02:46:23 +00:00
}
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* Get the port database entity for this index.
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
r = isp_getpdb(isp, chan, handle, &pdb, 1);
if (r != 0) {
isp_prt(isp, ISP_LOGDEBUG1,
"Chan %d FC scan loop handle %d returned %x",
chan, handle, r);
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
if (fcp->isp_loopstate < LOOP_SCANNING_LOOP) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MARK_PORTDB(isp, chan, 1);
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
"Chan %d FC scan loop DONE (bad)", chan);
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
return (-1);
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
}
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
continue;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
if (fcp->isp_loopstate < LOOP_SCANNING_LOOP) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MARK_PORTDB(isp, chan, 1);
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
"Chan %d FC scan loop DONE (bad)", chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (-1);
}
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* On *very* old 2100 firmware we would end up sometimes
* with the firmware returning the port database entry
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
* for something else. We used to restart this, but
* now we just punt.
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
*/
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
if (IS_2100(isp) && pdb.handle != handle) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_prt(isp, ISP_LOGWARN,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d cannot synchronize port database", chan);
ISP_MARK_PORTDB(isp, chan, 1);
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
"Chan %d FC scan loop DONE (bad)", chan);
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
return (-1);
}
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
If the HBA is already 'touched', still set maxluns. Othewise for CAM_QUIRK_HILUN devices we loop thru 32bits of lun. Oops. Switch to using USEC_DELAY rather than USEC_SLEEP at isp_reset time. Try to paper around a defect in clients that don't correctly registers themeselves with the fabric nameserver. Minor updates for Mirapoint support- they still use code that is not HANDLE_LOOPSTATE_IN_OUTER_LAYERS, and, surprise surprise, this old stuff had some bugs in it. Clean up some target mode stuff. MFC after: 1 week
2002-06-16 05:18:22 +00:00
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* Save the pertinent info locally.
If the HBA is already 'touched', still set maxluns. Othewise for CAM_QUIRK_HILUN devices we loop thru 32bits of lun. Oops. Switch to using USEC_DELAY rather than USEC_SLEEP at isp_reset time. Try to paper around a defect in clients that don't correctly registers themeselves with the fabric nameserver. Minor updates for Mirapoint support- they still use code that is not HANDLE_LOOPSTATE_IN_OUTER_LAYERS, and, surprise surprise, this old stuff had some bugs in it. Clean up some target mode stuff. MFC after: 1 week
2002-06-16 05:18:22 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
MAKE_WWN_FROM_NODE_NAME(tmp.node_wwn, pdb.nodename);
MAKE_WWN_FROM_NODE_NAME(tmp.port_wwn, pdb.portname);
tmp.roles = (pdb.s3_role & SVC3_ROLE_MASK) >> SVC3_ROLE_SHIFT;
tmp.portid = pdb.portid;
tmp.handle = pdb.handle;
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* Check to make sure it's still a valid entry. The 24XX seems
* to return a portid but not a WWPN/WWNN or role for devices
Don't attempt to load illegal hard loop addresses into an ICB. This shows up on card restarts, and usually for 2200-2300 cards. What happens is that we start up, attempting to acquire a hard address. We end up instead being an F-port topology, which reports out a loop id of 0xff (or 0xffff for 2K Login f/w). Then, if we restart, we end up telling the card to go off an acquire this loop address, which the card then rejects. Bah. Compilation fixes from Solaris port.
2007-02-23 21:59:21 +00:00
* which shift on a loop.
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (tmp.node_wwn == 0 || tmp.port_wwn == 0 || tmp.portid == 0) {
Don't attempt to load illegal hard loop addresses into an ICB. This shows up on card restarts, and usually for 2200-2300 cards. What happens is that we start up, attempting to acquire a hard address. We end up instead being an F-port topology, which reports out a loop id of 0xff (or 0xffff for 2K Login f/w). Then, if we restart, we end up telling the card to go off an acquire this loop address, which the card then rejects. Bah. Compilation fixes from Solaris port.
2007-02-23 21:59:21 +00:00
int a, b, c;
a = (tmp.node_wwn == 0);
b = (tmp.port_wwn == 0);
c = (tmp.portid == 0);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (a == 0 && b == 0) {
tmp.node_wwn =
isp_get_wwn(isp, chan, handle, 1);
tmp.port_wwn =
isp_get_wwn(isp, chan, handle, 0);
if (tmp.node_wwn && tmp.port_wwn) {
isp_prt(isp, ISP_LOGINFO, "DODGED!");
goto cont;
}
}
Don't attempt to load illegal hard loop addresses into an ICB. This shows up on card restarts, and usually for 2200-2300 cards. What happens is that we start up, attempting to acquire a hard address. We end up instead being an F-port topology, which reports out a loop id of 0xff (or 0xffff for 2K Login f/w). Then, if we restart, we end up telling the card to go off an acquire this loop address, which the card then rejects. Bah. Compilation fixes from Solaris port.
2007-02-23 21:59:21 +00:00
isp_prt(isp, ISP_LOGWARN,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d bad pdb (%1d%1d%1d) @ handle 0x%x", chan,
a, b, c, handle);
isp_dump_portdb(isp, chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
continue;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
cont:
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* Now search the entire port database
* for the same Port and Node WWN.
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
for (i = 0; i < MAX_FC_TARG; i++) {
lp = &fcp->portdb[i];
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
Remove file unused in freebsd.
2009-09-21 01:38:22 +00:00
if (lp->state == FC_PORTDB_STATE_NIL ||
lp->target_mode) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
continue;
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (lp->node_wwn != tmp.node_wwn) {
continue;
}
if (lp->port_wwn != tmp.port_wwn) {
continue;
}
/*
* Okay- we've found a non-nil entry that matches.
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
* Check to make sure it's probational or a zombie.
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*/
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
if (lp->state != FC_PORTDB_STATE_PROBATIONAL &&
lp->state != FC_PORTDB_STATE_ZOMBIE) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_prt(isp, ISP_LOGERR,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d [%d] not probational/zombie (0x%x)",
chan, i, lp->state);
isp_dump_portdb(isp, chan);
ISP_MARK_PORTDB(isp, chan, 1);
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
"Chan %d FC scan loop DONE (bad)", chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (-1);
}
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
/*
* Mark the device as something the f/w logs into
* automatically.
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
lp->autologin = 1;
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
* Check to make see if really still the same
* device. If it is, we mark it pending valid.
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*/
if (lp->portid == tmp.portid &&
lp->handle == tmp.handle &&
lp->roles == tmp.roles) {
lp->new_portid = tmp.portid;
lp->new_roles = tmp.roles;
lp->state = FC_PORTDB_STATE_PENDING_VALID;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
"Chan %d Loop Port 0x%06x@0x%04x Pending "
"Valid", chan, tmp.portid, tmp.handle);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
* We can wipe out the old handle value
* here because it's no longer valid.
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*/
lp->handle = tmp.handle;
/*
* Claim that this has changed and let somebody else
* decide what to do.
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
"Chan %d Loop Port 0x%06x@0x%04x changed",
chan, tmp.portid, tmp.handle);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
lp->state = FC_PORTDB_STATE_CHANGED;
lp->new_portid = tmp.portid;
lp->new_roles = tmp.roles;
break;
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
* Did we find and update an old entry?
*/
if (i < MAX_FC_TARG) {
continue;
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
* Ah. A new device entry. Find an empty slot
* for it and save info for later disposition.
*/
for (i = 0; i < MAX_FC_TARG; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (fcp->portdb[i].target_mode) {
continue;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (fcp->portdb[i].state == FC_PORTDB_STATE_NIL) {
Protect against deranged fabric nameservers that spit out 10000 identical port numbers. MFC after: 1 day
2001-10-18 17:26:52 +00:00
break;
}
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (i == MAX_FC_TARG) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGERR,
"Chan %d out of portdb entries", chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
continue;
}
lp = &fcp->portdb[i];
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(lp, sizeof (fcportdb_t));
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
lp->autologin = 1;
lp->state = FC_PORTDB_STATE_NEW;
lp->new_portid = tmp.portid;
lp->new_roles = tmp.roles;
lp->handle = tmp.handle;
lp->port_wwn = tmp.port_wwn;
lp->node_wwn = tmp.node_wwn;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
"Chan %d Loop Port 0x%06x@0x%04x is New Entry",
chan, tmp.portid, tmp.handle);
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
fcp->isp_loopstate = LOOP_LSCAN_DONE;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
"Chan %d FC scan loop DONE", chan);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
return (0);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
* Scan the fabric for devices and add them to our port database.
*
* Use the GID_FT command to get all Port IDs for FC4 SCSI devices it knows.
*
* For 2100-23XX cards, we can use the SNS mailbox command to pass simple
* name server commands to the switch management server via the QLogic f/w.
*
* For the 24XX card, we have to use CT-Pass through run via the Execute IOCB
* mailbox command.
*
* The net result is to leave the list of Port IDs setting untranslated in
* offset IGPOFF of the FC scratch area, whereupon we'll canonicalize it to
* host order at OGPOFF.
*/
/*
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* Take less than half of our scratch area to store Port IDs
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
#define GIDLEN ((ISP_FC_SCRLEN >> 1) - 16 - SNS_GID_FT_REQ_SIZE)
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
#define NGENT ((GIDLEN - 16) >> 2)
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
#define IGPOFF (2 * QENTRY_LEN)
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
#define OGPOFF (ISP_FC_SCRLEN >> 1)
#define ZTXOFF (ISP_FC_SCRLEN - (1 * QENTRY_LEN))
#define CTXOFF (ISP_FC_SCRLEN - (2 * QENTRY_LEN))
#define XTXOFF (ISP_FC_SCRLEN - (3 * QENTRY_LEN))
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
static int
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_gid_ft_sns(ispsoftc_t *isp, int chan)
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
{
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
union {
sns_gid_ft_req_t _x;
uint8_t _y[SNS_GID_FT_REQ_SIZE];
} un;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcparam *fcp = FCPARAM(isp, chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
sns_gid_ft_req_t *rq = &un._x;
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
mbreg_t mbs;
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGDEBUG0,
"Chan %d scanning fabric (GID_FT) via SNS", chan);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(rq, SNS_GID_FT_REQ_SIZE);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
rq->snscb_rblen = GIDLEN >> 1;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
rq->snscb_addr[RQRSP_ADDR0015] = DMA_WD0(fcp->isp_scdma + IGPOFF);
rq->snscb_addr[RQRSP_ADDR1631] = DMA_WD1(fcp->isp_scdma + IGPOFF);
rq->snscb_addr[RQRSP_ADDR3247] = DMA_WD2(fcp->isp_scdma + IGPOFF);
rq->snscb_addr[RQRSP_ADDR4863] = DMA_WD3(fcp->isp_scdma + IGPOFF);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
rq->snscb_sblen = 6;
rq->snscb_cmd = SNS_GID_FT;
rq->snscb_mword_div_2 = NGENT;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
rq->snscb_fc4_type = FC4_SCSI;
isp_put_gid_ft_request(isp, rq, fcp->isp_scratch);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
MEMORYBARRIER(isp, SYNC_SFORDEV, 0, SNS_GID_FT_REQ_SIZE);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_SEND_SNS, MBLOGALL, 10000000);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
mbs.param[0] = MBOX_SEND_SNS;
mbs.param[1] = SNS_GID_FT_REQ_SIZE >> 1;
mbs.param[2] = DMA_WD1(fcp->isp_scdma);
mbs.param[3] = DMA_WD0(fcp->isp_scdma);
mbs.param[6] = DMA_WD3(fcp->isp_scdma);
mbs.param[7] = DMA_WD2(fcp->isp_scdma);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
if (mbs.param[0] == MBOX_INVALID_COMMAND) {
return (1);
} else {
return (-1);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
return (0);
}
static int
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_gid_ft_ct_passthru(ispsoftc_t *isp, int chan)
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
{
mbreg_t mbs;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcparam *fcp = FCPARAM(isp, chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
union {
isp_ct_pt_t plocal;
ct_hdr_t clocal;
uint8_t q[QENTRY_LEN];
} un;
isp_ct_pt_t *pt;
ct_hdr_t *ct;
uint32_t *rp;
uint8_t *scp = fcp->isp_scratch;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGDEBUG0,
"Chan %d scanning fabric (GID_FT) via CT", chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (!IS_24XX(isp)) {
return (1);
}
/*
* Build a Passthrough IOCB in memory.
*/
pt = &un.plocal;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(un.q, QENTRY_LEN);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
pt->ctp_header.rqs_entry_count = 1;
pt->ctp_header.rqs_entry_type = RQSTYPE_CT_PASSTHRU;
pt->ctp_handle = 0xffffffff;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
pt->ctp_nphdl = fcp->isp_sns_hdl;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
pt->ctp_cmd_cnt = 1;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
pt->ctp_vpidx = ISP_GET_VPIDX(isp, chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
pt->ctp_time = 30;
pt->ctp_rsp_cnt = 1;
pt->ctp_rsp_bcnt = GIDLEN;
pt->ctp_cmd_bcnt = sizeof (*ct) + sizeof (uint32_t);
pt->ctp_dataseg[0].ds_base = DMA_LO32(fcp->isp_scdma+XTXOFF);
pt->ctp_dataseg[0].ds_basehi = DMA_HI32(fcp->isp_scdma+XTXOFF);
pt->ctp_dataseg[0].ds_count = sizeof (*ct) + sizeof (uint32_t);
pt->ctp_dataseg[1].ds_base = DMA_LO32(fcp->isp_scdma+IGPOFF);
pt->ctp_dataseg[1].ds_basehi = DMA_HI32(fcp->isp_scdma+IGPOFF);
pt->ctp_dataseg[1].ds_count = GIDLEN;
if (isp->isp_dblev & ISP_LOGDEBUG1) {
isp_print_bytes(isp, "ct IOCB", QENTRY_LEN, pt);
}
isp_put_ct_pt(isp, pt, (isp_ct_pt_t *) &scp[CTXOFF]);
/*
* Build the CT header and command in memory.
*
* Note that the CT header has to end up as Big Endian format in memory.
*/
ct = &un.clocal;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(ct, sizeof (*ct));
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ct->ct_revision = CT_REVISION;
ct->ct_fcs_type = CT_FC_TYPE_FC;
ct->ct_fcs_subtype = CT_FC_SUBTYPE_NS;
ct->ct_cmd_resp = SNS_GID_FT;
ct->ct_bcnt_resid = (GIDLEN - 16) >> 2;
isp_put_ct_hdr(isp, ct, (ct_hdr_t *) &scp[XTXOFF]);
rp = (uint32_t *) &scp[XTXOFF+sizeof (*ct)];
ISP_IOZPUT_32(isp, FC4_SCSI, rp);
Make ISPCTL_PLOGX find a handle to log into the management server with- not hope for the best. Change some things which were gated off of 24XX to be gated off of 2K login support. Convert some isp_prt calls to xpt_print calls.
2006-12-05 07:50:23 +00:00
if (isp->isp_dblev & ISP_LOGDEBUG1) {
isp_print_bytes(isp, "CT HDR + payload after put",
sizeof (*ct) + sizeof (uint32_t), &scp[XTXOFF]);
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(&scp[ZTXOFF], QENTRY_LEN);
MBSINIT(&mbs, MBOX_EXEC_COMMAND_IOCB_A64, MBLOGALL, 500000);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.param[1] = QENTRY_LEN;
mbs.param[2] = DMA_WD1(fcp->isp_scdma + CTXOFF);
mbs.param[3] = DMA_WD0(fcp->isp_scdma + CTXOFF);
mbs.param[6] = DMA_WD3(fcp->isp_scdma + CTXOFF);
mbs.param[7] = DMA_WD2(fcp->isp_scdma + CTXOFF);
MEMORYBARRIER(isp, SYNC_SFORDEV, XTXOFF, 2 * QENTRY_LEN);
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
return (-1);
}
MEMORYBARRIER(isp, SYNC_SFORCPU, ZTXOFF, QENTRY_LEN);
pt = &un.plocal;
isp_get_ct_pt(isp, (isp_ct_pt_t *) &scp[ZTXOFF], pt);
if (isp->isp_dblev & ISP_LOGDEBUG1) {
isp_print_bytes(isp, "IOCB response", QENTRY_LEN, pt);
}
if (pt->ctp_status && pt->ctp_status != RQCS_DATA_UNDERRUN) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGWARN,
"Chan %d ISP GID FT CT Passthrough returned 0x%x",
chan, pt->ctp_status);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (-1);
}
MEMORYBARRIER(isp, SYNC_SFORCPU, IGPOFF, GIDLEN + 16);
Make ISPCTL_PLOGX find a handle to log into the management server with- not hope for the best. Change some things which were gated off of 24XX to be gated off of 2K login support. Convert some isp_prt calls to xpt_print calls.
2006-12-05 07:50:23 +00:00
if (isp->isp_dblev & ISP_LOGDEBUG1) {
isp_print_bytes(isp, "CT response", GIDLEN+16, &scp[IGPOFF]);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (0);
}
static int
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_scan_fabric(ispsoftc_t *isp, int chan)
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
{
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcparam *fcp = FCPARAM(isp, chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
uint32_t portid;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
uint16_t handle, oldhandle, loopid;
isp_pdb_t pdb;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
int portidx, portlim, r;
sns_gid_ft_rsp_t *rs0, *rs1;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
"Chan %d FC Scan Fabric", chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (fcp->isp_fwstate != FW_READY ||
fcp->isp_loopstate < LOOP_LSCAN_DONE) {
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
return (-1);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (fcp->isp_loopstate > LOOP_SCANNING_FABRIC) {
return (0);
}
if (fcp->isp_topo != TOPO_FL_PORT && fcp->isp_topo != TOPO_F_PORT) {
fcp->isp_loopstate = LOOP_FSCAN_DONE;
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d FC Scan Fabric Done (no fabric)", chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (0);
}
fcp->isp_loopstate = LOOP_SCANNING_FABRIC;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (FC_SCRATCH_ACQUIRE(isp, chan)) {
isp_prt(isp, ISP_LOGERR, sacq);
ISP_MARK_PORTDB(isp, chan, 1);
return (-1);
}
if (fcp->isp_loopstate < LOOP_SCANNING_FABRIC) {
FC_SCRATCH_RELEASE(isp, chan);
ISP_MARK_PORTDB(isp, chan, 1);
return (-1);
}
/*
* Make sure we still are logged into the fabric controller.
*/
if (IS_24XX(isp)) { /* XXX SHOULDN'T THIS BE TRUE FOR 2K F/W? XXX */
loopid = NPH_FL_ID;
} else {
loopid = FL_ID;
}
r = isp_getpdb(isp, chan, loopid, &pdb, 0);
if (r == MBOX_NOT_LOGGED_IN) {
isp_dump_chip_portdb(isp, chan, 0);
}
if (r) {
fcp->isp_loopstate = LOOP_PDB_RCVD;
FC_SCRATCH_RELEASE(isp, chan);
ISP_MARK_PORTDB(isp, chan, 1);
return (-1);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
r = isp_gid_ft_ct_passthru(isp, chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
r = isp_gid_ft_sns(isp, chan);
}
if (fcp->isp_loopstate < LOOP_SCANNING_FABRIC) {
FC_SCRATCH_RELEASE(isp, chan);
ISP_MARK_PORTDB(isp, chan, 1);
return (-1);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
if (r > 0) {
fcp->isp_loopstate = LOOP_FSCAN_DONE;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FC_SCRATCH_RELEASE(isp, chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (0);
} else if (r < 0) {
fcp->isp_loopstate = LOOP_PDB_RCVD; /* try again */
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FC_SCRATCH_RELEASE(isp, chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (0);
}
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
MEMORYBARRIER(isp, SYNC_SFORCPU, IGPOFF, GIDLEN);
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
rs0 = (sns_gid_ft_rsp_t *) ((uint8_t *)fcp->isp_scratch+IGPOFF);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
rs1 = (sns_gid_ft_rsp_t *) ((uint8_t *)fcp->isp_scratch+OGPOFF);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
isp_get_gid_ft_response(isp, rs0, rs1, NGENT);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (fcp->isp_loopstate < LOOP_SCANNING_FABRIC) {
FC_SCRATCH_RELEASE(isp, chan);
ISP_MARK_PORTDB(isp, chan, 1);
return (-1);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (rs1->snscb_cthdr.ct_cmd_resp != LS_ACC) {
If the HBA is already 'touched', still set maxluns. Othewise for CAM_QUIRK_HILUN devices we loop thru 32bits of lun. Oops. Switch to using USEC_DELAY rather than USEC_SLEEP at isp_reset time. Try to paper around a defect in clients that don't correctly registers themeselves with the fabric nameserver. Minor updates for Mirapoint support- they still use code that is not HANDLE_LOOPSTATE_IN_OUTER_LAYERS, and, surprise surprise, this old stuff had some bugs in it. Clean up some target mode stuff. MFC after: 1 week
2002-06-16 05:18:22 +00:00
int level;
if (rs1->snscb_cthdr.ct_reason == 9 &&
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
rs1->snscb_cthdr.ct_explanation == 7) {
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
level = ISP_LOGSANCFG|ISP_LOGDEBUG0;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else {
If the HBA is already 'touched', still set maxluns. Othewise for CAM_QUIRK_HILUN devices we loop thru 32bits of lun. Oops. Switch to using USEC_DELAY rather than USEC_SLEEP at isp_reset time. Try to paper around a defect in clients that don't correctly registers themeselves with the fabric nameserver. Minor updates for Mirapoint support- they still use code that is not HANDLE_LOOPSTATE_IN_OUTER_LAYERS, and, surprise surprise, this old stuff had some bugs in it. Clean up some target mode stuff. MFC after: 1 week
2002-06-16 05:18:22 +00:00
level = ISP_LOGWARN;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, level, "Chan %d Fabric Nameserver rejected GID_FT"
" (Reason=0x%x Expl=0x%x)", chan,
rs1->snscb_cthdr.ct_reason,
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
rs1->snscb_cthdr.ct_explanation);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FC_SCRATCH_RELEASE(isp, chan);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
fcp->isp_loopstate = LOOP_FSCAN_DONE;
return (0);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
* If we get this far, we certainly still have the fabric controller.
*/
fcp->portdb[FL_ID].state = FC_PORTDB_STATE_PENDING_VALID;
/*
* Prime the handle we will start using.
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
oldhandle = FCPARAM(isp, 0)->isp_lasthdl;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* Go through the list and remove duplicate port ids.
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*/
portlim = 0;
portidx = 0;
for (portidx = 0; portidx < NGENT-1; portidx++) {
if (rs1->snscb_ports[portidx].control & 0x80) {
break;
}
}
/*
* If we're not at the last entry, our list wasn't big enough.
*/
if ((rs1->snscb_ports[portidx].control & 0x80) == 0) {
isp_prt(isp, ISP_LOGWARN,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"fabric too big for scratch area: increase ISP_FC_SCRLEN");
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
portlim = portidx + 1;
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
isp_prt(isp, ISP_LOGSANCFG,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d got %d ports back from name server", chan, portlim);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
for (portidx = 0; portidx < portlim; portidx++) {
int npidx;
portid =
((rs1->snscb_ports[portidx].portid[0]) << 16) |
((rs1->snscb_ports[portidx].portid[1]) << 8) |
((rs1->snscb_ports[portidx].portid[2]));
for (npidx = portidx + 1; npidx < portlim; npidx++) {
uint32_t new_portid =
((rs1->snscb_ports[npidx].portid[0]) << 16) |
((rs1->snscb_ports[npidx].portid[1]) << 8) |
((rs1->snscb_ports[npidx].portid[2]));
if (new_portid == portid) {
break;
}
}
if (npidx < portlim) {
rs1->snscb_ports[npidx].portid[0] = 0;
rs1->snscb_ports[npidx].portid[1] = 0;
rs1->snscb_ports[npidx].portid[2] = 0;
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d removing duplicate PortID 0x%06x"
" entry from list", chan, portid);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
}
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
/*
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* We now have a list of Port IDs for all FC4 SCSI devices
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* that the Fabric Name server knows about.
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* For each entry on this list go through our port database looking
* for probational entries- if we find one, then an old entry is
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* maybe still this one. We get some information to find out.
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*
* Otherwise, it's a new fabric device, and we log into it
* (unconditionally). After searching the entire database
* again to make sure that we never ever ever ever have more
* than one entry that has the same PortID or the same
* WWNN/WWPN duple, we enter the device into our database.
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
for (portidx = 0; portidx < portlim; portidx++) {
fcportdb_t *lp;
uint64_t wwnn, wwpn;
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
int dbidx, nr;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
portid =
((rs1->snscb_ports[portidx].portid[0]) << 16) |
((rs1->snscb_ports[portidx].portid[1]) << 8) |
((rs1->snscb_ports[portidx].portid[2]));
if (portid == 0) {
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
isp_prt(isp, ISP_LOGSANCFG,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d skipping null PortID at idx %d",
chan, portidx);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
continue;
}
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
/*
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* Skip ourselves here and on other channels. If we're
* multi-id, we can't check the portids in other FCPARAM
* arenas because the resolutions here aren't synchronized.
* The best way to do this is to exclude looking at portids
* that have the same domain and area code as our own
* portid.
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (ISP_CAP_MULTI_ID(isp)) {
if ((portid >> 8) == (fcp->isp_portid >> 8)) {
isp_prt(isp, ISP_LOGSANCFG,
"Chan %d skip PortID 0x%06x",
chan, portid);
continue;
}
} else if (portid == fcp->isp_portid) {
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
isp_prt(isp, ISP_LOGSANCFG,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d skip ourselves on @ PortID 0x%06x",
chan, portid);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
continue;
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
isp_prt(isp, ISP_LOGSANCFG,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d Checking Fabric Port 0x%06x", chan, portid);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
* We now search our Port Database for any
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
* probational entries with this PortID. We don't
* look for zombies here- only probational
* entries (we've already logged out of zombies).
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*/
for (dbidx = 0; dbidx < MAX_FC_TARG; dbidx++) {
lp = &fcp->portdb[dbidx];
Remove file unused in freebsd.
2009-09-21 01:38:22 +00:00
if (lp->state != FC_PORTDB_STATE_PROBATIONAL ||
lp->target_mode) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
continue;
}
if (lp->portid == portid) {
break;
}
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
* We found a probational entry with this Port ID.
*/
if (dbidx < MAX_FC_TARG) {
int handle_changed = 0;
lp = &fcp->portdb[dbidx];
/*
* See if we're still logged into it.
*
* If we aren't, mark it as a dead device and
* leave the new portid in the database entry
* for somebody further along to decide what to
* do (policy choice).
*
* If we are, check to see if it's the same
* device still (it should be). If for some
* reason it isn't, mark it as a changed device
* and leave the new portid and role in the
* database entry for somebody further along to
* decide what to do (policy choice).
*
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
r = isp_getpdb(isp, chan, lp->handle, &pdb, 0);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
if (fcp->isp_loopstate != LOOP_SCANNING_FABRIC) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FC_SCRATCH_RELEASE(isp, chan);
ISP_MARK_PORTDB(isp, chan, 1);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
return (-1);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (r != 0) {
lp->new_portid = portid;
lp->state = FC_PORTDB_STATE_DEAD;
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d Fabric Port 0x%06x is dead",
chan, portid);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
continue;
}
/*
* Check to make sure that handle, portid, WWPN and
* WWNN agree. If they don't, then the association
* between this PortID and the stated handle has been
* broken by the firmware.
*/
MAKE_WWN_FROM_NODE_NAME(wwnn, pdb.nodename);
MAKE_WWN_FROM_NODE_NAME(wwpn, pdb.portname);
if (pdb.handle != lp->handle ||
pdb.portid != portid ||
wwpn != lp->port_wwn ||
wwnn != lp->node_wwn) {
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fconf, chan, dbidx, pdb.handle, pdb.portid,
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
(uint32_t) (wwnn >> 32), (uint32_t) wwnn,
(uint32_t) (wwpn >> 32), (uint32_t) wwpn,
lp->handle, portid,
(uint32_t) (lp->node_wwn >> 32),
(uint32_t) lp->node_wwn,
(uint32_t) (lp->port_wwn >> 32),
(uint32_t) lp->port_wwn);
/*
* Try to re-login to this device using a
* new handle. If that fails, mark it dead.
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* isp_login_device will check for handle and
* portid consistency after re-login.
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (isp_login_device(isp, chan, portid, &pdb,
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
&oldhandle)) {
lp->new_portid = portid;
lp->state = FC_PORTDB_STATE_DEAD;
if (fcp->isp_loopstate !=
LOOP_SCANNING_FABRIC) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FC_SCRATCH_RELEASE(isp, chan);
ISP_MARK_PORTDB(isp, chan, 1);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (-1);
}
continue;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (fcp->isp_loopstate !=
LOOP_SCANNING_FABRIC) {
FC_SCRATCH_RELEASE(isp, chan);
ISP_MARK_PORTDB(isp, chan, 1);
return (-1);
}
FCPARAM(isp, 0)->isp_lasthdl = oldhandle;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
MAKE_WWN_FROM_NODE_NAME(wwnn, pdb.nodename);
MAKE_WWN_FROM_NODE_NAME(wwpn, pdb.portname);
if (wwpn != lp->port_wwn ||
wwnn != lp->node_wwn) {
isp_prt(isp, ISP_LOGWARN, "changed WWN"
" after relogin");
lp->new_portid = portid;
lp->state = FC_PORTDB_STATE_DEAD;
continue;
}
lp->handle = pdb.handle;
handle_changed++;
}
nr = (pdb.s3_role & SVC3_ROLE_MASK) >> SVC3_ROLE_SHIFT;
/*
* Check to see whether the portid and roles have
* stayed the same. If they have stayed the same,
* we believe that this is the same device and it
* hasn't become disconnected and reconnected, so
* mark it as pending valid.
*
* If they aren't the same, mark the device as a
* changed device and save the new port id and role
* and let somebody else decide.
*/
lp->new_portid = portid;
lp->new_roles = nr;
if (pdb.portid != lp->portid || nr != lp->roles ||
handle_changed) {
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
isp_prt(isp, ISP_LOGSANCFG,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d Fabric Port 0x%06x changed",
chan, portid);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
lp->state = FC_PORTDB_STATE_CHANGED;
} else {
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
isp_prt(isp, ISP_LOGSANCFG,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d Fabric Port 0x%06x "
"Now Pending Valid", chan, portid);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
lp->state = FC_PORTDB_STATE_PENDING_VALID;
}
continue;
}
/*
* Ah- a new entry. Search the database again for all non-NIL
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
* entries to make sure we never ever make a new database entry
* with the same port id. While we're at it, mark where the
* last free entry was.
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
dbidx = MAX_FC_TARG;
for (lp = fcp->portdb; lp < &fcp->portdb[MAX_FC_TARG]; lp++) {
if (lp >= &fcp->portdb[FL_ID] &&
lp <= &fcp->portdb[SNS_ID]) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
continue;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* Skip any target mode entries.
*/
if (lp->target_mode) {
continue;
}
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
if (lp->state == FC_PORTDB_STATE_NIL) {
if (dbidx == MAX_FC_TARG) {
dbidx = lp - fcp->portdb;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
continue;
}
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
if (lp->state == FC_PORTDB_STATE_ZOMBIE) {
continue;
}
if (lp->portid == portid) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
}
}
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
if (lp < &fcp->portdb[MAX_FC_TARG]) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGWARN, "Chan %d PortID 0x%06x "
"already at %d handle %d state %d",
chan, portid, dbidx, lp->handle, lp->state);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
continue;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
/*
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
* We should have the index of the first free entry seen.
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (dbidx == MAX_FC_TARG) {
isp_prt(isp, ISP_LOGERR,
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
"port database too small to login PortID 0x%06x"
"- increase MAX_FC_TARG", portid);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
continue;
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
/*
* Otherwise, point to our new home.
*/
lp = &fcp->portdb[dbidx];
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
* Try to see if we are logged into this device,
* and maybe log into it.
*
* isp_login_device will check for handle and
* portid consistency after login.
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (isp_login_device(isp, chan, portid, &pdb, &oldhandle)) {
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
if (fcp->isp_loopstate != LOOP_SCANNING_FABRIC) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FC_SCRATCH_RELEASE(isp, chan);
ISP_MARK_PORTDB(isp, chan, 1);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
return (-1);
}
continue;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (fcp->isp_loopstate != LOOP_SCANNING_FABRIC) {
FC_SCRATCH_RELEASE(isp, chan);
ISP_MARK_PORTDB(isp, chan, 1);
return (-1);
}
FCPARAM(isp, 0)->isp_lasthdl = oldhandle;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
handle = pdb.handle;
MAKE_WWN_FROM_NODE_NAME(wwnn, pdb.nodename);
MAKE_WWN_FROM_NODE_NAME(wwpn, pdb.portname);
nr = (pdb.s3_role & SVC3_ROLE_MASK) >> SVC3_ROLE_SHIFT;
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* And go through the database *one* more time to make sure
* that we do not make more than one entry that has the same
* WWNN/WWPN duple
Scale back # of luns supported for SCC to 16384- oops- top 3 bits are a lun address modifier of sorts. Only an HP XP-512 seems to have cared. Fix a few misplaced pointers for the new fabric goop, which has been demonstrated to work on newer Brocades and McData switches now. Put in commented out code which would run GFF_ID if the QLogic f/w allowed it. Don't whine about not being able to find a handle for a command if it was a command aborted (by us).
2002-04-16 19:55:35 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
for (dbidx = 0; dbidx < MAX_FC_TARG; dbidx++) {
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
if (dbidx >= FL_ID && dbidx <= SNS_ID) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
continue;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (fcp->portdb[dbidx].target_mode) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
continue;
}
Remove file unused in freebsd.
2009-09-21 01:38:22 +00:00
if (fcp->portdb[dbidx].node_wwn == wwnn &&
fcp->portdb[dbidx].port_wwn == wwpn) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
}
}
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
if (dbidx == MAX_FC_TARG) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(lp, sizeof (fcportdb_t));
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
lp->handle = handle;
lp->node_wwn = wwnn;
lp->port_wwn = wwpn;
lp->new_portid = portid;
lp->new_roles = nr;
lp->state = FC_PORTDB_STATE_NEW;
isp_prt(isp, ISP_LOGSANCFG,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d Fabric Port 0x%06x is a New Entry",
chan, portid);
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
continue;
}
if (fcp->portdb[dbidx].state != FC_PORTDB_STATE_ZOMBIE) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_prt(isp, ISP_LOGWARN,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d PortID 0x%x 0x%08x%08x/0x%08x%08x %ld "
"already at idx %d, state 0x%x", chan, portid,
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
(uint32_t) (wwnn >> 32), (uint32_t) wwnn,
(uint32_t) (wwpn >> 32), (uint32_t) wwpn,
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
(long) (lp - fcp->portdb), dbidx,
fcp->portdb[dbidx].state);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
continue;
}
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
/*
* We found a zombie entry that matches us.
* Revive it. We know that WWN and WWPN
* are the same. For fabric devices, we
* don't care that handle is different
* as we assign that. If role or portid
* are different, it maybe a changed device.
*/
lp = &fcp->portdb[dbidx];
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
lp->handle = handle;
lp->new_portid = portid;
lp->new_roles = nr;
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
if (lp->portid != portid || lp->roles != nr) {
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d Zombie Fabric Port 0x%06x Now Changed",
chan, portid);
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
lp->state = FC_PORTDB_STATE_CHANGED;
} else {
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d Zombie Fabric Port 0x%06x "
"Now Pending Valid", chan, portid);
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
lp->state = FC_PORTDB_STATE_PENDING_VALID;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FC_SCRATCH_RELEASE(isp, chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (fcp->isp_loopstate != LOOP_SCANNING_FABRIC) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MARK_PORTDB(isp, chan, 1);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (-1);
}
fcp->isp_loopstate = LOOP_FSCAN_DONE;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
"Chan %d FC Scan Fabric Done", chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (0);
}
/*
* Find an unused handle and try and use to login to a port.
*/
static int
Don't try and re-use a handle, even if the firmware tells you that's what is logged in. PR: kern/144026 MFC after: 1 week
2010-02-18 18:35:09 +00:00
isp_login_device(ispsoftc_t *isp, int chan, uint32_t portid, isp_pdb_t *p, uint16_t *ohp)
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
{
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
int lim, i, r;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
uint16_t handle;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (ISP_CAP_2KLOGIN(isp)) {
Make ISPCTL_PLOGX find a handle to log into the management server with- not hope for the best. Change some things which were gated off of 24XX to be gated off of 2K login support. Convert some isp_prt calls to xpt_print calls.
2006-12-05 07:50:23 +00:00
lim = NPH_MAX_2K;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else {
lim = NPH_MAX;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
handle = isp_nxt_handle(isp, chan, *ohp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
for (i = 0; i < lim; i++) {
Scale back # of luns supported for SCC to 16384- oops- top 3 bits are a lun address modifier of sorts. Only an HP XP-512 seems to have cared. Fix a few misplaced pointers for the new fabric goop, which has been demonstrated to work on newer Brocades and McData switches now. Put in commented out code which would run GFF_ID if the QLogic f/w allowed it. Don't whine about not being able to find a handle for a command if it was a command aborted (by us).
2002-04-16 19:55:35 +00:00
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* See if we're still logged into something with
* this handle and that something agrees with this
* port id.
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
r = isp_getpdb(isp, chan, handle, p, 0);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (r == 0 && p->portid != portid) {
Don't try and re-use a handle, even if the firmware tells you that's what is logged in. PR: kern/144026 MFC after: 1 week
2010-02-18 18:35:09 +00:00
(void) isp_plogx(isp, chan, handle, portid, PLOGX_FLG_CMD_LOGO | PLOGX_FLG_IMPLICIT | PLOGX_FLG_FREE_NPHDL, 1);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else if (r == 0) {
break;
Scale back # of luns supported for SCC to 16384- oops- top 3 bits are a lun address modifier of sorts. Only an HP XP-512 seems to have cared. Fix a few misplaced pointers for the new fabric goop, which has been demonstrated to work on newer Brocades and McData switches now. Put in commented out code which would run GFF_ID if the QLogic f/w allowed it. Don't whine about not being able to find a handle for a command if it was a command aborted (by us).
2002-04-16 19:55:35 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (FCPARAM(isp, chan)->isp_loopstate != LOOP_SCANNING_FABRIC) {
Scale back # of luns supported for SCC to 16384- oops- top 3 bits are a lun address modifier of sorts. Only an HP XP-512 seems to have cared. Fix a few misplaced pointers for the new fabric goop, which has been demonstrated to work on newer Brocades and McData switches now. Put in commented out code which would run GFF_ID if the QLogic f/w allowed it. Don't whine about not being able to find a handle for a command if it was a command aborted (by us).
2002-04-16 19:55:35 +00:00
return (-1);
}
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* Now try and log into the device
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
*/
Don't try and re-use a handle, even if the firmware tells you that's what is logged in. PR: kern/144026 MFC after: 1 week
2010-02-18 18:35:09 +00:00
r = isp_plogx(isp, chan, handle, portid, PLOGX_FLG_CMD_PLOGI, 1);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (FCPARAM(isp, chan)->isp_loopstate != LOOP_SCANNING_FABRIC) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (-1);
}
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
if (r == 0) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*ohp = handle;
break;
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
} else if ((r & 0xffff) == MBOX_PORT_ID_USED) {
Don't try and re-use a handle, even if the firmware tells you that's what is logged in. PR: kern/144026 MFC after: 1 week
2010-02-18 18:35:09 +00:00
/*
* If we get here, then the firmwware still thinks we're logged into this device, but with a different
* handle. We need to break that association. We used to try and just substitute the handle, but then
* failed to get any data via isp_getpdb (below).
*/
if (isp_plogx(isp, chan, r >> 16, portid, PLOGX_FLG_CMD_LOGO | PLOGX_FLG_IMPLICIT | PLOGX_FLG_FREE_NPHDL, 1)) {
isp_prt(isp, ISP_LOGERR, "baw... logout of %x failed", r >> 16);
}
if (FCPARAM(isp, chan)->isp_loopstate != LOOP_SCANNING_FABRIC) {
return (-1);
}
r = isp_plogx(isp, chan, handle, portid, PLOGX_FLG_CMD_PLOGI, 1);
if (FCPARAM(isp, chan)->isp_loopstate != LOOP_SCANNING_FABRIC) {
return (-1);
}
if (r == 0) {
*ohp = handle;
} else {
i = lim;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
Various minor and not so minor fixes suggested by Coverity. In at least one case, it's amazing that target mode worked at all. Found by: Coverity. MFC after: 2 weeks
2010-06-02 23:31:27 +00:00
} else if ((r & 0xffff) == MBOX_LOOP_ID_USED) {
/*
* Try the next loop id.
*/
*ohp = handle;
handle = isp_nxt_handle(isp, chan, handle);
} else {
/*
* Give up.
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
i = lim;
break;
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (i == lim) {
Don't try and re-use a handle, even if the firmware tells you that's what is logged in. PR: kern/144026 MFC after: 1 week
2010-02-18 18:35:09 +00:00
isp_prt(isp, ISP_LOGWARN, "Chan %d PLOGI 0x%06x failed", chan, portid);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (-1);
}
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* If we successfully logged into it, get the PDB for it
* so we can crosscheck that it is still what we think it
* is and that we also have the role it plays
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
r = isp_getpdb(isp, chan, handle, p, 0);
if (FCPARAM(isp, chan)->isp_loopstate != LOOP_SCANNING_FABRIC) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (-1);
}
if (r != 0) {
Don't try and re-use a handle, even if the firmware tells you that's what is logged in. PR: kern/144026 MFC after: 1 week
2010-02-18 18:35:09 +00:00
isp_prt(isp, ISP_LOGERR, "Chan %d new device 0x%06x@0x%x disappeared", chan, portid, handle);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (-1);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (p->handle != handle || p->portid != portid) {
Don't try and re-use a handle, even if the firmware tells you that's what is logged in. PR: kern/144026 MFC after: 1 week
2010-02-18 18:35:09 +00:00
isp_prt(isp, ISP_LOGERR, "Chan %d new device 0x%06x@0x%x changed (0x%06x@0x%0x)",
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
chan, portid, handle, p->portid, p->handle);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (-1);
}
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
return (0);
}
Add a isp_register_fc4_type function so that we work with McData switches that require us to register our FC4 types of interest. Allow ourselves, in F-port topologies, to start logging in fabric devices in the target 0..125 range. Change ISPASYNC_PDB_CHANGED (misnamed) to ISPASYNC_LOGGED_INOUT. Fix (*SMACK*) again some default WWN stuff. This is *really* hard to get right across all the range of platforms.
