d/freebsd-nq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
00a8e1748b
freebsd-nq/sys/dev/isp/isp.c

4485 lines
119 KiB
C
Raw Normal View History

$Id$ -> $FreeBSD$
1999-08-28 01:08:13 +00:00
/* $FreeBSD$ */
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
* Machine and OS Independent (well, as best as possible)
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
* code for the Qlogic ISP SCSI adapters.
*
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
* Copyright (c) 1997, 1998, 1999 by Matthew Jacob
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
* NASA/Ames Research Center
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice immediately at the beginning of the file, without modification,
* 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.
* 3. The name of the author may not be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE 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 THE 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.
*/
/*
* 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__
#include <dev/ic/isp_netbsd.h>
#endif
#ifdef __FreeBSD__
#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
/*
* General defines
*/
#define MBOX_DELAY_COUNT 1000000 / 100
/*
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
*/
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.
*/
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
static int isp_parse_async __P((struct ispsoftc *, int));
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 isp_handle_other_response
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
__P((struct ispsoftc *, ispstatusreq_t *, u_int16_t *));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
static void isp_parse_status
__P((struct ispsoftc *, ispstatusreq_t *, ISP_SCSI_XFER_T *));
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
static void isp_fastpost_complete __P((struct ispsoftc *, u_int32_t));
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 isp_scsi_init __P((struct ispsoftc *));
static void isp_scsi_channel_init __P((struct ispsoftc *, int));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
static void isp_fibre_init __P((struct ispsoftc *));
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 isp_mark_getpdb_all __P((struct ispsoftc *));
static int isp_getpdb __P((struct ispsoftc *, int, isp_pdb_t *));
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 u_int64_t isp_get_portname __P((struct ispsoftc *, int, int));
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
static int isp_fclink_test __P((struct ispsoftc *, int));
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 isp_same_lportdb __P((struct lportdb *, struct lportdb *));
static int isp_pdb_sync __P((struct ispsoftc *, int));
#ifdef ISP2100_FABRIC
static int isp_scan_fabric __P((struct ispsoftc *));
#endif
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
static void isp_fw_state __P((struct ispsoftc *));
static void isp_dumpregs __P((struct ispsoftc *, const char *));
static void isp_mboxcmd __P((struct ispsoftc *, mbreg_t *));
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 isp_update __P((struct ispsoftc *));
static void isp_update_bus __P((struct ispsoftc *, int));
static void isp_setdfltparm __P((struct ispsoftc *, int));
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 isp_read_nvram __P((struct ispsoftc *));
static void isp_rdnvram_word __P((struct ispsoftc *, int, u_int16_t *));
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
*/
void
isp_reset(isp)
struct ispsoftc *isp;
{
mbreg_t mbs;
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 loops, i, dodnld = 1;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
char *revname;
isp->isp_state = ISP_NILSTATE;
/*
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
* here.
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
isp->isp_dblev = DFLT_DBLEVEL;
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
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
* for SCSI adapters and other settings for the 2100.
*/
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 the current running firmware revision out of the
* chip before we hit it over the head (if this is our
* first time through). Note that we store this as the
* 'ROM' firmware revision- which it may not be. In any
* case, we don't really use this yet, but we may in
* the future.
*/
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 (isp->isp_touched == 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
/*
* Just in case it was paused...
*/
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
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
mbs.param[0] = MBOX_ABOUT_FIRMWARE;
isp_mboxcmd(isp, &mbs);
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 this fails, it probably means we're running
* an old prom, if anything at all...
*/
if (mbs.param[0] == MBOX_COMMAND_COMPLETE) {
isp->isp_romfw_rev[0] = mbs.param[1];
isp->isp_romfw_rev[1] = mbs.param[2];
isp->isp_romfw_rev[2] = mbs.param[3];
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
}
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
isp->isp_touched = 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
}
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
DISABLE_INTS(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
/*
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
* Put the board into PAUSE mode (so we can read the SXP 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
*/
ISP_WRITE(isp, HCCR, HCCR_CMD_PAUSE);
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
revname = "2X00";
switch (isp->isp_type) {
case ISP_HA_FC_2100:
revname[1] = '1';
break;
case ISP_HA_FC_2200:
revname[1] = '2';
break;
default:
break;
}
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
} else if (IS_1240(isp)) {
sdparam *sdp = isp->isp_param;
revname = "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 Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
sdp->isp_ultramode = 1;
sdp++;
sdp->isp_ultramode = 1;
/*
* XXX: Should probably do some bus sensing.
*/
} else if (IS_ULTRA2(isp)) {
static char *m = "%s: bus %d is in %s Mode\n";
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
u_int16_t l;
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
sdparam *sdp = isp->isp_param;
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
revname = "1080";
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;
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
CFGPRINTF(m, isp->isp_name, 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;
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
CFGPRINTF(m, isp->isp_name, 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;
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
CFGPRINTF(m, isp->isp_name, 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:
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
CFGPRINTF("%s: unknown mode on bus %d (0x%x)\n",
isp->isp_name, 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
if (IS_1280(isp)) {
sdp++;
revname[1] = '2';
l = ISP_READ(isp, SXP_PINS_DIFF|SXP_BANK1_SELECT);
l &= ISP1080_MODE_MASK;
switch(l) {
case ISP1080_LVD_MODE:
sdp->isp_lvdmode = 1;
CFGPRINTF(m, isp->isp_name, 1, "LVD");
break;
case ISP1080_HVD_MODE:
sdp->isp_diffmode = 1;
CFGPRINTF(m, isp->isp_name, 1, "Differential");
break;
case ISP1080_SE_MODE:
sdp->isp_ultramode = 1;
CFGPRINTF(m, isp->isp_name, 1, "Single-Ended");
break;
default:
CFGPRINTF("%s: unknown mode on bus %d (0x%x)\n",
isp->isp_name, 1, l);
break;
}
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
} 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
sdparam *sdp = isp->isp_param;
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:
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
PRINTF("%s: unknown chip rev. 0x%x- assuming a 1020\n",
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
isp->isp_name, 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 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
revname = "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 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
revname = "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:
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
revname = "1040";
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 support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
revname = "1040A";
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:
revname = "1040B";
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;
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
case 6:
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
revname = "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;
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
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) {
turn some messages into CFGPRINT messages
1999-12-03 06:55:39 +00:00
CFGPRINTF("%s: Differential Mode\n", isp->isp_name);
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) {
turn some messages into CFGPRINT messages
1999-12-03 06:55:39 +00:00
CFGPRINTF("%s: Ultra Mode Capable\n", isp->isp_name);
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
/*
* If we're in Ultra Mode, we have to be 60Mhz clock-
* 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
}
/*
* 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
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
again:
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 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,
* and give the ISP a chance to recover.
*/
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...
*/
SYS_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
#if 0
PRINTF("%s: mbox0-5: 0x%x 0x%x 0x%x 0x%x 0x%x 0x%x\n",
isp->isp_name, ISP_READ(isp, OUTMAILBOX0),
ISP_READ(isp, OUTMAILBOX1), ISP_READ(isp, OUTMAILBOX2),
ISP_READ(isp, OUTMAILBOX3), ISP_READ(isp, OUTMAILBOX4),
ISP_READ(isp, OUTMAILBOX5));
#endif
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
* Clear data && control DMA engines.
*/
ISP_WRITE(isp, CDMA_CONTROL,
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
DMA_CNTRL_CLEAR_CHAN | DMA_CNTRL_RESET_INT);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
ISP_WRITE(isp, DDMA_CONTROL,
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
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 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...
*/
SYS_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 support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +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);
}
/*
* 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 support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (!(ISP_READ(isp, BIU_ICR) & BIU_ICR_SOFT_RESET))
break;
} else {
if (!(ISP_READ(isp, BIU2100_CSR) & BIU2100_SOFT_RESET))
break;
}
SYS_DELAY(100);
if (--loops < 0) {
isp_dumpregs(isp, "chip reset timed out");
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);
} else {
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 support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
ISP_WRITE(isp, HCCR, HCCR_CMD_RESET);
SYS_DELAY(100);
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
* Establish some initial burst rate 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
* (only for the 1XX0 boards). This really should
* be done later after fetching from NVRAM.
*/
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)) {
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
u_int16_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);
}
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
#ifdef PTI_CARDS
if (((sdparam *) isp->isp_param)->isp_ultramode) {
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
while (ISP_READ(isp, RISC_MTR) != 0x1313) {
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_WRITE(isp, RISC_MTR, 0x1313);
ISP_WRITE(isp, HCCR, HCCR_CMD_STEP);
}
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
ISP_WRITE(isp, RISC_MTR, 0x1212);
}
/*
* PTI specific register
*/
ISP_WRITE(isp, RISC_EMB, DUAL_BANK)
#else
ISP_WRITE(isp, RISC_MTR, 0x1212);
#endif
} else {
ISP_WRITE(isp, RISC_MTR2100, 0x1212);
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 support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
ISP_WRITE(isp, HCCR, HCCR_CMD_RELEASE); /* release paused processor */
/*
* 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
/*
* Wait for everything to finish firing up...
*/
loops = MBOX_DELAY_COUNT;
while (ISP_READ(isp, OUTMAILBOX0) == MBOX_BUSY) {
SYS_DELAY(100);
if (--loops < 0) {
PRINTF("%s: MBOX_BUSY never cleared on reset\n",
isp->isp_name);
return;
}
}
/*
* 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
/*
* Do some sanity checking.
*/
mbs.param[0] = MBOX_NO_OP;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
isp_dumpregs(isp, "NOP test failed");
return;
}
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 (IS_SCSI(isp)) {
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;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
isp_dumpregs(isp,
"Mailbox Register test didn't complete");
return;
}
if (mbs.param[1] != 0xdead || mbs.param[2] != 0xbeef ||
mbs.param[3] != 0xffff || mbs.param[4] != 0x1111 ||
mbs.param[5] != 0xa5a5) {
isp_dumpregs(isp, "Register Test Failed");
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.