2001-01-09 02:46:23 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
static int
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_register_fc4_type(ispsoftc_t *isp, int chan)
Add a isp_register_fc4_type function so that we work with McData switches that require us to register our FC4 types of interest. Allow ourselves, in F-port topologies, to start logging in fabric devices in the target 0..125 range. Change ISPASYNC_PDB_CHANGED (misnamed) to ISPASYNC_LOGGED_INOUT. Fix (*SMACK*) again some default WWN stuff. This is *really* hard to get right across all the range of platforms.
2001-01-09 02:46:23 +00:00
{
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcparam *fcp = FCPARAM(isp, chan);
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
uint8_t local[SNS_RFT_ID_REQ_SIZE];
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
sns_screq_t *reqp = (sns_screq_t *) local;
Add a isp_register_fc4_type function so that we work with McData switches that require us to register our FC4 types of interest. Allow ourselves, in F-port topologies, to start logging in fabric devices in the target 0..125 range. Change ISPASYNC_PDB_CHANGED (misnamed) to ISPASYNC_LOGGED_INOUT. Fix (*SMACK*) again some default WWN stuff. This is *really* hard to get right across all the range of platforms.
2001-01-09 02:46:23 +00:00
mbreg_t mbs;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO((void *) reqp, SNS_RFT_ID_REQ_SIZE);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
reqp->snscb_rblen = SNS_RFT_ID_RESP_SIZE >> 1;
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
reqp->snscb_addr[RQRSP_ADDR0015] = DMA_WD0(fcp->isp_scdma + 0x100);
reqp->snscb_addr[RQRSP_ADDR1631] = DMA_WD1(fcp->isp_scdma + 0x100);
reqp->snscb_addr[RQRSP_ADDR3247] = DMA_WD2(fcp->isp_scdma + 0x100);
reqp->snscb_addr[RQRSP_ADDR4863] = DMA_WD3(fcp->isp_scdma + 0x100);
Add a isp_register_fc4_type function so that we work with McData switches that require us to register our FC4 types of interest. Allow ourselves, in F-port topologies, to start logging in fabric devices in the target 0..125 range. Change ISPASYNC_PDB_CHANGED (misnamed) to ISPASYNC_LOGGED_INOUT. Fix (*SMACK*) again some default WWN stuff. This is *really* hard to get right across all the range of platforms.
2001-01-09 02:46:23 +00:00
reqp->snscb_sblen = 22;
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
reqp->snscb_data[0] = SNS_RFT_ID;
Add a isp_register_fc4_type function so that we work with McData switches that require us to register our FC4 types of interest. Allow ourselves, in F-port topologies, to start logging in fabric devices in the target 0..125 range. Change ISPASYNC_PDB_CHANGED (misnamed) to ISPASYNC_LOGGED_INOUT. Fix (*SMACK*) again some default WWN stuff. This is *really* hard to get right across all the range of platforms.
2001-01-09 02:46:23 +00:00
reqp->snscb_data[4] = fcp->isp_portid & 0xffff;
reqp->snscb_data[5] = (fcp->isp_portid >> 16) & 0xff;
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
reqp->snscb_data[6] = (1 << FC4_SCSI);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (FC_SCRATCH_ACQUIRE(isp, chan)) {
isp_prt(isp, ISP_LOGERR, sacq);
return (-1);
}
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
isp_put_sns_request(isp, reqp, (sns_screq_t *) fcp->isp_scratch);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_SEND_SNS, MBLOGALL, 1000000);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
mbs.param[1] = SNS_RFT_ID_REQ_SIZE >> 1;
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
mbs.param[2] = DMA_WD1(fcp->isp_scdma);
mbs.param[3] = DMA_WD0(fcp->isp_scdma);
mbs.param[6] = DMA_WD3(fcp->isp_scdma);
mbs.param[7] = DMA_WD2(fcp->isp_scdma);
Add some missing braces. Add MEMORY_BARRIER for the few scratch dma ops that were missing them plus add a couple of hi 32 bit dma ops (we could probably allow 64 bit scratch and request/response queue dma now).
2006-07-14 05:14:48 +00:00
MEMORYBARRIER(isp, SYNC_SFORDEV, 0, SNS_RFT_ID_REQ_SIZE);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FC_SCRATCH_RELEASE(isp, chan);
Add a isp_register_fc4_type function so that we work with McData switches that require us to register our FC4 types of interest. Allow ourselves, in F-port topologies, to start logging in fabric devices in the target 0..125 range. Change ISPASYNC_PDB_CHANGED (misnamed) to ISPASYNC_LOGGED_INOUT. Fix (*SMACK*) again some default WWN stuff. This is *really* hard to get right across all the range of platforms.
2001-01-09 02:46:23 +00:00
if (mbs.param[0] == MBOX_COMMAND_COMPLETE) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (0);
} else {
return (-1);
}
}
static int
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_register_fc4_type_24xx(ispsoftc_t *isp, int chan)
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
{
mbreg_t mbs;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcparam *fcp = FCPARAM(isp, chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
union {
isp_ct_pt_t plocal;
rft_id_t clocal;
uint8_t q[QENTRY_LEN];
} un;
isp_ct_pt_t *pt;
ct_hdr_t *ct;
rft_id_t *rp;
uint8_t *scp = fcp->isp_scratch;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (FC_SCRATCH_ACQUIRE(isp, chan)) {
isp_prt(isp, ISP_LOGERR, sacq);
return (-1);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
* Build a Passthrough IOCB in memory.
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(un.q, QENTRY_LEN);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
pt = &un.plocal;
pt->ctp_header.rqs_entry_count = 1;
pt->ctp_header.rqs_entry_type = RQSTYPE_CT_PASSTHRU;
pt->ctp_handle = 0xffffffff;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
pt->ctp_nphdl = fcp->isp_sns_hdl;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
pt->ctp_cmd_cnt = 1;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
pt->ctp_vpidx = ISP_GET_VPIDX(isp, chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
pt->ctp_time = 1;
pt->ctp_rsp_cnt = 1;
pt->ctp_rsp_bcnt = sizeof (ct_hdr_t);
pt->ctp_cmd_bcnt = sizeof (rft_id_t);
pt->ctp_dataseg[0].ds_base = DMA_LO32(fcp->isp_scdma+XTXOFF);
pt->ctp_dataseg[0].ds_basehi = DMA_HI32(fcp->isp_scdma+XTXOFF);
pt->ctp_dataseg[0].ds_count = sizeof (rft_id_t);
pt->ctp_dataseg[1].ds_base = DMA_LO32(fcp->isp_scdma+IGPOFF);
pt->ctp_dataseg[1].ds_basehi = DMA_HI32(fcp->isp_scdma+IGPOFF);
pt->ctp_dataseg[1].ds_count = sizeof (ct_hdr_t);
isp_put_ct_pt(isp, pt, (isp_ct_pt_t *) &scp[CTXOFF]);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (isp->isp_dblev & ISP_LOGDEBUG1) {
isp_print_bytes(isp, "IOCB CT Request", QENTRY_LEN, pt);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
* Build the CT header and command in memory.
*
* Note that the CT header has to end up as Big Endian format in memory.
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(&un.clocal, sizeof (un.clocal));
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ct = &un.clocal.rftid_hdr;
ct->ct_revision = CT_REVISION;
ct->ct_fcs_type = CT_FC_TYPE_FC;
ct->ct_fcs_subtype = CT_FC_SUBTYPE_NS;
ct->ct_cmd_resp = SNS_RFT_ID;
ct->ct_bcnt_resid = (sizeof (rft_id_t) - sizeof (ct_hdr_t)) >> 2;
rp = &un.clocal;
rp->rftid_portid[0] = fcp->isp_portid >> 16;
rp->rftid_portid[1] = fcp->isp_portid >> 8;
rp->rftid_portid[2] = fcp->isp_portid;
rp->rftid_fc4types[FC4_SCSI >> 5] = 1 << (FC4_SCSI & 0x1f);
isp_put_rft_id(isp, rp, (rft_id_t *) &scp[XTXOFF]);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (isp->isp_dblev & ISP_LOGDEBUG1) {
isp_print_bytes(isp, "CT Header", QENTRY_LEN, &scp[XTXOFF]);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(&scp[ZTXOFF], sizeof (ct_hdr_t));
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_EXEC_COMMAND_IOCB_A64, MBLOGALL, 1000000);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.param[1] = QENTRY_LEN;
mbs.param[2] = DMA_WD1(fcp->isp_scdma + CTXOFF);
mbs.param[3] = DMA_WD0(fcp->isp_scdma + CTXOFF);
mbs.param[6] = DMA_WD3(fcp->isp_scdma + CTXOFF);
mbs.param[7] = DMA_WD2(fcp->isp_scdma + CTXOFF);
MEMORYBARRIER(isp, SYNC_SFORDEV, XTXOFF, 2 * QENTRY_LEN);
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FC_SCRATCH_RELEASE(isp, chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (-1);
}
MEMORYBARRIER(isp, SYNC_SFORCPU, ZTXOFF, QENTRY_LEN);
pt = &un.plocal;
isp_get_ct_pt(isp, (isp_ct_pt_t *) &scp[ZTXOFF], pt);
if (isp->isp_dblev & ISP_LOGDEBUG1) {
isp_print_bytes(isp, "IOCB response", QENTRY_LEN, pt);
}
if (pt->ctp_status) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FC_SCRATCH_RELEASE(isp, chan);
isp_prt(isp, ISP_LOGWARN,
"Chan %d Register FC4 Type CT Passthrough returned 0x%x",
chan, pt->ctp_status);
return (1);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
isp_get_ct_hdr(isp, (ct_hdr_t *) &scp[IGPOFF], ct);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FC_SCRATCH_RELEASE(isp, chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (ct->ct_cmd_resp == LS_RJT) {
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d Register FC4 Type rejected", chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (-1);
} else if (ct->ct_cmd_resp == LS_ACC) {
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
isp_prt(isp, ISP_LOGSANCFG|ISP_LOGDEBUG0,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d Register FC4 Type accepted", chan);
return (0);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else {
isp_prt(isp, ISP_LOGWARN,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Chan %d Register FC4 Type: 0x%x",
chan, ct->ct_cmd_resp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (-1);
}
}
static uint16_t
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_nxt_handle(ispsoftc_t *isp, int chan, uint16_t handle)
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
{
MFP4: Move default setting to the end of isp_reset instead of the front of isp_init so we can read NVRAM even if we're role ISP_NONE. Prepare for reintroduction of channels (for FC) for N-Port Virtualization. Fix a botch in handle assignment that caused us to nuke one device when a new one arrives and end up with two devices with the same identity in the virtual target mapping table.
2007-01-20 04:00:21 +00:00
int i;
Make ISPCTL_PLOGX find a handle to log into the management server with- not hope for the best. Change some things which were gated off of 24XX to be gated off of 2K login support. Convert some isp_prt calls to xpt_print calls.
2006-12-05 07:50:23 +00:00
if (handle == NIL_HANDLE) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (FCPARAM(isp, chan)->isp_topo == TOPO_F_PORT) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
handle = 0;
} else {
handle = SNS_ID+1;
}
} else {
handle += 1;
if (handle >= FL_ID && handle <= SNS_ID) {
handle = SNS_ID+1;
Make ISPCTL_PLOGX find a handle to log into the management server with- not hope for the best. Change some things which were gated off of 24XX to be gated off of 2K login support. Convert some isp_prt calls to xpt_print calls.
2006-12-05 07:50:23 +00:00
}
if (handle >= NPH_RESERVED && handle <= NPH_FL_ID) {
handle = NPH_FL_ID+1;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (ISP_CAP_2KLOGIN(isp)) {
Make ISPCTL_PLOGX find a handle to log into the management server with- not hope for the best. Change some things which were gated off of 24XX to be gated off of 2K login support. Convert some isp_prt calls to xpt_print calls.
2006-12-05 07:50:23 +00:00
if (handle == NPH_MAX_2K) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
handle = 0;
}
} else {
Make ISPCTL_PLOGX find a handle to log into the management server with- not hope for the best. Change some things which were gated off of 24XX to be gated off of 2K login support. Convert some isp_prt calls to xpt_print calls.
2006-12-05 07:50:23 +00:00
if (handle == NPH_MAX) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
handle = 0;
}
}
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (handle == FCPARAM(isp, chan)->isp_loopid) {
return (isp_nxt_handle(isp, chan, handle));
Add a isp_register_fc4_type function so that we work with McData switches that require us to register our FC4 types of interest. Allow ourselves, in F-port topologies, to start logging in fabric devices in the target 0..125 range. Change ISPASYNC_PDB_CHANGED (misnamed) to ISPASYNC_LOGGED_INOUT. Fix (*SMACK*) again some default WWN stuff. This is *really* hard to get right across all the range of platforms.
2001-01-09 02:46:23 +00:00
}
MFP4: Move default setting to the end of isp_reset instead of the front of isp_init so we can read NVRAM even if we're role ISP_NONE. Prepare for reintroduction of channels (for FC) for N-Port Virtualization. Fix a botch in handle assignment that caused us to nuke one device when a new one arrives and end up with two devices with the same identity in the virtual target mapping table.
2007-01-20 04:00:21 +00:00
for (i = 0; i < MAX_FC_TARG; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (FCPARAM(isp, chan)->portdb[i].state ==
FC_PORTDB_STATE_NIL) {
MFP4: Move default setting to the end of isp_reset instead of the front of isp_init so we can read NVRAM even if we're role ISP_NONE. Prepare for reintroduction of channels (for FC) for N-Port Virtualization. Fix a botch in handle assignment that caused us to nuke one device when a new one arrives and end up with two devices with the same identity in the virtual target mapping table.
2007-01-20 04:00:21 +00:00
continue;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (FCPARAM(isp, chan)->portdb[i].handle == handle) {
return (isp_nxt_handle(isp, chan, handle));
MFP4: Move default setting to the end of isp_reset instead of the front of isp_init so we can read NVRAM even if we're role ISP_NONE. Prepare for reintroduction of channels (for FC) for N-Port Virtualization. Fix a botch in handle assignment that caused us to nuke one device when a new one arrives and end up with two devices with the same identity in the virtual target mapping table.
2007-01-20 04:00:21 +00:00
}
}
return (handle);
Add a isp_register_fc4_type function so that we work with McData switches that require us to register our FC4 types of interest. Allow ourselves, in F-port topologies, to start logging in fabric devices in the target 0..125 range. Change ISPASYNC_PDB_CHANGED (misnamed) to ISPASYNC_LOGGED_INOUT. Fix (*SMACK*) again some default WWN stuff. This is *really* hard to get right across all the range of platforms.
2001-01-09 02:46:23 +00:00
}
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
* Start a command. Locking is assumed done in the caller.
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
int
In order to save ourselves grief with the SUNPRO compiler under Solaris (which, for reasons unknown to me, chokes on u_int16_t as a typedef of unsigned short if used in a transitional (mixed K&R and ANSI) way), we'll go the extra mile and fully ANSIfy things.
2001-03-14 04:11:56 +00:00
isp_start(XS_T *xs)
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
{
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
ispsoftc_t *isp;
Be more specific about which CDB length we're going to use. Not really a likely bug but we might as well be clearer. Found with: Coverity Prevent(tm) CID: 3981 MFC after: 2 weeks
2010-06-05 20:37:40 +00:00
uint32_t handle, cdblen;
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
uint8_t local[QENTRY_LEN];
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ispreq_t *reqp;
void *cdbp, *qep;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
uint16_t *tptr;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
int target, dmaresult, hdlidx = 0;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
XS_INITERR(xs);
isp = XS_ISP(xs);
When resetting the Qlogic 2X00 units, reset the FPM (Fibre Protocol Module) and FBM (Fibre Buffer Modules). Also remember to clear the semaphore registers. Tell the RISC processor to not halt on FPM parity errors. Throw out the ISP_CFG_NOINIT silliness and instead go to the use of adapter 'roles' to see whether one completes initialization or not (mostly for Fibre Channel). The ultimate intent, btw, of all of this is to have a warm standby adapter for failover reasons. Because we do roles now, setting of Target Capable Class 3 service parameters in the ICB for the 2x00 cards reflects from role. Also, in isp_start, if we're not supporting an initiator role, we bounce outgoing commands with a Selection Timeout error. Also clean out the TOGGLE_TMODE goop for FC- there is no toggling of target mode like there is for parallel SCSI cards. Do more cleanup with respect to using target ids 0..125 in F-port topologies. Also keep track of things which *were* fabric devices so that when you rescan the fabric you can notify the outer layers when fabric devices go away. Only force a LOGOUT for fabric devices if they're still logged in (i.e., you cat their Port Database entry. Clean up the Get All Next scanning. Finally, use a new tag in the softc to store the opcode for the last mailbox command used so we can report which opcode timed out.
2001-01-15 18:33:08 +00:00
/*
* Now make sure we're running.
*/
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
if (isp->isp_state != ISP_RUNSTATE) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGERR, "Adapter not at RUNSTATE");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
XS_SETERR(xs, HBA_BOTCH);
return (CMD_COMPLETE);
}
/*
Clean out residual bogosity for fast posting stuff- ISP_NO_FASTPOST_SCSI is gone as a define. We just don't support fast posting for anything less than the 1240/1080/1280/12160 or Fibre Channel cards. Put in support for CDB's larger than 12 bytes for parallel SCSI (up to 44 bytes are allowed). Approved: jkh
2000-02-15 00:35:00 +00:00
* Check command CDB length, etc.. We really are limited to 16 bytes
* for Fibre Channel, but can do up to 44 bytes in parallel SCSI,
* but probably only if we're running fairly new firmware (we'll
* let the old f/w choke on an extended command queue entry).
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
*/
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
Clean out residual bogosity for fast posting stuff- ISP_NO_FASTPOST_SCSI is gone as a define. We just don't support fast posting for anything less than the 1240/1080/1280/12160 or Fibre Channel cards. Put in support for CDB's larger than 12 bytes for parallel SCSI (up to 44 bytes are allowed). Approved: jkh
2000-02-15 00:35:00 +00:00
if (XS_CDBLEN(xs) > (IS_FC(isp)? 16 : 44) || XS_CDBLEN(xs) == 0) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGERR, "unsupported cdb length (%d, CDB[0]=0x%x)", XS_CDBLEN(xs), XS_CDBP(xs)[0] & 0xff);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
XS_SETERR(xs, HBA_BOTCH);
return (CMD_COMPLETE);
}
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* Translate the target to device handle as appropriate, checking
* for correct device state as well.
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
*/
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
target = XS_TGT(xs);
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
if (IS_FC(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcparam *fcp = FCPARAM(isp, XS_CHANNEL(xs));
if ((fcp->role & ISP_ROLE_INITIATOR) == 0) {
XS_SETERR(xs, HBA_SELTIMEOUT);
return (CMD_COMPLETE);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
/*
* Try again later.
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (fcp->isp_fwstate != FW_READY || fcp->isp_loopstate != LOOP_READY) {
Switch to using 16 bit handles instead of 32 bit handles. This is a pretty invasive change, but there are three good reasons to do this: 1. We'll never have > 16 bits of handle. 2. We can (eventually) enable the RIO (Reduced Interrupt Operation) bits which return multiple completing 16 bit handles in mailbox registers. 3. The !)$*)$*~)@$*~)$* Qlogic target mode for parallel SCSI spec changed such that at_reserved (which was 32 bits) was split into two pieces- and one of which was a 16 bit handle id that functions like the at_rxid for Fibre Channel (a tag for the f/w to correlate CTIOs with a particular command). Since we had to muck with that and this changed the whole handler architecture, we might as well... Propagate new at_handle on through int ct_fwhandle. Follow implications of changing to 16 bit handles. These above changes at least get Qlogic 1040 cards working in target mode again. 1080/12160 cards don't work yet. In isp.c: Prepare for doing all loop management in outer layers.
2001-03-02 06:28:55 +00:00
return (CMD_RQLATER);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (XS_TGT(xs) >= MAX_FC_TARG) {
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
XS_SETERR(xs, HBA_SELTIMEOUT);
return (CMD_COMPLETE);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
hdlidx = fcp->isp_dev_map[XS_TGT(xs)] - 1;
isp_prt(isp, ISP_LOGDEBUG2, "XS_TGT(xs)=%d- hdlidx value %d", XS_TGT(xs), hdlidx);
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
if (hdlidx < 0 || hdlidx >= MAX_FC_TARG) {
XS_SETERR(xs, HBA_SELTIMEOUT);
return (CMD_COMPLETE);
}
if (fcp->portdb[hdlidx].state == FC_PORTDB_STATE_ZOMBIE) {
return (CMD_RQLATER);
}
if (fcp->portdb[hdlidx].state != FC_PORTDB_STATE_VALID) {
Roll revision levels. Add support for the Qlogic 2200 (warn about not having SCSI_ISP_SCCLUN config defined if we don't have f/w for the 2200- it's resident firmware uses SCCLUN (65535 luns)). Change the way the default LoopID is gathered (it's now a platform specific define so that some attempt at a synthetic WWN can be made in case NVRAM isn't readable). Change initialization of options a bit- don't use ADISC. Set FullDuplex mode if config options tells us to do so. Do not use FULL_LOGIN after LIP- it's the right thing to do but it causes too much loop disruption (Loop Resets). Sanity check some default values. Redo construction of port and node WWNs based upon what we have- if we have 2 in the top nibble, we can have distinct port and node WWNs. Clean up some SCCLUN related code that we obviously had never compiled (:-(). Audit commands coming int ispscsicmd and don't throw commands at Fibre devices that do not have Class 3 service parameters TARGET ROLE defined. Clean up f/w initialization a bit. Add Fabric support (or at least the first blush of it). Whew - way too much to describe here. Basically, after a LIP, hang out until we see a Loop Up or a Port DataBase Change async event, then see if we're on a Fabric (GET_PORT_NAME of FL_PORT_ID). If we are, try and scan the fabric controller for fabric devices using the GetAllNext SNS subcommand. As we find devices, announce them to the outer layer. Try and do some guard code for broken (Brocade) SNS servers (that get stuck in loops- gotta maybe do this a different way using the GP_ID3 cmd instead). Then do a scan of the lower (local loop) ids using a GET_PORT_NAME to see if the f/w has logged into anything at that loop id. If so, then do a GET_PORT_DATABASE command. Do this scan into a local database. At this point we can say the loop is 'Ready'. After this, we merge our local loop port database with our stored port database- in a as yet to be really fully exercised fashion we try and follow the logic of something having moved around. The first time we see something at a Loop ID, we fix it, for the purpose of this system instance, at that Loop ID. If things shift around so it ends up somewhere else, we still keep it at this Loop ID (our 'Target') but use the new (moved) Loop ID when we actually throw commands at it. Check for insane cases of different Loop IDs both claiming to have the same WWN- if that happens, invalidate both. Notify the outer layer of devices that have arrived and devices that have gone away. *Finally*, when this is done, search the softc's database of Fabric devices and perform logout/login actions. The Qlogic f/w maintains logout/login for all local loop devices. We have to maintain logout/login for fabric devices- total PITA. Expect to see this area undergo more change over time.
1999-07-02 23:06:38 +00:00
XS_SETERR(xs, HBA_SELTIMEOUT);
return (CMD_COMPLETE);
}
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
target = fcp->portdb[hdlidx].handle;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcp->portdb[hdlidx].dirty = 1;
} else {
sdparam *sdp = SDPARAM(isp, XS_CHANNEL(xs));
if ((sdp->role & ISP_ROLE_INITIATOR) == 0) {
XS_SETERR(xs, HBA_SELTIMEOUT);
return (CMD_COMPLETE);
}
if (sdp->update) {
isp_spi_update(isp, XS_CHANNEL(xs));
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
start_again:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
qep = isp_getrqentry(isp);
if (qep == NULL) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGDEBUG0, "Request Queue Overflow");
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
XS_SETERR(xs, HBA_BOTCH);
return (CMD_EAGAIN);
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
XS_SETERR(xs, HBA_NOERROR);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
/*
* Now see if we need to synchronize the ISP with respect to anything.
* We do dual duty here (cough) for synchronizing for busses other
* than which we got here to send a command to.
*/
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
reqp = (ispreq_t *) local;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(local, QENTRY_LEN);
if (ISP_TST_SENDMARKER(isp, XS_CHANNEL(xs))) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_marker_24xx_t *m = (isp_marker_24xx_t *) reqp;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
m->mrk_header.rqs_entry_count = 1;
m->mrk_header.rqs_entry_type = RQSTYPE_MARKER;
m->mrk_modifier = SYNC_ALL;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_put_marker_24xx(isp, m, qep);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_marker_t *m = (isp_marker_t *) reqp;
m->mrk_header.rqs_entry_count = 1;
m->mrk_header.rqs_entry_type = RQSTYPE_MARKER;
m->mrk_target = (XS_CHANNEL(xs) << 7); /* bus # */
m->mrk_modifier = SYNC_ALL;
isp_put_marker(isp, m, qep);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SYNC_REQUEST(isp);
ISP_SET_SENDMARKER(isp, XS_CHANNEL(xs), 0);
goto start_again;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
reqp->req_header.rqs_entry_count = 1;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
reqp->req_header.rqs_entry_type = RQSTYPE_T7RQS;
} else if (IS_FC(isp)) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
reqp->req_header.rqs_entry_type = RQSTYPE_T2RQS;
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (XS_CDBLEN(xs) > 12) {
Clean out residual bogosity for fast posting stuff- ISP_NO_FASTPOST_SCSI is gone as a define. We just don't support fast posting for anything less than the 1240/1080/1280/12160 or Fibre Channel cards. Put in support for CDB's larger than 12 bytes for parallel SCSI (up to 44 bytes are allowed). Approved: jkh
2000-02-15 00:35:00 +00:00
reqp->req_header.rqs_entry_type = RQSTYPE_CMDONLY;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
} else {
Clean out residual bogosity for fast posting stuff- ISP_NO_FASTPOST_SCSI is gone as a define. We just don't support fast posting for anything less than the 1240/1080/1280/12160 or Fibre Channel cards. Put in support for CDB's larger than 12 bytes for parallel SCSI (up to 44 bytes are allowed). Approved: jkh
2000-02-15 00:35:00 +00:00
reqp->req_header.rqs_entry_type = RQSTYPE_REQUEST;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Various minor and not so minor fixes suggested by Coverity. In at least one case, it's amazing that target mode worked at all. Found by: Coverity. MFC after: 2 weeks
2010-06-02 23:31:27 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
int ttype;
if (XS_TAG_P(xs)) {
ttype = XS_TAG_TYPE(xs);
} else {
if (XS_CDBP(xs)[0] == 0x3) {
ttype = REQFLAG_HTAG;
} else {
ttype = REQFLAG_STAG;
}
}
if (ttype == REQFLAG_OTAG) {
ttype = FCP_CMND_TASK_ATTR_ORDERED;
} else if (ttype == REQFLAG_HTAG) {
ttype = FCP_CMND_TASK_ATTR_HEAD;
} else {
ttype = FCP_CMND_TASK_ATTR_SIMPLE;
}
((ispreqt7_t *)reqp)->req_task_attribute = ttype;
} else if (IS_FC(isp)) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
* See comment in isp_intr
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/* XS_SET_RESID(xs, 0); */
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
I was misinformed. I cannot get away from specifying tags for FC. Some devices are happy w/o them- some are unhappy (IBM drives).
1999-10-28 02:48:42 +00:00
* Fibre Channel always requires some kind of tag.
* The Qlogic drivers seem be happy not to use a tag,
* but this breaks for some devices (IBM drives).
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (XS_TAG_P(xs)) {
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
((ispreqt2_t *)reqp)->req_flags = XS_TAG_TYPE(xs);
I was misinformed. I cannot get away from specifying tags for FC. Some devices are happy w/o them- some are unhappy (IBM drives).
1999-10-28 02:48:42 +00:00
} else {
various fixes
2000-08-27 23:38:44 +00:00
/*
* If we don't know what tag to use, use HEAD OF QUEUE
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
* for Request Sense or Simple.
various fixes
2000-08-27 23:38:44 +00:00
*/
I was misinformed. I cannot get away from specifying tags for FC. Some devices are happy w/o them- some are unhappy (IBM drives).
1999-10-28 02:48:42 +00:00
if (XS_CDBP(xs)[0] == 0x3) /* REQUEST SENSE */
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
((ispreqt2_t *)reqp)->req_flags = REQFLAG_HTAG;
I was misinformed. I cannot get away from specifying tags for FC. Some devices are happy w/o them- some are unhappy (IBM drives).
1999-10-28 02:48:42 +00:00
else
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
((ispreqt2_t *)reqp)->req_flags = REQFLAG_STAG;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdparam *sdp = SDPARAM(isp, XS_CHANNEL(xs));
if ((sdp->isp_devparam[target].actv_flags & DPARM_TQING) && XS_TAG_P(xs)) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
reqp->req_flags = XS_TAG_TYPE(xs);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
}
Various minor and not so minor fixes suggested by Coverity. In at least one case, it's amazing that target mode worked at all. Found by: Coverity. MFC after: 2 weeks
2010-06-02 23:31:27 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
tptr = &reqp->req_time;
Be more specific about which CDB length we're going to use. Not really a likely bug but we might as well be clearer. Found with: Coverity Prevent(tm) CID: 3981 MFC after: 2 weeks
2010-06-05 20:37:40 +00:00
/*
* NB: we do not support long CDBs
*/
cdblen = XS_CDBLEN(xs);
More code cleanup. Go back to using FULL_LOGIN Fibre Chan if f/w is less than 1.17.0 level. Change where we do the loop database init. Add in the CMD_RQLATER return. Add some register debounce.
1999-08-16 19:59:55 +00:00
if (IS_SCSI(isp)) {
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
reqp->req_target = target | (XS_CHANNEL(xs) << 7);
Remove the Target mode functions until they're in better shape. Implement some suggested compilation cleanups from Eklund. Wire down a hard loop id if we are not on a platform that has the ability to get to a PCI BIOS (it still will float to the ID it gets after a LIP but at least we can try). Clarify that the expanded lun is based upon SCCLUN defines (in f/w).
1998-12-05 01:33:57 +00:00
reqp->req_lun_trn = XS_LUN(xs);
Be more specific about which CDB length we're going to use. Not really a likely bug but we might as well be clearer. Found with: Coverity Prevent(tm) CID: 3981 MFC after: 2 weeks
2010-06-05 20:37:40 +00:00
cdblen = MIN(cdblen, sizeof (reqp->req_cdb));
Various minor and not so minor fixes suggested by Coverity. In at least one case, it's amazing that target mode worked at all. Found by: Coverity. MFC after: 2 weeks
2010-06-02 23:31:27 +00:00
cdbp = reqp->req_cdb;
Be more specific about which CDB length we're going to use. Not really a likely bug but we might as well be clearer. Found with: Coverity Prevent(tm) CID: 3981 MFC after: 2 weeks
2010-06-05 20:37:40 +00:00
reqp->req_cdblen = cdblen;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else if (IS_24XX(isp)) {
Various minor and not so minor fixes suggested by Coverity. In at least one case, it's amazing that target mode worked at all. Found by: Coverity. MFC after: 2 weeks
2010-06-02 23:31:27 +00:00
ispreqt7_t *t7 = (ispreqt7_t *)local;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
fcportdb_t *lp;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
lp = &FCPARAM(isp, XS_CHANNEL(xs))->portdb[hdlidx];
Various minor and not so minor fixes suggested by Coverity. In at least one case, it's amazing that target mode worked at all. Found by: Coverity. MFC after: 2 weeks
2010-06-02 23:31:27 +00:00
t7->req_nphdl = target;
t7->req_tidlo = lp->portid;
t7->req_tidhi = lp->portid >> 16;
t7->req_vpidx = ISP_GET_VPIDX(isp, XS_CHANNEL(xs));
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (XS_LUN(xs) > 256) {
Various minor and not so minor fixes suggested by Coverity. In at least one case, it's amazing that target mode worked at all. Found by: Coverity. MFC after: 2 weeks
2010-06-02 23:31:27 +00:00
t7->req_lun[0] = XS_LUN(xs) >> 8;
t7->req_lun[0] |= 0x40;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Various minor and not so minor fixes suggested by Coverity. In at least one case, it's amazing that target mode worked at all. Found by: Coverity. MFC after: 2 weeks
2010-06-02 23:31:27 +00:00
t7->req_lun[1] = XS_LUN(xs);
tptr = &t7->req_time;
Be more specific about which CDB length we're going to use. Not really a likely bug but we might as well be clearer. Found with: Coverity Prevent(tm) CID: 3981 MFC after: 2 weeks
2010-06-05 20:37:40 +00:00
cdbp = t7->req_cdb;
cdblen = MIN(cdblen, sizeof (t7->req_cdb));
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
} else if (ISP_CAP_2KLOGIN(isp)) {
Various minor and not so minor fixes suggested by Coverity. In at least one case, it's amazing that target mode worked at all. Found by: Coverity. MFC after: 2 weeks
2010-06-02 23:31:27 +00:00
ispreqt2e_t *t2e = (ispreqt2e_t *)local;
t2e->req_target = target;
t2e->req_scclun = XS_LUN(xs);
cdbp = t2e->req_cdb;
Be more specific about which CDB length we're going to use. Not really a likely bug but we might as well be clearer. Found with: Coverity Prevent(tm) CID: 3981 MFC after: 2 weeks
2010-06-05 20:37:40 +00:00
cdblen = MIN(cdblen, sizeof (t2e->req_cdb));
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
} else if (ISP_CAP_SCCFW(isp)) {
Various minor and not so minor fixes suggested by Coverity. In at least one case, it's amazing that target mode worked at all. Found by: Coverity. MFC after: 2 weeks
2010-06-02 23:31:27 +00:00
ispreqt2_t *t2 = (ispreqt2_t *)local;
t2->req_target = target;
t2->req_scclun = XS_LUN(xs);
cdbp = t2->req_cdb;
Be more specific about which CDB length we're going to use. Not really a likely bug but we might as well be clearer. Found with: Coverity Prevent(tm) CID: 3981 MFC after: 2 weeks
2010-06-05 20:37:40 +00:00
cdblen = MIN(cdblen, sizeof (t2->req_cdb));
Remove the Target mode functions until they're in better shape. Implement some suggested compilation cleanups from Eklund. Wire down a hard loop id if we are not on a platform that has the ability to get to a PCI BIOS (it still will float to the ID it gets after a LIP but at least we can try). Clarify that the expanded lun is based upon SCCLUN defines (in f/w).
1998-12-05 01:33:57 +00:00
} else {
Various minor and not so minor fixes suggested by Coverity. In at least one case, it's amazing that target mode worked at all. Found by: Coverity. MFC after: 2 weeks
2010-06-02 23:31:27 +00:00
ispreqt2_t *t2 = (ispreqt2_t *)local;
t2->req_target = target;
t2->req_lun_trn = XS_LUN(xs);
cdbp = t2->req_cdb;
Be more specific about which CDB length we're going to use. Not really a likely bug but we might as well be clearer. Found with: Coverity Prevent(tm) CID: 3981 MFC after: 2 weeks
2010-06-05 20:37:40 +00:00
cdblen = MIN(cdblen, sizeof (t2->req_cdb));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Be more specific about which CDB length we're going to use. Not really a likely bug but we might as well be clearer. Found with: Coverity Prevent(tm) CID: 3981 MFC after: 2 weeks
2010-06-05 20:37:40 +00:00
ISP_MEMCPY(cdbp, XS_CDBP(xs), cdblen);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*tptr = XS_TIME(xs) / 1000;
if (*tptr == 0 && XS_TIME(xs)) {
*tptr = 1;
}
if (IS_24XX(isp) && *tptr > 0x1999) {
*tptr = 0x1999;
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
}
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
if (isp_allocate_xs(isp, xs, &handle)) {
Tra-La, another QLogic f/w funny- this time with the 2300. If we get a completion status of RQCS_QUEUE_FULL, it means that the internal queues are full. Other QLogic boards set the QFULL SCSI status. But *nooooooooooo*, not the 2300. MFC after: 1 day
2001-10-23 23:05:20 +00:00
isp_prt(isp, ISP_LOGDEBUG0, "out of xflist pointers");
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
XS_SETERR(xs, HBA_BOTCH);
return (CMD_EAGAIN);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/* Whew. Thankfully the same for type 7 requests */
Switch to using 16 bit handles instead of 32 bit handles. This is a pretty invasive change, but there are three good reasons to do this: 1. We'll never have > 16 bits of handle. 2. We can (eventually) enable the RIO (Reduced Interrupt Operation) bits which return multiple completing 16 bit handles in mailbox registers. 3. The !)$*)$*~)@$*~)$* Qlogic target mode for parallel SCSI spec changed such that at_reserved (which was 32 bits) was split into two pieces- and one of which was a 16 bit handle id that functions like the at_rxid for Fibre Channel (a tag for the f/w to correlate CTIOs with a particular command). Since we had to muck with that and this changed the whole handler architecture, we might as well... Propagate new at_handle on through int ct_fwhandle. Follow implications of changing to 16 bit handles. These above changes at least get Qlogic 1040 cards working in target mode again. 1080/12160 cards don't work yet. In isp.c: Prepare for doing all loop management in outer layers.
2001-03-02 06:28:55 +00:00
reqp->req_handle = handle;
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
/*
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* Set up DMA and/or do any platform dependent swizzling of the request entry
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
* so that the Qlogic F/W understands what is being asked of it.
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
*
* The callee is responsible for adding all requests at this point.
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
dmaresult = ISP_DMASETUP(isp, xs, reqp);
if (dmaresult != CMD_QUEUED) {
Switch to using 16 bit handles instead of 32 bit handles. This is a pretty invasive change, but there are three good reasons to do this: 1. We'll never have > 16 bits of handle. 2. We can (eventually) enable the RIO (Reduced Interrupt Operation) bits which return multiple completing 16 bit handles in mailbox registers. 3. The !)$*)$*~)@$*~)$* Qlogic target mode for parallel SCSI spec changed such that at_reserved (which was 32 bits) was split into two pieces- and one of which was a 16 bit handle id that functions like the at_rxid for Fibre Channel (a tag for the f/w to correlate CTIOs with a particular command). Since we had to muck with that and this changed the whole handler architecture, we might as well... Propagate new at_handle on through int ct_fwhandle. Follow implications of changing to 16 bit handles. These above changes at least get Qlogic 1040 cards working in target mode again. 1080/12160 cards don't work yet. In isp.c: Prepare for doing all loop management in outer layers.
2001-03-02 06:28:55 +00:00
isp_destroy_handle(isp, handle);
ISP_DMASETUP now returns a value to be possibly punted to outer layers. Turn request queue overflow messages into debug messages. Ensure on isp_restarts that we nullify the xflist array.
1998-09-17 21:03:45 +00:00
/*
* dmasetup sets actual error in packet, and
* return what we were given to return.
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
return (dmaresult);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGDEBUG0, "START cmd cdb[0]=0x%x datalen %ld", XS_CDBP(xs)[0], (long) XS_XFRLEN(xs));
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
isp->isp_nactive++;
return (CMD_QUEUED);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
/*
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
* isp control
* Locks (ints blocked) assumed held.
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
int
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_control(ispsoftc_t *isp, ispctl_t ctl, ...)
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
{
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
XS_T *xs;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
mbreg_t *mbr, mbs;
int chan, tgt;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
uint32_t handle;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
va_list ap;
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
switch (ctl) {
case ISPCTL_RESET_BUS:
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
/*
* Issue a bus reset.