*/
correct moronic typo
1999-11-01 04:39:52 +00:00
if ((isp->isp_mdvec->dv_ispfw == NULL) ||
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
(isp->isp_confopts & ISP_CFG_NORELOAD)) {
dodnld = 0;
}
correct moronic typo
1999-11-01 04:39:52 +00:00
if (dodnld) {
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
u_int16_t fwlen = isp->isp_mdvec->dv_fwlen;
if (fwlen == 0)
fwlen = isp->isp_mdvec->dv_ispfw[3]; /* usually here */
for (i = 0; i < fwlen; i++) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
mbs.param[0] = MBOX_WRITE_RAM_WORD;
mbs.param[1] = isp->isp_mdvec->dv_codeorg + i;
mbs.param[2] = isp->isp_mdvec->dv_ispfw[i];
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
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
PRINTF("%s: F/W download failed at word %d\n",
isp->isp_name, i);
dodnld = 0;
goto again;
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
/*
* Verify that it downloaded correctly.
*/
mbs.param[0] = MBOX_VERIFY_CHECKSUM;
mbs.param[1] = isp->isp_mdvec->dv_codeorg;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
isp_dumpregs(isp, "ram checksum failure");
return;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
} else {
IDPRINTF(3, ("%s: skipping f/w download\n", isp->isp_name));
}
/*
* Now start it rolling.
*
* If we didn't actually download f/w,
* we still need to (re)start it.
*/
mbs.param[0] = MBOX_EXEC_FIRMWARE;
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->isp_mdvec->dv_codeorg)
mbs.param[1] = isp->isp_mdvec->dv_codeorg;
else
mbs.param[1] = 0x1000;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp_mboxcmd(isp, &mbs);
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 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
* Set CLOCK RATE, but only if asked to.
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
if (isp->isp_clock) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
mbs.param[0] = MBOX_SET_CLOCK_RATE;
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] = isp->isp_clock;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
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
PRINTF("failed to set clockrate (0x%x)\n",
mbs.param[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
/* but continue */
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
}
}
mbs.param[0] = MBOX_ABOUT_FIRMWARE;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
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
PRINTF("could not get f/w started (0x%x)\n", mbs.param[0]);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +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
CFGPRINTF("%s: Board Revision %s, %s F/W Revision %d.%d.%d\n",
isp->isp_name, revname, dodnld? "loaded" : "resident",
mbs.param[1], mbs.param[2], mbs.param[3]);
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)) {
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 (ISP_READ(isp, BIU2100_CSR) & BIU2100_PCI64) {
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
CFGPRINTF("%s: in 64-Bit PCI slot\n", isp->isp_name);
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
}
}
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 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_fwrev[0] = mbs.param[1];
isp->isp_fwrev[1] = mbs.param[2];
isp->isp_fwrev[2] = mbs.param[3];
if (isp->isp_romfw_rev[0] || isp->isp_romfw_rev[1] ||
isp->isp_romfw_rev[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
CFGPRINTF("%s: Last F/W revision was %d.%d.%d\n", isp->isp_name,
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_romfw_rev[0], isp->isp_romfw_rev[1],
isp->isp_romfw_rev[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
}
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
mbs.param[0] = MBOX_GET_FIRMWARE_STATUS;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
PRINTF("%s: could not GET FIRMWARE STATUS\n", isp->isp_name);
return;
}
isp->isp_maxcmds = mbs.param[2];
CFGPRINTF("%s: %d max I/O commands supported\n",
isp->isp_name, mbs.param[2]);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp_fw_state(isp);
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
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 up DMA for the request and result mailboxes.
*/
if (ISP_MBOXDMASETUP(isp) != 0) {
PRINTF("%s: can't setup dma mailboxes\n", isp->isp_name);
return;
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp->isp_state = ISP_RESETSTATE;
}
/*
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
isp_init(isp)
struct ispsoftc *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
/*
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
* Must do this first to get defaults established.
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
isp_setdfltparm(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_setdfltparm(isp, 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
}
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
if (IS_FC(isp)) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp_fibre_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
} else {
isp_scsi_init(isp);
}
}
static void
isp_scsi_init(isp)
struct ispsoftc *isp;
{
sdparam *sdp_chan0, *sdp_chan1;
mbreg_t mbs;
sdp_chan0 = isp->isp_param;
sdp_chan1 = sdp_chan0;
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
sdp_chan1++;
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
/* 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.
*/
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_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
*/
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[0] = MBOX_SET_RETRY_COUNT;
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 support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
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
PRINTF("%s: failed to set retry count and retry delay\n",
isp->isp_name);
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 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
*/
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[0] = MBOX_SET_ASYNC_DATA_SETUP_TIME;
mbs.param[1] = sdp_chan0->isp_async_data_setup;
mbs.param[2] = sdp_chan1->isp_async_data_setup;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
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
PRINTF("%s: failed to set asynchronous data setup time\n",
isp->isp_name);
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 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
*/
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[0] = MBOX_SET_ACT_NEG_STATE;
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);
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
PRINTF("%s: failed to set active negation state "
"(%d,%d),(%d,%d)\n", isp->isp_name,
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
*/
mbs.param[0] = MBOX_SET_TAG_AGE_LIMIT;
mbs.param[1] = sdp_chan0->isp_tag_aging;
mbs.param[2] = sdp_chan1->isp_tag_aging;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
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
PRINTF("%s: failed to set tag age limit (%d,%d)\n",
isp->isp_name, 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
*/
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[0] = MBOX_SET_SELECT_TIMEOUT;
mbs.param[1] = sdp_chan0->isp_selection_timeout;
mbs.param[2] = sdp_chan1->isp_selection_timeout;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
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
PRINTF("%s: failed to set selection timeout\n", isp->isp_name);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
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
*/
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[0] = MBOX_INIT_RES_QUEUE;
mbs.param[1] = RESULT_QUEUE_LEN;
mbs.param[2] = DMA_MSW(isp->isp_result_dma);
mbs.param[3] = DMA_LSW(isp->isp_result_dma);
mbs.param[4] = 0;
mbs.param[5] = 0;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
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
PRINTF("%s: set of response queue failed\n", isp->isp_name);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
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
isp->isp_residx = 0;
mbs.param[0] = MBOX_INIT_REQ_QUEUE;
mbs.param[1] = RQUEST_QUEUE_LEN;
mbs.param[2] = DMA_MSW(isp->isp_rquest_dma);
mbs.param[3] = DMA_LSW(isp->isp_rquest_dma);
mbs.param[4] = 0;
mbs.param[5] = 0;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
PRINTF("%s: set of request queue failed\n", isp->isp_name);
return;
}
isp->isp_reqidx = isp->isp_reqodx = 0;
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
/*
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
* Turn on Fast Posting, LVD transitions
*
* Ultra2 F/W always has had fast posting (and LVD transitions)
*
* Ultra and older (i.e., SBus) cards may not. Assume SBus cards
* do not, and only guess that 4.55.0 <= x < 5.0.0 (initiator
* only) and x >= 7.55 (initiator/target) has fast posting.
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 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
mbs.param[0] = MBOX_SET_FW_FEATURES;
mbs.param[1] = 0;
if (IS_ULTRA2(isp))
mbs.param[1] |= FW_FEATURE_LVD_NOTIFY;
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
#ifndef ISP_NO_FASTPOST_SCSI
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 ((ISP_FW_REVX(isp->isp_fwrev) >= ISP_FW_REV(4, 55, 0) &&
(ISP_FW_REVX(isp->isp_fwrev) < ISP_FW_REV(5, 0, 0))) ||
(ISP_FW_REVX(isp->isp_fwrev) >= ISP_FW_REV(7, 55, 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] |= FW_FEATURE_FAST_POST;
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
}
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
#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) {
u_int16_t sfeat = mbs.param[1];
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
PRINTF("%s: cannot enable FW features (0x%x)\n",
isp->isp_name, sfeat);
} else {
CFGPRINTF("%s: enabled FW features (0x%x)\n",
isp->isp_name, 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
}
}
/*
* Let the outer layers decide whether to issue a SCSI bus reset.
*/
isp->isp_state = ISP_INITSTATE;
}
static void
isp_scsi_channel_init(isp, channel)
struct ispsoftc *isp;
int channel;
{
sdparam *sdp;
mbreg_t mbs;
int tgt;
sdp = isp->isp_param;
sdp += channel;
/*
* Set (possibly new) Initiator ID.
*/
mbs.param[0] = MBOX_SET_INIT_SCSI_ID;
mbs.param[1] = (channel << 7) | sdp->isp_initiator_id;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
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
PRINTF("%s: cannot set initiator id on bus %d to %d\n",
isp->isp_name, channel, sdp->isp_initiator_id);
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
/*
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
* Set current per-target parameters to a 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++) {
int maxlun, lun;
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
u_int16_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) {
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
IDPRINTF(1, ("%s: skipping target %d bus %d settings\n",
isp->isp_name, tgt, channel));
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
}
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
/*
* If we're in LVD mode, then we pretty much should
* only disable tagged queuing.
*/
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
if (IS_ULTRA2(isp) && sdp->isp_lvdmode) {
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
sdf = DPARM_DEFAULT & ~DPARM_TQING;
} else {
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
int rvf = ISP_FW_REVX(isp->isp_fwrev);
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
sdf = DPARM_SAFE_DFLT;
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
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
/*
* It is not quite clear when this changed over so that
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
* we could force narrow and async, so assume >= 7.55
* for i/t F/W and = 4.55 for initiator f/w.
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
*/
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 ((ISP_FW_REV(4, 55, 0) <= rvf &&
(ISP_FW_REV(5, 0, 0) > rvf)) ||
(ISP_FW_REV(7, 55, 0) <= rvf)) {
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
sdf |= DPARM_NARROW | DPARM_ASYNC;
}
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
mbs.param[0] = MBOX_SET_TARGET_PARAMS;
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] = (tgt << 8) | (channel << 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;
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
mbs.param[3] =
(sdp->isp_devparam[tgt].sync_offset << 8) |
(sdp->isp_devparam[tgt].sync_period);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp_mboxcmd(isp, &mbs);
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 support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
mbs.param[0] = MBOX_SET_TARGET_PARAMS;
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] = (tgt << 8) | (channel << 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;
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
mbs.param[3] =
(sdp->isp_devparam[tgt].sync_offset << 8) |
(sdp->isp_devparam[tgt].sync_period);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp_mboxcmd(isp, &mbs);
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
PRINTF("%s: failed even to set defaults for "
"target %d\n", isp->isp_name, tgt);
continue;
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
#if 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
/*
* We don't update dev_flags with what we've set
* because that's not the ultimate goal setting.
* If we succeed with the command, we *do* update
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
* cur_dflags by getting target parameters.
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
*/
mbs.param[0] = MBOX_GET_TARGET_PARAMS;
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] = (tgt << 8) | (channel << 15);
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
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
/*
* Urrr.... We'll set cur_dflags to DPARM_SAFE_DFLT so
* we don't try and do tags if tags aren't enabled.
*/
sdp->isp_devparam[tgt].cur_dflags = DPARM_SAFE_DFLT;
} else {
sdp->isp_devparam[tgt].cur_dflags = mbs.param[2];
sdp->isp_devparam[tgt].cur_offset = mbs.param[3] >> 8;
sdp->isp_devparam[tgt].cur_period = mbs.param[3] & 0xff;
}
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
IDPRINTF(3, ("%s: set flags 0x%x got 0x%x back for target %d\n",
isp->isp_name, sdf, mbs.param[2], tgt));
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
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
/*
* We don't update any information because we need to run
* at least one command per target to cause a new state
* to be latched.
*/
#endif
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.
*/
sdp->isp_devparam[tgt].cur_dflags &= ~DPARM_TQING;
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 ((ISP_FW_REV(4, 55, 0) <= ISP_FW_REVX(isp->isp_fwrev) &&
(ISP_FW_REV(5, 0, 0) > ISP_FW_REVX(isp->isp_fwrev))) ||
(ISP_FW_REVX(isp->isp_fwrev) >= ISP_FW_REV(7, 55, 0)))
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
maxlun = 32;
else
maxlun = 8;
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 (lun = 0; lun < maxlun; lun++) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
mbs.param[0] = MBOX_SET_DEV_QUEUE_PARAMS;
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] = (channel << 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 support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp_mboxcmd(isp, &mbs);
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
PRINTF("%s: failed to set device queue "
"parameters for target %d, lun %d\n",
isp->isp_name, tgt, lun);
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
}
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.
*
* Locks are held before coming here.
*/
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
static void
isp_fibre_init(isp)
struct ispsoftc *isp;
{
fcparam *fcp;
isp_icb_t *icbp;
mbreg_t mbs;
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
int loopid;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
fcp = isp->isp_param;
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
/*
* For systems that don't have BIOS methods for which
* we can easily change the NVRAM based loopid, we'll
* override that here. Note that when we initialize
* the firmware we may get back a different loopid than
* we asked for anyway. XXX This is probably not the
* best way to figure this out XXX
*/
#ifndef __i386__
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
loopid = DEFAULT_LOOPID(isp);
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
loopid = fcp->isp_loopid;
#endif
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
icbp = (isp_icb_t *) fcp->isp_scratch;
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
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;
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
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
fcp->isp_fwoptions = ICBOPT_TGT_ENABLE;
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
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
fcp->isp_fwoptions = 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
#endif
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
fcp->isp_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
/*
* If this is a 2100 < revision 5, we have to turn off FAIRNESS.
*/
if ((isp->isp_type == ISP_HA_FC_2100) && isp->isp_revision < 5) {
fcp->isp_fwoptions &= ~ICBOPT_FAIRNESS;
}
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
fcp->isp_fwoptions |= ICBOPT_PDBCHANGE_AE;
fcp->isp_fwoptions |= ICBOPT_HARD_ADDRESS;
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
* a LIP- this is a known f/w bug.
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
if (ISP_FW_REVX(isp->isp_fwrev) < ISP_FW_REV(1, 17, 0)) {
fcp->isp_fwoptions |= ICBOPT_FULL_LOGIN;
}
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
#ifndef ISP_NO_FASTPOST_FC
fcp->isp_fwoptions |= ICBOPT_FAST_POST;
#endif
if (isp->isp_confopts & ISP_CFG_FULL_DUPLEX)
fcp->isp_fwoptions |= ICBOPT_FULL_DUPLEX;
/*
* We don't set ICBOPT_PORTNAME because we want our
* Node Name && Port Names to be distinct.
*/
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
icbp->icb_fwoptions = fcp->isp_fwoptions;
icbp->icb_maxfrmlen = fcp->isp_maxfrmlen;
if (icbp->icb_maxfrmlen < ICB_MIN_FRMLEN ||
icbp->icb_maxfrmlen > ICB_MAX_FRMLEN) {
PRINTF("%s: bad frame length (%d) from NVRAM- using %d\n",
isp->isp_name, fcp->isp_maxfrmlen, 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) {
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
PRINTF("%s: bad maximum allocation (%d)- using 16\n",
isp->isp_name, fcp->isp_maxalloc);
icbp->icb_maxalloc = 16;
}
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_execthrottle = fcp->isp_execthrottle;
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) {
PRINTF("%s: bad execution throttle of %d- using 16\n",
isp->isp_name, fcp->isp_execthrottle);
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;
oops on last
1998-12-05 01:46:40 +00:00
icbp->icb_hardaddr = loopid;
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
icbp->icb_logintime = 60; /* 60 second login 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
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 (fcp->isp_nodewwn) {
u_int64_t pn;
MAKE_NODE_NAME_FROM_WWN(icbp->icb_nodename, fcp->isp_nodewwn);
if (fcp->isp_portwwn) {
pn = fcp->isp_portwwn;
} else {
pn = fcp->isp_nodewwn |
(((u_int64_t)(isp->isp_unit+1)) << 56);
}
/*
* If the top nibble is 2, we can construct a port name
* from the node name by setting a nonzero instance in
* bits 56..59. Otherwise, we need to make it identical
* to Node name...
*/
if ((fcp->isp_nodewwn >> 60) == 2) {
MAKE_NODE_NAME_FROM_WWN(icbp->icb_portname, pn);
} else {
MAKE_NODE_NAME_FROM_WWN(icbp->icb_portname,
fcp->isp_nodewwn);
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 {
fcp->isp_fwoptions &= ~(ICBOPT_USE_PORTNAME|ICBOPT_FULL_LOGIN);
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
}
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_rqstqlen = RQUEST_QUEUE_LEN;
icbp->icb_rsltqlen = RESULT_QUEUE_LEN;
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_rqstaddr[RQRSP_ADDR0015] = DMA_LSW(isp->isp_rquest_dma);
icbp->icb_rqstaddr[RQRSP_ADDR1631] = DMA_MSW(isp->isp_rquest_dma);
icbp->icb_respaddr[RQRSP_ADDR0015] = DMA_LSW(isp->isp_result_dma);
icbp->icb_respaddr[RQRSP_ADDR1631] = DMA_MSW(isp->isp_result_dma);
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_SWIZZLE_ICB(isp, icbp);
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
/*
* Do this *before* initializing the firmware.
*/
isp_mark_getpdb_all(isp);
fcp->isp_fwstate = FW_CONFIG_WAIT;
fcp->isp_loopstate = LOOP_NIL;
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
MemoryBarrier();
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
for (;;) {
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[0] = MBOX_INIT_FIRMWARE;
mbs.param[1] = 0;
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
mbs.param[2] = DMA_MSW(fcp->isp_scdma);
mbs.param[3] = DMA_LSW(fcp->isp_scdma);
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[4] = 0;
mbs.param[5] = 0;
mbs.param[6] = 0;
mbs.param[7] = 0;
isp_mboxcmd(isp, &mbs);
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 (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
PRINTF("%s: INIT FIRMWARE failed (code 0x%x)\n",
isp->isp_name, mbs.param[0]);
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 (mbs.param[0] & 0x8000) {
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
SYS_DELAY(1000);
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
}
return;
}
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;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +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
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_reqidx = 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
isp->isp_sendmarker = 1;
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
* Whatever happens, we're now committed to being here.
*/
isp->isp_state = ISP_INITSTATE;
}
/*
* Fibre Channel Support- get the port database for the id.
*
* Locks are held before coming here. Return 0 if success,
* else failure.
*/
static void
isp_mark_getpdb_all(isp)
struct ispsoftc *isp;
{
fcparam *fcp = (fcparam *) isp->isp_param;
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
int i;
for (i = 0; i < MAX_FC_TARG; i++) {
fcp->portdb[i].valid = 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
}
}
static int
isp_getpdb(isp, id, pdbp)
struct ispsoftc *isp;
int id;
isp_pdb_t *pdbp;
{
fcparam *fcp = (fcparam *) isp->isp_param;
mbreg_t mbs;
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
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[0] = MBOX_GET_PORT_DB;
mbs.param[1] = id << 8;
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
mbs.param[2] = DMA_MSW(fcp->isp_scdma);
mbs.param[3] = DMA_LSW(fcp->isp_scdma);
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
/*
* Unneeded. For the 2100, except for initializing f/w, registers
* 4/5 have to not be written to.