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
some minor error message cleanups
2007-03-29 21:29:26 +00:00
isp_prt(isp, ISP_LOGWARN, "RESET BUS NOT IMPLEMENTED");
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
} else if (IS_FC(isp)) {
mbs.param[1] = 10;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
chan = 0;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
va_start(ap, ctl);
chan = va_arg(ap, int);
va_end(ap);
mbs.param[1] = SDPARAM(isp, chan)->isp_bus_reset_delay;
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
if (mbs.param[1] < 2) {
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
mbs.param[1] = 2;
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
mbs.param[2] = chan;
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 01:07:06 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_BUS_RESET, MBLOGALL, 0);
ISP_SET_SENDMARKER(isp, chan, 1);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
break;
}
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGINFO,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"driver initiated bus reset of bus %d", chan);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
return (0);
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
case ISPCTL_RESET_DEV:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
va_start(ap, ctl);
chan = va_arg(ap, int);
tgt = va_arg(ap, int);
va_end(ap);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
uint8_t local[QENTRY_LEN];
isp24xx_tmf_t *tmf;
isp24xx_statusreq_t *sp;
fcparam *fcp = FCPARAM(isp, chan);
fcportdb_t *lp;
int hdlidx;
hdlidx = fcp->isp_dev_map[tgt] - 1;
if (hdlidx < 0 || hdlidx >= MAX_FC_TARG) {
isp_prt(isp, ISP_LOGWARN,
"Chan %d bad handle %d trying to reset"
"target %d", chan, hdlidx, tgt);
break;
}
lp = &fcp->portdb[hdlidx];
if (lp->state != FC_PORTDB_STATE_VALID) {
isp_prt(isp, ISP_LOGWARN,
"Chan %d handle %d for abort of target %d "
"no longer valid", chan,
hdlidx, tgt);
break;
}
tmf = (isp24xx_tmf_t *) local;
ISP_MEMZERO(tmf, QENTRY_LEN);
tmf->tmf_header.rqs_entry_type = RQSTYPE_TSK_MGMT;
tmf->tmf_header.rqs_entry_count = 1;
tmf->tmf_nphdl = lp->handle;
tmf->tmf_delay = 2;
tmf->tmf_timeout = 2;
tmf->tmf_flags = ISP24XX_TMF_TARGET_RESET;
tmf->tmf_tidlo = lp->portid;
tmf->tmf_tidhi = lp->portid >> 16;
tmf->tmf_vpidx = ISP_GET_VPIDX(isp, chan);
isp_prt(isp, ISP_LOGALL, "Chan %d Reset N-Port Handle 0x%04x @ Port 0x%06x", chan, lp->handle, lp->portid);
MBSINIT(&mbs, MBOX_EXEC_COMMAND_IOCB_A64, MBLOGALL, 5000000);
mbs.param[1] = QENTRY_LEN;
mbs.param[2] = DMA_WD1(fcp->isp_scdma);
mbs.param[3] = DMA_WD0(fcp->isp_scdma);
mbs.param[6] = DMA_WD3(fcp->isp_scdma);
mbs.param[7] = DMA_WD2(fcp->isp_scdma);
if (FC_SCRATCH_ACQUIRE(isp, chan)) {
isp_prt(isp, ISP_LOGERR, sacq);
break;
}
isp_put_24xx_tmf(isp, tmf, fcp->isp_scratch);
MEMORYBARRIER(isp, SYNC_SFORDEV, 0, QENTRY_LEN);
fcp->sendmarker = 1;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
FC_SCRATCH_RELEASE(isp, chan);
break;
}
MEMORYBARRIER(isp, SYNC_SFORCPU, QENTRY_LEN,
QENTRY_LEN);
sp = (isp24xx_statusreq_t *) local;
isp_get_24xx_response(isp,
&((isp24xx_statusreq_t *)fcp->isp_scratch)[1], sp);
FC_SCRATCH_RELEASE(isp, chan);
if (sp->req_completion_status == 0) {
return (0);
}
isp_prt(isp, ISP_LOGWARN,
"Chan %d reset of target %d returned 0x%x",
chan, tgt, sp->req_completion_status);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
} else if (IS_FC(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (ISP_CAP_2KLOGIN(isp)) {
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
mbs.param[1] = tgt;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.ibits = (1 << 10);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
} else {
mbs.param[1] = (tgt << 8);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
mbs.param[1] = (chan << 15) | (tgt << 8);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_ABORT_TARGET, MBLOGALL, 0);
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 01:07:06 +00:00
mbs.param[2] = 3; /* 'delay', in seconds */
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
break;
}
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGINFO,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"Target %d on Bus %d Reset Succeeded", tgt, chan);
ISP_SET_SENDMARKER(isp, chan, 1);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
return (0);
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
case ISPCTL_ABORT_CMD:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
va_start(ap, ctl);
xs = va_arg(ap, XS_T *);
va_end(ap);
Clean up ISPCTL_ABORT_CMD function to not be too chatty if it succeeds, or even if it fails with INVALID_PARM (which just means that the handle doesn't refer to an active commane).
2000-07-05 06:41:36 +00:00
tgt = XS_TGT(xs);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
chan = XS_CHANNEL(xs);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
handle = isp_find_handle(isp, xs);
if (handle == 0) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGWARN,
"cannot find handle for command to abort");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp24xx_abrt_t local, *ab = &local, *ab2;
fcparam *fcp;
fcportdb_t *lp;
int hdlidx;
fcp = FCPARAM(isp, chan);
hdlidx = fcp->isp_dev_map[tgt] - 1;
if (hdlidx < 0 || hdlidx >= MAX_FC_TARG) {
isp_prt(isp, ISP_LOGWARN,
"Chan %d bad handle %d trying to abort"
"target %d", chan, hdlidx, tgt);
break;
}
lp = &fcp->portdb[hdlidx];
if (lp->state != FC_PORTDB_STATE_VALID) {
isp_prt(isp, ISP_LOGWARN,
"Chan %d handle %d for abort of target %d "
"no longer valid", chan, hdlidx, tgt);
break;
}
isp_prt(isp, ISP_LOGALL,
"Chan %d Abort Cmd for N-Port 0x%04x @ Port "
"0x%06x %p", chan, lp->handle, lp->portid, xs);
ISP_MEMZERO(ab, QENTRY_LEN);
ab->abrt_header.rqs_entry_type = RQSTYPE_ABORT_IO;
ab->abrt_header.rqs_entry_count = 1;
ab->abrt_handle = lp->handle;
ab->abrt_cmd_handle = handle;
ab->abrt_tidlo = lp->portid;
ab->abrt_tidhi = lp->portid >> 16;
ab->abrt_vpidx = ISP_GET_VPIDX(isp, chan);
ISP_MEMZERO(&mbs, sizeof (mbs));
MBSINIT(&mbs, MBOX_EXEC_COMMAND_IOCB_A64, MBLOGALL, 5000000);
mbs.param[1] = QENTRY_LEN;
mbs.param[2] = DMA_WD1(fcp->isp_scdma);
mbs.param[3] = DMA_WD0(fcp->isp_scdma);
mbs.param[6] = DMA_WD3(fcp->isp_scdma);
mbs.param[7] = DMA_WD2(fcp->isp_scdma);
if (FC_SCRATCH_ACQUIRE(isp, chan)) {
isp_prt(isp, ISP_LOGERR, sacq);
break;
}
isp_put_24xx_abrt(isp, ab, fcp->isp_scratch);
ab2 = (isp24xx_abrt_t *)
&((uint8_t *)fcp->isp_scratch)[QENTRY_LEN];
ab2->abrt_nphdl = 0xdeaf;
MEMORYBARRIER(isp, SYNC_SFORDEV, 0, 2 * QENTRY_LEN);
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
FC_SCRATCH_RELEASE(isp, chan);
break;
}
MEMORYBARRIER(isp, SYNC_SFORCPU, QENTRY_LEN,
QENTRY_LEN);
isp_get_24xx_abrt(isp, ab2, ab);
FC_SCRATCH_RELEASE(isp, chan);
if (ab->abrt_nphdl == ISP24XX_ABRT_OKAY) {
return (0);
}
isp_prt(isp, ISP_LOGWARN,
"Chan %d handle %d abort returned 0x%x", chan,
hdlidx, ab->abrt_nphdl);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
} else if (IS_FC(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (ISP_CAP_SCCFW(isp)) {
if (ISP_CAP_2KLOGIN(isp)) {
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
mbs.param[1] = tgt;
} else {
mbs.param[1] = tgt << 8;
}
Once we have firmware running (if isp_reset) and this is the first time through, establish what our LUN width is. Unfortunately, we can't ask the f/w. If we loaded the f/w, we'll now assume we have expanded LUNs (SCCLUN for fibre channel, just plain 32 LUN for SCSI). If we didn't load firmware, assume 8 LUNs for SCSI and 1 LUN for Fibre Channel. We have to assume only one LUN for Fibre Channel because the LUN setting in Request Queue entries is in different places whether we have SCCLUN firmware or not, so the only LUN guaranteed to work for both is LUN 0. Clean up the rest of isp.c so that ISP2100_SCCLUN defines aren't used- instead use run time determinants based upon isp->isp_maxluns. After starting firmware, delay 500us to give it a chance to get rolling. Fix the interrupt service routine to check for both isr && sema being zero before thinking this was a spurious interrupt. Following the manuals, allow for both Mailbox as well as Queue Reponse type interrupts for regular SCSI.
2000-06-18 04:56:17 +00:00
mbs.param[6] = XS_LUN(xs);
} else {
Clean up ISPCTL_ABORT_CMD function to not be too chatty if it succeeds, or even if it fails with INVALID_PARM (which just means that the handle doesn't refer to an active commane).
2000-07-05 06:41:36 +00:00
mbs.param[1] = tgt << 8 | XS_LUN(xs);
Once we have firmware running (if isp_reset) and this is the first time through, establish what our LUN width is. Unfortunately, we can't ask the f/w. If we loaded the f/w, we'll now assume we have expanded LUNs (SCCLUN for fibre channel, just plain 32 LUN for SCSI). If we didn't load firmware, assume 8 LUNs for SCSI and 1 LUN for Fibre Channel. We have to assume only one LUN for Fibre Channel because the LUN setting in Request Queue entries is in different places whether we have SCCLUN firmware or not, so the only LUN guaranteed to work for both is LUN 0. Clean up the rest of isp.c so that ISP2100_SCCLUN defines aren't used- instead use run time determinants based upon isp->isp_maxluns. After starting firmware, delay 500us to give it a chance to get rolling. Fix the interrupt service routine to check for both isr && sema being zero before thinking this was a spurious interrupt. Following the manuals, allow for both Mailbox as well as Queue Reponse type interrupts for regular SCSI.
2000-06-18 04:56:17 +00:00
}
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
mbs.param[1] = (chan << 15) | (tgt << 8) | XS_LUN(xs);
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_ABORT, MBLOGALL & ~MBOX_COMMAND_ERROR, 0);
Switch to using 16 bit handles instead of 32 bit handles. This is a pretty invasive change, but there are three good reasons to do this: 1. We'll never have > 16 bits of handle. 2. We can (eventually) enable the RIO (Reduced Interrupt Operation) bits which return multiple completing 16 bit handles in mailbox registers. 3. The !)$*)$*~)@$*~)$* Qlogic target mode for parallel SCSI spec changed such that at_reserved (which was 32 bits) was split into two pieces- and one of which was a 16 bit handle id that functions like the at_rxid for Fibre Channel (a tag for the f/w to correlate CTIOs with a particular command). Since we had to muck with that and this changed the whole handler architecture, we might as well... Propagate new at_handle on through int ct_fwhandle. Follow implications of changing to 16 bit handles. These above changes at least get Qlogic 1040 cards working in target mode again. 1080/12160 cards don't work yet. In isp.c: Prepare for doing all loop management in outer layers.
2001-03-02 06:28:55 +00:00
mbs.param[2] = handle;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
break;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return (0);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
case ISPCTL_UPDATE_PARAMS:
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
va_start(ap, ctl);
chan = va_arg(ap, int);
va_end(ap);
isp_spi_update(isp, chan);
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
return (0);
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
case ISPCTL_FCLINK_TEST:
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
if (IS_FC(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
int usdelay;
va_start(ap, ctl);
chan = va_arg(ap, int);
usdelay = va_arg(ap, int);
va_end(ap);
Redo some code based upon issues found by Coverity.
2006-04-21 18:46:35 +00:00
if (usdelay == 0) {
usdelay = 250000;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
return (isp_fclink_test(isp, chan, usdelay));
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
}
break;
Clean up stupidity in the isp_handle_other_response function- indexes of queue entries have to be at least 16 bits now! If we're running a 2100 less than rev 5, turn off loop fairness (per Qlogic errata). Fix typo in checking against 2200 F/W revision. Slightly fix/reorder fabric login stuff. Change to usage of isp_getrqentry for code clarity. Add some defensive dual bus assumptions. Various cleanups, etc...
1999-11-21 03:18:22 +00:00
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
case ISPCTL_SCAN_FABRIC:
if (IS_FC(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
va_start(ap, ctl);
chan = va_arg(ap, int);
va_end(ap);
return (isp_scan_fabric(isp, chan));
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
}
break;
case ISPCTL_SCAN_LOOP:
if (IS_FC(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
va_start(ap, ctl);
chan = va_arg(ap, int);
va_end(ap);
return (isp_scan_loop(isp, chan));
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
}
break;
Raise default FCP logintime to 60 seconds. Move the position of where we could have seen the loop up at least once so it makes sense. Change some stuff in ispscsicmd so we don't get stuck there if the loop has never come up yet. Add in some target mode support code.
2000-01-03 23:52:41 +00:00
case ISPCTL_PDB_SYNC:
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
if (IS_FC(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
va_start(ap, ctl);
chan = va_arg(ap, int);
va_end(ap);
return (isp_pdb_sync(isp, chan));
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
}
break;
case ISPCTL_SEND_LIP:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_FC(isp) && !IS_24XX(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_INIT_LIP, MBLOGALL, 0);
if (ISP_CAP_2KLOGIN(isp)) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.ibits = (1 << 10);
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
if (mbs.param[0] == MBOX_COMMAND_COMPLETE) {
return (0);
}
}
break;
Add case to handle ISPCTL_GET_PDB. MFC after: 1 week
2004-02-07 03:42:17 +00:00
case ISPCTL_GET_PDB:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (IS_FC(isp)) {
isp_pdb_t *pdb;
va_start(ap, ctl);
chan = va_arg(ap, int);
tgt = va_arg(ap, int);
pdb = va_arg(ap, isp_pdb_t *);
va_end(ap);
return (isp_getpdb(isp, chan, tgt, pdb, 1));
Add case to handle ISPCTL_GET_PDB. MFC after: 1 week
2004-02-07 03:42:17 +00:00
}
break;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
case ISPCTL_GET_NAMES:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
{
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
uint64_t *wwnn, *wwnp;
va_start(ap, ctl);
chan = va_arg(ap, int);
tgt = va_arg(ap, int);
wwnn = va_arg(ap, uint64_t *);
wwnp = va_arg(ap, uint64_t *);
va_end(ap);
if (wwnn == NULL && wwnp == NULL) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (wwnn) {
*wwnn = isp_get_wwn(isp, chan, tgt, 1);
if (*wwnn == INI_NONE) {
break;
}
}
if (wwnp) {
*wwnp = isp_get_wwn(isp, chan, tgt, 0);
if (*wwnp == INI_NONE) {
break;
}
}
return (0);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Fix a longstanding bug- we had the sense of what bit 14 for the ICB firmware options meant- *I* had taken it to mean that if you set it, Node Name would be ignored and derived from Port Name. Actually, it meant the opposite. As a consequence- change ICBOPT_USE_PORTNAME to the define ICBOPT_BOTH_WWNS- makes more sense. Fix wrong input bitmap for MBOX_DUMP_RAM command. Call ISP_DUMPREGS if we get a f/w crash. Add ISPCTL_RUN_MBOXCMD control command (so outer layers can run a mailbox command directly) and add a ISPASYNC_UNHANDLED_RESPONSE hook so outer layers can understand response queue entries we might not know about.
2001-02-23 05:35:50 +00:00
case ISPCTL_RUN_MBOXCMD:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
{
va_start(ap, ctl);
mbr = va_arg(ap, mbreg_t *);
va_end(ap);
isp_mboxcmd(isp, mbr);
return (0);
}
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
case ISPCTL_PLOGX:
{
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_plcmd_t *p;
Make ISPCTL_PLOGX find a handle to log into the management server with- not hope for the best. Change some things which were gated off of 24XX to be gated off of 2K login support. Convert some isp_prt calls to xpt_print calls.
2006-12-05 07:50:23 +00:00
int r;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
va_start(ap, ctl);
p = va_arg(ap, isp_plcmd_t *);
va_end(ap);
if ((p->flags & PLOGX_FLG_CMD_MASK) != PLOGX_FLG_CMD_PLOGI || (p->handle != NIL_HANDLE)) {
return (isp_plogx(isp, p->channel, p->handle, p->portid, p->flags, 0));
Make ISPCTL_PLOGX find a handle to log into the management server with- not hope for the best. Change some things which were gated off of 24XX to be gated off of 2K login support. Convert some isp_prt calls to xpt_print calls.
2006-12-05 07:50:23 +00:00
}
do {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
p->handle = isp_nxt_handle(isp, p->channel, p->handle);
r = isp_plogx(isp, p->channel, p->handle, p->portid, p->flags, 0);
Make ISPCTL_PLOGX find a handle to log into the management server with- not hope for the best. Change some things which were gated off of 24XX to be gated off of 2K login support. Convert some isp_prt calls to xpt_print calls.
2006-12-05 07:50:23 +00:00
if ((r & 0xffff) == MBOX_PORT_ID_USED) {
p->handle = r >> 16;
r = 0;
break;
}
} while ((r & 0xffff) == MBOX_LOOP_ID_USED);
return (r);
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
default:
isp_prt(isp, ISP_LOGERR, "Unknown Control Opcode 0x%x", ctl);
break;
When resetting the Qlogic 2X00 units, reset the FPM (Fibre Protocol Module) and FBM (Fibre Buffer Modules). Also remember to clear the semaphore registers. Tell the RISC processor to not halt on FPM parity errors. Throw out the ISP_CFG_NOINIT silliness and instead go to the use of adapter 'roles' to see whether one completes initialization or not (mostly for Fibre Channel). The ultimate intent, btw, of all of this is to have a warm standby adapter for failover reasons. Because we do roles now, setting of Target Capable Class 3 service parameters in the ICB for the 2x00 cards reflects from role. Also, in isp_start, if we're not supporting an initiator role, we bounce outgoing commands with a Selection Timeout error. Also clean out the TOGGLE_TMODE goop for FC- there is no toggling of target mode like there is for parallel SCSI cards. Do more cleanup with respect to using target ids 0..125 in F-port topologies. Also keep track of things which *were* fabric devices so that when you rescan the fabric you can notify the outer layers when fabric devices go away. Only force a LOGOUT for fabric devices if they're still logged in (i.e., you cat their Port Database entry. Clean up the Get All Next scanning. Finally, use a new tag in the softc to store the opcode for the last mailbox command used so we can report which opcode timed out.
2001-01-15 18:33:08 +00:00
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
return (-1);
}
/*
* Interrupt Service Routine(s).
*
* External (OS) framework has done the appropriate locking,
* and the locking will be held throughout this function.
*/
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
/*
* Limit our stack depth by sticking with the max likely number
* of completions on a request queue at any one time.
*/
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
#ifndef MAX_REQUESTQ_COMPLETIONS
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
#define MAX_REQUESTQ_COMPLETIONS 32
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
#endif
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_intr(ispsoftc_t *isp, uint32_t isr, uint16_t sema, uint16_t mbox)
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
{
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
XS_T *complist[MAX_REQUESTQ_COMPLETIONS], *xs;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
uint32_t iptr, optr, junk;
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 01:07:06 +00:00
int i, nlooked = 0, ndone = 0;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
again:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
optr = isp->isp_residx;
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
/*
* Is this a mailbox related interrupt?
* The mailbox semaphore will be nonzero if so.
*/
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
if (sema) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fmbox:
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
if (mbox & MBOX_COMMAND_COMPLETE) {
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
isp->isp_intmboxc++;
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
if (isp->isp_mboxbsy) {
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
int obits = isp->isp_obits;
isp->isp_mboxtmp[0] = mbox;
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
for (i = 1; i < MAX_MAILBOX(isp); i++) {
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
if ((obits & (1 << i)) == 0) {
continue;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp->isp_mboxtmp[i] = ISP_READ(isp, MBOX_OFF(i));
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
}
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
if (isp->isp_mbxwrk0) {
if (isp_mbox_continue(isp) == 0) {
return;
}
}
Change delay loop in new isp_mboxcmd to the use of the new MBOX_WAIT_COMPLETE macro. Change notification of completion of a mailbox command in isp_intr to MBOX_NOTIFY_COMPLETE macro.
2000-07-04 01:02:38 +00:00
MBOX_NOTIFY_COMPLETE(isp);
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGWARN, "mailbox cmd (0x%x) with no waiters", mbox);
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
}
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
} else {
i = IS_FC(isp)? isp_parse_async_fc(isp, mbox) : isp_parse_async(isp, mbox);
if (i < 0) {
return;
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
}
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
if ((IS_FC(isp) && mbox != ASYNC_RIOZIO_STALL) || isp->isp_state != ISP_RUNSTATE) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
goto out;
Once we have firmware running (if isp_reset) and this is the first time through, establish what our LUN width is. Unfortunately, we can't ask the f/w. If we loaded the f/w, we'll now assume we have expanded LUNs (SCCLUN for fibre channel, just plain 32 LUN for SCSI). If we didn't load firmware, assume 8 LUNs for SCSI and 1 LUN for Fibre Channel. We have to assume only one LUN for Fibre Channel because the LUN setting in Request Queue entries is in different places whether we have SCCLUN firmware or not, so the only LUN guaranteed to work for both is LUN 0. Clean up the rest of isp.c so that ISP2100_SCCLUN defines aren't used- instead use run time determinants based upon isp->isp_maxluns. After starting firmware, delay 500us to give it a chance to get rolling. Fix the interrupt service routine to check for both isr && sema being zero before thinking this was a spurious interrupt. Following the manuals, allow for both Mailbox as well as Queue Reponse type interrupts for regular SCSI.
2000-06-18 04:56:17 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
/*
* We can't be getting this now.
*/
if (isp->isp_state != ISP_RUNSTATE) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* This seems to happen to 23XX and 24XX cards- don't know why.
*/
if (isp->isp_mboxbsy && isp->isp_lastmbxcmd == MBOX_ABOUT_FIRMWARE) {
goto fmbox;
}
isp_prt(isp, ISP_LOGINFO, "interrupt (ISR=%x SEMA=%x) when not ready", isr, sema);
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
/*
* Thank you very much! *Burrrp*!
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_WRITE(isp, isp->isp_respoutrp, ISP_READ(isp, isp->isp_respinrp));
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
ISP_DISABLE_INTS(isp);
}
goto out;
}
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
#ifdef ISP_TARGET_MODE
/*
* Check for ATIO Queue entries.
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (IS_24XX(isp)) {
iptr = ISP_READ(isp, BIU2400_ATIO_RSPINP);
optr = ISP_READ(isp, BIU2400_ATIO_RSPOUTP);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
while (optr != iptr) {
uint8_t qe[QENTRY_LEN];
isphdr_t *hp;
uint32_t oop;
void *addr;
oop = optr;
MEMORYBARRIER(isp, SYNC_ATIOQ, oop, QENTRY_LEN);
addr = ISP_QUEUE_ENTRY(isp->isp_atioq, oop);
isp_get_hdr(isp, addr, (isphdr_t *)qe);
hp = (isphdr_t *)qe;
switch (hp->rqs_entry_type) {
case RQSTYPE_NOTIFY:
case RQSTYPE_ATIO:
(void) isp_target_notify(isp, addr, &oop);
break;
default:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_print_qentry(isp, "?ATIOQ entry?", oop, addr);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
}
optr = ISP_NXT_QENTRY(oop, RESULT_QUEUE_LEN(isp));
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_WRITE(isp, BIU2400_ATIO_RSPOUTP, optr);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Restore optr if you trash it for 24XX target mode. MFC after: 3 days
2007-03-13 06:44:07 +00:00
optr = isp->isp_residx;
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
#endif
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 02:28:29 +00:00
/*
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
* Get the current Response Queue Out Pointer.
*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* If we're a 2300 or 2400, we can ask what hardware what it thinks.
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 02:28:29 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_23XX(isp) || IS_24XX(isp)) {
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
optr = ISP_READ(isp, isp->isp_respoutrp);
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
/*
* Debug: to be taken out eventually
*/
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
if (isp->isp_residx != optr) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGINFO, "isp_intr: hard optr=%x, soft optr %x", optr, isp->isp_residx);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp->isp_residx = optr;
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
}
} else {
optr = isp->isp_residx;
}
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
/*
* You *must* read the Response Queue In Pointer
* prior to clearing the RISC interrupt.
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
*
* Debounce the 2300 if revision less than 2.
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
*/
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
if (IS_2100(isp) || (IS_2300(isp) && isp->isp_revision < 2)) {
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
i = 0;
do {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
iptr = ISP_READ(isp, isp->isp_respinrp);
junk = ISP_READ(isp, isp->isp_respinrp);
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
} while (junk != iptr && ++i < 1000);
if (iptr != junk) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGWARN, "Response Queue Out Pointer Unstable (%x, %x)", iptr, junk);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
goto out;
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
}
} else {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
iptr = ISP_READ(isp, isp->isp_respinrp);
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
}
Limit fabric search to a default 256 entries. This will all go away soon because it's just getting harder and harder to find switches that correctly implement the GET ALL NEXT subcommands for the SNS protocol. Latch up result out pointer and set a busy flag when we're looking at the response queue. This allows for a cleaner way to make sure we don't get multiple CPUs trying to read the same response queue entries. Change how isp_handle_other_response returns values (clarity). Make PORT UNAVAILABLE the same as PORT LOGOUT (force a LIP). Do some formatting changes. MFC after: 0 days
2002-03-21 21:10:16 +00:00
isp->isp_resodx = iptr;
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
Once we have firmware running (if isp_reset) and this is the first time through, establish what our LUN width is. Unfortunately, we can't ask the f/w. If we loaded the f/w, we'll now assume we have expanded LUNs (SCCLUN for fibre channel, just plain 32 LUN for SCSI). If we didn't load firmware, assume 8 LUNs for SCSI and 1 LUN for Fibre Channel. We have to assume only one LUN for Fibre Channel because the LUN setting in Request Queue entries is in different places whether we have SCCLUN firmware or not, so the only LUN guaranteed to work for both is LUN 0. Clean up the rest of isp.c so that ISP2100_SCCLUN defines aren't used- instead use run time determinants based upon isp->isp_maxluns. After starting firmware, delay 500us to give it a chance to get rolling. Fix the interrupt service routine to check for both isr && sema being zero before thinking this was a spurious interrupt. Following the manuals, allow for both Mailbox as well as Queue Reponse type interrupts for regular SCSI.
2000-06-18 04:56:17 +00:00
if (optr == iptr && sema == 0) {
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
/*
* There are a lot of these- reasons unknown- mostly on
* faster Alpha machines.
*
* I tried delaying after writing HCCR_CMD_CLEAR_RISC_INT to
* make sure the old interrupt went away (to avoid 'ringing'
* effects), but that didn't stop this from occurring.
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
junk = 0;
} else if (IS_23XX(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(100);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
iptr = ISP_READ(isp, isp->isp_respinrp);
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
junk = ISP_READ(isp, BIU_R2HSTSLO);
} else {
junk = ISP_READ(isp, BIU_ISR);
}
if (optr == iptr) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_23XX(isp) || IS_24XX(isp)) {
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
;
} else {
sema = ISP_READ(isp, BIU_SEMA);
mbox = ISP_READ(isp, OUTMAILBOX0);
if ((sema & 0x3) && (mbox & 0x8000)) {
goto again;
}
}
isp->isp_intbogus++;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGDEBUG1, "bogus intr- isr %x (%x) iptr %x optr %x", isr, junk, iptr, optr);
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Limit fabric search to a default 256 entries. This will all go away soon because it's just getting harder and harder to find switches that correctly implement the GET ALL NEXT subcommands for the SNS protocol. Latch up result out pointer and set a busy flag when we're looking at the response queue. This allows for a cleaner way to make sure we don't get multiple CPUs trying to read the same response queue entries. Change how isp_handle_other_response returns values (clarity). Make PORT UNAVAILABLE the same as PORT LOGOUT (force a LIP). Do some formatting changes. MFC after: 0 days
2002-03-21 21:10:16 +00:00
isp->isp_resodx = iptr;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
while (optr != iptr) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
uint8_t qe[QENTRY_LEN];
ispstatusreq_t *sp = (ispstatusreq_t *) qe;
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
isphdr_t *hp;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
int buddaboom, etype, scsi_status, completion_status;
int req_status_flags, req_state_flags;
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
uint8_t *snsp, *resp;
uint32_t rlen, slen;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
long resid;
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
uint16_t oop;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
hp = (isphdr_t *) ISP_QUEUE_ENTRY(isp->isp_result, optr);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
oop = optr;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
optr = ISP_NXT_QENTRY(optr, RESULT_QUEUE_LEN(isp));
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 01:07:06 +00:00
nlooked++;
Don't set ZIO for 23XX for target mode (use fast posting instead). Use the correct number of handles for multihandle returns. Very, very, rarely on some SMP systems we've seen an 'unstable' type in the response queue. I dunno whether or not it's a bug in our handling, or whether there's a cache incoherency issue, but try to guard against it. MFC after: 2 weeks
2005-01-23 06:26:45 +00:00
read_again:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
buddaboom = req_status_flags = req_state_flags = 0;
resid = 0L;
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
/*
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
* Synchronize our view of this response queue entry.
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
*/
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
MEMORYBARRIER(isp, SYNC_RESULT, oop, QENTRY_LEN);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_hdr(isp, hp, &sp->req_header);
etype = sp->req_header.rqs_entry_type;
Recover from some major omissions/problems with the 24XX port. First, we were never correctly checking for a 24XX Status Type 0 response- that cased us to fall through to evaluate status for commands as if this were a 2100/2200/2300 Status Type 0 response. This is *close*, but not quite the same. This has been reported to be apparent with some wierd lun configuration problems with some arrays. It became glaringly apparent on sparc64 where none of the correct byte swap things were done. Fixing this omission then caused a whole universe shifting debug cycle of endian issues for the 2400. The manual for 24XX f/w turns out to be wrong about the endianness of a couple of entities. The lun and cdb fields for the type 7 request are *not* unconditionally big endian- they happen to be opposite of whatever the endian of the current machine type is. Same with the sense data for the 24XX type 0 response. While we're at it investigate and resolve some NVRAM endian issues. Approved by: re (ken) MFC after: 3 days
2007-07-02 20:08:20 +00:00
if (IS_24XX(isp) && etype == RQSTYPE_RESPONSE) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp24xx_statusreq_t *sp2 = (isp24xx_statusreq_t *)qe;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_get_24xx_response(isp, (isp24xx_statusreq_t *)hp, sp2);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (isp->isp_dblev & ISP_LOGDEBUG1) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_print_bytes(isp, "Response Queue Entry", QENTRY_LEN, sp2);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
scsi_status = sp2->req_scsi_status;
completion_status = sp2->req_completion_status;
req_state_flags = 0;
resid = sp2->req_resid;
} else if (etype == RQSTYPE_RESPONSE) {
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
isp_get_response(isp, (ispstatusreq_t *) hp, sp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (isp->isp_dblev & ISP_LOGDEBUG1) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_print_bytes(isp, "Response Queue Entry", QENTRY_LEN, sp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
scsi_status = sp->req_scsi_status;
completion_status = sp->req_completion_status;
req_status_flags = sp->req_status_flags;
req_state_flags = sp->req_state_flags;
resid = sp->req_resid;
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
} else if (etype == RQSTYPE_RIO1) {
isp_rio1_t *rio = (isp_rio1_t *) qe;
isp_get_rio1(isp, (isp_rio1_t *) hp, rio);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (isp->isp_dblev & ISP_LOGDEBUG1) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_print_bytes(isp, "Response Queue Entry", QENTRY_LEN, rio);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
for (i = 0; i < rio->req_header.rqs_seqno; i++) {
isp_fastpost_complete(isp, rio->req_handles[i]);
}
if (isp->isp_fpcchiwater < rio->req_header.rqs_seqno) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp->isp_fpcchiwater = rio->req_header.rqs_seqno;
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(hp, QENTRY_LEN); /* PERF */
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
continue;
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
} else if (etype == RQSTYPE_RIO2) {
isp_prt(isp, ISP_LOGERR, "dropping RIO2 response\n");
ISP_MEMZERO(hp, QENTRY_LEN); /* PERF */
continue;
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
} else {
Disable RIO (reduced interrupt operation) for 2200 boards- it seemed like it worked- but I ran into a case with a 2204 where commands were being lost right and left. Best be safe. For target mode, or things called if we call isp_handle_other response- note that we might have dropped locks by changing the output pointer so we bail from the loop. It's the responsibility of the entity dropping the lock to make sure that we let the f/w know we've read thus far into the response queue (else we begin processing the same entries again- blech!). MFC after: 1 day
2002-03-07 17:32:45 +00:00
/*
* Somebody reachable via isp_handle_other_response
* may have updated the response queue pointers for
Limit fabric search to a default 256 entries. This will all go away soon because it's just getting harder and harder to find switches that correctly implement the GET ALL NEXT subcommands for the SNS protocol. Latch up result out pointer and set a busy flag when we're looking at the response queue. This allows for a cleaner way to make sure we don't get multiple CPUs trying to read the same response queue entries. Change how isp_handle_other_response returns values (clarity). Make PORT UNAVAILABLE the same as PORT LOGOUT (force a LIP). Do some formatting changes. MFC after: 0 days
2002-03-21 21:10:16 +00:00
* us, so we reload our goal index.
Disable RIO (reduced interrupt operation) for 2200 boards- it seemed like it worked- but I ran into a case with a 2204 where commands were being lost right and left. Best be safe. For target mode, or things called if we call isp_handle_other response- note that we might have dropped locks by changing the output pointer so we bail from the loop. It's the responsibility of the entity dropping the lock to make sure that we let the f/w know we've read thus far into the response queue (else we begin processing the same entries again- blech!). MFC after: 1 day
2002-03-07 17:32:45 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
int r;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
uint32_t tsto = oop;
r = isp_handle_other_response(isp, etype, hp, &tsto);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (r < 0) {
Don't set ZIO for 23XX for target mode (use fast posting instead). Use the correct number of handles for multihandle returns. Very, very, rarely on some SMP systems we've seen an 'unstable' type in the response queue. I dunno whether or not it's a bug in our handling, or whether there's a cache incoherency issue, but try to guard against it. MFC after: 2 weeks
2005-01-23 06:26:45 +00:00
goto read_again;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* If somebody updated the output pointer, then reset
* optr to be one more than the updated amount.
*/
while (tsto != oop) {
optr = ISP_NXT_QENTRY(tsto,
RESULT_QUEUE_LEN(isp));
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (r > 0) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_WRITE(isp, isp->isp_respoutrp, optr);
ISP_MEMZERO(hp, QENTRY_LEN); /* PERF */
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
continue;
}
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
/*
* After this point, we'll just look at the header as
* we don't know how to deal with the rest of the
* response.
*/
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
/*
* It really has to be a bounced request just copied
temp fix for internal queue overflow problem
1999-04-14 17:37:36 +00:00
* from the request queue to the response queue. If
* not, something bad has happened.
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (etype != RQSTYPE_REQUEST) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGERR, notresp,
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
etype, oop, optr, nlooked);
isp_print_bytes(isp,
Fix compilation issues found in RELENG_4 port and merge the diffs back to -current to keep versions identical.
2007-03-12 04:54:30 +00:00
"Request Queue Entry", QENTRY_LEN, sp);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(hp, QENTRY_LEN); /* PERF */
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
continue;
}
buddaboom = 1;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
scsi_status = sp->req_scsi_status;
completion_status = sp->req_completion_status;
req_status_flags = sp->req_status_flags;
req_state_flags = sp->req_state_flags;
resid = sp->req_resid;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (sp->req_header.rqs_flags & RQSFLAG_MASK) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (sp->req_header.rqs_flags & RQSFLAG_CONTINUATION) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_print_bytes(isp, "unexpected continuation segment", QENTRY_LEN, sp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_WRITE(isp, isp->isp_respoutrp, optr);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
continue;
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
if (sp->req_header.rqs_flags & RQSFLAG_FULL) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGDEBUG0, "internal queues full");
temp fix for internal queue overflow problem
1999-04-14 17:37:36 +00:00
/*
* We'll synthesize a QUEUE FULL message below.
*/
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
if (sp->req_header.rqs_flags & RQSFLAG_BADHEADER) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_print_bytes(isp, "bad header flag", QENTRY_LEN, sp);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
buddaboom++;
}
if (sp->req_header.rqs_flags & RQSFLAG_BADPACKET) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_print_bytes(isp, "bad request packet", QENTRY_LEN, sp);
buddaboom++;
}
if (sp->req_header.rqs_flags & RQSFLAG_BADCOUNT) {
isp_print_bytes(isp, "invalid entry count", QENTRY_LEN, sp);
temp fix for internal queue overflow problem
1999-04-14 17:37:36 +00:00
buddaboom++;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (sp->req_header.rqs_flags & RQSFLAG_BADORDER) {
isp_print_bytes(isp, "invalid IOCB ordering", QENTRY_LEN, sp);
ISP_WRITE(isp, isp->isp_respoutrp, optr);
continue;
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
if (!ISP_VALID_HANDLE(isp, sp->req_handle)) {
isp_prt(isp, ISP_LOGERR, "bad request handle 0x%x (iocb type 0x%x)", sp->req_handle, etype);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(hp, QENTRY_LEN); /* PERF */
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_WRITE(isp, isp->isp_respoutrp, optr);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
continue;
}
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
xs = isp_find_xs(isp, sp->req_handle);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (xs == NULL) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
uint8_t ts = completion_status & 0xff;
Scale back # of luns supported for SCC to 16384- oops- top 3 bits are a lun address modifier of sorts. Only an HP XP-512 seems to have cared. Fix a few misplaced pointers for the new fabric goop, which has been demonstrated to work on newer Brocades and McData switches now. Put in commented out code which would run GFF_ID if the QLogic f/w allowed it. Don't whine about not being able to find a handle for a command if it was a command aborted (by us).
2002-04-16 19:55:35 +00:00
/*
* Only whine if this isn't the expected fallout of
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* aborting the command or resetting the target.
Scale back # of luns supported for SCC to 16384- oops- top 3 bits are a lun address modifier of sorts. Only an HP XP-512 seems to have cared. Fix a few misplaced pointers for the new fabric goop, which has been demonstrated to work on newer Brocades and McData switches now. Put in commented out code which would run GFF_ID if the QLogic f/w allowed it. Don't whine about not being able to find a handle for a command if it was a command aborted (by us).
2002-04-16 19:55:35 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (etype != RQSTYPE_RESPONSE) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGERR, "cannot find handle 0x%x (type 0x%x)", sp->req_handle, etype);
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
} else if (ts != RQCS_ABORTED && ts != RQCS_RESET_OCCURRED) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGERR, "cannot find handle 0x%x (status 0x%x)", sp->req_handle, ts);
Scale back # of luns supported for SCC to 16384- oops- top 3 bits are a lun address modifier of sorts. Only an HP XP-512 seems to have cared. Fix a few misplaced pointers for the new fabric goop, which has been demonstrated to work on newer Brocades and McData switches now. Put in commented out code which would run GFF_ID if the QLogic f/w allowed it. Don't whine about not being able to find a handle for a command if it was a command aborted (by us).
2002-04-16 19:55:35 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(hp, QENTRY_LEN); /* PERF */
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_WRITE(isp, isp->isp_respoutrp, optr);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
continue;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (req_status_flags & RQSTF_BUS_RESET) {
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
XS_SETERR(xs, HBA_BUSRESET);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SET_SENDMARKER(isp, XS_CHANNEL(xs), 1);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
if (buddaboom) {
XS_SETERR(xs, HBA_BOTCH);
}
various fixes
2000-08-27 23:38:44 +00:00
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
resp = NULL;
rlen = 0;
snsp = NULL;
slen = 0;
Recover from some major omissions/problems with the 24XX port. First, we were never correctly checking for a 24XX Status Type 0 response- that cased us to fall through to evaluate status for commands as if this were a 2100/2200/2300 Status Type 0 response. This is *close*, but not quite the same. This has been reported to be apparent with some wierd lun configuration problems with some arrays. It became glaringly apparent on sparc64 where none of the correct byte swap things were done. Fixing this omission then caused a whole universe shifting debug cycle of endian issues for the 2400. The manual for 24XX f/w turns out to be wrong about the endianness of a couple of entities. The lun and cdb fields for the type 7 request are *not* unconditionally big endian- they happen to be opposite of whatever the endian of the current machine type is. Same with the sense data for the 24XX type 0 response. While we're at it investigate and resolve some NVRAM endian issues. Approved by: re (ken) MFC after: 3 days
2007-07-02 20:08:20 +00:00
if (IS_24XX(isp) && (scsi_status & (RQCS_RV|RQCS_SV)) != 0) {
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
resp = ((isp24xx_statusreq_t *)sp)->req_rsp_sense;
rlen = ((isp24xx_statusreq_t *)sp)->req_response_len;
} else if (IS_FC(isp) && (scsi_status & RQCS_RV) != 0) {
resp = sp->req_response;
rlen = sp->req_response_len;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_FC(isp) && (scsi_status & RQCS_SV) != 0) {
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
/*
various fixes
2000-08-27 23:38:44 +00:00
* Fibre Channel F/W doesn't say we got status
* if there's Sense Data instead. I guess they
* think it goes w/o saying.
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
*/
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
req_state_flags |= RQSF_GOT_STATUS|RQSF_GOT_SENSE;
if (IS_24XX(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
snsp = ((isp24xx_statusreq_t *)sp)->req_rsp_sense;
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
snsp += rlen;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
slen = ((isp24xx_statusreq_t *)sp)->req_sense_len;
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
} else {
snsp = sp->req_sense_data;
slen = sp->req_sense_len;
}
} else if (IS_SCSI(isp) && (req_state_flags & RQSF_GOT_SENSE)) {
snsp = sp->req_sense_data;
slen = sp->req_sense_len;
various fixes
2000-08-27 23:38:44 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (req_state_flags & RQSF_GOT_STATUS) {
*XS_STSP(xs) = scsi_status & 0xff;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
switch (etype) {
various fixes
2000-08-27 23:38:44 +00:00
case RQSTYPE_RESPONSE:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (resp && rlen >= 4 && resp[FCP_RSPNS_CODE_OFFSET] != 0) {
const char *ptr;
char lb[64];
const char *rnames[6] = {
"Task Management Function Done",
"Data Length Differs From Burst Length",
"Invalid FCP Cmnd",
"FCP DATA RO mismatch with FCP DATA_XFR_RDY RO",
"Task Management Function Rejected",
"Task Management Function Failed",
};
if (resp[FCP_RSPNS_CODE_OFFSET] > 5) {
ISP_SNPRINTF(lb, sizeof lb, "Unknown FCP Response Code 0x%x", resp[FCP_RSPNS_CODE_OFFSET]);
ptr = lb;
} else {
ptr = rnames[resp[FCP_RSPNS_CODE_OFFSET]];
}
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGWARN, "FCP RESPONSE, LENGTH %u: %s CDB0=0x%02x", rlen, ptr, XS_CDBP(xs)[0] & 0xff);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (resp[FCP_RSPNS_CODE_OFFSET] != 0) {
XS_SETERR(xs, HBA_BOTCH);
}
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_parse_status_24xx(isp, (isp24xx_statusreq_t *)sp, xs, &resid);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else {
isp_parse_status(isp, (void *)sp, xs, &resid);
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if ((XS_NOERR(xs) || XS_ERR(xs) == HBA_NOERROR) && (*XS_STSP(xs) == SCSI_BUSY)) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
XS_SETERR(xs, HBA_TGTBSY);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
various fixes
2000-08-27 23:38:44 +00:00
if (IS_SCSI(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
XS_SET_RESID(xs, resid);
various fixes
2000-08-27 23:38:44 +00:00
/*
* A new synchronous rate was negotiated for
* this target. Mark state such that we'll go
* look up that which has changed later.
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (req_status_flags & RQSTF_NEGOTIATION) {
various fixes
2000-08-27 23:38:44 +00:00
int t = XS_TGT(xs);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdparam *sdp = SDPARAM(isp, XS_CHANNEL(xs));
various fixes
2000-08-27 23:38:44 +00:00
sdp->isp_devparam[t].dev_refresh = 1;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdp->update = 1;
various fixes
2000-08-27 23:38:44 +00:00
}
} else {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (req_status_flags & RQSF_XFER_COMPLETE) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
XS_SET_RESID(xs, 0);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else if (scsi_status & RQCS_RESID) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
XS_SET_RESID(xs, resid);
various fixes
2000-08-27 23:38:44 +00:00
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
XS_SET_RESID(xs, 0);
various fixes
2000-08-27 23:38:44 +00:00
}
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
}
if (snsp && slen) {
XS_SAVE_SENSE(xs, snsp, slen);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
} else if ((req_status_flags & RQSF_GOT_STATUS) && (scsi_status & 0xff) == SCSI_CHECK && IS_FC(isp)) {
isp_prt(isp, ISP_LOGWARN, "CHECK CONDITION w/o sense data for CDB=0x%x", XS_CDBP(xs)[0] & 0xff);
isp_print_bytes(isp, "CC with no Sense", QENTRY_LEN, qe);
various fixes
2000-08-27 23:38:44 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGDEBUG2, "asked for %ld got raw resid %ld settled for %ld", (long) XS_XFRLEN(xs), resid, (long) XS_GET_RESID(xs));
various fixes
2000-08-27 23:38:44 +00:00
break;
case RQSTYPE_REQUEST:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
case RQSTYPE_A64:
case RQSTYPE_T2RQS:
case RQSTYPE_T3RQS:
case RQSTYPE_T7RQS:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (!IS_24XX(isp) && (sp->req_header.rqs_flags & RQSFLAG_FULL)) {
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
/*
* Force Queue Full status.
*/
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
*XS_STSP(xs) = SCSI_QFULL;
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
XS_SETERR(xs, HBA_NOERROR);
} else if (XS_NOERR(xs)) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
XS_SETERR(xs, HBA_BOTCH);
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
XS_SET_RESID(xs, XS_XFRLEN(xs));
various fixes
2000-08-27 23:38:44 +00:00
break;
default:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_print_bytes(isp, "Unhandled Response Type", QENTRY_LEN, qe);
Add some prototype deadchip detection. Set FIFO bursting (1XX0 only- it's already on for the 2XX0) and detect the broken 1040A FIFO. Change bzero to MEMZERO (portability with **nux). Use memcpy for same reason. Finally detect QUEUE FULL conditions and return this as an error that will get cam_periph_error to do it's 'tagged openings now XXX' dance.
1999-01-10 02:55:10 +00:00
if (XS_NOERR(xs)) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
XS_SETERR(xs, HBA_BOTCH);
Add some prototype deadchip detection. Set FIFO bursting (1XX0 only- it's already on for the 2XX0) and detect the broken 1040A FIFO. Change bzero to MEMZERO (portability with **nux). Use memcpy for same reason. Finally detect QUEUE FULL conditions and return this as an error that will get cam_periph_error to do it's 'tagged openings now XXX' dance.
1999-01-10 02:55:10 +00:00
}
various fixes
2000-08-27 23:38:44 +00:00
break;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
various fixes
2000-08-27 23:38:44 +00:00
/*
On reset, make sure that we have some parameters set correctly. This fixes a longstanding issue WRT resetting the chip after startup- it would fail if we were connected as an F-port to a switch. If we were connected as an F-port, we got assigned a hard loop ID of 255, which is really a bogus loop id. Then when we turned around to reset ourselves, the firmware would reject the ICB_INIT request because the loop id was bogus. *sputter* Minor fixlet from somebody in NetBSD with too much time on their hands (dma -> DMA).
2003-09-13 01:55:44 +00:00
* Free any DMA resources. As a side effect, this may
Recover from some major omissions/problems with the 24XX port. First, we were never correctly checking for a 24XX Status Type 0 response- that cased us to fall through to evaluate status for commands as if this were a 2100/2200/2300 Status Type 0 response. This is *close*, but not quite the same. This has been reported to be apparent with some wierd lun configuration problems with some arrays. It became glaringly apparent on sparc64 where none of the correct byte swap things were done. Fixing this omission then caused a whole universe shifting debug cycle of endian issues for the 2400. The manual for 24XX f/w turns out to be wrong about the endianness of a couple of entities. The lun and cdb fields for the type 7 request are *not* unconditionally big endian- they happen to be opposite of whatever the endian of the current machine type is. Same with the sense data for the 24XX type 0 response. While we're at it investigate and resolve some NVRAM endian issues. Approved by: re (ken) MFC after: 3 days
2007-07-02 20:08:20 +00:00
* also do any cache flushing necessary for data coherence.
*/
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (XS_XFRLEN(xs)) {
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
ISP_DMAFREE(isp, xs, sp->req_handle);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Amazingly we've been freeing a handle and using that which it refers to for years. Bad! MFC after: 1 week
2010-01-15 20:08:08 +00:00
isp_destroy_handle(isp, sp->req_handle);
various fixes
2000-08-27 23:38:44 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (((isp->isp_dblev & (ISP_LOGDEBUG1|ISP_LOGDEBUG2|ISP_LOGDEBUG3))) ||
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
((isp->isp_dblev & (ISP_LOGDEBUG0|ISP_LOG_CWARN) && ((!XS_NOERR(xs)) || (*XS_STSP(xs) != SCSI_GOOD))))) {
isp_prt_endcmd(isp, xs);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (isp->isp_nactive > 0) {
Remove the Target mode functions until they're in better shape. Implement some suggested compilation cleanups from Eklund. Wire down a hard loop id if we are not on a platform that has the ability to get to a PCI BIOS (it still will float to the ID it gets after a LIP but at least we can try). Clarify that the expanded lun is based upon SCCLUN defines (in f/w).
1998-12-05 01:33:57 +00:00
isp->isp_nactive--;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
complist[ndone++] = xs; /* defer completion call until later */
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(hp, QENTRY_LEN); /* PERF */
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (ndone == MAX_REQUESTQ_COMPLETIONS) {
break;
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 01:07:06 +00:00
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
/*
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 01:07:06 +00:00
* If we looked at any commands, then it's valid to find out
* what the outpointer is. It also is a trigger to update the
* ISP's notion of what we've seen so far.
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
*/
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
if (nlooked) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_WRITE(isp, isp->isp_respoutrp, optr);
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
/*
Disable RIO (reduced interrupt operation) for 2200 boards- it seemed like it worked- but I ran into a case with a 2204 where commands were being lost right and left. Best be safe. For target mode, or things called if we call isp_handle_other response- note that we might have dropped locks by changing the output pointer so we bail from the loop. It's the responsibility of the entity dropping the lock to make sure that we let the f/w know we've read thus far into the response queue (else we begin processing the same entries again- blech!). MFC after: 1 day
2002-03-07 17:32:45 +00:00
* While we're at it, read the requst queue out pointer.
Add 2 Gigabit Fibre Channel support (2300 && 2312 cards). This required some reworking (and consequent cleanup) of the interrupt service code. Also begin to start a cleanup of target mode support that will (eventually) not require more inforamtion routed with the ATIO to come back with the CTIO other than tag. MFC after: 4 weeks
2001-08-31 21:39:04 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp->isp_reqodx = ISP_READ(isp, isp->isp_rqstoutrp);
if (isp->isp_rscchiwater < ndone) {
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
isp->isp_rscchiwater = ndone;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
}
out:
if (IS_24XX(isp)) {
ISP_WRITE(isp, BIU2400_HCCR, HCCR_2400_CMD_CLEAR_RISC_INT);
} else {
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
ISP_WRITE(isp, BIU_SEMA, 0);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp->isp_residx = optr;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
for (i = 0; i < ndone; i++) {
xs = complist[i];
if (xs) {
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
isp->isp_rsltccmplt++;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_done(xs);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
/*
* Support routines.
*/
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
static void
isp_prt_endcmd(ispsoftc_t *isp, XS_T *xs)
{
char cdbstr[16 * 5 + 1];
int i, lim;
lim = XS_CDBLEN(xs) > 16? 16 : XS_CDBLEN(xs);
ISP_SNPRINTF(cdbstr, sizeof (cdbstr), "0x%02x ", XS_CDBP(xs)[0]);
for (i = 1; i < lim; i++) {
ISP_SNPRINTF(cdbstr, sizeof (cdbstr), "%s0x%02x ", cdbstr, XS_CDBP(xs)[i]);
}
if (XS_SENSE_VALID(xs)) {
isp_xs_prt(isp, xs, ISP_LOGALL, "FIN dl%d resid %ld CDB=%s KEY/ASC/ASCQ=0x%02x/0x%02x/0x%02x",
XS_XFRLEN(xs), (long) XS_GET_RESID(xs), cdbstr, XS_SNSKEY(xs), XS_SNSASC(xs), XS_SNSASCQ(xs));
} else {
isp_xs_prt(isp, xs, ISP_LOGALL, "FIN dl%d resid %ld CDB=%s STS 0x%x XS_ERR=0x%x", XS_XFRLEN(xs), (long) XS_GET_RESID(xs), cdbstr, *XS_STSP(xs), XS_ERR(xs));
}
}
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
/*
* Parse an ASYNC mailbox complete
*
* Return non-zero if the event has been acknowledged.
*/
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
static int
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
isp_parse_async(ispsoftc_t *isp, uint16_t mbox)
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
{
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
int acked = 0;
uint32_t h1 = 0, h2 = 0;
uint16_t chan = 0;
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
/*
* Pick up the channel, but not if this is a ASYNC_RIO32_2,
* where Mailboxes 6/7 have the second handle.
*/
if (mbox != ASYNC_RIO32_2) {
if (IS_DUALBUS(isp)) {
chan = ISP_READ(isp, OUTMAILBOX6);
}
Clean up stupidity in the isp_handle_other_response function- indexes of queue entries have to be at least 16 bits now! If we're running a 2100 less than rev 5, turn off loop fairness (per Qlogic errata). Fix typo in checking against 2200 F/W revision. Slightly fix/reorder fabric login stuff. Change to usage of isp_getrqentry for code clarity. Add some defensive dual bus assumptions. Various cleanups, etc...
1999-11-21 03:18:22 +00:00
}
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
isp_prt(isp, ISP_LOGDEBUG2, "Async Mbox 0x%x", mbox);
Clean up stupidity in the isp_handle_other_response function- indexes of queue entries have to be at least 16 bits now! If we're running a 2100 less than rev 5, turn off loop fairness (per Qlogic errata). Fix typo in checking against 2200 F/W revision. Slightly fix/reorder fabric login stuff. Change to usage of isp_getrqentry for code clarity. Add some defensive dual bus assumptions. Various cleanups, etc...
1999-11-21 03:18:22 +00:00
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
switch (mbox) {
case ASYNC_BUS_RESET:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SET_SENDMARKER(isp, chan, 1);
Remove the Target mode functions until they're in better shape. Implement some suggested compilation cleanups from Eklund. Wire down a hard loop id if we are not on a platform that has the ability to get to a PCI BIOS (it still will float to the ID it gets after a LIP but at least we can try). Clarify that the expanded lun is based upon SCCLUN defines (in f/w).
1998-12-05 01:33:57 +00:00
#ifdef ISP_TARGET_MODE
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (isp_target_async(isp, chan, mbox)) {
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
acked = 1;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Remove the Target mode functions until they're in better shape. Implement some suggested compilation cleanups from Eklund. Wire down a hard loop id if we are not on a platform that has the ability to get to a PCI BIOS (it still will float to the ID it gets after a LIP but at least we can try). Clarify that the expanded lun is based upon SCCLUN defines (in f/w).
1998-12-05 01:33:57 +00:00
#endif
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_async(isp, ISPASYNC_BUS_RESET, chan);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case ASYNC_SYSTEM_ERROR:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp->isp_dead = 1;
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
isp->isp_state = ISP_CRASHED;
/*
* Were we waiting for a mailbox command to complete?
* If so, it's dead, so wake up the waiter.
*/
if (isp->isp_mboxbsy) {
isp->isp_obits = 1;
isp->isp_mboxtmp[0] = MBOX_HOST_INTERFACE_ERROR;
MBOX_NOTIFY_COMPLETE(isp);
}
If we're using ancient (pre 1.17.0) 2100 f/w (for the cards that cannot load f/w images > 0x7fff words), set ISP_FW_ATTR_SCCLUN. We explicitly don't believe we can find attributes if f/w is < 1.17.0, so we have to set SCCLUN for the 1.15.37 f/w we're using manually- otherwise every target will replicate itself across all 16 supported luns for non-SCCLUN f/w. Correctly set things up for 23XX and either fast posting or ZIO. The 23XX, it turns out, does not support RIO. If you put a non-zero value in xfwoptions, this will disable fast posting. If you put ICBXOPT_ZIO in xfwoptions, then the 23XX will do interrupt delays but post to the response queue- apparently QLogic *now* believes that reading multiple handles from registers is less of a win than writing (and delaying) multiple 64 byte responses to the response queue. At the end of taking a a good f/w crash dump, send the ISPASYNC_FW_DUMPED event to the outer layers (who can then do things like wake a user daemon to *fetch* the crash image, etc.).
2002-08-17 17:29:15 +00:00
/*
Make this an MP safe driver but also still be multi-release. Seems to work on RELENG_4 through -current and also on sparc64 now. There may still be some issues with the auto attach/detach code to sort out. MFC after: 3 days
2007-05-05 20:17:23 +00:00
* It's up to the handler for isp_async to reinit stuff and
* restart the firmware
If we're using ancient (pre 1.17.0) 2100 f/w (for the cards that cannot load f/w images > 0x7fff words), set ISP_FW_ATTR_SCCLUN. We explicitly don't believe we can find attributes if f/w is < 1.17.0, so we have to set SCCLUN for the 1.15.37 f/w we're using manually- otherwise every target will replicate itself across all 16 supported luns for non-SCCLUN f/w. Correctly set things up for 23XX and either fast posting or ZIO. The 23XX, it turns out, does not support RIO. If you put a non-zero value in xfwoptions, this will disable fast posting. If you put ICBXOPT_ZIO in xfwoptions, then the 23XX will do interrupt delays but post to the response queue- apparently QLogic *now* believes that reading multiple handles from registers is less of a win than writing (and delaying) multiple 64 byte responses to the response queue. At the end of taking a a good f/w crash dump, send the ISPASYNC_FW_DUMPED event to the outer layers (who can then do things like wake a user daemon to *fetch* the crash image, etc.).
2002-08-17 17:29:15 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_async(isp, ISPASYNC_FW_CRASH);
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
acked = 1;
If the HBA is already 'touched', still set maxluns. Othewise for CAM_QUIRK_HILUN devices we loop thru 32bits of lun. Oops. Switch to using USEC_DELAY rather than USEC_SLEEP at isp_reset time. Try to paper around a defect in clients that don't correctly registers themeselves with the fabric nameserver. Minor updates for Mirapoint support- they still use code that is not HANDLE_LOOPSTATE_IN_OUTER_LAYERS, and, surprise surprise, this old stuff had some bugs in it. Clean up some target mode stuff. MFC after: 1 week
2002-06-16 05:18:22 +00:00
break;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
case ASYNC_RQS_XFER_ERR:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGERR, "Request Queue Transfer Error");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case ASYNC_RSP_XFER_ERR:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGERR, "Response Queue Transfer Error");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case ASYNC_QWAKEUP:
Clean up stupidity in the isp_handle_other_response function- indexes of queue entries have to be at least 16 bits now! If we're running a 2100 less than rev 5, turn off loop fairness (per Qlogic errata). Fix typo in checking against 2200 F/W revision. Slightly fix/reorder fabric login stuff. Change to usage of isp_getrqentry for code clarity. Add some defensive dual bus assumptions. Various cleanups, etc...
1999-11-21 03:18:22 +00:00
/*
* We've just been notified that the Queue has woken up.
* We don't need to be chatty about this- just unlatch things
* and move on.
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbox = ISP_READ(isp, isp->isp_rqstoutrp);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case ASYNC_TIMEOUT_RESET:
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
isp_prt(isp, ISP_LOGWARN, "timeout initiated SCSI bus reset of chan %d", chan);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SET_SENDMARKER(isp, chan, 1);
Remove the Target mode functions until they're in better shape. Implement some suggested compilation cleanups from Eklund. Wire down a hard loop id if we are not on a platform that has the ability to get to a PCI BIOS (it still will float to the ID it gets after a LIP but at least we can try). Clarify that the expanded lun is based upon SCCLUN defines (in f/w).
1998-12-05 01:33:57 +00:00
#ifdef ISP_TARGET_MODE
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (isp_target_async(isp, chan, mbox)) {
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
acked = 1;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Remove the Target mode functions until they're in better shape. Implement some suggested compilation cleanups from Eklund. Wire down a hard loop id if we are not on a platform that has the ability to get to a PCI BIOS (it still will float to the ID it gets after a LIP but at least we can try). Clarify that the expanded lun is based upon SCCLUN defines (in f/w).
1998-12-05 01:33:57 +00:00
#endif
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
Remove the Target mode functions until they're in better shape. Implement some suggested compilation cleanups from Eklund. Wire down a hard loop id if we are not on a platform that has the ability to get to a PCI BIOS (it still will float to the ID it gets after a LIP but at least we can try). Clarify that the expanded lun is based upon SCCLUN defines (in f/w).
1998-12-05 01:33:57 +00:00
case ASYNC_DEVICE_RESET:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGINFO, "device reset on chan %d", chan);
ISP_SET_SENDMARKER(isp, chan, 1);
Remove the Target mode functions until they're in better shape. Implement some suggested compilation cleanups from Eklund. Wire down a hard loop id if we are not on a platform that has the ability to get to a PCI BIOS (it still will float to the ID it gets after a LIP but at least we can try). Clarify that the expanded lun is based upon SCCLUN defines (in f/w).
1998-12-05 01:33:57 +00:00
#ifdef ISP_TARGET_MODE
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (isp_target_async(isp, chan, mbox)) {
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
acked = 1;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Remove the Target mode functions until they're in better shape. Implement some suggested compilation cleanups from Eklund. Wire down a hard loop id if we are not on a platform that has the ability to get to a PCI BIOS (it still will float to the ID it gets after a LIP but at least we can try). Clarify that the expanded lun is based upon SCCLUN defines (in f/w).
1998-12-05 01:33:57 +00:00
#endif
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case ASYNC_EXTMSG_UNDERRUN:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGWARN, "extended message underrun");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case ASYNC_SCAM_INT:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGINFO, "SCAM interrupt");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case ASYNC_HUNG_SCSI:
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
isp_prt(isp, ISP_LOGERR, "stalled SCSI Bus after DATA Overrun");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
/* XXX: Need to issue SCSI reset at this point */
break;
case ASYNC_KILLED_BUS:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGERR, "SCSI Bus reset after DATA Overrun");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case ASYNC_BUS_TRANSIT:
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
mbox = ISP_READ(isp, OUTMAILBOX2);
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
switch (mbox & SXP_PINS_MODE_MASK) {
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
case SXP_PINS_LVD_MODE:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGINFO, "Transition to LVD mode");
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
SDPARAM(isp, chan)->isp_diffmode = 0;
SDPARAM(isp, chan)->isp_ultramode = 0;
SDPARAM(isp, chan)->isp_lvdmode = 1;
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
break;
case SXP_PINS_HVD_MODE:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGINFO,
"Transition to Differential mode");
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
SDPARAM(isp, chan)->isp_diffmode = 1;
SDPARAM(isp, chan)->isp_ultramode = 0;
SDPARAM(isp, chan)->isp_lvdmode = 0;
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
break;
case SXP_PINS_SE_MODE:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGINFO,
"Transition to Single Ended mode");
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
SDPARAM(isp, chan)->isp_diffmode = 0;
SDPARAM(isp, chan)->isp_ultramode = 1;
SDPARAM(isp, chan)->isp_lvdmode = 0;
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
break;
default:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGWARN,
"Transition to Unknown Mode 0x%x", mbox);
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
break;
}
/*
* XXX: Set up to renegotiate again!
*/
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
/* Can only be for a 1080... */
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SET_SENDMARKER(isp, chan, 1);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
case ASYNC_CMD_CMPLT:
case ASYNC_RIO32_1:
if (!IS_ULTRA3(isp)) {
isp_prt(isp, ISP_LOGERR, "unexpected fast posting completion");
break;
}
/* FALLTHROUGH */
h1 = (ISP_READ(isp, OUTMAILBOX2) << 16) | ISP_READ(isp, OUTMAILBOX1);
break;
case ASYNC_RIO32_2:
h1 = (ISP_READ(isp, OUTMAILBOX2) << 16) | ISP_READ(isp, OUTMAILBOX1);
h2 = (ISP_READ(isp, OUTMAILBOX7) << 16) | ISP_READ(isp, OUTMAILBOX6);
break;
case ASYNC_RIO16_5:
case ASYNC_RIO16_4:
case ASYNC_RIO16_3:
case ASYNC_RIO16_2:
case ASYNC_RIO16_1:
isp_prt(isp, ISP_LOGERR, "unexpected 16 bit RIO handle");
break;
default:
isp_prt(isp, ISP_LOGWARN, "%s: unhandled async code 0x%x", __func__, mbox);
break;
}
if (h1 || h2) {
isp_prt(isp, ISP_LOGDEBUG3, "fast post/rio completion of 0x%08x", h1);
isp_fastpost_complete(isp, h1);
if (h2) {
isp_prt(isp, ISP_LOGDEBUG3, "fast post/rio completion of 0x%08x", h2);
isp_fastpost_complete(isp, h2);
if (isp->isp_fpcchiwater < 2) {
isp->isp_fpcchiwater = 2;
}
} else {
if (isp->isp_fpcchiwater < 1) {
isp->isp_fpcchiwater = 1;
}
}
} else {
isp->isp_intoasync++;
}
return (acked);
}
#define GET_24XX_BUS(isp, chan, msg) \
if (IS_24XX(isp)) { \
chan = ISP_READ(isp, OUTMAILBOX3) & 0xff; \
if (chan >= isp->isp_nchan) { \
isp_prt(isp, ISP_LOGERR, "bogus channel %u for %s at line %d", chan, msg, __LINE__); \
break; \
} \
}
static int
isp_parse_async_fc(ispsoftc_t *isp, uint16_t mbox)
{
int acked = 0;
uint16_t chan;
if (IS_DUALBUS(isp)) {
chan = ISP_READ(isp, OUTMAILBOX6);
} else {
chan = 0;
}
isp_prt(isp, ISP_LOGDEBUG2, "Async Mbox 0x%x", mbox);
switch (mbox) {
case ASYNC_SYSTEM_ERROR:
isp->isp_dead = 1;
isp->isp_state = ISP_CRASHED;
FCPARAM(isp, chan)->isp_loopstate = LOOP_NIL;
FCPARAM(isp, chan)->isp_fwstate = FW_CONFIG_WAIT;
/*
* Were we waiting for a mailbox command to complete?
* If so, it's dead, so wake up the waiter.
*/
if (isp->isp_mboxbsy) {
isp->isp_obits = 1;
isp->isp_mboxtmp[0] = MBOX_HOST_INTERFACE_ERROR;
MBOX_NOTIFY_COMPLETE(isp);
}
/*
* It's up to the handler for isp_async to reinit stuff and
* restart the firmware
*/
isp_async(isp, ISPASYNC_FW_CRASH);
acked = 1;
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
break;
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
case ASYNC_RQS_XFER_ERR:
isp_prt(isp, ISP_LOGERR, "Request Queue Transfer Error");
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
break;
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
case ASYNC_RSP_XFER_ERR:
isp_prt(isp, ISP_LOGERR, "Response Queue Transfer Error");
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
break;
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
case ASYNC_QWAKEUP:
#ifdef ISP_TARGET_MODE
if (IS_24XX(isp)) {
isp_prt(isp, ISP_LOGERR, "ATIO Queue Transfer Error");
break;
}
#endif
isp_prt(isp, ISP_LOGERR, "%s: unexpected ASYNC_QWAKEUP code", __func__);
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
break;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
case ASYNC_CMD_CMPLT:
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
isp_fastpost_complete(isp, (ISP_READ(isp, OUTMAILBOX2) << 16) | ISP_READ(isp, OUTMAILBOX1));
if (isp->isp_fpcchiwater < 1) {
isp->isp_fpcchiwater = 1;
}
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
break;
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
case ASYNC_RIOZIO_STALL:
break;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
case ASYNC_CTIO_DONE:
various fixes
2000-08-27 23:38:44 +00:00
#ifdef ISP_TARGET_MODE
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
if (isp_target_async(isp, (ISP_READ(isp, OUTMAILBOX2) << 16) | ISP_READ(isp, OUTMAILBOX1), mbox)) {
acked = 1;
Don't set ZIO for 23XX for target mode (use fast posting instead). Use the correct number of handles for multihandle returns. Very, very, rarely on some SMP systems we've seen an 'unstable' type in the response queue. I dunno whether or not it's a bug in our handling, or whether there's a cache incoherency issue, but try to guard against it. MFC after: 2 weeks
2005-01-23 06:26:45 +00:00
} else {
isp->isp_fphccmplt++;
}
various fixes
2000-08-27 23:38:44 +00:00
#else
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
isp_prt(isp, ISP_LOGWARN, "unexpected ASYNC CTIO done");
Don't set ZIO for 23XX for target mode (use fast posting instead). Use the correct number of handles for multihandle returns. Very, very, rarely on some SMP systems we've seen an 'unstable' type in the response queue. I dunno whether or not it's a bug in our handling, or whether there's a cache incoherency issue, but try to guard against it. MFC after: 2 weeks
2005-01-23 06:26:45 +00:00
#endif
If the HBA is already 'touched', still set maxluns. Othewise for CAM_QUIRK_HILUN devices we loop thru 32bits of lun. Oops. Switch to using USEC_DELAY rather than USEC_SLEEP at isp_reset time. Try to paper around a defect in clients that don't correctly registers themeselves with the fabric nameserver. Minor updates for Mirapoint support- they still use code that is not HANDLE_LOOPSTATE_IN_OUTER_LAYERS, and, surprise surprise, this old stuff had some bugs in it. Clean up some target mode stuff. MFC after: 1 week
2002-06-16 05:18:22 +00:00
break;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
case ASYNC_LIP_ERROR:
Enable LIP F8, LIP Reset async events. Be more chatty about SNS failures. Fix typo for skipped phase mesage. Correct MBOX_GET_PORT_QUEUE_PARAMS options in table. MFC after: 2 weeks
2001-08-16 17:25:08 +00:00
case ASYNC_LIP_F8:
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
case ASYNC_LIP_OCCURRED:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
case ASYNC_PTPMODE:
We've had problems with data corruption occuring on commands that complete (with no apparent error) after we receive a LIP. This has been observed mostly on Local Loop topologies. To be safe, let's just mark all active commands as dead if we get a LIP and we're on a private or public loop. MFC after: 4 weeks
2001-06-14 17:13:24 +00:00
/*
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* These are broadcast events that have to be sent across
* all active channels.
We've had problems with data corruption occuring on commands that complete (with no apparent error) after we receive a LIP. This has been observed mostly on Local Loop topologies. To be safe, let's just mark all active commands as dead if we get a LIP and we're on a private or public loop. MFC after: 4 weeks
2001-06-14 17:13:24 +00:00
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
for (chan = 0; chan < isp->isp_nchan; chan++) {
fcparam *fcp = FCPARAM(isp, chan);
int topo = fcp->isp_topo;
if (fcp->role == ISP_ROLE_NONE) {
continue;
}
fcp->isp_fwstate = FW_CONFIG_WAIT;
fcp->isp_loopstate = LOOP_LIP_RCVD;
ISP_SET_SENDMARKER(isp, chan, 1);
ISP_MARK_PORTDB(isp, chan, 1);
isp_async(isp, ISPASYNC_LIP, chan);
#ifdef ISP_TARGET_MODE
if (isp_target_async(isp, chan, mbox)) {
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
acked = 1;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
#endif
/*
* We've had problems with data corruption occuring on
* commands that complete (with no apparent error) after
* we receive a LIP. This has been observed mostly on
* Local Loop topologies. To be safe, let's just mark
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
* all active initiator commands as dead.
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
*/
if (topo == TOPO_NL_PORT || topo == TOPO_FL_PORT) {
int i, j;
for (i = j = 0; i < isp->isp_maxcmds; i++) {
XS_T *xs;
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
isp_hdl_t *hdp;
hdp = &isp->isp_xflist[i];
if (ISP_H2HT(hdp->handle) != ISP_HANDLE_INITIATOR) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
continue;
}
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
xs = hdp->cmd;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (XS_CHANNEL(xs) != chan) {
continue;
}
We've had problems with data corruption occuring on commands that complete (with no apparent error) after we receive a LIP. This has been observed mostly on Local Loop topologies. To be safe, let's just mark all active commands as dead if we get a LIP and we're on a private or public loop. MFC after: 4 weeks
2001-06-14 17:13:24 +00:00
j++;
XS_SETERR(xs, HBA_BUSRESET);
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (j) {
isp_prt(isp, ISP_LOGERR, lipd, chan, j);
}
More 2300 support prep- the Request/Response in/out pointers are part of the PCI block for the 2300- not software convention usage of the mailbox registers- so we macrosize in/out pointer usage. Only report that a LIP destroyed commands if it actually destroyed commands. Get the chan/tgt/lun order correct. Fix a longstanding stupid bug that caused us to try and issue a command with a tag on Channel B because we were checking the tagged capability for the target against Channel A. A firmware crash is now vectored out to platform specific code as an async event. Some minor formatting tweaks.
2001-07-04 18:42:41 +00:00
}
We've had problems with data corruption occuring on commands that complete (with no apparent error) after we receive a LIP. This has been observed mostly on Local Loop topologies. To be safe, let's just mark all active commands as dead if we get a LIP and we're on a private or public loop. MFC after: 4 weeks
2001-06-14 17:13:24 +00:00
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case ASYNC_LOOP_UP:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* This is a broadcast event that has to be sent across
* all active channels.
*/
for (chan = 0; chan < isp->isp_nchan; chan++) {
fcparam *fcp = FCPARAM(isp, chan);
if (fcp->role == ISP_ROLE_NONE) {
continue;
}
ISP_SET_SENDMARKER(isp, chan, 1);
fcp->isp_fwstate = FW_CONFIG_WAIT;
fcp->isp_loopstate = LOOP_LIP_RCVD;
ISP_MARK_PORTDB(isp, chan, 1);
isp_async(isp, ISPASYNC_LOOP_UP, chan);
#ifdef ISP_TARGET_MODE
if (isp_target_async(isp, chan, mbox)) {
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
acked = 1;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Raise default FCP logintime to 60 seconds. Move the position of where we could have seen the loop up at least once so it makes sense. Change some stuff in ispscsicmd so we don't get stuck there if the loop has never come up yet. Add in some target mode support code.
2000-01-03 23:52:41 +00:00
#endif
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case ASYNC_LOOP_DOWN:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* This is a broadcast event that has to be sent across
* all active channels.
*/
for (chan = 0; chan < isp->isp_nchan; chan++) {
fcparam *fcp = FCPARAM(isp, chan);
if (fcp->role == ISP_ROLE_NONE) {
continue;
}
ISP_SET_SENDMARKER(isp, chan, 1);
fcp->isp_fwstate = FW_CONFIG_WAIT;
fcp->isp_loopstate = LOOP_NIL;
ISP_MARK_PORTDB(isp, chan, 1);
isp_async(isp, ISPASYNC_LOOP_DOWN, chan);
#ifdef ISP_TARGET_MODE
if (isp_target_async(isp, chan, mbox)) {
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
acked = 1;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Raise default FCP logintime to 60 seconds. Move the position of where we could have seen the loop up at least once so it makes sense. Change some stuff in ispscsicmd so we don't get stuck there if the loop has never come up yet. Add in some target mode support code.
2000-01-03 23:52:41 +00:00
#endif
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case ASYNC_LOOP_RESET:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* This is a broadcast event that has to be sent across
* all active channels.
*/
for (chan = 0; chan < isp->isp_nchan; chan++) {
fcparam *fcp = FCPARAM(isp, chan);
if (fcp->role == ISP_ROLE_NONE) {
continue;
}
ISP_SET_SENDMARKER(isp, chan, 1);
fcp->isp_fwstate = FW_CONFIG_WAIT;
fcp->isp_loopstate = LOOP_NIL;
ISP_MARK_PORTDB(isp, chan, 1);
isp_async(isp, ISPASYNC_LOOP_RESET, chan);
#ifdef ISP_TARGET_MODE
if (isp_target_async(isp, chan, mbox)) {
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
acked = 1;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Remove the Target mode functions until they're in better shape. Implement some suggested compilation cleanups from Eklund. Wire down a hard loop id if we are not on a platform that has the ability to get to a PCI BIOS (it still will float to the ID it gets after a LIP but at least we can try). Clarify that the expanded lun is based upon SCCLUN defines (in f/w).
1998-12-05 01:33:57 +00:00
#endif
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case ASYNC_PDB_CHANGED:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
{
int nphdl, nlstate, reason;
/*
* We *should* get a channel out of the 24XX, but we don't seem
* to get more than a PDB CHANGED on channel 0, so turn it into
* a broadcast event.
*/
if (IS_24XX(isp)) {
nphdl = ISP_READ(isp, OUTMAILBOX1);
nlstate = ISP_READ(isp, OUTMAILBOX2);
reason = ISP_READ(isp, OUTMAILBOX3) >> 8;
} else {
nphdl = NIL_HANDLE;
nlstate = reason = 0;
}
for (chan = 0; chan < isp->isp_nchan; chan++) {
fcparam *fcp = FCPARAM(isp, chan);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (fcp->role == ISP_ROLE_NONE) {
continue;
}
ISP_SET_SENDMARKER(isp, chan, 1);
fcp->isp_loopstate = LOOP_PDB_RCVD;
ISP_MARK_PORTDB(isp, chan, 1);
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
isp_async(isp, ISPASYNC_CHANGE_NOTIFY, chan, ISPASYNC_CHANGE_PDB, nphdl, nlstate, reason);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
break;
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
case ASYNC_CHANGE_NOTIFY:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
{
int lochan, hichan;
if (ISP_FW_NEWER_THAN(isp, 4, 0, 25) && ISP_CAP_MULTI_ID(isp)) {
GET_24XX_BUS(isp, chan, "ASYNC_CHANGE_NOTIFY");
lochan = chan;
hichan = chan + 1;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
lochan = 0;
hichan = isp->isp_nchan;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
for (chan = lochan; chan < hichan; chan++) {
fcparam *fcp = FCPARAM(isp, chan);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (fcp->role == ISP_ROLE_NONE) {
continue;
}
if (fcp->isp_topo == TOPO_F_PORT) {
fcp->isp_loopstate = LOOP_LSCAN_DONE;
} else {
fcp->isp_loopstate = LOOP_PDB_RCVD;
}
ISP_MARK_PORTDB(isp, chan, 1);
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
isp_async(isp, ISPASYNC_CHANGE_NOTIFY, chan, ISPASYNC_CHANGE_SNS);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
break;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
case ASYNC_CONNMODE:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* This only applies to 2100 amd 2200 cards
*/
if (!IS_2200(isp) && !IS_2100(isp)) {
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
isp_prt(isp, ISP_LOGWARN, "bad card for ASYNC_CONNMODE event");
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
break;
}
chan = 0;
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
mbox = ISP_READ(isp, OUTMAILBOX1);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MARK_PORTDB(isp, chan, 1);
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
switch (mbox) {
case ISP_CONN_LOOP:
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
isp_prt(isp, ISP_LOGINFO,
"Point-to-Point -> Loop mode");
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
break;
case ISP_CONN_PTP:
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
isp_prt(isp, ISP_LOGINFO,
"Loop -> Point-to-Point mode");
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
break;
case ISP_CONN_BADLIP:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGWARN,
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
"Point-to-Point -> Loop mode (BAD LIP)");
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
break;
case ISP_CONN_FATAL:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp->isp_dead = 1;
isp->isp_state = ISP_CRASHED;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGERR, "FATAL CONNECTION ERROR");
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_async(isp, ISPASYNC_FW_CRASH);
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
return (-1);
case ISP_CONN_LOOPBACK:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGWARN,
"Looped Back in Point-to-Point mode");
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
break;
default:
isp_prt(isp, ISP_LOGWARN,
"Unknown connection mode (0x%x)", mbox);
break;
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
}
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
isp_async(isp, ISPASYNC_CHANGE_NOTIFY, chan, ISPASYNC_CHANGE_OTHER);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FCPARAM(isp, chan)->sendmarker = 1;
FCPARAM(isp, chan)->isp_fwstate = FW_CONFIG_WAIT;
FCPARAM(isp, chan)->isp_loopstate = LOOP_LIP_RCVD;
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
break;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
case ASYNC_RCV_ERR:
if (IS_24XX(isp)) {
isp_prt(isp, ISP_LOGWARN, "Receive Error");
} else {
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
isp_prt(isp, ISP_LOGWARN, "unexpected ASYNC_RCV_ERR");
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
break;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
case ASYNC_RJT_SENT: /* same as ASYNC_QFULL_SENT */
if (IS_24XX(isp)) {
isp_prt(isp, ISP_LOGTDEBUG0, "LS_RJT sent");
break;
} else if (IS_2200(isp)) {
isp_prt(isp, ISP_LOGTDEBUG0, "QFULL sent");
break;
}
/* FALLTHROUGH */
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
default:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGWARN, "Unknown Async Code 0x%x", mbox);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
}
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
if (mbox != ASYNC_CTIO_DONE && mbox != ASYNC_CMD_CMPLT) {
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
isp->isp_intoasync++;
}
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
return (acked);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
Clean up stupidity in the isp_handle_other_response function- indexes of queue entries have to be at least 16 bits now! If we're running a 2100 less than rev 5, turn off loop fairness (per Qlogic errata). Fix typo in checking against 2200 F/W revision. Slightly fix/reorder fabric login stuff. Change to usage of isp_getrqentry for code clarity. Add some defensive dual bus assumptions. Various cleanups, etc...
1999-11-21 03:18:22 +00:00
/*
* Handle other response entries. A pointer to the request queue output
* index is here in case we want to eat several entries at once, although
* this is not used currently.
*/
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
static int
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_handle_other_response(ispsoftc_t *isp, int type, isphdr_t *hp, uint32_t *optrp)
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
{
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
switch (type) {
various fixes
2000-08-27 23:38:44 +00:00
case RQSTYPE_STATUS_CONT:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_prt(isp, ISP_LOGDEBUG0, "Ignored Continuation Response");
return (1);
case RQSTYPE_MARKER:
isp_prt(isp, ISP_LOGDEBUG0, "Marker Response");
Limit fabric search to a default 256 entries. This will all go away soon because it's just getting harder and harder to find switches that correctly implement the GET ALL NEXT subcommands for the SNS protocol. Latch up result out pointer and set a busy flag when we're looking at the response queue. This allows for a cleaner way to make sure we don't get multiple CPUs trying to read the same response queue entries. Change how isp_handle_other_response returns values (clarity). Make PORT UNAVAILABLE the same as PORT LOGOUT (force a LIP). Do some formatting changes. MFC after: 0 days
2002-03-21 21:10:16 +00:00
return (1);
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
case RQSTYPE_ATIO:
Raise default FCP logintime to 60 seconds. Move the position of where we could have seen the loop up at least once so it makes sense. Change some stuff in ispscsicmd so we don't get stuck there if the loop has never come up yet. Add in some target mode support code.
2000-01-03 23:52:41 +00:00
case RQSTYPE_CTIO:
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
case RQSTYPE_ENABLE_LUN:
Remove the Target mode functions until they're in better shape. Implement some suggested compilation cleanups from Eklund. Wire down a hard loop id if we are not on a platform that has the ability to get to a PCI BIOS (it still will float to the ID it gets after a LIP but at least we can try). Clarify that the expanded lun is based upon SCCLUN defines (in f/w).
1998-12-05 01:33:57 +00:00
case RQSTYPE_MODIFY_LUN:
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
case RQSTYPE_NOTIFY:
case RQSTYPE_NOTIFY_ACK:
case RQSTYPE_CTIO1:
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
case RQSTYPE_ATIO2:
case RQSTYPE_CTIO2:
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
case RQSTYPE_CTIO3:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
case RQSTYPE_CTIO7:
case RQSTYPE_ABTS_RCVD:
case RQSTYPE_ABTS_RSP:
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
isp->isp_rsltccmplt++; /* count as a response completion */
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
#ifdef ISP_TARGET_MODE
Limit fabric search to a default 256 entries. This will all go away soon because it's just getting harder and harder to find switches that correctly implement the GET ALL NEXT subcommands for the SNS protocol. Latch up result out pointer and set a busy flag when we're looking at the response queue. This allows for a cleaner way to make sure we don't get multiple CPUs trying to read the same response queue entries. Change how isp_handle_other_response returns values (clarity). Make PORT UNAVAILABLE the same as PORT LOGOUT (force a LIP). Do some formatting changes. MFC after: 0 days
2002-03-21 21:10:16 +00:00
if (isp_target_notify(isp, (ispstatusreq_t *) hp, optrp)) {
return (1);
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
#endif
Revert previous commit. Violates Maintainer (O'Brien knows how to reach me directly), but more importantly, breaks compiles on non-FreeBSD platforms.
2003-08-25 17:58:23 +00:00
/* FALLTHROUGH */
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
case RQSTYPE_RPT_ID_ACQ:
if (IS_24XX(isp)) {
isp_ridacq_t rid;
isp_get_ridacq(isp, (isp_ridacq_t *)hp, &rid);
if (rid.ridacq_format == 0) {
}
return (1);
}
/* FALLTHROUGH */
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
case RQSTYPE_REQUEST:
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
default:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(100);
Don't set ZIO for 23XX for target mode (use fast posting instead). Use the correct number of handles for multihandle returns. Very, very, rarely on some SMP systems we've seen an 'unstable' type in the response queue. I dunno whether or not it's a bug in our handling, or whether there's a cache incoherency issue, but try to guard against it. MFC after: 2 weeks
2005-01-23 06:26:45 +00:00
if (type != isp_get_response_type(isp, hp)) {
/*
* This is questionable- we're just papering over
* something we've seen on SMP linux in target
* mode- we don't really know what's happening
* here that causes us to think we've gotten
* an entry, but that either the entry isn't
* filled out yet or our CPU read data is stale.