* mbs.param[4] = 0;
* mbs.param[5] = 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
mbs.param[6] = 0;
mbs.param[7] = 0;
isp_mboxcmd(isp, &mbs);
switch (mbs.param[0]) {
case MBOX_COMMAND_COMPLETE:
MemoryBarrier();
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_UNSWIZZLE_AND_COPY_PDBP(isp, pdbp, fcp->isp_scratch);
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
break;
case MBOX_HOST_INTERFACE_ERROR:
PRINTF("%s: DMA error getting port database\n", isp->isp_name);
return (-1);
case MBOX_COMMAND_PARAM_ERROR:
/* Not Logged In */
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
IDPRINTF(3, ("%s: Param Error on Get Port Database for id %d\n",
isp->isp_name, 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
return (-1);
default:
PRINTF("%s: error 0x%x getting port database for ID %d\n",
isp->isp_name, mbs.param[0], id);
return (-1);
}
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
static u_int64_t
isp_get_portname(isp, loopid, nodename)
struct ispsoftc *isp;
int loopid;
int nodename;
{
u_int64_t wwn = 0;
mbreg_t mbs;
mbs.param[0] = MBOX_GET_PORT_NAME;
mbs.param[1] = loopid << 8;
if (nodename)
mbs.param[1] |= 1;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] == MBOX_COMMAND_COMPLETE) {
wwn =
(((u_int64_t)(mbs.param[2] & 0xff)) << 56) |
(((u_int64_t)(mbs.param[2] >> 8)) << 48) |
(((u_int64_t)(mbs.param[3] & 0xff)) << 40) |
(((u_int64_t)(mbs.param[3] >> 8)) << 32) |
(((u_int64_t)(mbs.param[6] & 0xff)) << 24) |
(((u_int64_t)(mbs.param[6] >> 8)) << 16) |
(((u_int64_t)(mbs.param[7] & 0xff)) << 8) |
(((u_int64_t)(mbs.param[7] >> 8)));
}
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
/*
* Make sure we have good FC link and know our Loop ID.
*/
static int
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
isp_fclink_test(isp, waitdelay)
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
struct ispsoftc *isp;
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 waitdelay;
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
{
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
static char *toponames[] = {
"Private Loop",
"FL Port",
"N-Port to N-Port",
"F Port"
};
char *tname;
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;
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
int count, topo = -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
u_int8_t lwfs;
fcparam *fcp;
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 defined(ISP2100_FABRIC)
isp_pdb_t pdb;
#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
fcp = isp->isp_param;
/*
* 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;
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
for (count = 0; count < waitdelay; count += 100) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp_fw_state(isp);
if (lwfs != fcp->isp_fwstate) {
per bde (who is right about this) that an inlined fucntion with const char * strings being returned defined in a header file included several places but only used in one module, is, uh, silly.
1998-09-17 23:20:29 +00:00
PRINTF("%s: Firmware State %s -> %s\n",
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_name, isp2100_fw_statename((int)lwfs),
isp2100_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;
}
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
SYS_DELAY(100); /* wait 100 microseconds */
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) {
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
/*
* Get our Loop ID (if possible). We really need to have it.
*/
mbs.param[0] = MBOX_GET_LOOP_ID;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
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
PRINTF("%s: GET LOOP ID failed\n", isp->isp_name);
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);
}
fcp->isp_loopid = mbs.param[1];
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 (isp->isp_type == ISP_HA_FC_2200) {
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
if (ISP_FW_REVX(isp->isp_fwrev) >= ISP_FW_REV(2, 0, 14)) {
topo = (int) mbs.param[6];
}
} else if (isp->isp_type == ISP_HA_FC_2100) {
if (ISP_FW_REVX(isp->isp_fwrev) >= ISP_FW_REV(1, 17, 26)) {
topo = (int) mbs.param[6];
}
}
if (topo < 0 || topo > 3)
tname = "unknown";
else
tname = toponames[topo];
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 not on a fabric, the low 8 bits will be our AL_PA.
* If we're on a fabric, the low 8 bits will still be our AL_PA.
*/
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
fcp->isp_alpa = mbs.param[2];
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 defined(ISP2100_FABRIC)
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
fcp->isp_onfabric = 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 (isp_getpdb(isp, FL_PORT_ID, &pdb) == 0) {
fcp->isp_portid = mbs.param[2] | (((int)mbs.param[3]) << 16);
fcp->isp_onfabric = 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
CFGPRINTF("%s: Loop ID %d, AL_PA 0x%x, Port ID 0x%x Loop State "
"0x%x topology %s\n", isp->isp_name, fcp->isp_loopid,
fcp->isp_alpa, fcp->isp_portid, fcp->isp_loopstate, tname);
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
/*
* Make sure we're logged out of all fabric devices.
*/
for (count = FC_SNS_ID+1; count < MAX_FC_TARG; count++) {
struct lportdb *lp = &fcp->portdb[count];
if (lp->valid == 0 || lp->fabdev == 0)
continue;
PRINTF("%s: logging out target %d at Loop ID %d "
"(port id 0x%x)\n", isp->isp_name, count,
lp->loopid, lp->portid);
mbs.param[0] = MBOX_FABRIC_LOGOUT;
mbs.param[1] = lp->loopid << 8;
mbs.param[2] = 0;
mbs.param[3] = 0;
isp_mboxcmd(isp, &mbs);
}
} else
#endif
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
CFGPRINTF("%s: Loop ID %d, ALPA 0x%x Loop State 0x%x topology %s\n",
isp->isp_name, fcp->isp_loopid, fcp->isp_alpa, fcp->isp_loopstate,
tname);
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
}
/*
* Compare two local port db entities and return 1 if they're the same, else 0.
*/
static int
isp_same_lportdb(a, b)
struct lportdb *a, *b;
{
/*
* We decide two lports are the same if they have non-zero and
* identical port WWNs and identical loop IDs.
*/
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
if (a->port_wwn == 0 || a->port_wwn != b->port_wwn ||
a->loopid != b->loopid) {
return (0);
} else {
return (1);
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +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
/*
* Synchronize our soft copy of the port database with what the f/w thinks
* (with a view toward possibly for a specific target....)
*/
static int
isp_pdb_sync(isp, target)
struct ispsoftc *isp;
int target;
{
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
struct lportdb *lp, *tport;
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
fcparam *fcp = isp->isp_param;
isp_pdb_t pdb;
int loopid, lim;
#ifdef ISP2100_FABRIC
/*
* XXX: If we do this *after* building up our local port database,
* XXX: the commands simply don't work.
*/
/*
* (Re)discover all fabric devices
*/
if (fcp->isp_onfabric)
(void) isp_scan_fabric(isp);
#endif
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
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
/*
* Run through the local loop ports and get port database info
* for each loop ID.
*
* There's a somewhat unexplained situation where the f/w passes back
* the wrong database entity- if that happens, just restart (up to
* FL_PORT_ID times).
*/
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
tport = fcp->tport;
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
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
/*
* make sure the temp port database is clean...
*/
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
MEMZERO((void *) tport, sizeof (tport));
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
for (lim = loopid = 0; loopid < FL_PORT_ID; loopid++) {
lp = &tport[loopid];
lp->node_wwn = isp_get_portname(isp, loopid, 1);
if (lp->node_wwn == 0)
continue;
lp->port_wwn = isp_get_portname(isp, loopid, 0);
if (lp->port_wwn == 0) {
lp->node_wwn = 0;
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
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
/*
* Get an entry....
*/
if (isp_getpdb(isp, loopid, &pdb) != 0) {
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
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 the returned database element doesn't match what we
* asked for, restart the process entirely (up to a point...).
*/
if (pdb.pdb_loopid != loopid) {
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
IDPRINTF(1, ("%s: wankage (%d != %d)\n",
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->isp_name, pdb.pdb_loopid, loopid));
loopid = 0;
if (lim++ < FL_PORT_ID) {
continue;
}
PRINTF("%s: giving up on synchronizing the port "
"database\n", isp->isp_name);
return (-1);
}
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
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
/*
* Save the pertinent info locally.
*/
lp->node_wwn =
(((u_int64_t)pdb.pdb_nodename[0]) << 56) |
(((u_int64_t)pdb.pdb_nodename[1]) << 48) |
(((u_int64_t)pdb.pdb_nodename[2]) << 40) |
(((u_int64_t)pdb.pdb_nodename[3]) << 32) |
(((u_int64_t)pdb.pdb_nodename[4]) << 24) |
(((u_int64_t)pdb.pdb_nodename[5]) << 16) |
(((u_int64_t)pdb.pdb_nodename[6]) << 8) |
(((u_int64_t)pdb.pdb_nodename[7]));
lp->port_wwn =
(((u_int64_t)pdb.pdb_portname[0]) << 56) |
(((u_int64_t)pdb.pdb_portname[1]) << 48) |
(((u_int64_t)pdb.pdb_portname[2]) << 40) |
(((u_int64_t)pdb.pdb_portname[3]) << 32) |
(((u_int64_t)pdb.pdb_portname[4]) << 24) |
(((u_int64_t)pdb.pdb_portname[5]) << 16) |
(((u_int64_t)pdb.pdb_portname[6]) << 8) |
(((u_int64_t)pdb.pdb_portname[7]));
lp->roles =
(pdb.pdb_prli_svc3 & SVC3_ROLE_MASK) >> SVC3_ROLE_SHIFT;
lp->portid = BITS2WORD(pdb.pdb_portid_bits);
lp->loopid = pdb.pdb_loopid;
/*
* Do a quick check to see whether this matches the saved port
* database for the same loopid. We do this here to save
* searching later (if possible). Note that this fails over
* time as things shuffle on the loop- we get the current
* loop state (where loop id as an index matches loop id in
* use) and then compare it to our saved database which
* never shifts.