*/
isp_prt(isp, ISP_LOGINFO,
"unstable type in response queue");
return (-1);
Fix a longstanding bug- we had the sense of what bit 14 for the ICB firmware options meant- *I* had taken it to mean that if you set it, Node Name would be ignored and derived from Port Name. Actually, it meant the opposite. As a consequence- change ICBOPT_USE_PORTNAME to the define ICBOPT_BOTH_WWNS- makes more sense. Fix wrong input bitmap for MBOX_DUMP_RAM command. Call ISP_DUMPREGS if we get a f/w crash. Add ISPCTL_RUN_MBOXCMD control command (so outer layers can run a mailbox command directly) and add a ISPASYNC_UNHANDLED_RESPONSE hook so outer layers can understand response queue entries we might not know about.
2001-02-23 05:35:50 +00:00
}
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGWARN, "Unhandled Response Type 0x%x",
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
isp_get_response_type(isp, hp));
Limit fabric search to a default 256 entries. This will all go away soon because it's just getting harder and harder to find switches that correctly implement the GET ALL NEXT subcommands for the SNS protocol. Latch up result out pointer and set a busy flag when we're looking at the response queue. This allows for a cleaner way to make sure we don't get multiple CPUs trying to read the same response queue entries. Change how isp_handle_other_response returns values (clarity). Make PORT UNAVAILABLE the same as PORT LOGOUT (force a LIP). Do some formatting changes. MFC after: 0 days
2002-03-21 21:10:16 +00:00
return (0);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
}
Remove the Target mode functions until they're in better shape. Implement some suggested compilation cleanups from Eklund. Wire down a hard loop id if we are not on a platform that has the ability to get to a PCI BIOS (it still will float to the ID it gets after a LIP but at least we can try). Clarify that the expanded lun is based upon SCCLUN defines (in f/w).
1998-12-05 01:33:57 +00:00
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
static void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_parse_status(ispsoftc_t *isp, ispstatusreq_t *sp, XS_T *xs, long *rp)
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
{
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
switch (sp->req_completion_status & 0xff) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
case RQCS_COMPLETE:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_NOERROR);
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
return;
case RQCS_INCOMPLETE:
if ((sp->req_state_flags & RQSF_GOT_TARGET) == 0) {
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGDEBUG1, "Selection Timeout");
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_SELTIMEOUT);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*rp = XS_XFRLEN(xs);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
return;
}
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "Command Incomplete, state 0x%x", sp->req_state_flags);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case RQCS_DMA_ERROR:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "DMA Error");
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*rp = XS_XFRLEN(xs);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
case RQCS_TRANSPORT_ERROR:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
{
char buf[172];
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "states=>");
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (sp->req_state_flags & RQSF_GOT_BUS) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "%s GOT_BUS", buf);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
if (sp->req_state_flags & RQSF_GOT_TARGET) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "%s GOT_TGT", buf);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
if (sp->req_state_flags & RQSF_SENT_CDB) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "%s SENT_CDB", buf);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
if (sp->req_state_flags & RQSF_XFRD_DATA) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "%s XFRD_DATA", buf);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
if (sp->req_state_flags & RQSF_GOT_STATUS) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "%s GOT_STS", buf);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
if (sp->req_state_flags & RQSF_GOT_SENSE) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "%s GOT_SNS", buf);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
if (sp->req_state_flags & RQSF_XFER_COMPLETE) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "%s XFR_CMPLT", buf);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "%s\nstatus=>", buf);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (sp->req_status_flags & RQSTF_DISCONNECT) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "%s Disconnect", buf);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
if (sp->req_status_flags & RQSTF_SYNCHRONOUS) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "%s Sync_xfr", buf);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
if (sp->req_status_flags & RQSTF_PARITY_ERROR) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "%s Parity", buf);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
if (sp->req_status_flags & RQSTF_BUS_RESET) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "%s Bus_Reset", buf);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
if (sp->req_status_flags & RQSTF_DEVICE_RESET) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "%s Device_Reset", buf);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
if (sp->req_status_flags & RQSTF_ABORTED) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "%s Aborted", buf);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
if (sp->req_status_flags & RQSTF_TIMEOUT) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "%s Timeout", buf);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
if (sp->req_status_flags & RQSTF_NEGOTIATION) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, sizeof (buf), "%s Negotiation", buf);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "Transport Error: %s", buf);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*rp = XS_XFRLEN(xs);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
case RQCS_RESET_OCCURRED:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
{
int chan;
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGWARN, "Bus Reset destroyed command");
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
for (chan = 0; chan < isp->isp_nchan; chan++) {
FCPARAM(isp, chan)->sendmarker = 1;
}
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_BUSRESET);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*rp = XS_XFRLEN(xs);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
return;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
case RQCS_ABORTED:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "Command Aborted");
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SET_SENDMARKER(isp, XS_CHANNEL(xs), 1);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_ABORTED);
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
return;
case RQCS_TIMEOUT:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGWARN, "Command timed out");
If the HBA is already 'touched', still set maxluns. Othewise for CAM_QUIRK_HILUN devices we loop thru 32bits of lun. Oops. Switch to using USEC_DELAY rather than USEC_SLEEP at isp_reset time. Try to paper around a defect in clients that don't correctly registers themeselves with the fabric nameserver. Minor updates for Mirapoint support- they still use code that is not HANDLE_LOOPSTATE_IN_OUTER_LAYERS, and, surprise surprise, this old stuff had some bugs in it. Clean up some target mode stuff. MFC after: 1 week
2002-06-16 05:18:22 +00:00
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* XXX: Check to see if we logged out of the device.
If the HBA is already 'touched', still set maxluns. Othewise for CAM_QUIRK_HILUN devices we loop thru 32bits of lun. Oops. Switch to using USEC_DELAY rather than USEC_SLEEP at isp_reset time. Try to paper around a defect in clients that don't correctly registers themeselves with the fabric nameserver. Minor updates for Mirapoint support- they still use code that is not HANDLE_LOOPSTATE_IN_OUTER_LAYERS, and, surprise surprise, this old stuff had some bugs in it. Clean up some target mode stuff. MFC after: 1 week
2002-06-16 05:18:22 +00:00
*/
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_CMDTIMEOUT);
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
return;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
case RQCS_DATA_OVERRUN:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
XS_SET_RESID(xs, sp->req_resid);
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "data overrun (%ld)", (long) XS_GET_RESID(xs));
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_DATAOVR);
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
return;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
case RQCS_COMMAND_OVERRUN:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "command overrun");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case RQCS_STATUS_OVERRUN:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "status overrun");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case RQCS_BAD_MESSAGE:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "msg not COMMAND COMPLETE after status");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case RQCS_NO_MESSAGE_OUT:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "No MESSAGE OUT phase after selection");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case RQCS_EXT_ID_FAILED:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "EXTENDED IDENTIFY failed");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case RQCS_IDE_MSG_FAILED:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "INITIATOR DETECTED ERROR rejected");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case RQCS_ABORT_MSG_FAILED:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "ABORT OPERATION rejected");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case RQCS_REJECT_MSG_FAILED:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "MESSAGE REJECT rejected");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case RQCS_NOP_MSG_FAILED:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "NOP rejected");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case RQCS_PARITY_ERROR_MSG_FAILED:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "MESSAGE PARITY ERROR rejected");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case RQCS_DEVICE_RESET_MSG_FAILED:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGWARN, "BUS DEVICE RESET rejected");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case RQCS_ID_MSG_FAILED:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "IDENTIFY rejected");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case RQCS_UNEXP_BUS_FREE:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "Unexpected Bus Free");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
case RQCS_DATA_UNDERRUN:
If we get a DATA UNDERRUN error from QLogic FC cards, but the RQCS_RU bit is not set in the scsi completion status, or if the residual is clearly nonsense, then this was a command that suffered the loss of one or more FC frames in the middle of the exchange. Set HBA_BOTCH and hope it will get retried. It's the only thing we can do. MFC after: 1 day
2002-05-01 21:58:36 +00:00
{
if (IS_FC(isp)) {
int ru_marked = (sp->req_scsi_status & RQCS_RU) != 0;
if (!ru_marked || sp->req_resid > XS_XFRLEN(xs)) {
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGWARN, bun, XS_XFRLEN(xs), sp->req_resid, (ru_marked)? "marked" : "not marked");
If we get a DATA UNDERRUN error from QLogic FC cards, but the RQCS_RU bit is not set in the scsi completion status, or if the residual is clearly nonsense, then this was a command that suffered the loss of one or more FC frames in the middle of the exchange. Set HBA_BOTCH and hope it will get retried. It's the only thing we can do. MFC after: 1 day
2002-05-01 21:58:36 +00:00
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_BOTCH);
}
return;
}
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
XS_SET_RESID(xs, sp->req_resid);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_NOERROR);
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
return;
If we get a DATA UNDERRUN error from QLogic FC cards, but the RQCS_RU bit is not set in the scsi completion status, or if the residual is clearly nonsense, then this was a command that suffered the loss of one or more FC frames in the middle of the exchange. Set HBA_BOTCH and hope it will get retried. It's the only thing we can do. MFC after: 1 day
2002-05-01 21:58:36 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
case RQCS_XACT_ERR1:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "HBA attempted queued transaction with disconnect not set");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
case RQCS_XACT_ERR2:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "HBA attempted queued transaction to target routine %d", XS_LUN(xs));
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case RQCS_XACT_ERR3:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "HBA attempted queued cmd when queueing disabled");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case RQCS_BAD_ENTRY:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGERR, "Invalid IOCB entry type detected");
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
break;
case RQCS_QUEUE_FULL:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGDEBUG0, "internal queues full status 0x%x", *XS_STSP(xs));
Tra-La, another QLogic f/w funny- this time with the 2300. If we get a completion status of RQCS_QUEUE_FULL, it means that the internal queues are full. Other QLogic boards set the QFULL SCSI status. But *nooooooooooo*, not the 2300. MFC after: 1 day
2001-10-23 23:05:20 +00:00
Add some prototype deadchip detection. Set FIFO bursting (1XX0 only- it's already on for the 2XX0) and detect the broken 1040A FIFO. Change bzero to MEMZERO (portability with **nux). Use memcpy for same reason. Finally detect QUEUE FULL conditions and return this as an error that will get cam_periph_error to do it's 'tagged openings now XXX' dance.
1999-01-10 02:55:10 +00:00
/*
* If QFULL or some other status byte is set, then this
* isn't an error, per se.
Tra-La, another QLogic f/w funny- this time with the 2300. If we get a completion status of RQCS_QUEUE_FULL, it means that the internal queues are full. Other QLogic boards set the QFULL SCSI status. But *nooooooooooo*, not the 2300. MFC after: 1 day
2001-10-23 23:05:20 +00:00
*
* Unfortunately, some QLogic f/w writers have, in
* some cases, ommitted to *set* status to QFULL.
*
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (*XS_STSP(xs) != SCSI_GOOD && XS_NOERR(xs)) {
Add some prototype deadchip detection. Set FIFO bursting (1XX0 only- it's already on for the 2XX0) and detect the broken 1040A FIFO. Change bzero to MEMZERO (portability with **nux). Use memcpy for same reason. Finally detect QUEUE FULL conditions and return this as an error that will get cam_periph_error to do it's 'tagged openings now XXX' dance.
1999-01-10 02:55:10 +00:00
XS_SETERR(xs, HBA_NOERROR);
return;
}
Tra-La, another QLogic f/w funny- this time with the 2300. If we get a completion status of RQCS_QUEUE_FULL, it means that the internal queues are full. Other QLogic boards set the QFULL SCSI status. But *nooooooooooo*, not the 2300. MFC after: 1 day
2001-10-23 23:05:20 +00:00
*
*
*/
*XS_STSP(xs) = SCSI_QFULL;
XS_SETERR(xs, HBA_NOERROR);
return;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
case RQCS_PHASE_SKIPPED:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "SCSI phase skipped");
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case RQCS_ARQS_FAILED:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "Auto Request Sense Failed");
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_ARQFAIL);
}
return;
case RQCS_WIDE_FAILED:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "Wide Negotiation Failed");
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_SCSI(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdparam *sdp = SDPARAM(isp, XS_CHANNEL(xs));
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
sdp->isp_devparam[XS_TGT(xs)].goal_flags &= ~DPARM_WIDE;
sdp->isp_devparam[XS_TGT(xs)].dev_update = 1;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdp->update = 1;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_NOERROR);
}
return;
case RQCS_SYNCXFER_FAILED:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "SDTR Message Failed");
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_SCSI(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdparam *sdp = SDPARAM(isp, XS_CHANNEL(xs));
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
sdp += XS_CHANNEL(xs);
sdp->isp_devparam[XS_TGT(xs)].goal_flags &= ~DPARM_SYNC;
sdp->isp_devparam[XS_TGT(xs)].dev_update = 1;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdp->update = 1;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
break;
case RQCS_LVD_BUSERR:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "Bad LVD condition");
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case RQCS_PORT_UNAVAILABLE:
/*
* No such port on the loop. Moral equivalent of SELTIMEO
*/
case RQCS_PORT_LOGGED_OUT:
{
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
const char *reason;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
uint8_t sts = sp->req_completion_status & 0xff;
/*
* It was there (maybe)- treat as a selection timeout.
*/
if (sts == RQCS_PORT_UNAVAILABLE) {
reason = "unavailable";
} else {
reason = "logout";
}
isp_prt(isp, ISP_LOGINFO, "port %s for target %d",
reason, XS_TGT(xs));
/*
* If we're on a local loop, force a LIP (which is overkill)
* to force a re-login of this unit. If we're on fabric,
* then we'll have to log in again as a matter of course.
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (FCPARAM(isp, 0)->isp_topo == TOPO_NL_PORT ||
FCPARAM(isp, 0)->isp_topo == TOPO_FL_PORT) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbreg_t mbs;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_INIT_LIP, MBLOGALL, 0);
if (ISP_CAP_2KLOGIN(isp)) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.ibits = (1 << 10);
}
isp_mboxcmd_qnw(isp, &mbs, 1);
}
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_SELTIMEOUT);
}
return;
}
case RQCS_PORT_CHANGED:
isp_prt(isp, ISP_LOGWARN,
"port changed for target %d", XS_TGT(xs));
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_SELTIMEOUT);
}
return;
case RQCS_PORT_BUSY:
isp_prt(isp, ISP_LOGWARN,
"port busy for target %d", XS_TGT(xs));
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_TGTBSY);
}
return;
default:
isp_prt(isp, ISP_LOGERR, "Unknown Completion Status 0x%x",
sp->req_completion_status);
break;
}
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_BOTCH);
}
}
static void
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_parse_status_24xx(ispsoftc_t *isp, isp24xx_statusreq_t *sp, XS_T *xs, long *rp)
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
{
Recover from some major omissions/problems with the 24XX port. First, we were never correctly checking for a 24XX Status Type 0 response- that cased us to fall through to evaluate status for commands as if this were a 2100/2200/2300 Status Type 0 response. This is *close*, but not quite the same. This has been reported to be apparent with some wierd lun configuration problems with some arrays. It became glaringly apparent on sparc64 where none of the correct byte swap things were done. Fixing this omission then caused a whole universe shifting debug cycle of endian issues for the 2400. The manual for 24XX f/w turns out to be wrong about the endianness of a couple of entities. The lun and cdb fields for the type 7 request are *not* unconditionally big endian- they happen to be opposite of whatever the endian of the current machine type is. Same with the sense data for the 24XX type 0 response. While we're at it investigate and resolve some NVRAM endian issues. Approved by: re (ken) MFC after: 3 days
2007-07-02 20:08:20 +00:00
int ru_marked, sv_marked;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
int chan = XS_CHANNEL(xs);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
switch (sp->req_completion_status) {
case RQCS_COMPLETE:
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_NOERROR);
}
return;
case RQCS_DMA_ERROR:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "DMA error");
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case RQCS_TRANSPORT_ERROR:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "Transport Error");
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
case RQCS_RESET_OCCURRED:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGWARN, "reset destroyed command");
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FCPARAM(isp, chan)->sendmarker = 1;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_BUSRESET);
}
return;
case RQCS_ABORTED:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "Command Aborted");
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
FCPARAM(isp, chan)->sendmarker = 1;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_ABORTED);
}
return;
case RQCS_TIMEOUT:
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGWARN, "Command Timed Out");
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (XS_NOERR(xs)) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
XS_SETERR(xs, HBA_CMDTIMEOUT);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
return;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
case RQCS_DATA_OVERRUN:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
XS_SET_RESID(xs, sp->req_resid);
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGERR, "Data Overrun");
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_DATAOVR);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return;
case RQCS_24XX_DRE: /* data reassembly error */
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGERR,
"Chan %d data reassembly error for target %d",
chan, XS_TGT(xs));
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (XS_NOERR(xs)) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
XS_SETERR(xs, HBA_ABORTED);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*rp = XS_XFRLEN(xs);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
return;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
case RQCS_24XX_TABORT: /* aborted by target */
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGERR, "Chan %d target %d sent ABTS",
chan, XS_TGT(xs));
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_ABORTED);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
return;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
case RQCS_DATA_UNDERRUN:
Recover from some major omissions/problems with the 24XX port. First, we were never correctly checking for a 24XX Status Type 0 response- that cased us to fall through to evaluate status for commands as if this were a 2100/2200/2300 Status Type 0 response. This is *close*, but not quite the same. This has been reported to be apparent with some wierd lun configuration problems with some arrays. It became glaringly apparent on sparc64 where none of the correct byte swap things were done. Fixing this omission then caused a whole universe shifting debug cycle of endian issues for the 2400. The manual for 24XX f/w turns out to be wrong about the endianness of a couple of entities. The lun and cdb fields for the type 7 request are *not* unconditionally big endian- they happen to be opposite of whatever the endian of the current machine type is. Same with the sense data for the 24XX type 0 response. While we're at it investigate and resolve some NVRAM endian issues. Approved by: re (ken) MFC after: 3 days
2007-07-02 20:08:20 +00:00
ru_marked = (sp->req_scsi_status & RQCS_RU) != 0;
/*
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* We can get an underrun w/o things being marked
Recover from some major omissions/problems with the 24XX port. First, we were never correctly checking for a 24XX Status Type 0 response- that cased us to fall through to evaluate status for commands as if this were a 2100/2200/2300 Status Type 0 response. This is *close*, but not quite the same. This has been reported to be apparent with some wierd lun configuration problems with some arrays. It became glaringly apparent on sparc64 where none of the correct byte swap things were done. Fixing this omission then caused a whole universe shifting debug cycle of endian issues for the 2400. The manual for 24XX f/w turns out to be wrong about the endianness of a couple of entities. The lun and cdb fields for the type 7 request are *not* unconditionally big endian- they happen to be opposite of whatever the endian of the current machine type is. Same with the sense data for the 24XX type 0 response. While we're at it investigate and resolve some NVRAM endian issues. Approved by: re (ken) MFC after: 3 days
2007-07-02 20:08:20 +00:00
* if we got a non-zero status.
*/
sv_marked = (sp->req_scsi_status & (RQCS_SV|RQCS_RV)) != 0;
if ((ru_marked == 0 && sv_marked == 0) ||
(sp->req_resid > XS_XFRLEN(xs))) {
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGWARN, bun, XS_XFRLEN(xs), sp->req_resid, (ru_marked)? "marked" : "not marked");
Recover from some major omissions/problems with the 24XX port. First, we were never correctly checking for a 24XX Status Type 0 response- that cased us to fall through to evaluate status for commands as if this were a 2100/2200/2300 Status Type 0 response. This is *close*, but not quite the same. This has been reported to be apparent with some wierd lun configuration problems with some arrays. It became glaringly apparent on sparc64 where none of the correct byte swap things were done. Fixing this omission then caused a whole universe shifting debug cycle of endian issues for the 2400. The manual for 24XX f/w turns out to be wrong about the endianness of a couple of entities. The lun and cdb fields for the type 7 request are *not* unconditionally big endian- they happen to be opposite of whatever the endian of the current machine type is. Same with the sense data for the 24XX type 0 response. While we're at it investigate and resolve some NVRAM endian issues. Approved by: re (ken) MFC after: 3 days
2007-07-02 20:08:20 +00:00
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_BOTCH);
}
return;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
XS_SET_RESID(xs, sp->req_resid);
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_xs_prt(isp, xs, ISP_LOGDEBUG0, "Data Underrun (%d) for command 0x%x", sp->req_resid, XS_CDBP(xs)[0] & 0xff);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_NOERROR);
}
return;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
case RQCS_PORT_UNAVAILABLE:
/*
* No such port on the loop. Moral equivalent of SELTIMEO
*/
case RQCS_PORT_LOGGED_OUT:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
{
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
const char *reason;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
uint8_t sts = sp->req_completion_status & 0xff;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
/*
* It was there (maybe)- treat as a selection timeout.
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (sts == RQCS_PORT_UNAVAILABLE) {
reason = "unavailable";
} else {
reason = "logout";
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGINFO, "Chan %d port %s for target %d",
chan, reason, XS_TGT(xs));
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Fix a problem where a local loop disk logs out- and we get a PORT LOGGED OUT status. We are, apparently, required to force the f/w to log back in if we want to try and talk to that disk again. This means either issuing a LOGIN LOCAL LOOP PORT mailbox command, or by issuing a LIP. I've elected to issue a LIP because this has a better chance of waking up the disk which clearly just crashed and burned. These should not occur at all. If they do, they should be darned rare. MFC after: 1 week
2002-02-21 01:56:08 +00:00
/*
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* There is no MBOX_INIT_LIP for the 24XX.
Fix a problem where a local loop disk logs out- and we get a PORT LOGGED OUT status. We are, apparently, required to force the f/w to log back in if we want to try and talk to that disk again. This means either issuing a LOGIN LOCAL LOOP PORT mailbox command, or by issuing a LIP. I've elected to issue a LIP because this has a better chance of waking up the disk which clearly just crashed and burned. These should not occur at all. If they do, they should be darned rare. MFC after: 1 week
2002-02-21 01:56:08 +00:00
*/
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_SELTIMEOUT);
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
return;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
case RQCS_PORT_CHANGED:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGWARN,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"port changed for target %d chan %d", XS_TGT(xs), chan);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_SELTIMEOUT);
}
Slightly cleaner fabric support (whiter whites! redder reds!).. No, seriously- only attempt to logout a previously logged in fabric device. Fix a longstanding bug for aborting overtime commands- handle halves have always been reversed. Clean up some error messages to indicate channel number. Approved:jkh
2000-02-29 05:52:14 +00:00
return;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
case RQCS_24XX_ENOMEM: /* f/w resource unavailable */
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGWARN,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"f/w resource unavailable for target %d chan %d",
XS_TGT(xs), chan);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (XS_NOERR(xs)) {
*XS_STSP(xs) = SCSI_BUSY;
XS_SETERR(xs, HBA_TGTBSY);
}
return;
case RQCS_24XX_TMO: /* task management overrun */
isp_prt(isp, ISP_LOGWARN,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"command for target %d overlapped task management for "
"chan %d", XS_TGT(xs), chan);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (XS_NOERR(xs)) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
*XS_STSP(xs) = SCSI_BUSY;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
XS_SETERR(xs, HBA_TGTBSY);
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
return;
default:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGERR,
"Unknown Completion Status 0x%x on chan %d",
sp->req_completion_status, chan);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
}
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_BOTCH);
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
static void
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
isp_fastpost_complete(ispsoftc_t *isp, uint32_t fph)
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
{
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
XS_T *xs;
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
Switch to using 16 bit handles instead of 32 bit handles. This is a pretty invasive change, but there are three good reasons to do this: 1. We'll never have > 16 bits of handle. 2. We can (eventually) enable the RIO (Reduced Interrupt Operation) bits which return multiple completing 16 bit handles in mailbox registers. 3. The !)$*)$*~)@$*~)$* Qlogic target mode for parallel SCSI spec changed such that at_reserved (which was 32 bits) was split into two pieces- and one of which was a 16 bit handle id that functions like the at_rxid for Fibre Channel (a tag for the f/w to correlate CTIOs with a particular command). Since we had to muck with that and this changed the whole handler architecture, we might as well... Propagate new at_handle on through int ct_fwhandle. Follow implications of changing to 16 bit handles. These above changes at least get Qlogic 1040 cards working in target mode again. 1080/12160 cards don't work yet. In isp.c: Prepare for doing all loop management in outer layers.
2001-03-02 06:28:55 +00:00
if (fph == 0) {
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
return;
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
}
xs = isp_find_xs(isp, fph);
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
if (xs == NULL) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGWARN,
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
"Command for fast post handle 0x%x not found", fph);
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
return;
}
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
isp_destroy_handle(isp, fph);
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
/*
* Since we don't have a result queue entry item,
* we must believe that SCSI status is zero and
* that all data transferred.
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
XS_SET_RESID(xs, 0);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
*XS_STSP(xs) = SCSI_GOOD;
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
if (XS_XFRLEN(xs)) {
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
ISP_DMAFREE(isp, xs, fph);
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (isp->isp_nactive) {
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
isp->isp_nactive--;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
isp->isp_fphccmplt++;
various fixes
2000-08-27 23:38:44 +00:00
isp_done(xs);
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
}
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
static int
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
isp_mbox_continue(ispsoftc_t *isp)
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
{
mbreg_t mbs;
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
uint16_t *ptr;
Fix 2KLOGIN code to specify *ibits* (not *obits*) so that the options field in register 10 will be deterministic, not random. Correct the number of input bits for EXECUTE_FIRMWARE 0..1 to 0..2- the 2322 and 24XX cards use mailbox register 2 to specify whether the f/w being executed is freshly loaded or not. Correct the number of input bits for {READ,WRITE}_RAM_WORD_EXTENDED so that register 8 gets picked up. Fix the indexing and offset for the 2322 f/w download so that it correctly puts the different code segments where they belong. Move VERIFY_CHECKSUM to be the 'else' clause to 2322 f/w downloads- the EXECUTE_FIRMWARE command for 2322 and 24XX cards will tell you if the f/w checksum is incorrect and VERIFY_CHECKSUM only works for RISC SRAM address < 64K so you can only do a VERIFY_CHECKSUM on the first of the 3 f/w segments for the 2322. Shorten the delay for the continuation mailbox commands- 1ms is ridiculous (100us is more likely). All of the more or less is really only for the 2322/6322 cards.
2006-08-14 05:42:46 +00:00
uint32_t offset;
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
Support for f/w crash dumps (2200 && 23XX). If you want QLogic to look at a potential f/w problem for FC cards, you really have to provide them info in the format they expect. This involves dumping a lot of hardware registers (> 300 16 bit registers) and a lot of SRAM (> 128KB minimum). Thus all of this code is #ifdef protected which will become an option so that the memory allocation of where to dump the crash image is pretty expensive. It's worth it if you have a reproducible problem because they have some tools that can tell them, given the f/w version, the precise state of everything. MFC after: 1 week
2002-02-17 06:38:22 +00:00
switch (isp->isp_lastmbxcmd) {
case MBOX_WRITE_RAM_WORD:
case MBOX_READ_RAM_WORD:
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
case MBOX_WRITE_RAM_WORD_EXTENDED:
Support for f/w crash dumps (2200 && 23XX). If you want QLogic to look at a potential f/w problem for FC cards, you really have to provide them info in the format they expect. This involves dumping a lot of hardware registers (> 300 16 bit registers) and a lot of SRAM (> 128KB minimum). Thus all of this code is #ifdef protected which will become an option so that the memory allocation of where to dump the crash image is pretty expensive. It's worth it if you have a reproducible problem because they have some tools that can tell them, given the f/w version, the precise state of everything. MFC after: 1 week
2002-02-17 06:38:22 +00:00
case MBOX_READ_RAM_WORD_EXTENDED:
break;
default:
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
return (1);
}
if (isp->isp_mboxtmp[0] != MBOX_COMMAND_COMPLETE) {
isp->isp_mbxwrk0 = 0;
return (-1);
}
/*
* Clear the previous interrupt.
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
ISP_WRITE(isp, BIU2400_HCCR, HCCR_2400_CMD_CLEAR_RISC_INT);
} else {
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
ISP_WRITE(isp, BIU_SEMA, 0);
}
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
/*
* Continue with next word.
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(&mbs, sizeof (mbs));
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
ptr = isp->isp_mbxworkp;
Support for f/w crash dumps (2200 && 23XX). If you want QLogic to look at a potential f/w problem for FC cards, you really have to provide them info in the format they expect. This involves dumping a lot of hardware registers (> 300 16 bit registers) and a lot of SRAM (> 128KB minimum). Thus all of this code is #ifdef protected which will become an option so that the memory allocation of where to dump the crash image is pretty expensive. It's worth it if you have a reproducible problem because they have some tools that can tell them, given the f/w version, the precise state of everything. MFC after: 1 week
2002-02-17 06:38:22 +00:00
switch (isp->isp_lastmbxcmd) {
case MBOX_WRITE_RAM_WORD:
Remove extraneous semicolons, no functional changes. Submitted by: Marc Balmer <marc@msys.ch> MFC after: 1 week
2010-01-07 21:01:37 +00:00
mbs.param[1] = isp->isp_mbxwrk1++;
mbs.param[2] = *ptr++;
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
break;
Support for f/w crash dumps (2200 && 23XX). If you want QLogic to look at a potential f/w problem for FC cards, you really have to provide them info in the format they expect. This involves dumping a lot of hardware registers (> 300 16 bit registers) and a lot of SRAM (> 128KB minimum). Thus all of this code is #ifdef protected which will become an option so that the memory allocation of where to dump the crash image is pretty expensive. It's worth it if you have a reproducible problem because they have some tools that can tell them, given the f/w version, the precise state of everything. MFC after: 1 week
2002-02-17 06:38:22 +00:00
case MBOX_READ_RAM_WORD:
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
*ptr++ = isp->isp_mboxtmp[2];
mbs.param[1] = isp->isp_mbxwrk1++;
break;
Fix 2KLOGIN code to specify *ibits* (not *obits*) so that the options field in register 10 will be deterministic, not random. Correct the number of input bits for EXECUTE_FIRMWARE 0..1 to 0..2- the 2322 and 24XX cards use mailbox register 2 to specify whether the f/w being executed is freshly loaded or not. Correct the number of input bits for {READ,WRITE}_RAM_WORD_EXTENDED so that register 8 gets picked up. Fix the indexing and offset for the 2322 f/w download so that it correctly puts the different code segments where they belong. Move VERIFY_CHECKSUM to be the 'else' clause to 2322 f/w downloads- the EXECUTE_FIRMWARE command for 2322 and 24XX cards will tell you if the f/w checksum is incorrect and VERIFY_CHECKSUM only works for RISC SRAM address < 64K so you can only do a VERIFY_CHECKSUM on the first of the 3 f/w segments for the 2322. Shorten the delay for the continuation mailbox commands- 1ms is ridiculous (100us is more likely). All of the more or less is really only for the 2322/6322 cards.
2006-08-14 05:42:46 +00:00
case MBOX_WRITE_RAM_WORD_EXTENDED:
offset = isp->isp_mbxwrk1;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
offset |= isp->isp_mbxwrk8 << 16;
Fix 2KLOGIN code to specify *ibits* (not *obits*) so that the options field in register 10 will be deterministic, not random. Correct the number of input bits for EXECUTE_FIRMWARE 0..1 to 0..2- the 2322 and 24XX cards use mailbox register 2 to specify whether the f/w being executed is freshly loaded or not. Correct the number of input bits for {READ,WRITE}_RAM_WORD_EXTENDED so that register 8 gets picked up. Fix the indexing and offset for the 2322 f/w download so that it correctly puts the different code segments where they belong. Move VERIFY_CHECKSUM to be the 'else' clause to 2322 f/w downloads- the EXECUTE_FIRMWARE command for 2322 and 24XX cards will tell you if the f/w checksum is incorrect and VERIFY_CHECKSUM only works for RISC SRAM address < 64K so you can only do a VERIFY_CHECKSUM on the first of the 3 f/w segments for the 2322. Shorten the delay for the continuation mailbox commands- 1ms is ridiculous (100us is more likely). All of the more or less is really only for the 2322/6322 cards.
2006-08-14 05:42:46 +00:00
Remove extraneous semicolons, no functional changes. Submitted by: Marc Balmer <marc@msys.ch> MFC after: 1 week
2010-01-07 21:01:37 +00:00
mbs.param[2] = *ptr++;
Fix 2KLOGIN code to specify *ibits* (not *obits*) so that the options field in register 10 will be deterministic, not random. Correct the number of input bits for EXECUTE_FIRMWARE 0..1 to 0..2- the 2322 and 24XX cards use mailbox register 2 to specify whether the f/w being executed is freshly loaded or not. Correct the number of input bits for {READ,WRITE}_RAM_WORD_EXTENDED so that register 8 gets picked up. Fix the indexing and offset for the 2322 f/w download so that it correctly puts the different code segments where they belong. Move VERIFY_CHECKSUM to be the 'else' clause to 2322 f/w downloads- the EXECUTE_FIRMWARE command for 2322 and 24XX cards will tell you if the f/w checksum is incorrect and VERIFY_CHECKSUM only works for RISC SRAM address < 64K so you can only do a VERIFY_CHECKSUM on the first of the 3 f/w segments for the 2322. Shorten the delay for the continuation mailbox commands- 1ms is ridiculous (100us is more likely). All of the more or less is really only for the 2322/6322 cards.
2006-08-14 05:42:46 +00:00
mbs.param[1] = offset;
mbs.param[8] = offset >> 16;
isp->isp_mbxwrk1 = ++offset;
isp->isp_mbxwrk8 = offset >> 16;
break;
Support for f/w crash dumps (2200 && 23XX). If you want QLogic to look at a potential f/w problem for FC cards, you really have to provide them info in the format they expect. This involves dumping a lot of hardware registers (> 300 16 bit registers) and a lot of SRAM (> 128KB minimum). Thus all of this code is #ifdef protected which will become an option so that the memory allocation of where to dump the crash image is pretty expensive. It's worth it if you have a reproducible problem because they have some tools that can tell them, given the f/w version, the precise state of everything. MFC after: 1 week
2002-02-17 06:38:22 +00:00
case MBOX_READ_RAM_WORD_EXTENDED:
Fix 2KLOGIN code to specify *ibits* (not *obits*) so that the options field in register 10 will be deterministic, not random. Correct the number of input bits for EXECUTE_FIRMWARE 0..1 to 0..2- the 2322 and 24XX cards use mailbox register 2 to specify whether the f/w being executed is freshly loaded or not. Correct the number of input bits for {READ,WRITE}_RAM_WORD_EXTENDED so that register 8 gets picked up. Fix the indexing and offset for the 2322 f/w download so that it correctly puts the different code segments where they belong. Move VERIFY_CHECKSUM to be the 'else' clause to 2322 f/w downloads- the EXECUTE_FIRMWARE command for 2322 and 24XX cards will tell you if the f/w checksum is incorrect and VERIFY_CHECKSUM only works for RISC SRAM address < 64K so you can only do a VERIFY_CHECKSUM on the first of the 3 f/w segments for the 2322. Shorten the delay for the continuation mailbox commands- 1ms is ridiculous (100us is more likely). All of the more or less is really only for the 2322/6322 cards.
2006-08-14 05:42:46 +00:00
offset = isp->isp_mbxwrk1;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
offset |= isp->isp_mbxwrk8 << 16;
Fix 2KLOGIN code to specify *ibits* (not *obits*) so that the options field in register 10 will be deterministic, not random. Correct the number of input bits for EXECUTE_FIRMWARE 0..1 to 0..2- the 2322 and 24XX cards use mailbox register 2 to specify whether the f/w being executed is freshly loaded or not. Correct the number of input bits for {READ,WRITE}_RAM_WORD_EXTENDED so that register 8 gets picked up. Fix the indexing and offset for the 2322 f/w download so that it correctly puts the different code segments where they belong. Move VERIFY_CHECKSUM to be the 'else' clause to 2322 f/w downloads- the EXECUTE_FIRMWARE command for 2322 and 24XX cards will tell you if the f/w checksum is incorrect and VERIFY_CHECKSUM only works for RISC SRAM address < 64K so you can only do a VERIFY_CHECKSUM on the first of the 3 f/w segments for the 2322. Shorten the delay for the continuation mailbox commands- 1ms is ridiculous (100us is more likely). All of the more or less is really only for the 2322/6322 cards.
2006-08-14 05:42:46 +00:00
Support for f/w crash dumps (2200 && 23XX). If you want QLogic to look at a potential f/w problem for FC cards, you really have to provide them info in the format they expect. This involves dumping a lot of hardware registers (> 300 16 bit registers) and a lot of SRAM (> 128KB minimum). Thus all of this code is #ifdef protected which will become an option so that the memory allocation of where to dump the crash image is pretty expensive. It's worth it if you have a reproducible problem because they have some tools that can tell them, given the f/w version, the precise state of everything. MFC after: 1 week
2002-02-17 06:38:22 +00:00
*ptr++ = isp->isp_mboxtmp[2];
Fix 2KLOGIN code to specify *ibits* (not *obits*) so that the options field in register 10 will be deterministic, not random. Correct the number of input bits for EXECUTE_FIRMWARE 0..1 to 0..2- the 2322 and 24XX cards use mailbox register 2 to specify whether the f/w being executed is freshly loaded or not. Correct the number of input bits for {READ,WRITE}_RAM_WORD_EXTENDED so that register 8 gets picked up. Fix the indexing and offset for the 2322 f/w download so that it correctly puts the different code segments where they belong. Move VERIFY_CHECKSUM to be the 'else' clause to 2322 f/w downloads- the EXECUTE_FIRMWARE command for 2322 and 24XX cards will tell you if the f/w checksum is incorrect and VERIFY_CHECKSUM only works for RISC SRAM address < 64K so you can only do a VERIFY_CHECKSUM on the first of the 3 f/w segments for the 2322. Shorten the delay for the continuation mailbox commands- 1ms is ridiculous (100us is more likely). All of the more or less is really only for the 2322/6322 cards.
2006-08-14 05:42:46 +00:00
mbs.param[1] = offset;
mbs.param[8] = offset >> 16;
isp->isp_mbxwrk1 = ++offset;
isp->isp_mbxwrk8 = offset >> 16;
Support for f/w crash dumps (2200 && 23XX). If you want QLogic to look at a potential f/w problem for FC cards, you really have to provide them info in the format they expect. This involves dumping a lot of hardware registers (> 300 16 bit registers) and a lot of SRAM (> 128KB minimum). Thus all of this code is #ifdef protected which will become an option so that the memory allocation of where to dump the crash image is pretty expensive. It's worth it if you have a reproducible problem because they have some tools that can tell them, given the f/w version, the precise state of everything. MFC after: 1 week
2002-02-17 06:38:22 +00:00
break;
}
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
isp->isp_mbxworkp = ptr;
Fix 2KLOGIN code to specify *ibits* (not *obits*) so that the options field in register 10 will be deterministic, not random. Correct the number of input bits for EXECUTE_FIRMWARE 0..1 to 0..2- the 2322 and 24XX cards use mailbox register 2 to specify whether the f/w being executed is freshly loaded or not. Correct the number of input bits for {READ,WRITE}_RAM_WORD_EXTENDED so that register 8 gets picked up. Fix the indexing and offset for the 2322 f/w download so that it correctly puts the different code segments where they belong. Move VERIFY_CHECKSUM to be the 'else' clause to 2322 f/w downloads- the EXECUTE_FIRMWARE command for 2322 and 24XX cards will tell you if the f/w checksum is incorrect and VERIFY_CHECKSUM only works for RISC SRAM address < 64K so you can only do a VERIFY_CHECKSUM on the first of the 3 f/w segments for the 2322. Shorten the delay for the continuation mailbox commands- 1ms is ridiculous (100us is more likely). All of the more or less is really only for the 2322/6322 cards.
2006-08-14 05:42:46 +00:00
isp->isp_mbxwrk0--;
Support for f/w crash dumps (2200 && 23XX). If you want QLogic to look at a potential f/w problem for FC cards, you really have to provide them info in the format they expect. This involves dumping a lot of hardware registers (> 300 16 bit registers) and a lot of SRAM (> 128KB minimum). Thus all of this code is #ifdef protected which will become an option so that the memory allocation of where to dump the crash image is pretty expensive. It's worth it if you have a reproducible problem because they have some tools that can tell them, given the f/w version, the precise state of everything. MFC after: 1 week
2002-02-17 06:38:22 +00:00
mbs.param[0] = isp->isp_lastmbxcmd;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
mbs.logval = MBLOGALL;
Fix a problem where a local loop disk logs out- and we get a PORT LOGGED OUT status. We are, apparently, required to force the f/w to log back in if we want to try and talk to that disk again. This means either issuing a LOGIN LOCAL LOOP PORT mailbox command, or by issuing a LIP. I've elected to issue a LIP because this has a better chance of waking up the disk which clearly just crashed and burned. These should not occur at all. If they do, they should be darned rare. MFC after: 1 week
2002-02-21 01:56:08 +00:00
isp_mboxcmd_qnw(isp, &mbs, 0);
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
return (0);
}
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
#define HIWRD(x) ((x) >> 16)
#define LOWRD(x) ((x) & 0xffff)
#define ISPOPMAP(a, b) (((a) << 16) | (b))
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
static const uint32_t mbpscsi[] = {
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
ISPOPMAP(0x01, 0x01), /* 0x00: MBOX_NO_OP */
ISPOPMAP(0x1f, 0x01), /* 0x01: MBOX_LOAD_RAM */
ISPOPMAP(0x03, 0x01), /* 0x02: MBOX_EXEC_FIRMWARE */
ISPOPMAP(0x1f, 0x01), /* 0x03: MBOX_DUMP_RAM */
ISPOPMAP(0x07, 0x07), /* 0x04: MBOX_WRITE_RAM_WORD */
ISPOPMAP(0x03, 0x07), /* 0x05: MBOX_READ_RAM_WORD */
ISPOPMAP(0x3f, 0x3f), /* 0x06: MBOX_MAILBOX_REG_TEST */
Fix 2KLOGIN code to specify *ibits* (not *obits*) so that the options field in register 10 will be deterministic, not random. Correct the number of input bits for EXECUTE_FIRMWARE 0..1 to 0..2- the 2322 and 24XX cards use mailbox register 2 to specify whether the f/w being executed is freshly loaded or not. Correct the number of input bits for {READ,WRITE}_RAM_WORD_EXTENDED so that register 8 gets picked up. Fix the indexing and offset for the 2322 f/w download so that it correctly puts the different code segments where they belong. Move VERIFY_CHECKSUM to be the 'else' clause to 2322 f/w downloads- the EXECUTE_FIRMWARE command for 2322 and 24XX cards will tell you if the f/w checksum is incorrect and VERIFY_CHECKSUM only works for RISC SRAM address < 64K so you can only do a VERIFY_CHECKSUM on the first of the 3 f/w segments for the 2322. Shorten the delay for the continuation mailbox commands- 1ms is ridiculous (100us is more likely). All of the more or less is really only for the 2322/6322 cards.