*/
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
if (target >= 0 && isp_same_lportdb(lp, &fcp->portdb[target])) {
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
lp->valid = 1;
}
}
/*
* If we get this far, we've settled our differences with the f/w
* and we can say that the loop state is ready.
*/
fcp->isp_loopstate = LOOP_READY;
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
/*
* Mark all of the permanent local loop database entries as invalid.
*/
for (loopid = 0; loopid < FL_PORT_ID; loopid++) {
fcp->portdb[loopid].valid = 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
/*
* Now merge our local copy of the port database into our saved copy.
* Notify the outer layers of new devices arriving.
*/
for (loopid = 0; loopid < FL_PORT_ID; loopid++) {
int i;
/*
* If we don't have a non-zero Port WWN, we're not here.
*/
if (tport[loopid].port_wwn == 0) {
continue;
}
/*
* If we've already marked our tmp copy as valid,
* this means that we've decided that it's the
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
* same as our saved data base. This didn't include
* the 'valid' marking so we have set that here.
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 (tport[loopid].valid) {
fcp->portdb[loopid].valid = 1;
continue;
}
/*
* For the purposes of deciding whether this is the
* 'same' device or not, we only search for an identical
* Port WWN. Node WWNs may or may not be the same as
* the Port WWN, and there may be multiple different
* Port WWNs with the same Node WWN. It would be chaos
* to have multiple identical Port WWNs, so we don't
* allow that.
*/
for (i = 0; i < FL_PORT_ID; i++) {
int j;
if (fcp->portdb[i].port_wwn == 0)
continue;
if (fcp->portdb[i].port_wwn != tport[loopid].port_wwn)
continue;
/*
* We found this WWN elsewhere- it's changed
* loopids then. We don't change it's actual
* position in our cached port database- we
* just change the actual loop ID we'd use.
*/
if (fcp->portdb[i].loopid != loopid) {
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
PRINTF("%s: Target ID %d Loop 0x%x (Port 0x%x) "
"=> Loop 0x%x (Port 0x%x) \n",
isp->isp_name, i, fcp->portdb[i].loopid,
fcp->portdb[i].portid, loopid,
tport[loopid].portid);
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
fcp->portdb[i].portid = tport[loopid].portid;
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
fcp->portdb[i].loopid = loopid;
fcp->portdb[i].valid = 1;
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
/*
* XXX: Should we also propagate roles in case they
* XXX: changed?
*/
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
/*
* Now make sure this Port WWN doesn't exist elsewhere
* in the port database.
*/
for (j = i+1; j < FL_PORT_ID; j++) {
if (fcp->portdb[i].port_wwn !=
fcp->portdb[j].port_wwn) {
continue;
}
PRINTF("%s: Target ID %d Duplicates Target ID "
"%d- killing off both\n",
isp->isp_name, j, i);
/*
* Invalidate the 'old' *and* 'new' ones.
* This is really harsh and not quite right,
* but if this happens, we really don't know
* who is what at this point.
*/
fcp->portdb[i].valid = 0;
fcp->portdb[j].valid = 0;
}
break;
}
/*
* If we didn't traverse the entire port database,
* then we found (and remapped) an existing entry.
* No need to notify anyone- go for the next one.
*/
if (i < FL_PORT_ID) {
continue;
}
/*
* We've not found this Port WWN anywhere. It's a new entry.
* See if we can leave it where it is (with target == loopid).
*/
if (fcp->portdb[loopid].port_wwn != 0) {
for (lim = 0; lim < FL_PORT_ID; lim++) {
if (fcp->portdb[lim].port_wwn == 0)
break;
}
/* "Cannot Happen" */
if (lim == FL_PORT_ID) {
PRINTF("%s: remap overflow?\n", isp->isp_name);
continue;
}
i = lim;
} else {
i = loopid;
}
/*
* NB: The actual loopid we use here is loopid- we may
* in fact be at a completely different index (target).
*/
fcp->portdb[i].loopid = loopid;
fcp->portdb[i].port_wwn = tport[loopid].port_wwn;
fcp->portdb[i].node_wwn = tport[loopid].node_wwn;
fcp->portdb[i].roles = tport[loopid].roles;
fcp->portdb[i].portid = tport[loopid].portid;
fcp->portdb[i].valid = 1;
/*
* Tell the outside world we've arrived.
*/
(void) isp_async(isp, ISPASYNC_PDB_CHANGED, &i);
}
/*
* Now find all previously used targets that are now invalid and
* notify the outer layers that they're gone.
*/
for (lp = fcp->portdb; lp < &fcp->portdb[FL_PORT_ID]; lp++) {
if (lp->valid || lp->port_wwn == 0)
continue;
/*
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
* Tell the outside world we've gone away.
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
*/
loopid = lp - fcp->portdb;
(void) isp_async(isp, ISPASYNC_PDB_CHANGED, &loopid);
MEMZERO((void *) lp, sizeof (*lp));
}
#ifdef ISP2100_FABRIC
/*
* Now log in any fabric devices
*/
for (lp = &fcp->portdb[FC_SNS_ID+1];
lp < &fcp->portdb[MAX_FC_TARG]; lp++) {
mbreg_t mbs;
/*
* Nothing here?
*/
if (lp->port_wwn == 0)
continue;
/*
* Don't try to log into yourself.
*/
if (lp->portid == fcp->isp_portid)
continue;
/*
* Force a logout.
*/
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
lp->loopid = loopid = lp - fcp->portdb;
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
mbs.param[0] = MBOX_FABRIC_LOGOUT;
mbs.param[1] = lp->loopid << 8;
mbs.param[2] = 0;
mbs.param[3] = 0;
isp_mboxcmd(isp, &mbs);
/*
* And log in....
*/
mbs.param[0] = MBOX_FABRIC_LOGIN;
mbs.param[1] = lp->loopid << 8;
mbs.param[2] = lp->portid >> 16;
mbs.param[3] = lp->portid & 0xffff;
isp_mboxcmd(isp, &mbs);
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
switch (mbs.param[0]) {
case MBOX_COMMAND_COMPLETE:
break;
case MBOX_COMMAND_ERROR:
switch (mbs.param[1]) {
case 1:
PRINTF("%s: no loop\n", isp->isp_name);
break;
case 2:
PRINTF("%s: IOCB buffer could not be alloced\n",
isp->isp_name);
break;
case 3:
PRINTF("%s: could not alloc xchange resource\n",
isp->isp_name);
break;
case 4:
PRINTF("%s: ELS timeout\n", isp->isp_name);
break;
case 5:
PRINTF("%s: no fabric port\n", isp->isp_name);
break;
case 6:
PRINTF("%s: remote device cannot be a target\n",
isp->isp_name);
break;
default:
break;
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
}
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
continue;
default:
continue;
}
lp->valid = 1;
lp->fabdev = 1;
if (isp_getpdb(isp, loopid, &pdb) != 0) {
/*
* Be kind...
*/
lp->roles = (SVC3_TGT_ROLE >> SVC3_ROLE_SHIFT);
PRINTF("%s: Faked PortID 0x%x into LoopID %d\n",
isp->isp_name, lp->portid, lp->loopid);
continue;
}
if (pdb.pdb_loopid != lp->loopid) {
lp->roles = (SVC3_TGT_ROLE >> SVC3_ROLE_SHIFT);
PRINTF("%s: Wanked PortID 0x%x to LoopID %d\n",
isp->isp_name, lp->portid, lp->loopid);
continue;
}
lp->roles =
(pdb.pdb_prli_svc3 & SVC3_ROLE_MASK) >> SVC3_ROLE_SHIFT;
lp->portid = BITS2WORD(pdb.pdb_portid_bits);
lp->loopid = loopid;
lp->node_wwn =
(((u_int64_t)pdb.pdb_nodename[0]) << 56) |
(((u_int64_t)pdb.pdb_nodename[1]) << 48) |
(((u_int64_t)pdb.pdb_nodename[2]) << 40) |
(((u_int64_t)pdb.pdb_nodename[3]) << 32) |
(((u_int64_t)pdb.pdb_nodename[4]) << 24) |
(((u_int64_t)pdb.pdb_nodename[5]) << 16) |
(((u_int64_t)pdb.pdb_nodename[6]) << 8) |
(((u_int64_t)pdb.pdb_nodename[7]));
lp->port_wwn =
(((u_int64_t)pdb.pdb_portname[0]) << 56) |
(((u_int64_t)pdb.pdb_portname[1]) << 48) |
(((u_int64_t)pdb.pdb_portname[2]) << 40) |
(((u_int64_t)pdb.pdb_portname[3]) << 32) |
(((u_int64_t)pdb.pdb_portname[4]) << 24) |
(((u_int64_t)pdb.pdb_portname[5]) << 16) |
(((u_int64_t)pdb.pdb_portname[6]) << 8) |
(((u_int64_t)pdb.pdb_portname[7]));
/*
* Check to make sure this all makes sense.
*/
if (lp->node_wwn && lp->port_wwn) {
(void) isp_async(isp, ISPASYNC_PDB_CHANGED, &loopid);
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
}
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
lp->fabdev = lp->valid = 0;
PRINTF("%s: Target %d (Loop 0x%x) Port ID 0x%x lost its WWN\n",
isp->isp_name, loopid, lp->loopid, lp->portid);
mbs.param[0] = MBOX_FABRIC_LOGOUT;
mbs.param[1] = lp->loopid << 8;
mbs.param[2] = 0;
mbs.param[3] = 0;
isp_mboxcmd(isp, &mbs);
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
}
#endif
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
/*
* 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 ispscsicmd.