2006-08-14 05:42:46 +00:00
ISPOPMAP(0x07, 0x07), /* 0x07: MBOX_VERIFY_CHECKSUM */
Clarify issues about whether we have SCCLUN (65535 luns) or non-SCCLUN (16 luns) firmware for the Fibre Channel cards. We used to assume that if we didn't download firmware, we couldn't know what the firmware capability with respect to SCCLUNs is- and it's important because the lun field changes in the request queue entry based upon which firmware it is. At any rate, we *do* get back firmware attributes in mailbox register 6 when we do ABOUT FIRMWARE for all 2200/2300 cards- and for 2100 cards with at least 1.17.0 firmware. So- we now assume non-SCCLUN behaviour for 2100 cards with firmware < 1.17.0- and we check the firmware attributes for other cards (loaded firmware or not). This also allows us to get rid of the crappy test of isp_maxluns > 16- we simply can check firmware attributes for SCCLUN behaviour. This required an 'oops' fix to the outgoing mailbox count field for ABOUT FIRMWARE for FC cards. Also- while here, hardwire firmware revisions for loaded code for SBus cards. Apparently the 1.35 or 1.37 f/w we've been loading into isp1000 just doesn't report firmware revisions out to mailbox regs 1, 2 and 3 like everyone else. Grumble. Not that this fix hardly matters for FreeBSD. MFC after: 4 weeks
2001-09-03 03:09:48 +00:00
ISPOPMAP(0x01, 0x0f), /* 0x08: MBOX_ABOUT_FIRMWARE */
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
ISPOPMAP(0x00, 0x00), /* 0x09: */
ISPOPMAP(0x00, 0x00), /* 0x0a: */
ISPOPMAP(0x00, 0x00), /* 0x0b: */
ISPOPMAP(0x00, 0x00), /* 0x0c: */
ISPOPMAP(0x00, 0x00), /* 0x0d: */
ISPOPMAP(0x01, 0x05), /* 0x0e: MBOX_CHECK_FIRMWARE */
ISPOPMAP(0x00, 0x00), /* 0x0f: */
ISPOPMAP(0x1f, 0x1f), /* 0x10: MBOX_INIT_REQ_QUEUE */
ISPOPMAP(0x3f, 0x3f), /* 0x11: MBOX_INIT_RES_QUEUE */
ISPOPMAP(0x0f, 0x0f), /* 0x12: MBOX_EXECUTE_IOCB */
ISPOPMAP(0x03, 0x03), /* 0x13: MBOX_WAKE_UP */
ISPOPMAP(0x01, 0x3f), /* 0x14: MBOX_STOP_FIRMWARE */
ISPOPMAP(0x0f, 0x0f), /* 0x15: MBOX_ABORT */
ISPOPMAP(0x03, 0x03), /* 0x16: MBOX_ABORT_DEVICE */
ISPOPMAP(0x07, 0x07), /* 0x17: MBOX_ABORT_TARGET */
ISPOPMAP(0x07, 0x07), /* 0x18: MBOX_BUS_RESET */
ISPOPMAP(0x03, 0x07), /* 0x19: MBOX_STOP_QUEUE */
ISPOPMAP(0x03, 0x07), /* 0x1a: MBOX_START_QUEUE */
ISPOPMAP(0x03, 0x07), /* 0x1b: MBOX_SINGLE_STEP_QUEUE */
ISPOPMAP(0x03, 0x07), /* 0x1c: MBOX_ABORT_QUEUE */
ISPOPMAP(0x03, 0x4f), /* 0x1d: MBOX_GET_DEV_QUEUE_STATUS */
ISPOPMAP(0x00, 0x00), /* 0x1e: */
ISPOPMAP(0x01, 0x07), /* 0x1f: MBOX_GET_FIRMWARE_STATUS */
ISPOPMAP(0x01, 0x07), /* 0x20: MBOX_GET_INIT_SCSI_ID */
ISPOPMAP(0x01, 0x07), /* 0x21: MBOX_GET_SELECT_TIMEOUT */
ISPOPMAP(0x01, 0xc7), /* 0x22: MBOX_GET_RETRY_COUNT */
ISPOPMAP(0x01, 0x07), /* 0x23: MBOX_GET_TAG_AGE_LIMIT */
ISPOPMAP(0x01, 0x03), /* 0x24: MBOX_GET_CLOCK_RATE */
ISPOPMAP(0x01, 0x07), /* 0x25: MBOX_GET_ACT_NEG_STATE */
ISPOPMAP(0x01, 0x07), /* 0x26: MBOX_GET_ASYNC_DATA_SETUP_TIME */
ISPOPMAP(0x01, 0x07), /* 0x27: MBOX_GET_PCI_PARAMS */
ISPOPMAP(0x03, 0x4f), /* 0x28: MBOX_GET_TARGET_PARAMS */
ISPOPMAP(0x03, 0x0f), /* 0x29: MBOX_GET_DEV_QUEUE_PARAMS */
ISPOPMAP(0x01, 0x07), /* 0x2a: MBOX_GET_RESET_DELAY_PARAMS */
ISPOPMAP(0x00, 0x00), /* 0x2b: */
ISPOPMAP(0x00, 0x00), /* 0x2c: */
ISPOPMAP(0x00, 0x00), /* 0x2d: */
ISPOPMAP(0x00, 0x00), /* 0x2e: */
ISPOPMAP(0x00, 0x00), /* 0x2f: */
ISPOPMAP(0x03, 0x03), /* 0x30: MBOX_SET_INIT_SCSI_ID */
ISPOPMAP(0x07, 0x07), /* 0x31: MBOX_SET_SELECT_TIMEOUT */
ISPOPMAP(0xc7, 0xc7), /* 0x32: MBOX_SET_RETRY_COUNT */
ISPOPMAP(0x07, 0x07), /* 0x33: MBOX_SET_TAG_AGE_LIMIT */
ISPOPMAP(0x03, 0x03), /* 0x34: MBOX_SET_CLOCK_RATE */
ISPOPMAP(0x07, 0x07), /* 0x35: MBOX_SET_ACT_NEG_STATE */
ISPOPMAP(0x07, 0x07), /* 0x36: MBOX_SET_ASYNC_DATA_SETUP_TIME */
ISPOPMAP(0x07, 0x07), /* 0x37: MBOX_SET_PCI_CONTROL_PARAMS */
ISPOPMAP(0x4f, 0x4f), /* 0x38: MBOX_SET_TARGET_PARAMS */
ISPOPMAP(0x0f, 0x0f), /* 0x39: MBOX_SET_DEV_QUEUE_PARAMS */
ISPOPMAP(0x07, 0x07), /* 0x3a: MBOX_SET_RESET_DELAY_PARAMS */
ISPOPMAP(0x00, 0x00), /* 0x3b: */
ISPOPMAP(0x00, 0x00), /* 0x3c: */
ISPOPMAP(0x00, 0x00), /* 0x3d: */
ISPOPMAP(0x00, 0x00), /* 0x3e: */
ISPOPMAP(0x00, 0x00), /* 0x3f: */
ISPOPMAP(0x01, 0x03), /* 0x40: MBOX_RETURN_BIOS_BLOCK_ADDR */
ISPOPMAP(0x3f, 0x01), /* 0x41: MBOX_WRITE_FOUR_RAM_WORDS */
ISPOPMAP(0x03, 0x07), /* 0x42: MBOX_EXEC_BIOS_IOCB */
ISPOPMAP(0x00, 0x00), /* 0x43: */
ISPOPMAP(0x00, 0x00), /* 0x44: */
ISPOPMAP(0x03, 0x03), /* 0x45: SET SYSTEM PARAMETER */
ISPOPMAP(0x01, 0x03), /* 0x46: GET SYSTEM PARAMETER */
ISPOPMAP(0x00, 0x00), /* 0x47: */
ISPOPMAP(0x01, 0xcf), /* 0x48: GET SCAM CONFIGURATION */
ISPOPMAP(0xcf, 0xcf), /* 0x49: SET SCAM CONFIGURATION */
ISPOPMAP(0x03, 0x03), /* 0x4a: MBOX_SET_FIRMWARE_FEATURES */
ISPOPMAP(0x01, 0x03), /* 0x4b: MBOX_GET_FIRMWARE_FEATURES */
ISPOPMAP(0x00, 0x00), /* 0x4c: */
ISPOPMAP(0x00, 0x00), /* 0x4d: */
ISPOPMAP(0x00, 0x00), /* 0x4e: */
ISPOPMAP(0x00, 0x00), /* 0x4f: */
ISPOPMAP(0xdf, 0xdf), /* 0x50: LOAD RAM A64 */
ISPOPMAP(0xdf, 0xdf), /* 0x51: DUMP RAM A64 */
If we have a 1240 or an ULTRA2 or better card, use MBOX_INIT_RES_QUEUE_A64 (preparation for DAC/A64 support)
2002-09-23 04:59:42 +00:00
ISPOPMAP(0xdf, 0xff), /* 0x52: INITIALIZE REQUEST QUEUE A64 */
ISPOPMAP(0xef, 0xff), /* 0x53: INITIALIZE RESPONSE QUEUE A64 */
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISPOPMAP(0xcf, 0x01), /* 0x54: EXECUCUTE COMMAND IOCB A64 */
After loading f/w, for FC cards print out Firmware Attributes. Redo establishment of default SCSI parameters whether or not we've been compiled for target mode. Unfortunately, the Qlogic f/w is confused so that if we set all targets to be 'safe' (i.e., narrow/async), it will also then report narrow, async if we're contacted in target mode from that target (acting in initiator role). D'oh! Fix ISPCTL_TOGGLE_TMODE to correctly enable the right channel for dual channel cards. Add some more opcodes. Fix a stupid NULL pointer bug.
2001-04-04 21:42:59 +00:00
ISPOPMAP(0x07, 0x01), /* 0x55: ENABLE TARGET MODE */
ISPOPMAP(0x03, 0x0f), /* 0x56: GET TARGET STATUS */
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
ISPOPMAP(0x00, 0x00), /* 0x57: */
ISPOPMAP(0x00, 0x00), /* 0x58: */
ISPOPMAP(0x00, 0x00), /* 0x59: */
ISPOPMAP(0x03, 0x03), /* 0x5a: SET DATA OVERRUN RECOVERY MODE */
ISPOPMAP(0x01, 0x03), /* 0x5b: GET DATA OVERRUN RECOVERY MODE */
ISPOPMAP(0x0f, 0x0f), /* 0x5c: SET HOST DATA */
ISPOPMAP(0x01, 0x01) /* 0x5d: GET NOST DATA */
};
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
static const char *scsi_mbcmd_names[] = {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
"NO-OP",
"LOAD RAM",
"EXEC FIRMWARE",
"DUMP RAM",
"WRITE RAM WORD",
"READ RAM WORD",
"MAILBOX REG TEST",
"VERIFY CHECKSUM",
"ABOUT FIRMWARE",
NULL,
NULL,
NULL,
NULL,
NULL,
"CHECK FIRMWARE",
NULL,
"INIT REQUEST QUEUE",
"INIT RESULT QUEUE",
"EXECUTE IOCB",
"WAKE UP",
"STOP FIRMWARE",
"ABORT",
"ABORT DEVICE",
"ABORT TARGET",
"BUS RESET",
"STOP QUEUE",
"START QUEUE",
"SINGLE STEP QUEUE",
"ABORT QUEUE",
"GET DEV QUEUE STATUS",
NULL,
"GET FIRMWARE STATUS",
"GET INIT SCSI ID",
"GET SELECT TIMEOUT",
"GET RETRY COUNT",
"GET TAG AGE LIMIT",
"GET CLOCK RATE",
"GET ACT NEG STATE",
"GET ASYNC DATA SETUP TIME",
"GET PCI PARAMS",
"GET TARGET PARAMS",
"GET DEV QUEUE PARAMS",
"GET RESET DELAY PARAMS",
NULL,
NULL,
NULL,
NULL,
NULL,
"SET INIT SCSI ID",
"SET SELECT TIMEOUT",
"SET RETRY COUNT",
"SET TAG AGE LIMIT",
"SET CLOCK RATE",
"SET ACT NEG STATE",
"SET ASYNC DATA SETUP TIME",
"SET PCI CONTROL PARAMS",
"SET TARGET PARAMS",
"SET DEV QUEUE PARAMS",
"SET RESET DELAY PARAMS",
NULL,
NULL,
NULL,
NULL,
NULL,
"RETURN BIOS BLOCK ADDR",
"WRITE FOUR RAM WORDS",
"EXEC BIOS IOCB",
NULL,
NULL,
"SET SYSTEM PARAMETER",
"GET SYSTEM PARAMETER",
NULL,
"GET SCAM CONFIGURATION",
"SET SCAM CONFIGURATION",
"SET FIRMWARE FEATURES",
"GET FIRMWARE FEATURES",
NULL,
NULL,
NULL,
NULL,
"LOAD RAM A64",
"DUMP RAM A64",
"INITIALIZE REQUEST QUEUE A64",
"INITIALIZE RESPONSE QUEUE A64",
"EXECUTE IOCB A64",
"ENABLE TARGET MODE",
After loading f/w, for FC cards print out Firmware Attributes. Redo establishment of default SCSI parameters whether or not we've been compiled for target mode. Unfortunately, the Qlogic f/w is confused so that if we set all targets to be 'safe' (i.e., narrow/async), it will also then report narrow, async if we're contacted in target mode from that target (acting in initiator role). D'oh! Fix ISPCTL_TOGGLE_TMODE to correctly enable the right channel for dual channel cards. Add some more opcodes. Fix a stupid NULL pointer bug.
2001-04-04 21:42:59 +00:00
"GET TARGET MODE STATE",
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
NULL,
NULL,
NULL,
"SET DATA OVERRUN RECOVERY MODE",
"GET DATA OVERRUN RECOVERY MODE",
"SET HOST DATA",
"GET NOST DATA",
};
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
static const uint32_t mbpfc[] = {
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
ISPOPMAP(0x01, 0x01), /* 0x00: MBOX_NO_OP */
ISPOPMAP(0x1f, 0x01), /* 0x01: MBOX_LOAD_RAM */
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISPOPMAP(0x0f, 0x01), /* 0x02: MBOX_EXEC_FIRMWARE */
Fix a longstanding bug- we had the sense of what bit 14 for the ICB firmware options meant- *I* had taken it to mean that if you set it, Node Name would be ignored and derived from Port Name. Actually, it meant the opposite. As a consequence- change ICBOPT_USE_PORTNAME to the define ICBOPT_BOTH_WWNS- makes more sense. Fix wrong input bitmap for MBOX_DUMP_RAM command. Call ISP_DUMPREGS if we get a f/w crash. Add ISPCTL_RUN_MBOXCMD control command (so outer layers can run a mailbox command directly) and add a ISPASYNC_UNHANDLED_RESPONSE hook so outer layers can understand response queue entries we might not know about.
2001-02-23 05:35:50 +00:00
ISPOPMAP(0xdf, 0x01), /* 0x03: MBOX_DUMP_RAM */
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
ISPOPMAP(0x07, 0x07), /* 0x04: MBOX_WRITE_RAM_WORD */
ISPOPMAP(0x03, 0x07), /* 0x05: MBOX_READ_RAM_WORD */
ISPOPMAP(0xff, 0xff), /* 0x06: MBOX_MAILBOX_REG_TEST */
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISPOPMAP(0x07, 0x07), /* 0x07: MBOX_VERIFY_CHECKSUM */
Clarify issues about whether we have SCCLUN (65535 luns) or non-SCCLUN (16 luns) firmware for the Fibre Channel cards. We used to assume that if we didn't download firmware, we couldn't know what the firmware capability with respect to SCCLUNs is- and it's important because the lun field changes in the request queue entry based upon which firmware it is. At any rate, we *do* get back firmware attributes in mailbox register 6 when we do ABOUT FIRMWARE for all 2200/2300 cards- and for 2100 cards with at least 1.17.0 firmware. So- we now assume non-SCCLUN behaviour for 2100 cards with firmware < 1.17.0- and we check the firmware attributes for other cards (loaded firmware or not). This also allows us to get rid of the crappy test of isp_maxluns > 16- we simply can check firmware attributes for SCCLUN behaviour. This required an 'oops' fix to the outgoing mailbox count field for ABOUT FIRMWARE for FC cards. Also- while here, hardwire firmware revisions for loaded code for SBus cards. Apparently the 1.35 or 1.37 f/w we've been loading into isp1000 just doesn't report firmware revisions out to mailbox regs 1, 2 and 3 like everyone else. Grumble. Not that this fix hardly matters for FreeBSD. MFC after: 4 weeks
2001-09-03 03:09:48 +00:00
ISPOPMAP(0x01, 0x4f), /* 0x08: MBOX_ABOUT_FIRMWARE */
Try an experiment with using DMA to load firmware into a 2200- VERIFY CHECKSUM fails. Oh well, but keep a couple of the changes. Avoid overflow in usec counters when waiting for mailbox completion.
2006-12-17 16:59:19 +00:00
ISPOPMAP(0xdf, 0x01), /* 0x09: MBOX_LOAD_RISC_RAM_2100 */
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
ISPOPMAP(0xdf, 0x01), /* 0x0a: DUMP RAM */
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISPOPMAP(0x1ff, 0x01), /* 0x0b: MBOX_LOAD_RISC_RAM */
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
ISPOPMAP(0x00, 0x00), /* 0x0c: */
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISPOPMAP(0x10f, 0x01), /* 0x0d: MBOX_WRITE_RAM_WORD_EXTENDED */
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
ISPOPMAP(0x01, 0x05), /* 0x0e: MBOX_CHECK_FIRMWARE */
Fix 2KLOGIN code to specify *ibits* (not *obits*) so that the options field in register 10 will be deterministic, not random. Correct the number of input bits for EXECUTE_FIRMWARE 0..1 to 0..2- the 2322 and 24XX cards use mailbox register 2 to specify whether the f/w being executed is freshly loaded or not. Correct the number of input bits for {READ,WRITE}_RAM_WORD_EXTENDED so that register 8 gets picked up. Fix the indexing and offset for the 2322 f/w download so that it correctly puts the different code segments where they belong. Move VERIFY_CHECKSUM to be the 'else' clause to 2322 f/w downloads- the EXECUTE_FIRMWARE command for 2322 and 24XX cards will tell you if the f/w checksum is incorrect and VERIFY_CHECKSUM only works for RISC SRAM address < 64K so you can only do a VERIFY_CHECKSUM on the first of the 3 f/w segments for the 2322. Shorten the delay for the continuation mailbox commands- 1ms is ridiculous (100us is more likely). All of the more or less is really only for the 2322/6322 cards.
2006-08-14 05:42:46 +00:00
ISPOPMAP(0x10f, 0x05), /* 0x0f: MBOX_READ_RAM_WORD_EXTENDED */
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
ISPOPMAP(0x1f, 0x11), /* 0x10: MBOX_INIT_REQ_QUEUE */
ISPOPMAP(0x2f, 0x21), /* 0x11: MBOX_INIT_RES_QUEUE */
ISPOPMAP(0x0f, 0x01), /* 0x12: MBOX_EXECUTE_IOCB */
ISPOPMAP(0x03, 0x03), /* 0x13: MBOX_WAKE_UP */
ISPOPMAP(0x01, 0xff), /* 0x14: MBOX_STOP_FIRMWARE */
ISPOPMAP(0x4f, 0x01), /* 0x15: MBOX_ABORT */
ISPOPMAP(0x07, 0x01), /* 0x16: MBOX_ABORT_DEVICE */
ISPOPMAP(0x07, 0x01), /* 0x17: MBOX_ABORT_TARGET */
ISPOPMAP(0x03, 0x03), /* 0x18: MBOX_BUS_RESET */
ISPOPMAP(0x07, 0x05), /* 0x19: MBOX_STOP_QUEUE */
ISPOPMAP(0x07, 0x05), /* 0x1a: MBOX_START_QUEUE */
ISPOPMAP(0x07, 0x05), /* 0x1b: MBOX_SINGLE_STEP_QUEUE */
ISPOPMAP(0x07, 0x05), /* 0x1c: MBOX_ABORT_QUEUE */
ISPOPMAP(0x07, 0x03), /* 0x1d: MBOX_GET_DEV_QUEUE_STATUS */
ISPOPMAP(0x00, 0x00), /* 0x1e: */
ISPOPMAP(0x01, 0x07), /* 0x1f: MBOX_GET_FIRMWARE_STATUS */
ISPOPMAP(0x01, 0x4f), /* 0x20: MBOX_GET_LOOP_ID */
ISPOPMAP(0x00, 0x00), /* 0x21: */
ISPOPMAP(0x01, 0x07), /* 0x22: MBOX_GET_RETRY_COUNT */
ISPOPMAP(0x00, 0x00), /* 0x23: */
ISPOPMAP(0x00, 0x00), /* 0x24: */
ISPOPMAP(0x00, 0x00), /* 0x25: */
ISPOPMAP(0x00, 0x00), /* 0x26: */
ISPOPMAP(0x00, 0x00), /* 0x27: */
Implement a call to get the actual link data rate (if 23XX) so we can set whether it's a 2Gps or 1Gps link. MFC after: 1 week
2001-10-01 03:45:54 +00:00
ISPOPMAP(0x01, 0x03), /* 0x28: MBOX_GET_FIRMWARE_OPTIONS */
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
ISPOPMAP(0x03, 0x07), /* 0x29: MBOX_GET_PORT_QUEUE_PARAMS */
ISPOPMAP(0x00, 0x00), /* 0x2a: */
ISPOPMAP(0x00, 0x00), /* 0x2b: */
ISPOPMAP(0x00, 0x00), /* 0x2c: */
ISPOPMAP(0x00, 0x00), /* 0x2d: */
ISPOPMAP(0x00, 0x00), /* 0x2e: */
ISPOPMAP(0x00, 0x00), /* 0x2f: */
ISPOPMAP(0x00, 0x00), /* 0x30: */
ISPOPMAP(0x00, 0x00), /* 0x31: */
ISPOPMAP(0x07, 0x07), /* 0x32: MBOX_SET_RETRY_COUNT */
ISPOPMAP(0x00, 0x00), /* 0x33: */
ISPOPMAP(0x00, 0x00), /* 0x34: */
ISPOPMAP(0x00, 0x00), /* 0x35: */
ISPOPMAP(0x00, 0x00), /* 0x36: */
ISPOPMAP(0x00, 0x00), /* 0x37: */
ISPOPMAP(0x0f, 0x01), /* 0x38: MBOX_SET_FIRMWARE_OPTIONS */
ISPOPMAP(0x0f, 0x07), /* 0x39: MBOX_SET_PORT_QUEUE_PARAMS */
ISPOPMAP(0x00, 0x00), /* 0x3a: */
ISPOPMAP(0x00, 0x00), /* 0x3b: */
ISPOPMAP(0x00, 0x00), /* 0x3c: */
ISPOPMAP(0x00, 0x00), /* 0x3d: */
ISPOPMAP(0x00, 0x00), /* 0x3e: */
ISPOPMAP(0x00, 0x00), /* 0x3f: */
ISPOPMAP(0x03, 0x01), /* 0x40: MBOX_LOOP_PORT_BYPASS */
ISPOPMAP(0x03, 0x01), /* 0x41: MBOX_LOOP_PORT_ENABLE */
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISPOPMAP(0x03, 0x07), /* 0x42: MBOX_GET_RESOURCE_COUNT */
ISPOPMAP(0x01, 0x01), /* 0x43: MBOX_REQUEST_OFFLINE_MODE */
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
ISPOPMAP(0x00, 0x00), /* 0x44: */
ISPOPMAP(0x00, 0x00), /* 0x45: */
ISPOPMAP(0x00, 0x00), /* 0x46: */
ISPOPMAP(0xcf, 0x03), /* 0x47: GET PORT_DATABASE ENHANCED */
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISPOPMAP(0xcd, 0x01), /* 0x48: MBOX_INIT_FIRMWARE_MULTI_ID */
ISPOPMAP(0xcd, 0x01), /* 0x49: MBOX_GET_VP_DATABASE */
ISPOPMAP(0x2cd, 0x01), /* 0x4a: MBOX_GET_VP_DATABASE_ENTRY */
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
ISPOPMAP(0x00, 0x00), /* 0x4b: */
ISPOPMAP(0x00, 0x00), /* 0x4c: */
ISPOPMAP(0x00, 0x00), /* 0x4d: */
ISPOPMAP(0x00, 0x00), /* 0x4e: */
ISPOPMAP(0x00, 0x00), /* 0x4f: */
ISPOPMAP(0x00, 0x00), /* 0x50: */
ISPOPMAP(0x00, 0x00), /* 0x51: */
ISPOPMAP(0x00, 0x00), /* 0x52: */
ISPOPMAP(0x00, 0x00), /* 0x53: */
ISPOPMAP(0xcf, 0x01), /* 0x54: EXECUTE IOCB A64 */
ISPOPMAP(0x00, 0x00), /* 0x55: */
ISPOPMAP(0x00, 0x00), /* 0x56: */
ISPOPMAP(0x00, 0x00), /* 0x57: */
ISPOPMAP(0x00, 0x00), /* 0x58: */
ISPOPMAP(0x00, 0x00), /* 0x59: */
ISPOPMAP(0x00, 0x00), /* 0x5a: */
If we get a DATA UNDERRUN error from QLogic FC cards, but the RQCS_RU bit is not set in the scsi completion status, or if the residual is clearly nonsense, then this was a command that suffered the loss of one or more FC frames in the middle of the exchange. Set HBA_BOTCH and hope it will get retried. It's the only thing we can do. MFC after: 1 day
2002-05-01 21:58:36 +00:00
ISPOPMAP(0x03, 0x01), /* 0x5b: MBOX_DRIVER_HEARTBEAT */
ISPOPMAP(0xcf, 0x01), /* 0x5c: MBOX_FW_HEARTBEAT */
Implement a call to get the actual link data rate (if 23XX) so we can set whether it's a 2Gps or 1Gps link. MFC after: 1 week
2001-10-01 03:45:54 +00:00
ISPOPMAP(0x07, 0x03), /* 0x5d: MBOX_GET_SET_DATA_RATE */
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
ISPOPMAP(0x00, 0x00), /* 0x5e: */
ISPOPMAP(0x00, 0x00), /* 0x5f: */
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISPOPMAP(0xcd, 0x01), /* 0x60: MBOX_INIT_FIRMWARE */
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
ISPOPMAP(0x00, 0x00), /* 0x61: */
ISPOPMAP(0x01, 0x01), /* 0x62: MBOX_INIT_LIP */
ISPOPMAP(0xcd, 0x03), /* 0x63: MBOX_GET_FC_AL_POSITION_MAP */
ISPOPMAP(0xcf, 0x01), /* 0x64: MBOX_GET_PORT_DB */
ISPOPMAP(0x07, 0x01), /* 0x65: MBOX_CLEAR_ACA */
ISPOPMAP(0x07, 0x01), /* 0x66: MBOX_TARGET_RESET */
ISPOPMAP(0x07, 0x01), /* 0x67: MBOX_CLEAR_TASK_SET */
ISPOPMAP(0x07, 0x01), /* 0x68: MBOX_ABORT_TASK_SET */
ISPOPMAP(0x01, 0x07), /* 0x69: MBOX_GET_FW_STATE */
ISPOPMAP(0x03, 0xcf), /* 0x6a: MBOX_GET_PORT_NAME */
ISPOPMAP(0xcf, 0x01), /* 0x6b: MBOX_GET_LINK_STATUS */
ISPOPMAP(0x0f, 0x01), /* 0x6c: MBOX_INIT_LIP_RESET */
ISPOPMAP(0x00, 0x00), /* 0x6d: */
ISPOPMAP(0xcf, 0x03), /* 0x6e: MBOX_SEND_SNS */
ISPOPMAP(0x0f, 0x07), /* 0x6f: MBOX_FABRIC_LOGIN */
ISPOPMAP(0x03, 0x01), /* 0x70: MBOX_SEND_CHANGE_REQUEST */
ISPOPMAP(0x03, 0x03), /* 0x71: MBOX_FABRIC_LOGOUT */
ISPOPMAP(0x0f, 0x0f), /* 0x72: MBOX_INIT_LIP_LOGIN */
ISPOPMAP(0x00, 0x00), /* 0x73: */
ISPOPMAP(0x07, 0x01), /* 0x74: LOGIN LOOP PORT */
ISPOPMAP(0xcf, 0x03), /* 0x75: GET PORT/NODE NAME LIST */
ISPOPMAP(0x4f, 0x01), /* 0x76: SET VENDOR ID */
ISPOPMAP(0xcd, 0x01), /* 0x77: INITIALIZE IP MAILBOX */
ISPOPMAP(0x00, 0x00), /* 0x78: */
ISPOPMAP(0x00, 0x00), /* 0x79: */
ISPOPMAP(0x00, 0x00), /* 0x7a: */
ISPOPMAP(0x00, 0x00), /* 0x7b: */
ISPOPMAP(0x4f, 0x03), /* 0x7c: Get ID List */
ISPOPMAP(0xcf, 0x01), /* 0x7d: SEND LFA */
Add an ioctl framework for doing FC task management functions from a user space tool- useful for doing FC target mode certification.
2005-10-29 02:46:59 +00:00
ISPOPMAP(0x0f, 0x01) /* 0x7e: LUN RESET */
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
};
Support for f/w crash dumps (2200 && 23XX). If you want QLogic to look at a potential f/w problem for FC cards, you really have to provide them info in the format they expect. This involves dumping a lot of hardware registers (> 300 16 bit registers) and a lot of SRAM (> 128KB minimum). Thus all of this code is #ifdef protected which will become an option so that the memory allocation of where to dump the crash image is pretty expensive. It's worth it if you have a reproducible problem because they have some tools that can tell them, given the f/w version, the precise state of everything. MFC after: 1 week
2002-02-17 06:38:22 +00:00
/*
* Footnotes
*
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* (1): this sets bits 21..16 in mailbox register #8, which we nominally
Support for f/w crash dumps (2200 && 23XX). If you want QLogic to look at a potential f/w problem for FC cards, you really have to provide them info in the format they expect. This involves dumping a lot of hardware registers (> 300 16 bit registers) and a lot of SRAM (> 128KB minimum). Thus all of this code is #ifdef protected which will become an option so that the memory allocation of where to dump the crash image is pretty expensive. It's worth it if you have a reproducible problem because they have some tools that can tell them, given the f/w version, the precise state of everything. MFC after: 1 week
2002-02-17 06:38:22 +00:00
* do not access at this time in the core driver. The caller is
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
* responsible for setting this register first (Gross!). The assumption
* is that we won't overflow.
Support for f/w crash dumps (2200 && 23XX). If you want QLogic to look at a potential f/w problem for FC cards, you really have to provide them info in the format they expect. This involves dumping a lot of hardware registers (> 300 16 bit registers) and a lot of SRAM (> 128KB minimum). Thus all of this code is #ifdef protected which will become an option so that the memory allocation of where to dump the crash image is pretty expensive. It's worth it if you have a reproducible problem because they have some tools that can tell them, given the f/w version, the precise state of everything. MFC after: 1 week
2002-02-17 06:38:22 +00:00
*/
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
static const char *fc_mbcmd_names[] = {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
"NO-OP",
"LOAD RAM",
"EXEC FIRMWARE",
"DUMP RAM",
"WRITE RAM WORD",
"READ RAM WORD",
"MAILBOX REG TEST",
"VERIFY CHECKSUM",
"ABOUT FIRMWARE",
"LOAD RAM",
"DUMP RAM",
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
"WRITE RAM WORD EXTENDED",
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
NULL,
Support for f/w crash dumps (2200 && 23XX). If you want QLogic to look at a potential f/w problem for FC cards, you really have to provide them info in the format they expect. This involves dumping a lot of hardware registers (> 300 16 bit registers) and a lot of SRAM (> 128KB minimum). Thus all of this code is #ifdef protected which will become an option so that the memory allocation of where to dump the crash image is pretty expensive. It's worth it if you have a reproducible problem because they have some tools that can tell them, given the f/w version, the precise state of everything. MFC after: 1 week
2002-02-17 06:38:22 +00:00
"READ RAM WORD EXTENDED",
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
"CHECK FIRMWARE",
NULL,
"INIT REQUEST QUEUE",
"INIT RESULT QUEUE",
"EXECUTE IOCB",
"WAKE UP",
"STOP FIRMWARE",
"ABORT",
"ABORT DEVICE",
"ABORT TARGET",
"BUS RESET",
"STOP QUEUE",
"START QUEUE",
"SINGLE STEP QUEUE",
"ABORT QUEUE",
"GET DEV QUEUE STATUS",
NULL,
"GET FIRMWARE STATUS",
"GET LOOP ID",
NULL,
"GET RETRY COUNT",
NULL,
NULL,
NULL,
NULL,
NULL,
"GET FIRMWARE OPTIONS",
"GET PORT QUEUE PARAMS",
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
"SET RETRY COUNT",
NULL,
NULL,
NULL,
NULL,
NULL,
"SET FIRMWARE OPTIONS",
"SET PORT QUEUE PARAMS",
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
"LOOP PORT BYPASS",
"LOOP PORT ENABLE",
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
"GET RESOURCE COUNT",
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
"REQUEST NON PARTICIPATING MODE",
NULL,
NULL,
NULL,
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
"GET PORT DATABASE ENHANCED",
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
"INIT FIRMWARE MULTI ID",
"GET VP DATABASE",
"GET VP DATABASE ENTRY",
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
"EXECUTE IOCB A64",
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
Pick up some compilation warning fixes from NetBSD. If we don't have ISP_FW_CRASH_DUMP defined, we have to do a isp_reinit in the core code- not the platform code- so fix the ISP_CONN_FATAL case.
2003-02-16 01:32:52 +00:00
"DRIVER HEARTBEAT",
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
NULL,
Implement a call to get the actual link data rate (if 23XX) so we can set whether it's a 2Gps or 1Gps link. MFC after: 1 week
2001-10-01 03:45:54 +00:00
"GET/SET DATA RATE",
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
NULL,
NULL,
"INIT FIRMWARE",
NULL,
"INIT LIP",
"GET FC-AL POSITION MAP",
"GET PORT DATABASE",
"CLEAR ACA",
"TARGET RESET",
"CLEAR TASK SET",
"ABORT TASK SET",
"GET FW STATE",
"GET PORT NAME",
"GET LINK STATUS",
"INIT LIP RESET",
NULL,
"SEND SNS",
"FABRIC LOGIN",
"SEND CHANGE REQUEST",
"FABRIC LOGOUT",
"INIT LIP LOGIN",
NULL,
"LOGIN LOOP PORT",
"GET PORT/NODE NAME LIST",
"SET VENDOR ID",
"INITIALIZE IP MAILBOX",
NULL,
NULL,
NULL,
NULL,
"Get ID List",
"SEND LFA",
"Lun RESET"
};
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
static void
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
isp_mboxcmd_qnw(ispsoftc_t *isp, mbreg_t *mbp, int nodelay)
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
{
Remove unused variables Add /* FALLTHROUGH */ Found by: FlexeLint
2003-05-31 19:49:49 +00:00
unsigned int ibits, obits, box, opcode;
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
const uint32_t *mcp;
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
Restore parentheses removed inappropriately in last commit.
2003-06-01 19:01:01 +00:00
if (IS_FC(isp)) {
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
mcp = mbpfc;
Restore parentheses removed inappropriately in last commit.
2003-06-01 19:01:01 +00:00
} else {
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
mcp = mbpscsi;
Restore parentheses removed inappropriately in last commit.
2003-06-01 19:01:01 +00:00
}
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
opcode = mbp->param[0];
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
ibits = HIWRD(mcp[opcode]) & NMBOX_BMASK(isp);
obits = LOWRD(mcp[opcode]) & NMBOX_BMASK(isp);
ibits |= mbp->ibits;
obits |= mbp->obits;
for (box = 0; box < MAX_MAILBOX(isp); box++) {
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
if (ibits & (1 << box)) {
ISP_WRITE(isp, MBOX_OFF(box), mbp->param[box]);
}
Fix a problem where a local loop disk logs out- and we get a PORT LOGGED OUT status. We are, apparently, required to force the f/w to log back in if we want to try and talk to that disk again. This means either issuing a LOGIN LOCAL LOOP PORT mailbox command, or by issuing a LIP. I've elected to issue a LIP because this has a better chance of waking up the disk which clearly just crashed and burned. These should not occur at all. If they do, they should be darned rare. MFC after: 1 week
2002-02-21 01:56:08 +00:00
if (nodelay == 0) {
isp->isp_mboxtmp[box] = mbp->param[box] = 0;
}
}
if (nodelay == 0) {
isp->isp_lastmbxcmd = opcode;
isp->isp_obits = obits;
isp->isp_mboxbsy = 1;
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
ISP_WRITE(isp, BIU2400_HCCR, HCCR_2400_CMD_SET_HOST_INT);
} else {
ISP_WRITE(isp, HCCR, HCCR_CMD_SET_HOST_INT);
}
Fix a problem where a local loop disk logs out- and we get a PORT LOGGED OUT status. We are, apparently, required to force the f/w to log back in if we want to try and talk to that disk again. This means either issuing a LOGIN LOCAL LOOP PORT mailbox command, or by issuing a LIP. I've elected to issue a LIP because this has a better chance of waking up the disk which clearly just crashed and burned. These should not occur at all. If they do, they should be darned rare. MFC after: 1 week
2002-02-21 01:56:08 +00:00
/*
* Oddly enough, if we're not delaying for an answer,
* delay a bit to give the f/w a chance to pick up the
* command.
*/
if (nodelay) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(1000);
Fix a problem where a local loop disk logs out- and we get a PORT LOGGED OUT status. We are, apparently, required to force the f/w to log back in if we want to try and talk to that disk again. This means either issuing a LOGIN LOCAL LOOP PORT mailbox command, or by issuing a LIP. I've elected to issue a LIP because this has a better chance of waking up the disk which clearly just crashed and burned. These should not occur at all. If they do, they should be darned rare. MFC after: 1 week
2002-02-21 01:56:08 +00:00
}
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
static void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(ispsoftc_t *isp, mbreg_t *mbp)
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
{
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
const char *cname, *xname;
char tname[16], mname[16];
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
unsigned int lim, ibits, obits, box, opcode;
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
const uint32_t *mcp;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
if (IS_FC(isp)) {
mcp = mbpfc;
lim = (sizeof (mbpfc) / sizeof (mbpfc[0]));
} else {
mcp = mbpscsi;
lim = (sizeof (mbpscsi) / sizeof (mbpscsi[0]));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
if ((opcode = mbp->param[0]) >= lim) {
mbp->param[0] = MBOX_INVALID_COMMAND;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGERR, "Unknown Command 0x%x", opcode);
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
return;
Remove the Target mode functions until they're in better shape. Implement some suggested compilation cleanups from Eklund. Wire down a hard loop id if we are not on a platform that has the ability to get to a PCI BIOS (it still will float to the ID it gets after a LIP but at least we can try). Clarify that the expanded lun is based upon SCCLUN defines (in f/w).
1998-12-05 01:33:57 +00:00
}
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
ibits = HIWRD(mcp[opcode]) & NMBOX_BMASK(isp);
obits = LOWRD(mcp[opcode]) & NMBOX_BMASK(isp);
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
/*
* Pick up any additional bits that the caller might have set.
*/
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
ibits |= mbp->ibits;
obits |= mbp->obits;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
if (ibits == 0 && obits == 0) {
mbp->param[0] = MBOX_COMMAND_PARAM_ERROR;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGERR, "no parameters for 0x%x", opcode);
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
return;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
/*
* Get exclusive usage of mailbox registers.
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (MBOX_ACQUIRE(isp)) {
mbp->param[0] = MBOX_REGS_BUSY;
goto out;
}
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
for (box = 0; box < MAX_MAILBOX(isp); box++) {
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
if (ibits & (1 << box)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGDEBUG3, "IN mbox %d = 0x%04x", box,
Some rearrangement of headers to minimize diffs with outside of FreeBSD repository and to clean up the license header so as to not pollute the license with file function. Zero all mailbox structures prior to use (just in case). Change the outgoing mailbox count for INIT_FIRMWARE to be correct.
2006-07-16 20:11:50 +00:00
mbp->param[box]);
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
ISP_WRITE(isp, MBOX_OFF(box), mbp->param[box]);
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
}
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
isp->isp_mboxtmp[box] = mbp->param[box] = 0;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
When resetting the Qlogic 2X00 units, reset the FPM (Fibre Protocol Module) and FBM (Fibre Buffer Modules). Also remember to clear the semaphore registers. Tell the RISC processor to not halt on FPM parity errors. Throw out the ISP_CFG_NOINIT silliness and instead go to the use of adapter 'roles' to see whether one completes initialization or not (mostly for Fibre Channel). The ultimate intent, btw, of all of this is to have a warm standby adapter for failover reasons. Because we do roles now, setting of Target Capable Class 3 service parameters in the ICB for the 2x00 cards reflects from role. Also, in isp_start, if we're not supporting an initiator role, we bounce outgoing commands with a Selection Timeout error. Also clean out the TOGGLE_TMODE goop for FC- there is no toggling of target mode like there is for parallel SCSI cards. Do more cleanup with respect to using target ids 0..125 in F-port topologies. Also keep track of things which *were* fabric devices so that when you rescan the fabric you can notify the outer layers when fabric devices go away. Only force a LOGOUT for fabric devices if they're still logged in (i.e., you cat their Port Database entry. Clean up the Get All Next scanning. Finally, use a new tag in the softc to store the opcode for the last mailbox command used so we can report which opcode timed out.