*/
fcp->loop_seen_once = 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);
}
#ifdef ISP2100_FABRIC
static int
isp_scan_fabric(isp)
struct ispsoftc *isp;
{
fcparam *fcp = isp->isp_param;
u_int32_t portid, first_nz_portid;
sns_screq_t *reqp;
sns_scrsp_t *resp;
mbreg_t mbs;
int hicap;
reqp = (sns_screq_t *) fcp->isp_scratch;
resp = (sns_scrsp_t *) (&((char *)fcp->isp_scratch)[0x100]);
first_nz_portid = portid = fcp->isp_portid;
for (hicap = 0; hicap < 1024; hicap++) {
MEMZERO((void *) reqp, SNS_GAN_REQ_SIZE);
reqp->snscb_rblen = SNS_GAN_RESP_SIZE >> 1;
reqp->snscb_addr[RQRSP_ADDR0015] =
DMA_LSW(fcp->isp_scdma + 0x100);
reqp->snscb_addr[RQRSP_ADDR1631] =
DMA_MSW(fcp->isp_scdma + 0x100);
reqp->snscb_sblen = 6;
reqp->snscb_data[0] = SNS_GAN;
reqp->snscb_data[4] = portid & 0xffff;
reqp->snscb_data[5] = (portid >> 16) & 0xff;
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_SWIZZLE_SNS_REQ(isp, reqp);
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
mbs.param[0] = MBOX_SEND_SNS;
mbs.param[1] = SNS_GAN_REQ_SIZE >> 1;
mbs.param[2] = DMA_MSW(fcp->isp_scdma);
mbs.param[3] = DMA_LSW(fcp->isp_scdma);
mbs.param[6] = 0;
mbs.param[7] = 0;
MemoryBarrier();
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
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
ISP_UNSWIZZLE_SNS_RSP(isp, resp, SNS_GAN_RESP_SIZE >> 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
portid = (((u_int32_t) resp->snscb_port_id[0]) << 16) |
(((u_int32_t) resp->snscb_port_id[1]) << 8) |
(((u_int32_t) resp->snscb_port_id[2]));
if (isp_async(isp, ISPASYNC_FABRIC_DEV, resp)) {
return (-1);
}
if (first_nz_portid == 0 && portid) {
first_nz_portid = portid;
}
if (first_nz_portid == portid) {
return (0);
}
}
/*
* We either have a broken name server or a huge fabric if we get here.
*/
return (0);
}
#endif
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
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
int32_t
ispscsicmd(xs)
ISP_SCSI_XFER_T *xs;
{
struct ispsoftc *isp;
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
u_int16_t iptr, optr;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
union {
ispreq_t *_reqp;
ispreqt2_t *_t2reqp;
} _u;
#define reqp _u._reqp
#define t2reqp _u._t2reqp
#define UZSIZE max(sizeof (ispreq_t), sizeof (ispreqt2_t))
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
int target, i;
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);
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) {
PRINTF("%s: adapter not ready\n", isp->isp_name);
XS_SETERR(xs, HBA_BOTCH);
return (CMD_COMPLETE);
}
/*
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
* We *could* do the different sequence type that has close
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
* to the whole Queue Entry for the command...
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
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 (XS_CDBLEN(xs) > (IS_FC(isp) ? 16 : 12) || XS_CDBLEN(xs) == 0) {
PRINTF("%s: unsupported cdb length (%d, CDB[0]=0x%x)\n",
isp->isp_name, XS_CDBLEN(xs), XS_CDBP(xs)[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
XS_SETERR(xs, HBA_BOTCH);
return (CMD_COMPLETE);
}
/*
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
* Check to see whether we have good firmware state still or
* need to refresh our port database for this target.
*/
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)) {
fcparam *fcp = isp->isp_param;
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
struct lportdb *lp;
#if defined(ISP2100_FABRIC)
if (target >= FL_PORT_ID) {
/*
* If we're not on a Fabric, we can't have a target
* above FL_PORT_ID-1. If we're on a fabric, we
* can't have a target less than FC_SNS_ID+1.
*/
if (fcp->isp_onfabric == 0 || target <= FC_SNS_ID) {
XS_SETERR(xs, HBA_SELTIMEOUT);
return (CMD_COMPLETE);
}
}
#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
/*
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
* Check for f/w being in ready state. If the f/w
* isn't in ready state, then we don't know our
* loop ID and the f/w hasn't completed logging
* into all targets on the loop. If this is the
* case, then bounce the command. We pretend this is
* a SELECTION TIMEOUT error if we've never gone to
* FW_READY state at all- in this case we may not
* be hooked to a loop at all and we shouldn't hang
* the machine for this. Otherwise, defer this command
* until later.
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 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 (isp_fclink_test(isp, FC_FW_READY_DELAY)) {
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
XS_SETERR(xs, HBA_SELTIMEOUT);
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 (fcp->loop_seen_once) {
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
return (CMD_RQLATER);
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
} else {
return (CMD_COMPLETE);
}
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
/*
* If our loop state is such that we haven't yet received
* a "Port Database Changed" notification (after a LIP or
* a Loop Reset or firmware initialization), then defer
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
* sending commands for a little while, but only if we've
* seen a valid loop at one point (otherwise we can get
* stuck at initialization time).
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 (fcp->isp_loopstate < LOOP_PDB_RCVD) {
XS_SETERR(xs, HBA_SELTIMEOUT);
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
if (fcp->loop_seen_once) {
return (CMD_RQLATER);
} else {
return (CMD_COMPLETE);
}
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
}
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
* If our loop state is now such that we've just now
* received a Port Database Change notification, then
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 have to go off and (re)synchronize our port
* database.
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
if (fcp->isp_loopstate == LOOP_PDB_RCVD) {
if (isp_pdb_sync(isp, target)) {
XS_SETERR(xs, HBA_SELTIMEOUT);
return (CMD_COMPLETE);
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
/*
* Now check whether we should even think about pursuing this.
*/
lp = &fcp->portdb[target];
if (lp->valid == 0) {
XS_SETERR(xs, HBA_SELTIMEOUT);
return (CMD_COMPLETE);
}
if ((lp->roles & (SVC3_TGT_ROLE >> SVC3_ROLE_SHIFT)) == 0) {
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
IDPRINTF(3, ("%s: target %d is not a target\n",
isp->isp_name, target));
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);
}
/*
* Now turn target into what the actual loop ID is.
*/
target = lp->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
}
/*
* Next check to see if any HBA or Device
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
* parameters need to be updated.
*/
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_update != 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
isp_update(isp);
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
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 (isp_getrqentry(isp, &iptr, &optr, (void **) &reqp)) {
IDPRINTF(1, ("%s: Request Queue Overflow\n", isp->isp_name));
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);
}
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.
*/
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (isp->isp_sendmarker) {
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
u_int8_t n = (IS_DUALBUS(isp)? 2: 1);
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
/*
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
* Check ports to send markers for...
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
*/
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
for (i = 0; i < n; i++) {
if ((isp->isp_sendmarker & (1 << i)) == 0) {
continue;
}
MEMZERO((void *) reqp, sizeof (*reqp));
reqp->req_header.rqs_entry_count = 1;
reqp->req_header.rqs_entry_type = RQSTYPE_MARKER;
reqp->req_modifier = SYNC_ALL;
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
reqp->req_target = i << 7; /* insert bus number */
ISP_SWIZZLE_REQUEST(isp, reqp);
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
MemoryBarrier();
ISP_ADD_REQUEST(isp, iptr);
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
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 (isp_getrqentry(isp, &iptr, &optr, (void **)&reqp)) {
IDPRINTF(1, ("%s: Request Queue Overflow+\n",
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_name));
XS_SETERR(xs, HBA_BOTCH);
return (CMD_EAGAIN);
}
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
MEMZERO((void *) reqp, UZSIZE);
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;
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)) {
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 {
reqp->req_header.rqs_entry_type = RQSTYPE_REQUEST;
}
reqp->req_header.rqs_flags = 0;
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
reqp->req_header.rqs_seqno = 0;
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)) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
/*
* See comment in isp_intr
*/
XS_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
*/
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 (XS_CANTAG(xs)) {
t2reqp->req_flags = XS_KINDOF_TAG(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 {
if (XS_CDBP(xs)[0] == 0x3) /* REQUEST SENSE */
t2reqp->req_flags = REQFLAG_HTAG;
else
t2reqp->req_flags = REQFLAG_OTAG;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
} else {
sdparam *sdp = (sdparam *)isp->isp_param;
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 ((sdp->isp_devparam[target].cur_dflags & DPARM_TQING) &&
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_CANTAG(xs)) {
reqp->req_flags = XS_KINDOF_TAG(xs);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +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
reqp->req_target = target | (XS_CHANNEL(xs) << 7);
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)) {
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);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
reqp->req_cdblen = XS_CDBLEN(xs);
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 {
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 ISP2100_SCCLUN
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
t2reqp->req_scclun = XS_LUN(xs);
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
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
t2reqp->req_lun_trn = XS_LUN(xs);
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 support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
Suggested by bde@freebsd.org- memcpy not necessarily good to use. D'oh- not in the BSD DKI. Stop being lazy and finish the defines so MEMCPY becomes bzero for FreeBSD.
1999-01-10 11:15:23 +00:00
MEMCPY(reqp->req_cdb, XS_CDBP(xs), XS_CDBLEN(xs));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
reqp->req_time = XS_TIME(xs) / 1000;
if (reqp->req_time == 0 && XS_TIME(xs))
reqp->req_time = 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
/*
* Always give a bit more leeway to commands after a bus reset.
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
* XXX: DOES NOT DISTINGUISH WHICH PORT MAY HAVE BEEN SYNCED
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
*/
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
if (isp->isp_sendmarker && reqp->req_time < 5) {
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
reqp->req_time = 5;
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
}
if (isp_save_xs(isp, xs, &reqp->req_handle)) {
IDPRINTF(2, ("%s: out of xflist pointers\n", isp->isp_name));
XS_SETERR(xs, HBA_BOTCH);
return (CMD_EAGAIN);
}
/*
* Set up DMA and/or do any bus swizzling of the request entry
* so that the Qlogic F/W understands what is being asked of it.