2001-01-15 18:33:08 +00:00
isp->isp_lastmbxcmd = opcode;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
* We assume that we can't overwrite a previous command.
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
isp->isp_obits = obits;
isp->isp_mboxbsy = 1;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
* Set Host Interrupt condition so that RISC will pick up mailbox regs.
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
ISP_WRITE(isp, BIU2400_HCCR, HCCR_2400_CMD_SET_HOST_INT);
} else {
ISP_WRITE(isp, HCCR, HCCR_CMD_SET_HOST_INT);
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
/*
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
* While we haven't finished the command, spin our wheels here.
A wad of changes- prepping for 1080/1240 support (which caused a massive thwank in register layout goop). A different mboxcmd approach. Some PDB change infrastructure. Some better management of loopdown/loopup events (keep them distinct from resource starvation for simq freeze/unfreeze actions).
1999-03-17 05:04:39 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
MBOX_WAIT_COMPLETE(isp, mbp);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
/*
* Did the command time out?
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (mbp->param[0] == MBOX_TIMEOUT) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp->isp_mboxbsy = 0;
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
MBOX_RELEASE(isp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
goto out;
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
* Copy back output registers.
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
for (box = 0; box < MAX_MAILBOX(isp); box++) {
Fix usage of DELAY (SYS_DELAY is the platform independent local define). Fix stupidity wrt checking whether we've gone to LOOP_PDB_RCVD loopstate- it's okay to be greater than this state. D'oh! Protect calls to isp_pdb_sync and isp_fclink_state with IS_FC macros. Completely redo mailbox command routine (in preparation to make this possibly wait rather than poll for completion). Make a major attempt to solve the 'lost interrupt' problem 1. Problem The Qlogic cards would appear to 'lose' interrupts, i.e., a legitimate regular SCSI command placed on the request queue would never complete and the watchdog routine in the driver would eventually wakeup and catch it. This would typically only happen on Alphas, although a couple folks with 700MHz Intel platforms have also seen this. For a long time I thought it was a foulup with f/w negotiations of SYNC and/or WIDE as it always seemed to happen right after the platform it was running on had done a SET TARGET PARAMETERS mailbox command to (re)enable sync && wide (after initially forcing ASYNC/NARROW at startup). However, occasionally, the same thing would also occur for the Fibre Channel cards as well (which, ahem, have no SET TARGET PARAMETERS for transfer mode). After finally putting in a better set of watchdog routines for the platforms for this driver, it seemed to be the case that the command in question (usually a READ CAPACITY) just had up and died- the watchdog routine would catch it after ~10 seconds. For some platforms (NetBSD/OpenBSD)- an ABORT COMMAND mailbox command was sent (which would always fail- indicating that the f/w denied knowledge of this command, i.e., the f/w thought it was a done command). In any case, retrying the command worked. But this whole problem needed to be really fixed. 2. A False Step That Went in The Right Direction The mailbox code was completely rewritten to no longer try and grab the mailbox semaphore register and to try and 'by hand' complete async fast posting completions. It was also rewritten to now have separate in && out bitpatterns for registers to load to start and retrieve to complete. This means that isp_intr now handles mailbox completions. This substantially simplifies the mailbox handling code, and carries things 90% toward getting this to be a non-polled routine for this driver. This did not solve the problem, though. 3. Register Debouncing I saw some comments in some errata sheets and some notes in a Qlogic produced Linux driver (for the Qlogic 2100) that seemed to indicate that debouncing of reads of the mailbox registers might be needed, so I added this. This did not affect the problem. In fact, it made the problem worse for non-2100 cards. 5. Interrupt masking/unmasking The driver *used* to do a substantial amount of masking/unmasking of the interrupt control register. This was done to make sure that the core common code could just assume it would never get pre-empted. This apparently substantially contributed to the lost interrupt problem. The rewrite of the ICR (Interrupt Control Register), which is a separate register from the ISR (Interrupt Status Register) should not have caused any change to interrupt assertions pending. The manual does not state that it will, and the register layout seems to imply that the ICR is just an active route gate. We only enable PCI Interrupts and RISC Interrupts- this should mean that when the f/w asserts a RISC interrupt and (and the ICR allows RISC Interrupts) and we have PCI Interrupts enabled, we should get a PCI interrupt. Apparently this is a latch- not a signal route. Removing this got rid of *most* but not all, lost interrupts. 5. Watchdog Smartening I made sure that the watchdog routine would catch cases where the Qlogic's ISR showed an interrupt assertion. The watchdog routine now calls the interrupt service routine if it sees this. Some additional internal state flags were added so that the watchdog routine could then know whether the command it was in the middle of burying (because we had time it out) was in fact completed by the interrupt service routine. 6. Occasional Constipation Of Commands.. In running some very strenous high IOPs tests (generating about 11000 interrupts/second across one Qlogic 1040, one Qlogic 1080 and one Qlogic 2200 on an Alpha PC164), I found that I would get occasional but regular 'watchdog timeouts' on both the 1080 and the 2100 cards. This is under FreeBSD, and the watchdog timeout routine just marks the command in error and retries it. Invariably, right after this 'watchdog timeout' error, I'd get a command completion for the command that I had thought timed out. That is, I'd get a command completion, but the handle returned by the firmware mapped to no current command. The frequency of this problem is low under such a load- it would usually take an 30 minutes per 'lost' interrupt. I doubled the timeout for commands to see if it just was an edge case of waiting too short a period. This has no effect. I gathered and printed out microtimes for the watchdog completed command and the completion that couldn't find a command- it was always the case that the order of occurrence was "timeout, completion" separated by a time on the order of 100 to 150 ms. This caused me to consider 'firmware constipation' as to be a possible culprit. That is, resubmission of a command to the device that had suffered a watchdog timeout seemed to cause the presumed dead command to show back up. I added code in the watchdog routine that, when first entered for the command, marks the command with a flag, reissues a local timeout call for one second later, but also then issues a MARKER Request Queue entry to the Qlogic f/w. A MARKER entry is used typically after a Bus Reset to cause the f/w to get synchronized with respect to either a Bus, a Nexus or a Target. Since I've added this code, I always now see the occasional watchdog timeout, but the command that was about to be terminated always now seems to be completed after the MARKER entry is issued (and before the timeout extension fires, which would come back and *really* terminate the command).
2000-06-27 19:44:31 +00:00
if (obits & (1 << box)) {
mbp->param[box] = isp->isp_mboxtmp[box];
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGDEBUG3, "OUT mbox %d = 0x%04x", box,
Some rearrangement of headers to minimize diffs with outside of FreeBSD repository and to clean up the license header so as to not pollute the license with file function. Zero all mailbox structures prior to use (just in case). Change the outgoing mailbox count for INIT_FIRMWARE to be correct.
2006-07-16 20:11:50 +00:00
mbp->param[box]);
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
}
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp->isp_mboxbsy = 0;
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
MBOX_RELEASE(isp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
out:
if (mbp->logval == 0 || opcode == MBOX_EXEC_FIRMWARE) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
return;
}
cname = (IS_FC(isp))? fc_mbcmd_names[opcode] : scsi_mbcmd_names[opcode];
if (cname == NULL) {
After loading f/w, for FC cards print out Firmware Attributes. Redo establishment of default SCSI parameters whether or not we've been compiled for target mode. Unfortunately, the Qlogic f/w is confused so that if we set all targets to be 'safe' (i.e., narrow/async), it will also then report narrow, async if we're contacted in target mode from that target (acting in initiator role). D'oh! Fix ISPCTL_TOGGLE_TMODE to correctly enable the right channel for dual channel cards. Add some more opcodes. Fix a stupid NULL pointer bug.
2001-04-04 21:42:59 +00:00
cname = tname;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(tname, sizeof tname, "opcode %x", opcode);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
* Just to be chatty here...
*/
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
xname = NULL;
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
switch (mbp->param[0]) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
case MBOX_COMMAND_COMPLETE:
break;
case MBOX_INVALID_COMMAND:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (mbp->logval & MBLOGMASK(MBOX_COMMAND_COMPLETE)) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
xname = "INVALID COMMAND";
Add some missing braces. Add MEMORY_BARRIER for the few scratch dma ops that were missing them plus add a couple of hi 32 bit dma ops (we could probably allow 64 bit scratch and request/response queue dma now).
2006-07-14 05:14:48 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
case MBOX_HOST_INTERFACE_ERROR:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (mbp->logval & MBLOGMASK(MBOX_HOST_INTERFACE_ERROR)) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
xname = "HOST INTERFACE ERROR";
Add some missing braces. Add MEMORY_BARRIER for the few scratch dma ops that were missing them plus add a couple of hi 32 bit dma ops (we could probably allow 64 bit scratch and request/response queue dma now).
2006-07-14 05:14:48 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
case MBOX_TEST_FAILED:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (mbp->logval & MBLOGMASK(MBOX_TEST_FAILED)) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
xname = "TEST FAILED";
Add some missing braces. Add MEMORY_BARRIER for the few scratch dma ops that were missing them plus add a couple of hi 32 bit dma ops (we could probably allow 64 bit scratch and request/response queue dma now).
2006-07-14 05:14:48 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
case MBOX_COMMAND_ERROR:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (mbp->logval & MBLOGMASK(MBOX_COMMAND_ERROR)) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
xname = "COMMAND ERROR";
Add some missing braces. Add MEMORY_BARRIER for the few scratch dma ops that were missing them plus add a couple of hi 32 bit dma ops (we could probably allow 64 bit scratch and request/response queue dma now).
2006-07-14 05:14:48 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
case MBOX_COMMAND_PARAM_ERROR:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (mbp->logval & MBLOGMASK(MBOX_COMMAND_PARAM_ERROR)) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
xname = "COMMAND PARAMETER ERROR";
Add some missing braces. Add MEMORY_BARRIER for the few scratch dma ops that were missing them plus add a couple of hi 32 bit dma ops (we could probably allow 64 bit scratch and request/response queue dma now).
2006-07-14 05:14:48 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
Slightly cleaner fabric support (whiter whites! redder reds!).. No, seriously- only attempt to logout a previously logged in fabric device. Fix a longstanding bug for aborting overtime commands- handle halves have always been reversed. Clean up some error messages to indicate channel number. Approved:jkh
2000-02-29 05:52:14 +00:00
case MBOX_LOOP_ID_USED:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (mbp->logval & MBLOGMASK(MBOX_LOOP_ID_USED)) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
xname = "LOOP ID ALREADY IN USE";
Add some missing braces. Add MEMORY_BARRIER for the few scratch dma ops that were missing them plus add a couple of hi 32 bit dma ops (we could probably allow 64 bit scratch and request/response queue dma now).
2006-07-14 05:14:48 +00:00
}
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
break;
Slightly cleaner fabric support (whiter whites! redder reds!).. No, seriously- only attempt to logout a previously logged in fabric device. Fix a longstanding bug for aborting overtime commands- handle halves have always been reversed. Clean up some error messages to indicate channel number. Approved:jkh
2000-02-29 05:52:14 +00:00
case MBOX_PORT_ID_USED:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (mbp->logval & MBLOGMASK(MBOX_PORT_ID_USED)) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
xname = "PORT ID ALREADY IN USE";
Add some missing braces. Add MEMORY_BARRIER for the few scratch dma ops that were missing them plus add a couple of hi 32 bit dma ops (we could probably allow 64 bit scratch and request/response queue dma now).
2006-07-14 05:14:48 +00:00
}
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
break;
Slightly cleaner fabric support (whiter whites! redder reds!).. No, seriously- only attempt to logout a previously logged in fabric device. Fix a longstanding bug for aborting overtime commands- handle halves have always been reversed. Clean up some error messages to indicate channel number. Approved:jkh
2000-02-29 05:52:14 +00:00
case MBOX_ALL_IDS_USED:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (mbp->logval & MBLOGMASK(MBOX_ALL_IDS_USED)) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
xname = "ALL LOOP IDS IN USE";
Add some missing braces. Add MEMORY_BARRIER for the few scratch dma ops that were missing them plus add a couple of hi 32 bit dma ops (we could probably allow 64 bit scratch and request/response queue dma now).
2006-07-14 05:14:48 +00:00
}
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
break;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
case MBOX_REGS_BUSY:
xname = "REGISTERS BUSY";
break;
case MBOX_TIMEOUT:
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
xname = "TIMEOUT";
Slightly cleaner fabric support (whiter whites! redder reds!).. No, seriously- only attempt to logout a previously logged in fabric device. Fix a longstanding bug for aborting overtime commands- handle halves have always been reversed. Clean up some error messages to indicate channel number. Approved:jkh
2000-02-29 05:52:14 +00:00
break;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
default:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(mname, sizeof mname, "error 0x%x", mbp->param[0]);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
xname = mname;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
}
Add some missing braces. Add MEMORY_BARRIER for the few scratch dma ops that were missing them plus add a couple of hi 32 bit dma ops (we could probably allow 64 bit scratch and request/response queue dma now).
2006-07-14 05:14:48 +00:00
if (xname) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGALL, "Mailbox Command '%s' failed (%s)",
cname, xname);
Add some missing braces. Add MEMORY_BARRIER for the few scratch dma ops that were missing them plus add a couple of hi 32 bit dma ops (we could probably allow 64 bit scratch and request/response queue dma now).
2006-07-14 05:14:48 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
static void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_fw_state(ispsoftc_t *isp, int chan)
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
{
More code cleanup. Go back to using FULL_LOGIN Fibre Chan if f/w is less than 1.17.0 level. Change where we do the loop database init. Add in the CMD_RQLATER return. Add some register debounce.
1999-08-16 19:59:55 +00:00
if (IS_FC(isp)) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
mbreg_t mbs;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcparam *fcp = FCPARAM(isp, chan);
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, MBOX_GET_FW_STATE, MBLOGALL, 0);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
if (mbs.param[0] == MBOX_COMMAND_COMPLETE) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
fcp->isp_fwstate = mbs.param[1];
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
}
static void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_spi_update(ispsoftc_t *isp, int chan)
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
{
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
int tgt;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
mbreg_t mbs;
sdparam *sdp;
More code cleanup. Go back to using FULL_LOGIN Fibre Chan if f/w is less than 1.17.0 level. Change where we do the loop database init. Add in the CMD_RQLATER return. Add some register debounce.
1999-08-16 19:59:55 +00:00
if (IS_FC(isp)) {
various fixes
2000-08-27 23:38:44 +00:00
/*
* There are no 'per-bus' settings for Fibre Channel.
*/
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
return;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdp = SDPARAM(isp, chan);
sdp->update = 0;
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
for (tgt = 0; tgt < MAX_TARGETS; tgt++) {
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
uint16_t flags, period, offset;
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
int get;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
if (sdp->isp_devparam[tgt].dev_enable == 0) {
various fixes
2000-08-27 23:38:44 +00:00
sdp->isp_devparam[tgt].dev_update = 0;
sdp->isp_devparam[tgt].dev_refresh = 0;
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_prt(isp, ISP_LOGDEBUG0, "skipping target %d bus %d update", tgt, chan);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
continue;
}
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
/*
* If the goal is to update the status of the device,
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records either what's in NVRAM or what the safe defaults would be if we lack NVRAM. Then we rename cur_XXXX to actv_XXXX (these are the currently active settings) and the dev_XXX settings to goal_XXXX (these are the settings which we want cur_XXXX to converge to).
2001-07-30 00:59:06 +00:00
* take what's in goal_flags and try and set the device
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
* toward that. Otherwise, if we're just refreshing the
* current device state, get the current parameters.
*/
various fixes
2000-08-27 23:38:44 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
MBSINIT(&mbs, 0, MBLOGALL, 0);
Some rearrangement of headers to minimize diffs with outside of FreeBSD repository and to clean up the license header so as to not pollute the license with file function. Zero all mailbox structures prior to use (just in case). Change the outgoing mailbox count for INIT_FIRMWARE to be correct.
2006-07-16 20:11:50 +00:00
various fixes
2000-08-27 23:38:44 +00:00
/*
* Refresh overrides set
*/
if (sdp->isp_devparam[tgt].dev_refresh) {
mbs.param[0] = MBOX_GET_TARGET_PARAMS;
get = 1;
} else if (sdp->isp_devparam[tgt].dev_update) {
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
mbs.param[0] = MBOX_SET_TARGET_PARAMS;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
various fixes
2000-08-27 23:38:44 +00:00
/*
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records either what's in NVRAM or what the safe defaults would be if we lack NVRAM. Then we rename cur_XXXX to actv_XXXX (these are the currently active settings) and the dev_XXX settings to goal_XXXX (these are the settings which we want cur_XXXX to converge to).
2001-07-30 00:59:06 +00:00
* Make sure goal_flags has "Renegotiate on Error"
various fixes
2000-08-27 23:38:44 +00:00
* on and "Freeze Queue on Error" off.
*/
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records either what's in NVRAM or what the safe defaults would be if we lack NVRAM. Then we rename cur_XXXX to actv_XXXX (these are the currently active settings) and the dev_XXX settings to goal_XXXX (these are the settings which we want cur_XXXX to converge to).
2001-07-30 00:59:06 +00:00
sdp->isp_devparam[tgt].goal_flags |= DPARM_RENEG;
sdp->isp_devparam[tgt].goal_flags &= ~DPARM_QFRZ;
mbs.param[2] = sdp->isp_devparam[tgt].goal_flags;
More 2300 support prep- the Request/Response in/out pointers are part of the PCI block for the 2300- not software convention usage of the mailbox registers- so we macrosize in/out pointer usage. Only report that a LIP destroyed commands if it actually destroyed commands. Get the chan/tgt/lun order correct. Fix a longstanding stupid bug that caused us to try and issue a command with a tag on Channel B because we were checking the tagged capability for the target against Channel A. A firmware crash is now vectored out to platform specific code as an async event. Some minor formatting tweaks.
2001-07-04 18:42:41 +00:00
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
/*
various fixes
2000-08-27 23:38:44 +00:00
* Insist that PARITY must be enabled
* if SYNC or WIDE is enabled.
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
*/
various fixes
2000-08-27 23:38:44 +00:00
if ((mbs.param[2] & (DPARM_SYNC|DPARM_WIDE)) != 0) {
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
mbs.param[2] |= DPARM_PARITY;
}
various fixes
2000-08-27 23:38:44 +00:00
Some rearrangement of headers to minimize diffs with outside of FreeBSD repository and to clean up the license header so as to not pollute the license with file function. Zero all mailbox structures prior to use (just in case). Change the outgoing mailbox count for INIT_FIRMWARE to be correct.
2006-07-16 20:11:50 +00:00
if (mbs.param[2] & DPARM_SYNC) {
various fixes
2000-08-27 23:38:44 +00:00
mbs.param[3] =
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records either what's in NVRAM or what the safe defaults would be if we lack NVRAM. Then we rename cur_XXXX to actv_XXXX (these are the currently active settings) and the dev_XXX settings to goal_XXXX (these are the settings which we want cur_XXXX to converge to).
2001-07-30 00:59:06 +00:00
(sdp->isp_devparam[tgt].goal_offset << 8) |
(sdp->isp_devparam[tgt].goal_period);
various fixes
2000-08-27 23:38:44 +00:00
}
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
/*
* A command completion later that has
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records either what's in NVRAM or what the safe defaults would be if we lack NVRAM. Then we rename cur_XXXX to actv_XXXX (these are the currently active settings) and the dev_XXX settings to goal_XXXX (these are the settings which we want cur_XXXX to converge to).
2001-07-30 00:59:06 +00:00
* RQSTF_NEGOTIATION set can cause
various fixes
2000-08-27 23:38:44 +00:00
* the dev_refresh/announce cycle also.
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
*
* Note: It is really important to update our current
* flags with at least the state of TAG capabilities-
* otherwise we might try and send a tagged command
* when we have it all turned off. So change it here
* to say that current already matches goal.
*/
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records either what's in NVRAM or what the safe defaults would be if we lack NVRAM. Then we rename cur_XXXX to actv_XXXX (these are the currently active settings) and the dev_XXX settings to goal_XXXX (these are the settings which we want cur_XXXX to converge to).
2001-07-30 00:59:06 +00:00
sdp->isp_devparam[tgt].actv_flags &= ~DPARM_TQING;
sdp->isp_devparam[tgt].actv_flags |=
(sdp->isp_devparam[tgt].goal_flags & DPARM_TQING);
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_prt(isp, ISP_LOGDEBUG0, "bus %d set tgt %d flags 0x%x off 0x%x period 0x%x",
chan, tgt, mbs.param[2], mbs.param[3] >> 8, mbs.param[3] & 0xff);
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
get = 0;
} else {
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
continue;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
mbs.param[1] = (chan << 15) | (tgt << 8);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
continue;
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
if (get == 0) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdp->sendmarker = 1;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
sdp->isp_devparam[tgt].dev_update = 0;
sdp->isp_devparam[tgt].dev_refresh = 1;
} else {
sdp->isp_devparam[tgt].dev_refresh = 0;
flags = mbs.param[2];
period = mbs.param[3] & 0xff;
offset = mbs.param[3] >> 8;
sdp->isp_devparam[tgt].actv_flags = flags;
sdp->isp_devparam[tgt].actv_period = period;
sdp->isp_devparam[tgt].actv_offset = offset;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_async(isp, ISPASYNC_NEW_TGT_PARAMS, chan, tgt);
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 07:29:00 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
various fixes
2000-08-27 23:38:44 +00:00
for (tgt = 0; tgt < MAX_TARGETS; tgt++) {
if (sdp->isp_devparam[tgt].dev_update ||
sdp->isp_devparam[tgt].dev_refresh) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdp->update = 1;
various fixes
2000-08-27 23:38:44 +00:00
break;
}
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
static void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_setdfltsdparm(ispsoftc_t *isp)
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
{
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
int tgt;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdparam *sdp, *sdp1;
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdp = SDPARAM(isp, 0);
sdp->role = GET_DEFAULT_ROLE(isp, 0);
if (IS_DUALBUS(isp)) {
sdp1 = sdp + 1;
sdp1->role = GET_DEFAULT_ROLE(isp, 1);
} else {
sdp1 = NULL;
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
After loading f/w, for FC cards print out Firmware Attributes. Redo establishment of default SCSI parameters whether or not we've been compiled for target mode. Unfortunately, the Qlogic f/w is confused so that if we set all targets to be 'safe' (i.e., narrow/async), it will also then report narrow, async if we're contacted in target mode from that target (acting in initiator role). D'oh! Fix ISPCTL_TOGGLE_TMODE to correctly enable the right channel for dual channel cards. Add some more opcodes. Fix a stupid NULL pointer bug.
2001-04-04 21:42:59 +00:00
/*
* Establish some default parameters.
*/
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
sdp->isp_cmd_dma_burst_enable = 0;
After loading f/w, for FC cards print out Firmware Attributes. Redo establishment of default SCSI parameters whether or not we've been compiled for target mode. Unfortunately, the Qlogic f/w is confused so that if we set all targets to be 'safe' (i.e., narrow/async), it will also then report narrow, async if we're contacted in target mode from that target (acting in initiator role). D'oh! Fix ISPCTL_TOGGLE_TMODE to correctly enable the right channel for dual channel cards. Add some more opcodes. Fix a stupid NULL pointer bug.
2001-04-04 21:42:59 +00:00
sdp->isp_data_dma_burst_enabl = 1;
sdp->isp_fifo_threshold = 0;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdp->isp_initiator_id = DEFAULT_IID(isp, 0);
After loading f/w, for FC cards print out Firmware Attributes. Redo establishment of default SCSI parameters whether or not we've been compiled for target mode. Unfortunately, the Qlogic f/w is confused so that if we set all targets to be 'safe' (i.e., narrow/async), it will also then report narrow, async if we're contacted in target mode from that target (acting in initiator role). D'oh! Fix ISPCTL_TOGGLE_TMODE to correctly enable the right channel for dual channel cards. Add some more opcodes. Fix a stupid NULL pointer bug.
2001-04-04 21:42:59 +00:00
if (isp->isp_type >= ISP_HA_SCSI_1040) {
sdp->isp_async_data_setup = 9;
} else {
sdp->isp_async_data_setup = 6;
}
sdp->isp_selection_timeout = 250;
sdp->isp_max_queue_depth = MAXISPREQUEST(isp);
sdp->isp_tag_aging = 8;
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
sdp->isp_bus_reset_delay = 5;
/*
* Don't retry selection, busy or queue full automatically- reflect
* these back to us.
*/
sdp->isp_retry_count = 0;
sdp->isp_retry_delay = 0;
After loading f/w, for FC cards print out Firmware Attributes. Redo establishment of default SCSI parameters whether or not we've been compiled for target mode. Unfortunately, the Qlogic f/w is confused so that if we set all targets to be 'safe' (i.e., narrow/async), it will also then report narrow, async if we're contacted in target mode from that target (acting in initiator role). D'oh! Fix ISPCTL_TOGGLE_TMODE to correctly enable the right channel for dual channel cards. Add some more opcodes. Fix a stupid NULL pointer bug.
2001-04-04 21:42:59 +00:00
for (tgt = 0; tgt < MAX_TARGETS; tgt++) {
sdp->isp_devparam[tgt].exc_throttle = ISP_EXEC_THROTTLE;
sdp->isp_devparam[tgt].dev_enable = 1;
}
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
/*
* The trick here is to establish a default for the default (honk!)
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records either what's in NVRAM or what the safe defaults would be if we lack NVRAM. Then we rename cur_XXXX to actv_XXXX (these are the currently active settings) and the dev_XXX settings to goal_XXXX (these are the settings which we want cur_XXXX to converge to).
2001-07-30 00:59:06 +00:00
* state (goal_flags). Then try and get the current status from
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
* the card to fill in the current state. We don't, in fact, set
* the default to the SAFE default state- that's not the goal state.
*/
for (tgt = 0; tgt < MAX_TARGETS; tgt++) {
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
uint8_t off, per;
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records either what's in NVRAM or what the safe defaults would be if we lack NVRAM. Then we rename cur_XXXX to actv_XXXX (these are the currently active settings) and the dev_XXX settings to goal_XXXX (these are the settings which we want cur_XXXX to converge to).
2001-07-30 00:59:06 +00:00
sdp->isp_devparam[tgt].actv_offset = 0;
sdp->isp_devparam[tgt].actv_period = 0;
sdp->isp_devparam[tgt].actv_flags = 0;
sdp->isp_devparam[tgt].goal_flags =
sdp->isp_devparam[tgt].nvrm_flags = DPARM_DEFAULT;
clean up for SBus Ultra (yes, we do not do that here yet)
2000-01-15 01:52:01 +00:00
/*
* We default to Wide/Fast for versions less than a 1040
* (unless it's SBus).
*/
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records either what's in NVRAM or what the safe defaults would be if we lack NVRAM. Then we rename cur_XXXX to actv_XXXX (these are the currently active settings) and the dev_XXX settings to goal_XXXX (these are the settings which we want cur_XXXX to converge to).
2001-07-30 00:59:06 +00:00
if (IS_ULTRA3(isp)) {
off = ISP_80M_SYNCPARMS >> 8;
per = ISP_80M_SYNCPARMS & 0xff;
} else if (IS_ULTRA2(isp)) {
off = ISP_40M_SYNCPARMS >> 8;
per = ISP_40M_SYNCPARMS & 0xff;
} else if (IS_1240(isp)) {
off = ISP_20M_SYNCPARMS >> 8;
per = ISP_20M_SYNCPARMS & 0xff;
} else if ((isp->isp_bustype == ISP_BT_SBUS &&
clean up for SBus Ultra (yes, we do not do that here yet)
2000-01-15 01:52:01 +00:00
isp->isp_type < ISP_HA_SCSI_1020A) ||
(isp->isp_bustype == ISP_BT_PCI &&
isp->isp_type < ISP_HA_SCSI_1040) ||
Fix some breakage about how we build WWNs. Do some other fabric related changes: consider a new PDB entry different if Class 3 service parameter roles change (!!!). Do some checking as we're getting a port database that traps whether things change while we're doing so. Handle N-port and F-ports correctly. Fix the fabric login loop to retain a login/binding if things haven't changed (I mean, why logout a device only to log it back in). No longer accept, after fabric logins, garbage if we can't get a PDB entry that matches the device we've just logged into- if it doesn't, log it out as it is very unlikely to still be what we thought it was. Get rid of some of the debounce loops because we could get stuck there.
2000-05-09 01:14:43 +00:00
(isp->isp_clock && isp->isp_clock < 60) ||
(sdp->isp_ultramode == 0)) {
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records either what's in NVRAM or what the safe defaults would be if we lack NVRAM. Then we rename cur_XXXX to actv_XXXX (these are the currently active settings) and the dev_XXX settings to goal_XXXX (these are the settings which we want cur_XXXX to converge to).
2001-07-30 00:59:06 +00:00
off = ISP_10M_SYNCPARMS >> 8;
per = ISP_10M_SYNCPARMS & 0xff;
various fixes
2000-08-27 23:38:44 +00:00
} else {
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records either what's in NVRAM or what the safe defaults would be if we lack NVRAM. Then we rename cur_XXXX to actv_XXXX (these are the currently active settings) and the dev_XXX settings to goal_XXXX (these are the settings which we want cur_XXXX to converge to).
2001-07-30 00:59:06 +00:00
off = ISP_20M_SYNCPARMS_1040 >> 8;
per = ISP_20M_SYNCPARMS_1040 & 0xff;
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 01:07:06 +00:00
}
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records either what's in NVRAM or what the safe defaults would be if we lack NVRAM. Then we rename cur_XXXX to actv_XXXX (these are the currently active settings) and the dev_XXX settings to goal_XXXX (these are the settings which we want cur_XXXX to converge to).
2001-07-30 00:59:06 +00:00
sdp->isp_devparam[tgt].goal_offset =
sdp->isp_devparam[tgt].nvrm_offset = off;
sdp->isp_devparam[tgt].goal_period =
Oops- don't set 'goal' twice when you mean to set 'nvrm' as well. This breaks bogus NVRAM boards. MFC after: 1 day
2001-08-02 00:34:56 +00:00
sdp->isp_devparam[tgt].nvrm_period = per;
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 01:07:06 +00:00
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* If we're a dual bus card, just copy the data over
*/
if (sdp1) {
*sdp1 = *sdp;
sdp1->isp_initiator_id = DEFAULT_IID(isp, 1);
}
/*
* If we've not been told to avoid reading NVRAM, try and read it.
* If we're successful reading it, we can then return because NVRAM
* will tell us what the desired settings are. Otherwise, we establish
* some reasonable 'fake' nvram and goal defaults.
*/
if ((isp->isp_confopts & ISP_CFG_NONVRAM) == 0) {
mbreg_t mbs;
if (isp_read_nvram(isp, 0) == 0) {
if (IS_DUALBUS(isp)) {
if (isp_read_nvram(isp, 1) == 0) {
return;
}
}
}
MBSINIT(&mbs, MBOX_GET_ACT_NEG_STATE, MBLOGNONE, 0);
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
sdp->isp_req_ack_active_neg = 1;
sdp->isp_data_line_active_neg = 1;
if (sdp1) {
sdp1->isp_req_ack_active_neg = 1;
sdp1->isp_data_line_active_neg = 1;
}
} else {
sdp->isp_req_ack_active_neg =
(mbs.param[1] >> 4) & 0x1;
sdp->isp_data_line_active_neg =
(mbs.param[1] >> 5) & 0x1;
if (sdp1) {
sdp1->isp_req_ack_active_neg =
(mbs.param[2] >> 4) & 0x1;
sdp1->isp_data_line_active_neg =
(mbs.param[2] >> 5) & 0x1;
}
}
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
static void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_setdfltfcparm(ispsoftc_t *isp, int chan)
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
{
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcparam *fcp = FCPARAM(isp, chan);
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* Establish some default parameters.
*/
fcp->role = GET_DEFAULT_ROLE(isp, chan);
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
fcp->isp_maxalloc = ICB_DFLT_ALLOC;
fcp->isp_retry_delay = ICB_DFLT_RDELAY;
fcp->isp_retry_count = ICB_DFLT_RCOUNT;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcp->isp_loopid = DEFAULT_LOOPID(isp, chan);
fcp->isp_wwnn_nvram = DEFAULT_NODEWWN(isp, chan);
fcp->isp_wwpn_nvram = DEFAULT_PORTWWN(isp, chan);
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
fcp->isp_fwoptions = 0;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcp->isp_lasthdl = NIL_HANDLE;
if (IS_24XX(isp)) {
fcp->isp_fwoptions |= ICB2400_OPT1_FAIRNESS;
fcp->isp_fwoptions |= ICB2400_OPT1_HARD_ADDRESS;
if (isp->isp_confopts & ISP_CFG_FULL_DUPLEX) {
fcp->isp_fwoptions |= ICB2400_OPT1_FULL_DUPLEX;
}
fcp->isp_fwoptions |= ICB2400_OPT1_BOTH_WWNS;
} else {
fcp->isp_fwoptions |= ICBOPT_FAIRNESS;
fcp->isp_fwoptions |= ICBOPT_PDBCHANGE_AE;
fcp->isp_fwoptions |= ICBOPT_HARD_ADDRESS;
if (isp->isp_confopts & ISP_CFG_FULL_DUPLEX) {
fcp->isp_fwoptions |= ICBOPT_FULL_DUPLEX;
}
/*
* Make sure this is turned off now until we get
* extended options from NVRAM
*/
fcp->isp_fwoptions &= ~ICBOPT_EXTENDED;
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
}
/*
* Now try and read NVRAM unless told to not do so.
* This will set fcparam's isp_wwnn_nvram && isp_wwpn_nvram.
*/
if ((isp->isp_confopts & ISP_CFG_NONVRAM) == 0) {
int i, j = 0;
/*
* Give a couple of tries at reading NVRAM.
*/
for (i = 0; i < 2; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
j = isp_read_nvram(isp, chan);
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
if (j == 0) {
break;
}
}
if (j) {
isp->isp_confopts |= ISP_CFG_NONVRAM;
}
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcp->isp_wwnn = ACTIVE_NODEWWN(isp, chan);
fcp->isp_wwpn = ACTIVE_PORTWWN(isp, chan);
isp_prt(isp, ISP_LOGCONFIG, "Chan %d 0x%08x%08x/0x%08x%08x Role %s",
chan, (uint32_t) (fcp->isp_wwnn >> 32), (uint32_t) (fcp->isp_wwnn),
(uint32_t) (fcp->isp_wwpn >> 32), (uint32_t) (fcp->isp_wwpn),
isp_class3_roles[fcp->role]);
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
}
Add in 1080 LVD support and some basis also for the 1240. The port database printout is now enabled.
1999-03-25 22:52:45 +00:00
/*
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
* Re-initialize the ISP and complete all orphaned commands
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
* with a 'botched' notice. The reset/init routines should
* not disturb an already active list of commands.
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
*/
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_reinit(ispsoftc_t *isp, int do_load_defaults)
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
{
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
int i;
isp_reset(isp, do_load_defaults);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
Minor stuff: Remove ISP2100_FABRIC defines- we always handle fabric now. Insert isp_getmap helper function (for getting Loop Position map). Make sure we (for our own benefit) mark req_state_flags with RQSF_GOT_SENSE for Fibre Channel if we got sense data- the !*$)!*$)~*$)*$ Qlogic f/w doesn't do so. Add ISPCTL_SCAN_FABRIC, ISPCTL_SCAN_LOOP, ISPCTL_SEND_LIP, and ISPCTL_GET_POSMAP isp_control functions. Correctly send async notifications upstream for changes in the name server, changes in the port database, and f/w crashes. Correctly set topology when we get a ASYNC_PTPMODE event. Major stuff: Quite massively redo how we handle Loop events- we've now added several intermediate states between LOOP_PDB_RCVD and LOOP_READY. This allows us a lot finer control about how we scan fabric, whether we go further than scanning fabric, how we look at the local loop, and whether we merge entries at the level or not. This is the next to last step for moving managing loop state out of the core module entirely (whereupon loop && fabric events will simply freeze the command queue and a thread will run to figure out what's changed and *it* will re-enable the queu). This fine amount of control also gets us closer to having an external policy engine decide which fabric devices we really want to log into.
2001-02-11 03:44:43 +00:00
if (isp->isp_state != ISP_RESETSTATE) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_prt(isp, ISP_LOGERR, "%s: cannot reset card", __func__);
ISP_DISABLE_INTS(isp);
goto cleanup;
}
isp_init(isp);
if (isp->isp_state == ISP_INITSTATE) {
isp->isp_state = ISP_RUNSTATE;
}
if (isp->isp_state != ISP_RUNSTATE) {
#ifndef ISP_TARGET_MODE
isp_prt(isp, ISP_LOGWARN, "%s: not at runstate", __func__);
#endif
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_DISABLE_INTS(isp);
if (IS_FC(isp)) {
/*
* If we're in ISP_ROLE_NONE, turn off the lasers.
*/
if (!IS_24XX(isp)) {
ISP_WRITE(isp, BIU2100_CSR, BIU2100_FPM0_REGS);
ISP_WRITE(isp, FPM_DIAG_CONFIG, FPM_SOFT_RESET);
ISP_WRITE(isp, BIU2100_CSR, BIU2100_FB_REGS);
ISP_WRITE(isp, FBM_CMD, FBMCMD_FIFO_RESET_ALL);
ISP_WRITE(isp, BIU2100_CSR, BIU2100_RISC_REGS);
}
}
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
cleanup:
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp->isp_nactive = 0;
isp_clear_commands(isp);
if (IS_FC(isp)) {
for (i = 0; i < isp->isp_nchan; i++) {
ISP_MARK_PORTDB(isp, i, -1);
Remove some target mode stuff. It will get re-introduced in a different file later. Do some pencil-sharpening types of minor changes. Change how active commands are remembered (using new inline functions to get handles, etc..). Now do a GET FIRMWARE STATUS after firing up the f/w as outgoing mailbox 2 will tell you the f/w's notion of the max commands that can be supported. Attempt to retrieve loop topology. Add in the appropriate SWIZZLE/UNSWIZZLE macros calls (this is a no-op on Little Endian machines but is needed for sparc (on other platforms)). Move the temp port database we use to find out where things have moved to after a LIP to the softc and off the kernel stack. Follow Qlogic's hint and don't bother setting a tag for commands that don't have this enabled (presumably the f/w will do it's own selection then). Use an INT_PENDING macro to check for an interrupt. The call to ISP_DMAFREE now just takes the handle- not the 'handle-1' which was a layering violation. Use CFGPRINTF in a couple of places to make things less chatty if not booting verbose, or CAMDEBUG compiles, etc..
1999-10-17 18:58:22 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
/*
* NVRAM Routines
*/
static int
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_read_nvram(ispsoftc_t *isp, int bus)
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
{
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
int i, amt, retval;
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
uint8_t csum, minversion;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
union {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
uint8_t _x[ISP2400_NVRAM_SIZE];
uint16_t _s[ISP2400_NVRAM_SIZE>>1];
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
} _n;
#define nvram_data _n._x
#define nvram_words _n._s
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (IS_24XX(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
return (isp_read_nvram_2400(isp, nvram_data));
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else if (IS_FC(isp)) {
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
amt = ISP2100_NVRAM_SIZE;
minversion = 1;
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
} else if (IS_ULTRA2(isp)) {
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
amt = ISP1080_NVRAM_SIZE;
minversion = 0;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
} else {
amt = ISP_NVRAM_SIZE;
minversion = 2;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
for (i = 0; i < amt>>1; i++) {
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
isp_rdnvram_word(isp, i, &nvram_words[i]);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
if (nvram_data[0] != 'I' || nvram_data[1] != 'S' ||
nvram_data[2] != 'P') {
if (isp->isp_bustype != ISP_BT_SBUS) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGWARN, "invalid NVRAM header");
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_prt(isp, ISP_LOGDEBUG0, "%x %x %x", nvram_data[0], nvram_data[1], nvram_data[2]);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
retval = -1;
goto out;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
for (csum = 0, i = 0; i < amt; i++) {
csum += nvram_data[i];
}
if (csum != 0) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGWARN, "invalid NVRAM checksum");
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
retval = -1;
goto out;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
if (ISP_NVRAM_VERSION(nvram_data) < minversion) {
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
isp_prt(isp, ISP_LOGWARN, "version %d NVRAM not understood",
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
ISP_NVRAM_VERSION(nvram_data));
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
retval = -1;
goto out;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
if (IS_ULTRA3(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_parse_nvram_12160(isp, bus, nvram_data);
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
} else if (IS_1080(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_parse_nvram_1080(isp, bus, nvram_data);
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
} else if (IS_1280(isp) || IS_1240(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_parse_nvram_1080(isp, bus, nvram_data);
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
} else if (IS_SCSI(isp)) {
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
isp_parse_nvram_1020(isp, nvram_data);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
} else {
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
isp_parse_nvram_2100(isp, nvram_data);
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 01:07:06 +00:00
}
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
retval = 0;
out:
return (retval);
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
#undef nvram_data
#undef nvram_words
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
static int
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_read_nvram_2400(ispsoftc_t *isp, uint8_t *nvram_data)
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
{
int retval = 0;
uint32_t addr, csum, lwrds, *dptr;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (isp->isp_port) {
addr = ISP2400_NVRAM_PORT1_ADDR;
} else {
addr = ISP2400_NVRAM_PORT0_ADDR;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
dptr = (uint32_t *) nvram_data;
for (lwrds = 0; lwrds < ISP2400_NVRAM_SIZE >> 2; lwrds++) {
isp_rd_2400_nvram(isp, addr++, dptr++);
}
if (nvram_data[0] != 'I' || nvram_data[1] != 'S' ||
nvram_data[2] != 'P') {
Recover from some major omissions/problems with the 24XX port. First, we were never correctly checking for a 24XX Status Type 0 response- that cased us to fall through to evaluate status for commands as if this were a 2100/2200/2300 Status Type 0 response. This is *close*, but not quite the same. This has been reported to be apparent with some wierd lun configuration problems with some arrays. It became glaringly apparent on sparc64 where none of the correct byte swap things were done. Fixing this omission then caused a whole universe shifting debug cycle of endian issues for the 2400. The manual for 24XX f/w turns out to be wrong about the endianness of a couple of entities. The lun and cdb fields for the type 7 request are *not* unconditionally big endian- they happen to be opposite of whatever the endian of the current machine type is. Same with the sense data for the 24XX type 0 response. While we're at it investigate and resolve some NVRAM endian issues. Approved by: re (ken) MFC after: 3 days
2007-07-02 20:08:20 +00:00
isp_prt(isp, ISP_LOGWARN, "invalid NVRAM header (%x %x %x)",
nvram_data[0], nvram_data[1], nvram_data[2]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
retval = -1;
goto out;
}
dptr = (uint32_t *) nvram_data;
for (csum = 0, lwrds = 0; lwrds < ISP2400_NVRAM_SIZE >> 2; lwrds++) {
Recover from some major omissions/problems with the 24XX port. First, we were never correctly checking for a 24XX Status Type 0 response- that cased us to fall through to evaluate status for commands as if this were a 2100/2200/2300 Status Type 0 response. This is *close*, but not quite the same. This has been reported to be apparent with some wierd lun configuration problems with some arrays. It became glaringly apparent on sparc64 where none of the correct byte swap things were done. Fixing this omission then caused a whole universe shifting debug cycle of endian issues for the 2400. The manual for 24XX f/w turns out to be wrong about the endianness of a couple of entities. The lun and cdb fields for the type 7 request are *not* unconditionally big endian- they happen to be opposite of whatever the endian of the current machine type is. Same with the sense data for the 24XX type 0 response. While we're at it investigate and resolve some NVRAM endian issues. Approved by: re (ken) MFC after: 3 days
2007-07-02 20:08:20 +00:00
uint32_t tmp;
ISP_IOXGET_32(isp, &dptr[lwrds], tmp);
csum += tmp;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
if (csum != 0) {
isp_prt(isp, ISP_LOGWARN, "invalid NVRAM checksum");
retval = -1;
goto out;
}
isp_parse_nvram_2400(isp, nvram_data);
out:
return (retval);
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
static void
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
isp_rdnvram_word(ispsoftc_t *isp, int wo, uint16_t *rp)
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
{
int i, cbits;
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
uint16_t bit, rqst, junk;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
ISP_WRITE(isp, BIU_NVRAM, BIU_NVRAM_SELECT);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(10);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
ISP_WRITE(isp, BIU_NVRAM, BIU_NVRAM_SELECT|BIU_NVRAM_CLOCK);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(10);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
if (IS_FC(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
wo &= ((ISP2100_NVRAM_SIZE >> 1) - 1);
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
if (IS_2312(isp) && isp->isp_port) {
wo += 128;
}
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
rqst = (ISP_NVRAM_READ << 8) | wo;
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
cbits = 10;
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
} else if (IS_ULTRA2(isp)) {
Some massive thwunking in initialization to handle dual bus adapters. More massive thwunking to include an XS_CHANNEL value. Some changes of how parameters are reported to outer layers (including bus, e.g.). Yet more stirring around in isp_mboxcmd to try and get it right. Decode of 1080/1240 NVRAM.