*/
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
i = ISP_DMASETUP(isp, xs, reqp, &iptr, optr);
if (i != CMD_QUEUED) {
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, reqp->req_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.
*/
return (i);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
XS_SETERR(xs, HBA_NOERROR);
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
IDPRINTF(5, ("%s(%d.%d.%d): START cmd 0x%x datalen %d\n",
isp->isp_name, XS_CHANNEL(xs), target, XS_LUN(xs),
reqp->req_cdb[0], XS_XFRLEN(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
MemoryBarrier();
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
ISP_ADD_REQUEST(isp, iptr);
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++;
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 (isp->isp_sendmarker)
isp->isp_sendmarker = 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
return (CMD_QUEUED);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
#undef reqp
#undef t2reqp
}
/*
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
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(isp, ctl, arg)
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
struct ispsoftc *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
ispctl_t ctl;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
void *arg;
{
ISP_SCSI_XFER_T *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
mbreg_t mbs;
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
int bus, tgt;
u_int32_t handle;
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
switch (ctl) {
default:
PRINTF("%s: isp_control unknown control op %x\n",
isp->isp_name, ctl);
break;
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.
*/
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[0] = MBOX_BUS_RESET;
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)) {
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[1] =
((sdparam *) isp->isp_param)->isp_bus_reset_delay;
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[1] < 2)
mbs.param[1] = 2;
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
bus = *((int *) arg);
mbs.param[2] = bus;
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
} else {
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
mbs.param[1] = 10;
remember to initialize mailbox 2 for FC isp bus resets
1999-10-22 17:03:03 +00:00
mbs.param[2] = 0;
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
bus = 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
}
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_sendmarker = 1 << 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
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
isp_dumpregs(isp, "isp_control SCSI bus reset failed");
break;
}
turn some messages into CFGPRINT messages
1999-12-03 06:55:39 +00:00
CFGPRINTF("%s: driver initiated bus reset of bus %d\n",
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_name, 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
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:
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
tgt = (*((int *) arg)) & 0xffff;
bus = (*((int *) arg)) >> 16;
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[0] = MBOX_ABORT_TARGET;
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] = (tgt << 8) | (bus << 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] = 3; /* 'delay', in seconds */
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_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
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
PRINTF("%s: isp_control MBOX_RESET_DEV failure (code "
"%x)\n", isp->isp_name, mbs.param[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
break;
}
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
PRINTF("%s: Target %d on Bus %d Reset Succeeded\n",
isp->isp_name, tgt, bus);
isp->isp_sendmarker = 1 << 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
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:
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 = (ISP_SCSI_XFER_T *) arg;
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) {
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
PRINTF("%s: isp_control- cannot find command to abort "
"in active list\n", isp->isp_name);
break;
}
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
bus = XS_CHANNEL(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
mbs.param[0] = MBOX_ABORT;
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)) {
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 ISP2100_SCCLUN
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
mbs.param[1] = XS_TGT(xs) << 8;
mbs.param[4] = 0;
mbs.param[5] = 0;
mbs.param[6] = XS_LUN(xs);
#else
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
mbs.param[1] = XS_TGT(xs) << 8 | XS_LUN(xs);
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
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 {
mbs.param[1] =
(bus << 15) | (XS_TGT(xs) << 8) | XS_LUN(xs);
}
mbs.param[2] = handle >> 16;
mbs.param[3] = handle & 0xffff;
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_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
PRINTF("%s: isp_control MBOX_ABORT failure (code %x)\n",
isp->isp_name, mbs.param[0]);
break;
}
PRINTF("%s: command for target %d lun %d was aborted\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
return (0);
case ISPCTL_UPDATE_PARAMS:
isp_update(isp);
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:
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 (isp_fclink_test(isp, FC_FW_READY_DELAY));
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
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:
return (isp_pdb_sync(isp, -1));
#ifdef ISP_TARGET_MODE
case ISPCTL_TOGGLE_TMODE:
if (IS_SCSI(isp)) {
int ena = *(int *)arg;
mbs.param[0] = MBOX_ENABLE_TARGET_MODE;
mbs.param[1] = (ena)? ENABLE_TARGET_FLAG : 0;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
PRINTF("%s: cannot %sable target mode (0x%x)\n",
isp->isp_name, ena? "en" : "dis",
mbs.param[0]);
break;
}
}
return (0);
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
#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
}
return (-1);
}
/*
* Interrupt Service Routine(s).
*
* External (OS) framework has done the appropriate locking,
* and the locking will be held throughout this function.
*/
int
isp_intr(arg)
void *arg;
{
ISP_SCSI_XFER_T *complist[RESULT_QUEUE_LEN], *xs;
struct ispsoftc *isp = arg;
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
u_int16_t iptr, optr;
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
u_int16_t isr, isrb, sema;
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
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
/*
* Well, if we've disabled interrupts, we may get a case where
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
* isr isn't set, but sema is. In any case, debounce isr reads.
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
*/
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
do {
isr = ISP_READ(isp, BIU_ISR);
isrb = ISP_READ(isp, BIU_ISR);
} while (isr != isrb);
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
sema = ISP_READ(isp, BIU_SEMA) & 0x1;
IDPRINTF(5, ("%s: isp_intr isr %x sem %x\n", isp->isp_name, isr, sema));
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
if (isr == 0) {
return (0);
}
if (!INT_PENDING(isp, isr)) {
IDPRINTF(4, ("%s: isp_intr isr=%x\n", isp->isp_name, isr));
return (0);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +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
if (isp->isp_state != ISP_RUNSTATE) {
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
IDPRINTF(3, ("%s: interrupt (isr=%x,sema=%x) when not ready\n",
isp->isp_name, isr, sema));
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_WRITE(isp, INMAILBOX5, ISP_READ(isp, OUTMAILBOX5));
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
ISP_WRITE(isp, BIU_SEMA, 0);
ENABLE_INTS(isp);
return (1);
}
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
if (sema) {
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
u_int16_t mbox = ISP_READ(isp, OUTMAILBOX0);
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 (mbox & 0x4000) {
change debug printout lefvels for a couple of places
2000-01-09 21:47:39 +00:00
IDPRINTF(4, ("%s: Command Mbox 0x%x\n",
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
isp->isp_name, mbox));
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
} else {
u_int32_t fhandle = isp_parse_async(isp, (int) mbox);
change debug printout lefvels for a couple of places
2000-01-09 21:47:39 +00:00
IDPRINTF(4, ("%s: Async Mbox 0x%x\n",
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_name, mbox));
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 (fhandle > 0) {
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_fastpost_complete(isp, fhandle);
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
}
}
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_WRITE(isp, BIU_SEMA, 0);
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
ENABLE_INTS(isp);
return (1);
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
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
/*
* You *must* read OUTMAILBOX5 prior to clearing the RISC 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
optr = isp->isp_residx;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
iptr = ISP_READ(isp, OUTMAILBOX5);
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_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (optr == iptr) {
IDPRINTF(4, ("why intr? isr %x iptr %x optr %x\n",
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
isr, optr, iptr));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
while (optr != iptr) {
ispstatusreq_t *sp;
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
u_int16_t oop;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
int buddaboom = 0;
sp = (ispstatusreq_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;
optr = ISP_NXT_QENTRY(optr, RESULT_QUEUE_LEN);
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++;
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
MemoryBarrier();
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
/*
* Do any appropriate unswizzling of what the Qlogic f/w has
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
* written into memory so it makes sense to us. This is a
* per-platform thing.
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_UNSWIZZLE_RESPONSE(isp, sp);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (sp->req_header.rqs_entry_type != RQSTYPE_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
if (isp_handle_other_response(isp, sp, &optr) == 0) {
continue;
}
/*
* 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 support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (sp->req_header.rqs_entry_type != RQSTYPE_REQUEST) {
temp fix for internal queue overflow problem
1999-04-14 17:37:36 +00:00
PRINTF("%s: not RESPONSE in RESPONSE Queue "
"(type 0x%x) @ idx %d (next %d)\n",
isp->isp_name,
sp->req_header.rqs_entry_type, oop, optr);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
continue;
}
buddaboom = 1;
}
if (sp->req_header.rqs_flags & 0xf) {
temp fix for internal queue overflow problem
1999-04-14 17:37:36 +00:00
#define _RQS_OFLAGS \
~(RQSFLAG_CONTINUATION|RQSFLAG_FULL|RQSFLAG_BADHEADER|RQSFLAG_BADPACKET)
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) {
change debug printout lefvels for a couple of places
2000-01-09 21:47:39 +00:00
IDPRINTF(4, ("%s: continuation segment\n",
temp fix for internal queue overflow problem
1999-04-14 17:37:36 +00:00
isp->isp_name));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
ISP_WRITE(isp, INMAILBOX5, optr);
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) {
temp fix for internal queue overflow problem
1999-04-14 17:37:36 +00:00
IDPRINTF(2, ("%s: internal queues full\n",
isp->isp_name));
/*
* 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) {
PRINTF("%s: bad header\n", isp->isp_name);
buddaboom++;
}
if (sp->req_header.rqs_flags & RQSFLAG_BADPACKET) {
PRINTF("%s: bad request packet\n",
isp->isp_name);
buddaboom++;
}
temp fix for internal queue overflow problem
1999-04-14 17:37:36 +00:00
if (sp->req_header.rqs_flags & _RQS_OFLAGS) {
PRINTF("%s: unknown flags in response (0x%x)\n",
isp->isp_name, sp->req_header.rqs_flags);
buddaboom++;
}
#undef _RQS_OFLAGS
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +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
if (sp->req_handle > isp->isp_maxcmds || sp->req_handle < 1) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
PRINTF("%s: bad request handle %d\n", isp->isp_name,
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
sp->req_handle);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
ISP_WRITE(isp, INMAILBOX5, optr);
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) {
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
PRINTF("%s: NULL xs in xflist (handle 0x%x)\n",
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_name, sp->req_handle);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
ISP_WRITE(isp, INMAILBOX5, optr);
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
isp_destroy_handle(isp, sp->req_handle);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (sp->req_status_flags & RQSTF_BUS_RESET) {
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_sendmarker |= (1 << XS_CHANNEL(xs));
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);
}
XS_STS(xs) = sp->req_scsi_status & 0xff;
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_SCSI(isp)) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (sp->req_state_flags & RQSF_GOT_SENSE) {
Suggested by bde@freebsd.org- memcpy not necessarily good to use. D'oh- not in the BSD DKI. Stop being lazy and finish the defines so MEMCPY becomes bzero for FreeBSD.