1999-05-11 05:06:55 +00:00
wo &= ((ISP1080_NVRAM_SIZE >> 1) - 1);
rqst = (ISP_NVRAM_READ << 8) | wo;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
cbits = 10;
} else {
wo &= ((ISP_NVRAM_SIZE >> 1) - 1);
rqst = (ISP_NVRAM_READ << 6) | wo;
cbits = 8;
}
/*
* Clock the word select request out...
*/
for (i = cbits; i >= 0; i--) {
if ((rqst >> i) & 1) {
bit = BIU_NVRAM_SELECT | BIU_NVRAM_DATAOUT;
} else {
bit = BIU_NVRAM_SELECT;
}
ISP_WRITE(isp, BIU_NVRAM, bit);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(10);
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
junk = ISP_READ(isp, BIU_NVRAM); /* force PCI flush */
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
ISP_WRITE(isp, BIU_NVRAM, bit | BIU_NVRAM_CLOCK);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(10);
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
junk = ISP_READ(isp, BIU_NVRAM); /* force PCI flush */
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
ISP_WRITE(isp, BIU_NVRAM, bit);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(10);
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
junk = ISP_READ(isp, BIU_NVRAM); /* force PCI flush */
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
/*
* Now read the result back in (bits come back in MSB format).
*/
*rp = 0;
for (i = 0; i < 16; i++) {
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
uint16_t rv;
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
*rp <<= 1;
ISP_WRITE(isp, BIU_NVRAM, BIU_NVRAM_SELECT|BIU_NVRAM_CLOCK);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(10);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
rv = ISP_READ(isp, BIU_NVRAM);
if (rv & BIU_NVRAM_DATAIN) {
*rp |= 1;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(10);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
ISP_WRITE(isp, BIU_NVRAM, BIU_NVRAM_SELECT);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(10);
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
junk = ISP_READ(isp, BIU_NVRAM); /* force PCI flush */
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
ISP_WRITE(isp, BIU_NVRAM, 0);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(10);
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
junk = ISP_READ(isp, BIU_NVRAM); /* force PCI flush */
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
ISP_SWIZZLE_NVRAM_WORD(isp, rp);
Update QLogic ISP support for CAM. Add preliminary target mode support. Submitted by: Matthew Jacob <mjacob@feral.com>
1998-09-15 08:42:56 +00:00
}
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
static void
isp_rd_2400_nvram(ispsoftc_t *isp, uint32_t addr, uint32_t *rp)
{
int loops = 0;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
uint32_t base = 0x7ffe0000;
Fix compilation issues found in RELENG_4 port and merge the diffs back to -current to keep versions identical.
2007-03-12 04:54:30 +00:00
uint32_t tmp = 0;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (IS_25XX(isp)) {
base = 0x7ff00000 | 0x48000;
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_WRITE(isp, BIU2400_FLASH_ADDR, base | addr);
for (loops = 0; loops < 5000; loops++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_DELAY(10);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
tmp = ISP_READ(isp, BIU2400_FLASH_ADDR);
Don't attempt to load illegal hard loop addresses into an ICB. This shows up on card restarts, and usually for 2200-2300 cards. What happens is that we start up, attempting to acquire a hard address. We end up instead being an F-port topology, which reports out a loop id of 0xff (or 0xffff for 2K Login f/w). Then, if we restart, we end up telling the card to go off an acquire this loop address, which the card then rejects. Bah. Compilation fixes from Solaris port.
2007-02-23 21:59:21 +00:00
if ((tmp & (1U << 31)) != 0) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
break;
}
}
Don't attempt to load illegal hard loop addresses into an ICB. This shows up on card restarts, and usually for 2200-2300 cards. What happens is that we start up, attempting to acquire a hard address. We end up instead being an F-port topology, which reports out a loop id of 0xff (or 0xffff for 2K Login f/w). Then, if we restart, we end up telling the card to go off an acquire this loop address, which the card then rejects. Bah. Compilation fixes from Solaris port.
2007-02-23 21:59:21 +00:00
if (tmp & (1U << 31)) {
Recover from some major omissions/problems with the 24XX port. First, we were never correctly checking for a 24XX Status Type 0 response- that cased us to fall through to evaluate status for commands as if this were a 2100/2200/2300 Status Type 0 response. This is *close*, but not quite the same. This has been reported to be apparent with some wierd lun configuration problems with some arrays. It became glaringly apparent on sparc64 where none of the correct byte swap things were done. Fixing this omission then caused a whole universe shifting debug cycle of endian issues for the 2400. The manual for 24XX f/w turns out to be wrong about the endianness of a couple of entities. The lun and cdb fields for the type 7 request are *not* unconditionally big endian- they happen to be opposite of whatever the endian of the current machine type is. Same with the sense data for the 24XX type 0 response. While we're at it investigate and resolve some NVRAM endian issues. Approved by: re (ken) MFC after: 3 days
2007-07-02 20:08:20 +00:00
*rp = ISP_READ(isp, BIU2400_FLASH_DATA);
ISP_SWIZZLE_NVRAM_LONG(isp, rp);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else {
*rp = 0xffffffff;
}
}
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
static void
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
isp_parse_nvram_1020(ispsoftc_t *isp, uint8_t *nvram_data)
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
{
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdparam *sdp = SDPARAM(isp, 0);
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
int tgt;
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
sdp->isp_fifo_threshold =
ISP_NVRAM_FIFO_THRESHOLD(nvram_data) |
(ISP_NVRAM_FIFO_THRESHOLD_128(nvram_data) << 2);
Remove the 'bogus registrant' hack for fabric searches. It really turns out that there's something of a hole in our new fabric name server stuff. We ask the name server for entities that have registered as a specific type. That type is FC-SCSI. If the entity hasn't performed a REGISTER FC4 TYPES, the fabric nameserver won't return it. This brings this driver to a bit of a fork in the road as to what the right thing to do is. For servicing the needs of accessing FC-SCSI devices, this method is fine, and to be preferred. It is extremely unlikely we're interested in fabric devices that *don't* register correctly. If I ever get around to adding an FC-IP stack, then asking for devices that have registers as FC-IP types is also the right thing to do. So- asking the fabric nameserver for a specific type is fine, *as long as you are only interested in specific types*. If, on the other hand, you want to create (as for management tool support) a picture of everything on the fabric, this is *not* so fine. There are a large class of FC-SCSI *initiators* who *don't* correctly register, so we never will *see* them. Is this a problem? Yes, but only a little one. If we want to do such management tool support, we should probably run a *different* fabric nameserver query algorithm. Better yet, we should talk to the management nameserver in Brocade switches instead of the standard FC-GS-2 fabric nameserver (which can be unwieldy). Other changes: if we've overrrides marked, don't set some default values from reading NVRAM. This allows us to override things like EXEC throttle without having to ignore NVRAM entirely. MFC after: 1 week
2002-07-08 17:33:37 +00:00
if ((isp->isp_confopts & ISP_CFG_OWNLOOPID) == 0)
sdp->isp_initiator_id =
ISP_NVRAM_INITIATOR_ID(nvram_data);
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
sdp->isp_bus_reset_delay =
ISP_NVRAM_BUS_RESET_DELAY(nvram_data);
sdp->isp_retry_count =
ISP_NVRAM_BUS_RETRY_COUNT(nvram_data);
sdp->isp_retry_delay =
ISP_NVRAM_BUS_RETRY_DELAY(nvram_data);
sdp->isp_async_data_setup =
ISP_NVRAM_ASYNC_DATA_SETUP_TIME(nvram_data);
if (isp->isp_type >= ISP_HA_SCSI_1040) {
if (sdp->isp_async_data_setup < 9) {
sdp->isp_async_data_setup = 9;
}
} else {
if (sdp->isp_async_data_setup != 6) {
sdp->isp_async_data_setup = 6;
}
}
sdp->isp_req_ack_active_neg =
ISP_NVRAM_REQ_ACK_ACTIVE_NEGATION(nvram_data);
sdp->isp_data_line_active_neg =
ISP_NVRAM_DATA_LINE_ACTIVE_NEGATION(nvram_data);
sdp->isp_data_dma_burst_enabl =
ISP_NVRAM_DATA_DMA_BURST_ENABLE(nvram_data);
sdp->isp_cmd_dma_burst_enable =
ISP_NVRAM_CMD_DMA_BURST_ENABLE(nvram_data);
sdp->isp_tag_aging =
ISP_NVRAM_TAG_AGE_LIMIT(nvram_data);
sdp->isp_selection_timeout =
ISP_NVRAM_SELECTION_TIMEOUT(nvram_data);
sdp->isp_max_queue_depth =
ISP_NVRAM_MAX_QUEUE_DEPTH(nvram_data);
various fixes
2000-08-27 23:38:44 +00:00
sdp->isp_fast_mttr = ISP_NVRAM_FAST_MTTR_ENABLE(nvram_data);
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
for (tgt = 0; tgt < MAX_TARGETS; tgt++) {
sdp->isp_devparam[tgt].dev_enable =
ISP_NVRAM_TGT_DEVICE_ENABLE(nvram_data, tgt);
sdp->isp_devparam[tgt].exc_throttle =
ISP_NVRAM_TGT_EXEC_THROTTLE(nvram_data, tgt);
sdp->isp_devparam[tgt].nvrm_offset =
ISP_NVRAM_TGT_SYNC_OFFSET(nvram_data, tgt);
sdp->isp_devparam[tgt].nvrm_period =
ISP_NVRAM_TGT_SYNC_PERIOD(nvram_data, tgt);
Redo how we manage SCSI device settings- have a 3rd flags (nvram) that records either what's in NVRAM or what the safe defaults would be if we lack NVRAM. Then we rename cur_XXXX to actv_XXXX (these are the currently active settings) and the dev_XXX settings to goal_XXXX (these are the settings which we want cur_XXXX to converge to).
2001-07-30 00:59:06 +00:00
/*
* We probably shouldn't lie about this, but it
* it makes it much safer if we limit NVRAM values
* to sanity.
*/
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
if (isp->isp_type < ISP_HA_SCSI_1040) {
/*
* If we're not ultra, we can't possibly
* be a shorter period than this.
*/
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
if (sdp->isp_devparam[tgt].nvrm_period < 0x19) {
sdp->isp_devparam[tgt].nvrm_period = 0x19;
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
}
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
if (sdp->isp_devparam[tgt].nvrm_offset > 0xc) {
sdp->isp_devparam[tgt].nvrm_offset = 0x0c;
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
}
} else {
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
if (sdp->isp_devparam[tgt].nvrm_offset > 0x8) {
sdp->isp_devparam[tgt].nvrm_offset = 0x8;
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
}
}
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
sdp->isp_devparam[tgt].nvrm_flags = 0;
if (ISP_NVRAM_TGT_RENEG(nvram_data, tgt))
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_RENEG;
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_ARQ;
if (ISP_NVRAM_TGT_TQING(nvram_data, tgt))
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_TQING;
if (ISP_NVRAM_TGT_SYNC(nvram_data, tgt))
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_SYNC;
if (ISP_NVRAM_TGT_WIDE(nvram_data, tgt))
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_WIDE;
if (ISP_NVRAM_TGT_PARITY(nvram_data, tgt))
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_PARITY;
if (ISP_NVRAM_TGT_DISC(nvram_data, tgt))
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_DISC;
sdp->isp_devparam[tgt].actv_flags = 0; /* we don't know */
sdp->isp_devparam[tgt].goal_offset =
sdp->isp_devparam[tgt].nvrm_offset;
sdp->isp_devparam[tgt].goal_period =
sdp->isp_devparam[tgt].nvrm_period;
sdp->isp_devparam[tgt].goal_flags =
sdp->isp_devparam[tgt].nvrm_flags;
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
}
}
static void
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
isp_parse_nvram_1080(ispsoftc_t *isp, int bus, uint8_t *nvram_data)
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
{
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdparam *sdp = SDPARAM(isp, bus);
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
int tgt;
More 2300 support prep- the Request/Response in/out pointers are part of the PCI block for the 2300- not software convention usage of the mailbox registers- so we macrosize in/out pointer usage. Only report that a LIP destroyed commands if it actually destroyed commands. Get the chan/tgt/lun order correct. Fix a longstanding stupid bug that caused us to try and issue a command with a tag on Channel B because we were checking the tagged capability for the target against Channel A. A firmware crash is now vectored out to platform specific code as an async event. Some minor formatting tweaks.
2001-07-04 18:42:41 +00:00
sdp->isp_fifo_threshold =
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
ISP1080_NVRAM_FIFO_THRESHOLD(nvram_data);
Remove the 'bogus registrant' hack for fabric searches. It really turns out that there's something of a hole in our new fabric name server stuff. We ask the name server for entities that have registered as a specific type. That type is FC-SCSI. If the entity hasn't performed a REGISTER FC4 TYPES, the fabric nameserver won't return it. This brings this driver to a bit of a fork in the road as to what the right thing to do is. For servicing the needs of accessing FC-SCSI devices, this method is fine, and to be preferred. It is extremely unlikely we're interested in fabric devices that *don't* register correctly. If I ever get around to adding an FC-IP stack, then asking for devices that have registers as FC-IP types is also the right thing to do. So- asking the fabric nameserver for a specific type is fine, *as long as you are only interested in specific types*. If, on the other hand, you want to create (as for management tool support) a picture of everything on the fabric, this is *not* so fine. There are a large class of FC-SCSI *initiators* who *don't* correctly register, so we never will *see* them. Is this a problem? Yes, but only a little one. If we want to do such management tool support, we should probably run a *different* fabric nameserver query algorithm. Better yet, we should talk to the management nameserver in Brocade switches instead of the standard FC-GS-2 fabric nameserver (which can be unwieldy). Other changes: if we've overrrides marked, don't set some default values from reading NVRAM. This allows us to override things like EXEC throttle without having to ignore NVRAM entirely. MFC after: 1 week
2002-07-08 17:33:37 +00:00
if ((isp->isp_confopts & ISP_CFG_OWNLOOPID) == 0)
sdp->isp_initiator_id =
ISP1080_NVRAM_INITIATOR_ID(nvram_data, bus);
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
sdp->isp_bus_reset_delay =
ISP1080_NVRAM_BUS_RESET_DELAY(nvram_data, bus);
sdp->isp_retry_count =
ISP1080_NVRAM_BUS_RETRY_COUNT(nvram_data, bus);
sdp->isp_retry_delay =
ISP1080_NVRAM_BUS_RETRY_DELAY(nvram_data, bus);
sdp->isp_async_data_setup =
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
ISP1080_NVRAM_ASYNC_DATA_SETUP_TIME(nvram_data, bus);
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
sdp->isp_req_ack_active_neg =
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
ISP1080_NVRAM_REQ_ACK_ACTIVE_NEGATION(nvram_data, bus);
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
sdp->isp_data_line_active_neg =
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
ISP1080_NVRAM_DATA_LINE_ACTIVE_NEGATION(nvram_data, bus);
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
sdp->isp_data_dma_burst_enabl =
ISP1080_NVRAM_BURST_ENABLE(nvram_data);
sdp->isp_cmd_dma_burst_enable =
ISP1080_NVRAM_BURST_ENABLE(nvram_data);
sdp->isp_selection_timeout =
ISP1080_NVRAM_SELECTION_TIMEOUT(nvram_data, bus);
sdp->isp_max_queue_depth =
ISP1080_NVRAM_MAX_QUEUE_DEPTH(nvram_data, bus);
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
for (tgt = 0; tgt < MAX_TARGETS; tgt++) {
sdp->isp_devparam[tgt].dev_enable =
ISP1080_NVRAM_TGT_DEVICE_ENABLE(nvram_data, tgt, bus);
sdp->isp_devparam[tgt].exc_throttle =
ISP1080_NVRAM_TGT_EXEC_THROTTLE(nvram_data, tgt, bus);
sdp->isp_devparam[tgt].nvrm_offset =
ISP1080_NVRAM_TGT_SYNC_OFFSET(nvram_data, tgt, bus);
sdp->isp_devparam[tgt].nvrm_period =
ISP1080_NVRAM_TGT_SYNC_PERIOD(nvram_data, tgt, bus);
sdp->isp_devparam[tgt].nvrm_flags = 0;
if (ISP1080_NVRAM_TGT_RENEG(nvram_data, tgt, bus))
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_RENEG;
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_ARQ;
if (ISP1080_NVRAM_TGT_TQING(nvram_data, tgt, bus))
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_TQING;
if (ISP1080_NVRAM_TGT_SYNC(nvram_data, tgt, bus))
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_SYNC;
if (ISP1080_NVRAM_TGT_WIDE(nvram_data, tgt, bus))
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_WIDE;
if (ISP1080_NVRAM_TGT_PARITY(nvram_data, tgt, bus))
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_PARITY;
if (ISP1080_NVRAM_TGT_DISC(nvram_data, tgt, bus))
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_DISC;
sdp->isp_devparam[tgt].actv_flags = 0;
sdp->isp_devparam[tgt].goal_offset =
sdp->isp_devparam[tgt].nvrm_offset;
sdp->isp_devparam[tgt].goal_period =
sdp->isp_devparam[tgt].nvrm_period;
sdp->isp_devparam[tgt].goal_flags =
sdp->isp_devparam[tgt].nvrm_flags;
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
}
}
static void
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
isp_parse_nvram_12160(ispsoftc_t *isp, int bus, uint8_t *nvram_data)
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
{
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
sdparam *sdp = SDPARAM(isp, bus);
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
int tgt;
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
sdp->isp_fifo_threshold =
ISP12160_NVRAM_FIFO_THRESHOLD(nvram_data);
Remove the 'bogus registrant' hack for fabric searches. It really turns out that there's something of a hole in our new fabric name server stuff. We ask the name server for entities that have registered as a specific type. That type is FC-SCSI. If the entity hasn't performed a REGISTER FC4 TYPES, the fabric nameserver won't return it. This brings this driver to a bit of a fork in the road as to what the right thing to do is. For servicing the needs of accessing FC-SCSI devices, this method is fine, and to be preferred. It is extremely unlikely we're interested in fabric devices that *don't* register correctly. If I ever get around to adding an FC-IP stack, then asking for devices that have registers as FC-IP types is also the right thing to do. So- asking the fabric nameserver for a specific type is fine, *as long as you are only interested in specific types*. If, on the other hand, you want to create (as for management tool support) a picture of everything on the fabric, this is *not* so fine. There are a large class of FC-SCSI *initiators* who *don't* correctly register, so we never will *see* them. Is this a problem? Yes, but only a little one. If we want to do such management tool support, we should probably run a *different* fabric nameserver query algorithm. Better yet, we should talk to the management nameserver in Brocade switches instead of the standard FC-GS-2 fabric nameserver (which can be unwieldy). Other changes: if we've overrrides marked, don't set some default values from reading NVRAM. This allows us to override things like EXEC throttle without having to ignore NVRAM entirely. MFC after: 1 week
2002-07-08 17:33:37 +00:00
if ((isp->isp_confopts & ISP_CFG_OWNLOOPID) == 0)
sdp->isp_initiator_id =
ISP12160_NVRAM_INITIATOR_ID(nvram_data, bus);
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
sdp->isp_bus_reset_delay =
ISP12160_NVRAM_BUS_RESET_DELAY(nvram_data, bus);
sdp->isp_retry_count =
ISP12160_NVRAM_BUS_RETRY_COUNT(nvram_data, bus);
sdp->isp_retry_delay =
ISP12160_NVRAM_BUS_RETRY_DELAY(nvram_data, bus);
sdp->isp_async_data_setup =
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
ISP12160_NVRAM_ASYNC_DATA_SETUP_TIME(nvram_data, bus);
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
sdp->isp_req_ack_active_neg =
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
ISP12160_NVRAM_REQ_ACK_ACTIVE_NEGATION(nvram_data, bus);
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
sdp->isp_data_line_active_neg =
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
ISP12160_NVRAM_DATA_LINE_ACTIVE_NEGATION(nvram_data, bus);
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
sdp->isp_data_dma_burst_enabl =
ISP12160_NVRAM_BURST_ENABLE(nvram_data);
sdp->isp_cmd_dma_burst_enable =
ISP12160_NVRAM_BURST_ENABLE(nvram_data);
sdp->isp_selection_timeout =
ISP12160_NVRAM_SELECTION_TIMEOUT(nvram_data, bus);
sdp->isp_max_queue_depth =
ISP12160_NVRAM_MAX_QUEUE_DEPTH(nvram_data, bus);
Clean up some ways in which we set defaults for SCSI cards that do not have valid NVRAM. In particular, we were leaving a retry count set (to retry selection timeouts) when thats not really what we want. Do some constant string additions so that LOGDEBUG0 info is useful across all cards. MFC after: 2 weeks
2001-08-20 17:28:32 +00:00
for (tgt = 0; tgt < MAX_TARGETS; tgt++) {
sdp->isp_devparam[tgt].dev_enable =
ISP12160_NVRAM_TGT_DEVICE_ENABLE(nvram_data, tgt, bus);
sdp->isp_devparam[tgt].exc_throttle =
ISP12160_NVRAM_TGT_EXEC_THROTTLE(nvram_data, tgt, bus);
sdp->isp_devparam[tgt].nvrm_offset =
ISP12160_NVRAM_TGT_SYNC_OFFSET(nvram_data, tgt, bus);
sdp->isp_devparam[tgt].nvrm_period =
ISP12160_NVRAM_TGT_SYNC_PERIOD(nvram_data, tgt, bus);
sdp->isp_devparam[tgt].nvrm_flags = 0;
if (ISP12160_NVRAM_TGT_RENEG(nvram_data, tgt, bus))
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_RENEG;
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_ARQ;
if (ISP12160_NVRAM_TGT_TQING(nvram_data, tgt, bus))
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_TQING;
if (ISP12160_NVRAM_TGT_SYNC(nvram_data, tgt, bus))
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_SYNC;
if (ISP12160_NVRAM_TGT_WIDE(nvram_data, tgt, bus))
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_WIDE;
if (ISP12160_NVRAM_TGT_PARITY(nvram_data, tgt, bus))
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_PARITY;
if (ISP12160_NVRAM_TGT_DISC(nvram_data, tgt, bus))
sdp->isp_devparam[tgt].nvrm_flags |= DPARM_DISC;
sdp->isp_devparam[tgt].actv_flags = 0;
sdp->isp_devparam[tgt].goal_offset =
sdp->isp_devparam[tgt].nvrm_offset;
sdp->isp_devparam[tgt].goal_period =
sdp->isp_devparam[tgt].nvrm_period;
sdp->isp_devparam[tgt].goal_flags =
sdp->isp_devparam[tgt].nvrm_flags;
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
}
}
static void
Some more gratuitous format and name changes. Pull in some target mode changes from a private branch. Pull in some more RELENG_4 compilation changes. A lot of lines changed, but not much content change yet.
2006-04-21 18:30:01 +00:00
isp_parse_nvram_2100(ispsoftc_t *isp, uint8_t *nvram_data)
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
{
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcparam *fcp = FCPARAM(isp, 0);
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
uint64_t wwn;
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
/*
Redo how default Node and Port WWNs are determined (again!). This is so we don't stomp on the differences between ports for a Qlogic 2202.
2000-10-12 23:49:09 +00:00
* There is NVRAM storage for both Port and Node entities-
* but the Node entity appears to be unused on all the cards
* I can find. However, we should account for this being set
* at some point in the future.
*
* Qlogic WWNs have an NAA of 2, but usually nothing shows up in
* bits 48..60. In the case of the 2202, it appears that they do
* use bit 48 to distinguish between the two instances on the card.
* The 2204, which I've never seen, *probably* extends this method.
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
*/
Major whacking for core version 2.0. A major motivator for 2.0 and these changes is that there's now a Solaris port of this driver, so some things in the core version had to change (not much, but some). In order, from the top.....: A lot of error strings are gathered in one place at the head of the file. This caused me to rewrite them to look consistent (with respect to things like 'Port 0x%' and 'Target %d' and 'Loop ID 0x%x'. The major mailbox function, isp_mboxcmd, now takes a third argument, which is a mask that selectively says whether mailbox command failures will be logged. This will substantially reduce a lot of spurious noise from the driver. At the first run through isp_reset we used to try and get the current running firmware's revision by issuing a mailbox command. This would invariably fail on alpha's with anything but a Qlogic 1040 since SRM doesn't *start* the f/w on these cards. Instead, we now see whether we're sitting ROM state before trying to get a running BIOS loaded f/w version. All CFGPRINTF/PRINTF/IDPRINTF macros have been replaced with calls to isp_prt. There are seperate print levels that can be independently set (see ispvar.h), which include debugging, etc. All SYS_DELAY macros are now USEC_DELAY macros. RQUEST_QUEUE_LEN and RESULT_QUEUE_LEN now take ispsoftc as a parameter- the Fibre Channel cards and the Ultra2/Ultra3 cards can have 16 bit request queue entry indices, so we can make a 1024 entry index for them instead of the 256 entries we've had until now. A major change it to fix isp_fclink_test to actually only wait the delay of time specified in the microsecond argument being passed. The problem has always been that a call to isp_mboxcmd to get he current firmware state takes an unknown (sometimes long) amount of time- this is if the firmware is busy doing PLOGIs while we ask it what's up. So, up until now, the usdelay argument has been a joke. The net effect has been that if you boot without being plugged into a good loop or into a switch, you hang. Massively annonying, and hard to fix because the actual time delta was impossible to know from just guessing. Now, using the new GET_NANOTIME macros, a precise and measured amount of USEC_DELAY calls are done so that only the specified usecdelay is allowed to pass. This means that if the initial startup of the firmware if followed by a call from isp_freebsd.c:isp_attach to isp_control(isp, ISP_FCLINK_TEST, &tdelay) where tdelay is 2 * 1000000, no more than two seconds will actually elapse before we leave concluding that the cable is unhooked. Jeez. About time.... Change the ispscsicmd entry point to isp_start, and the XS_CMD_DONE macro to a call to the platform supplied isp_done (sane naming). Limit our size of request queue completions we'll look at at interrupt time. Since we've increased the size of the Request Queue (and the size of the Response Queue proportionally), let's not create an interrupt stack overflow by having to keep a max completion list (forw links are not an option because this is common code with some platforms that don't have link space in their XS_T structures). A limit of 32 is not unreasonable- I doubt there'd be even this many request queue completions at a time- remember, most boards now use fast posting for normal command completion instead of filling out response queue entries. In the isp_mboxcmd cleanup, also create an array of command names so that "ABOUT FIRMWARE" can be printed instead of "CMD #8". Remove the isp_lostcmd function- it's been deprecated for a while. Remove isp_dumpregs- the ISP_DUMPREGS goes to the specific bus register dump fucntion. Various other cleanups.
2000-08-01 06:51:05 +00:00
wwn = ISP2100_NVRAM_PORT_NAME(nvram_data);
Redo how default Node and Port WWNs are determined (again!). This is so we don't stomp on the differences between ports for a Qlogic 2202.
2000-10-12 23:49:09 +00:00
if (wwn) {
isp_prt(isp, ISP_LOGCONFIG, "NVRAM Port WWN 0x%08x%08x",
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
(uint32_t) (wwn >> 32), (uint32_t) (wwn));
Redo how default Node and Port WWNs are determined (again!). This is so we don't stomp on the differences between ports for a Qlogic 2202.
2000-10-12 23:49:09 +00:00
if ((wwn >> 60) == 0) {
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
wwn |= (((uint64_t) 2)<< 60);
Redo how default Node and Port WWNs are determined (again!). This is so we don't stomp on the differences between ports for a Qlogic 2202.
2000-10-12 23:49:09 +00:00
}
}
MFP4: Move default setting to the end of isp_reset instead of the front of isp_init so we can read NVRAM even if we're role ISP_NONE. Prepare for reintroduction of channels (for FC) for N-Port Virtualization. Fix a botch in handle assignment that caused us to nuke one device when a new one arrives and end up with two devices with the same identity in the virtual target mapping table.
2007-01-20 04:00:21 +00:00
fcp->isp_wwpn_nvram = wwn;
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
if (IS_2200(isp) || IS_23XX(isp)) {
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
wwn = ISP2100_NVRAM_NODE_NAME(nvram_data);
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
if (wwn) {
isp_prt(isp, ISP_LOGCONFIG, "NVRAM Node WWN 0x%08x%08x",
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
(uint32_t) (wwn >> 32),
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
(uint32_t) (wwn));
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
if ((wwn >> 60) == 0) {
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
wwn |= (((uint64_t) 2)<< 60);
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
}
Make sure that the WWNN is also created for 2100..2300 cards. MFC after: 1 day
2010-01-03 02:43:46 +00:00
} else {
wwn = fcp->isp_wwpn_nvram & ~((uint64_t) 0xfff << 48);
Fix some breakage about how we build WWNs. Do some other fabric related changes: consider a new PDB entry different if Class 3 service parameter roles change (!!!). Do some checking as we're getting a port database that traps whether things change while we're doing so. Handle N-port and F-ports correctly. Fix the fabric login loop to retain a login/binding if things haven't changed (I mean, why logout a device only to log it back in). No longer accept, after fabric logins, garbage if we can't get a PDB entry that matches the device we've just logged into- if it doesn't, log it out as it is very unlikely to still be what we thought it was. Get rid of some of the debounce loops because we could get stuck there.
2000-05-09 01:14:43 +00:00
}
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
} else {
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
wwn &= ~((uint64_t) 0xfff << 48);
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
}
MFP4: Move default setting to the end of isp_reset instead of the front of isp_init so we can read NVRAM even if we're role ISP_NONE. Prepare for reintroduction of channels (for FC) for N-Port Virtualization. Fix a botch in handle assignment that caused us to nuke one device when a new one arrives and end up with two devices with the same identity in the virtual target mapping table.
2007-01-20 04:00:21 +00:00
fcp->isp_wwnn_nvram = wwn;
Redo how default Node and Port WWNs are determined (again!). This is so we don't stomp on the differences between ports for a Qlogic 2202.
2000-10-12 23:49:09 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
fcp->isp_maxalloc = ISP2100_NVRAM_MAXIOCBALLOCATION(nvram_data);
if ((isp->isp_confopts & ISP_CFG_OWNFSZ) == 0) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
DEFAULT_FRAMESIZE(isp) =
ISP2100_NVRAM_MAXFRAMELENGTH(nvram_data);
Redo how default Node and Port WWNs are determined (again!). This is so we don't stomp on the differences between ports for a Qlogic 2202.
2000-10-12 23:49:09 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
fcp->isp_retry_delay = ISP2100_NVRAM_RETRY_DELAY(nvram_data);
fcp->isp_retry_count = ISP2100_NVRAM_RETRY_COUNT(nvram_data);
if ((isp->isp_confopts & ISP_CFG_OWNLOOPID) == 0) {
fcp->isp_loopid = ISP2100_NVRAM_HARDLOOPID(nvram_data);
}
if ((isp->isp_confopts & ISP_CFG_OWNEXCTHROTTLE) == 0) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
DEFAULT_EXEC_THROTTLE(isp) =
If the HBA is already 'touched', still set maxluns. Othewise for CAM_QUIRK_HILUN devices we loop thru 32bits of lun. Oops. Switch to using USEC_DELAY rather than USEC_SLEEP at isp_reset time. Try to paper around a defect in clients that don't correctly registers themeselves with the fabric nameserver. Minor updates for Mirapoint support- they still use code that is not HANDLE_LOOPSTATE_IN_OUTER_LAYERS, and, surprise surprise, this old stuff had some bugs in it. Clean up some target mode stuff. MFC after: 1 week
2002-06-16 05:18:22 +00:00
ISP2100_NVRAM_EXECUTION_THROTTLE(nvram_data);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
fcp->isp_fwoptions = ISP2100_NVRAM_OPTIONS(nvram_data);
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
isp_prt(isp, ISP_LOGDEBUG0,
"NVRAM 0x%08x%08x 0x%08x%08x maxalloc %d maxframelen %d",
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
(uint32_t) (fcp->isp_wwnn_nvram >> 32),
(uint32_t) fcp->isp_wwnn_nvram,
(uint32_t) (fcp->isp_wwpn_nvram >> 32),
(uint32_t) fcp->isp_wwpn_nvram,
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
ISP2100_NVRAM_MAXIOCBALLOCATION(nvram_data),
ISP2100_NVRAM_MAXFRAMELENGTH(nvram_data));
isp_prt(isp, ISP_LOGDEBUG0,
"execthrottle %d fwoptions 0x%x hardloop %d tov %d",
ISP2100_NVRAM_EXECUTION_THROTTLE(nvram_data),
ISP2100_NVRAM_OPTIONS(nvram_data),
ISP2100_NVRAM_HARDLOOPID(nvram_data),
ISP2100_NVRAM_TOV(nvram_data));
fcp->isp_xfwoptions = ISP2100_XFW_OPTIONS(nvram_data);
fcp->isp_zfwoptions = ISP2100_ZFW_OPTIONS(nvram_data);
Clean up some printing stuff so that we can have a bit finer control on debug output. Add a new platform function requirement to allow for printing based upon the ITL nexus instead of the isp unit plus channel, target and lun. This allows some printouts and error messages from the core code to appear in the same format as the platform's subsystem (in FreeBSD's case, CAM path). MFC after: 1 week
2010-03-26 15:13:31 +00:00
isp_prt(isp, ISP_LOGDEBUG0, "xfwoptions 0x%x zfw options 0x%x",
Do various fixes to support firmware loading for the 2322 (and by extension, the 2422). One peculiar thing I've found with the 2322 is that if you don't force it to do Hard LoopID acquisition, the firmware crashes. This took a while to figure out. While we're at it, fix various bugs having to do with NVRAM reading and option setting with respect to pieces of NVRAM.
2006-07-03 08:24:09 +00:00
ISP2100_XFW_OPTIONS(nvram_data), ISP2100_ZFW_OPTIONS(nvram_data));
Restructure nvram reading routine to split out to separate functions for 1020/1X80/12160/2X00- for readability. Add in 12160 (Ultra3) support- but not with PPR just yet. Fix and clarify fetching of return parameter for getting firmware rev which for the 2200 contains the connection topology (Private Loop (NL-port), N-port, FL-port, F-port). Synthesize the connection topology for the 2100 which can only be Private Loop or FL-port. Handle a couple of new async mailbox commands which signify connection in Point-to-Point mode (N-port or F-port) or indicate various toe stubbing getting to same. Approved: jkh@freebsd.org
2000-02-11 19:31:32 +00:00
}
Support for f/w crash dumps (2200 && 23XX). If you want QLogic to look at a potential f/w problem for FC cards, you really have to provide them info in the format they expect. This involves dumping a lot of hardware registers (> 300 16 bit registers) and a lot of SRAM (> 128KB minimum). Thus all of this code is #ifdef protected which will become an option so that the memory allocation of where to dump the crash image is pretty expensive. It's worth it if you have a reproducible problem because they have some tools that can tell them, given the f/w version, the precise state of everything. MFC after: 1 week
2002-02-17 06:38:22 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
static void
isp_parse_nvram_2400(ispsoftc_t *isp, uint8_t *nvram_data)
{
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcparam *fcp = FCPARAM(isp, 0);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
uint64_t wwn;
isp_prt(isp, ISP_LOGDEBUG0,
"NVRAM 0x%08x%08x 0x%08x%08x exchg_cnt %d maxframelen %d",
(uint32_t) (ISP2400_NVRAM_NODE_NAME(nvram_data) >> 32),
(uint32_t) (ISP2400_NVRAM_NODE_NAME(nvram_data)),
(uint32_t) (ISP2400_NVRAM_PORT_NAME(nvram_data) >> 32),
(uint32_t) (ISP2400_NVRAM_PORT_NAME(nvram_data)),
ISP2400_NVRAM_EXCHANGE_COUNT(nvram_data),
ISP2400_NVRAM_MAXFRAMELENGTH(nvram_data));
isp_prt(isp, ISP_LOGDEBUG0,
"NVRAM execthr %d loopid %d fwopt1 0x%x fwopt2 0x%x fwopt3 0x%x",
ISP2400_NVRAM_EXECUTION_THROTTLE(nvram_data),
ISP2400_NVRAM_HARDLOOPID(nvram_data),
ISP2400_NVRAM_FIRMWARE_OPTIONS1(nvram_data),
ISP2400_NVRAM_FIRMWARE_OPTIONS2(nvram_data),
ISP2400_NVRAM_FIRMWARE_OPTIONS3(nvram_data));
wwn = ISP2400_NVRAM_PORT_NAME(nvram_data);
MFP4: Move default setting to the end of isp_reset instead of the front of isp_init so we can read NVRAM even if we're role ISP_NONE. Prepare for reintroduction of channels (for FC) for N-Port Virtualization. Fix a botch in handle assignment that caused us to nuke one device when a new one arrives and end up with two devices with the same identity in the virtual target mapping table.
2007-01-20 04:00:21 +00:00
fcp->isp_wwpn_nvram = wwn;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
wwn = ISP2400_NVRAM_NODE_NAME(nvram_data);
if (wwn) {
if ((wwn >> 60) != 2 && (wwn >> 60) != 5) {
wwn = 0;
}
}
Create a Node WWN from the *Port* WWN, not vice versa, for 2400s. If the NAA is type 2, the Node WWN is the Port WWN with the 12 bits of port (48..60) cleared. This iff a wwn fetched from NVRAM is zero. MFC after: 1 week
2009-12-31 04:16:18 +00:00
if (wwn == 0 && (fcp->isp_wwpn_nvram >> 60) == 2) {
wwn = fcp->isp_wwpn_nvram;
wwn &= ~((uint64_t) 0xfff << 48);
}
MFP4: Move default setting to the end of isp_reset instead of the front of isp_init so we can read NVRAM even if we're role ISP_NONE. Prepare for reintroduction of channels (for FC) for N-Port Virtualization. Fix a botch in handle assignment that caused us to nuke one device when a new one arrives and end up with two devices with the same identity in the virtual target mapping table.
2007-01-20 04:00:21 +00:00
fcp->isp_wwnn_nvram = wwn;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (ISP2400_NVRAM_EXCHANGE_COUNT(nvram_data)) {
fcp->isp_maxalloc = ISP2400_NVRAM_EXCHANGE_COUNT(nvram_data);
}
if ((isp->isp_confopts & ISP_CFG_OWNFSZ) == 0) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
DEFAULT_FRAMESIZE(isp) =
ISP2400_NVRAM_MAXFRAMELENGTH(nvram_data);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
if ((isp->isp_confopts & ISP_CFG_OWNLOOPID) == 0) {
fcp->isp_loopid = ISP2400_NVRAM_HARDLOOPID(nvram_data);
}
if ((isp->isp_confopts & ISP_CFG_OWNEXCTHROTTLE) == 0) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
DEFAULT_EXEC_THROTTLE(isp) =
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP2400_NVRAM_EXECUTION_THROTTLE(nvram_data);
}
fcp->isp_fwoptions = ISP2400_NVRAM_FIRMWARE_OPTIONS1(nvram_data);
fcp->isp_xfwoptions = ISP2400_NVRAM_FIRMWARE_OPTIONS2(nvram_data);
fcp->isp_zfwoptions = ISP2400_NVRAM_FIRMWARE_OPTIONS3(nvram_data);
}
Reference in New Issue Copy Permalink