1999-01-10 11:15:23 +00:00
MEMCPY(XS_SNSP(xs), sp->req_sense_data,
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
XS_SNSLEN(xs));
XS_SNS_IS_VALID(xs);
}
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 new synchronous rate was negotiated for this
* target. Mark state such that we'll go look up
* that which has changed later.
*/
if (sp->req_status_flags & RQSTF_NEGOTIATION) {
sdparam *sdp = isp->isp_param;
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 += XS_CHANNEL(xs);
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_devparam[XS_TGT(xs)].dev_refresh = 1;
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_update |= (1 << XS_CHANNEL(xs));
temp fix for internal queue overflow problem
1999-04-14 17:37:36 +00:00
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
} else {
Make Fibre Channel cards correctly note the presence/absence of ARQ data and punt the dealing with its presence/absence to the platform layers.
2000-01-04 03:44:21 +00:00
if (sp->req_scsi_status & RQCS_SV) {
int amt = min(XS_SNSLEN(xs), sp->req_sense_len);
MEMCPY(XS_SNSP(xs), sp->req_sense_data, amt);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
XS_SNS_IS_VALID(xs);
sp->req_state_flags |= RQSF_GOT_SENSE;
Make Fibre Channel cards correctly note the presence/absence of ARQ data and punt the dealing with its presence/absence to the platform layers.
2000-01-04 03:44:21 +00:00
} else if (XS_STS(xs) == SCSI_CHECK) {
IDPRINTF(1, ("%s: check condition with no sense"
" data\n", isp->isp_name));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
}
if (XS_NOERR(xs) && XS_STS(xs) == SCSI_BUSY) {
XS_SETERR(xs, HBA_TGTBSY);
}
if (sp->req_header.rqs_entry_type == RQSTYPE_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
if (XS_NOERR(xs)) {
if (sp->req_completion_status != RQCS_COMPLETE) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp_parse_status(isp, sp, 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
} else {
XS_SETERR(xs, HBA_NOERROR);
}
}
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 (sp->req_header.rqs_entry_type == RQSTYPE_REQUEST) {
if (sp->req_header.rqs_flags & RQSFLAG_FULL) {
/*
* Force Queue Full status.
*/
XS_STS(xs) = SCSI_QFULL;
XS_SETERR(xs, HBA_NOERROR);
} else 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
} 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
PRINTF("%s: unhandled respose queue type 0x%x\n",
isp->isp_name, sp->req_header.rqs_entry_type);
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
}
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_SCSI(isp)) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
XS_RESID(xs) = sp->req_resid;
} else if (sp->req_scsi_status & RQCS_RU) {
XS_RESID(xs) = sp->req_resid;
IDPRINTF(4, ("%s: cnt %d rsd %d\n", isp->isp_name,
XS_XFRLEN(xs), sp->req_resid));
}
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
}
Make Fibre Channel cards correctly note the presence/absence of ARQ data and punt the dealing with its presence/absence to the platform layers.
2000-01-04 03:44:21 +00:00
/*
* Let the platforms cope.
*/
#if 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
/*
* XXX: If we have a check condition, but no Sense Data,
* XXX: mark it as an error (ARQ failed). We need to
* XXX: to do a more distinct job because there may
* XXX: cases where ARQ is disabled.
*/
if (XS_STS(xs) == SCSI_CHECK && !(XS_IS_SNS_VALID(xs))) {
if (XS_NOERR(xs)) {
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
PRINTF("%s: ARQ failure for target %d lun %d\n",
isp->isp_name, XS_TGT(xs), 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
XS_SETERR(xs, HBA_ARQFAIL);
}
}
Make Fibre Channel cards correctly note the presence/absence of ARQ data and punt the dealing with its presence/absence to the platform layers.
2000-01-04 03:44:21 +00:00
#endif
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if ((isp->isp_dblev >= 5) ||
(isp->isp_dblev > 2 && !XS_NOERR(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
PRINTF("%s(%d.%d): FIN dl%d resid%d STS %x",
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp->isp_name, XS_TGT(xs), 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
XS_XFRLEN(xs), XS_RESID(xs), XS_STS(xs));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (sp->req_state_flags & RQSF_GOT_SENSE) {
PRINTF(" Skey: %x", XS_SNSKEY(xs));
if (!(XS_IS_SNS_VALID(xs))) {
PRINTF(" BUT NOT SET");
}
}
fix reported compile error flying blind- I do not have the new compiler yet
1998-09-15 22:44:51 +00:00
PRINTF(" XS_ERR=0x%x\n", (unsigned int) XS_ERR(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
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
if (isp->isp_nactive > 0)
isp->isp_nactive--;
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 */
}
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) {
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, INMAILBOX5, optr);
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
isp->isp_reqodx = ISP_READ(isp, OUTMAILBOX4);
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) {
XS_CMD_DONE(xs);
}
}
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
ENABLE_INTS(isp);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
return (1);
}
/*
* Support routines.
*/
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
isp_parse_async(isp, mbox)
struct ispsoftc *isp;
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
int 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
{
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
int bus;
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
u_int32_t fast_post_handle = 0;
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
if (IS_DUALBUS(isp)) {
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
bus = ISP_READ(isp, OUTMAILBOX6);
} else {
bus = 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
switch (mbox) {
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 MBOX_COMMAND_COMPLETE: /* sometimes these show up */
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_BUS_RESET:
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_sendmarker = (1 << bus);
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
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
isp_target_async(isp, bus, mbox);
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
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
isp_async(isp, ISPASYNC_BUS_RESET, &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
break;
case ASYNC_SYSTEM_ERROR:
mbox = ISP_READ(isp, OUTMAILBOX1);
PRINTF("%s: Internal FW Error @ RISC Addr 0x%x\n",
isp->isp_name, mbox);
isp_restart(isp);
/* no point continuing after this */
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
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
case ASYNC_RQS_XFER_ERR:
PRINTF("%s: Request Queue Transfer Error\n", isp->isp_name);
break;
case ASYNC_RSP_XFER_ERR:
PRINTF("%s: Response Queue Transfer Error\n", isp->isp_name);
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.
*/
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
mbox = ISP_READ(isp, OUTMAILBOX4);
break;
case ASYNC_TIMEOUT_RESET:
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
PRINTF("%s: timeout initiated SCSI bus reset of bus %d\n",
isp->isp_name, bus);
isp->isp_sendmarker = (1 << bus);
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
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
isp_target_async(isp, bus, mbox);
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:
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
PRINTF("%s: device reset on bus %d\n", isp->isp_name, bus);
isp->isp_sendmarker = 1 << bus;
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
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
isp_target_async(isp, bus, mbox);
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:
PRINTF("%s: extended message underrun\n", isp->isp_name);
break;
case ASYNC_SCAM_INT:
PRINTF("%s: SCAM interrupt\n", isp->isp_name);
break;
case ASYNC_HUNG_SCSI:
PRINTF("%s: stalled SCSI Bus after DATA Overrun\n",
isp->isp_name);
/* XXX: Need to issue SCSI reset at this point */
break;
case ASYNC_KILLED_BUS:
PRINTF("%s: SCSI Bus reset after DATA Overrun\n",
isp->isp_name);
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);
switch (mbox & 0x1c00) {
case SXP_PINS_LVD_MODE:
PRINTF("%s: Transition to LVD mode\n", isp->isp_name);
((sdparam *)isp->isp_param)->isp_diffmode = 0;
((sdparam *)isp->isp_param)->isp_ultramode = 0;
((sdparam *)isp->isp_param)->isp_lvdmode = 1;
break;
case SXP_PINS_HVD_MODE:
PRINTF("%s: Transition to Differential mode\n",
isp->isp_name);
((sdparam *)isp->isp_param)->isp_diffmode = 1;
((sdparam *)isp->isp_param)->isp_ultramode = 0;
((sdparam *)isp->isp_param)->isp_lvdmode = 0;
break;
case SXP_PINS_SE_MODE:
PRINTF("%s: Transition to Single Ended mode\n",
isp->isp_name);
((sdparam *)isp->isp_param)->isp_diffmode = 0;
((sdparam *)isp->isp_param)->isp_ultramode = 1;
((sdparam *)isp->isp_param)->isp_lvdmode = 0;
break;
default:
PRINTF("%s: Transition to unknown mode 0x%x\n",
isp->isp_name, mbox);
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... */
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
isp->isp_sendmarker = (1 << 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
break;
case ASYNC_CMD_CMPLT:
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
fast_post_handle = (ISP_READ(isp, OUTMAILBOX2) << 16) |
ISP_READ(isp, OUTMAILBOX1);
change debug printout lefvels for a couple of places
2000-01-09 21:47:39 +00:00
IDPRINTF(4, ("%s: fast post completion of %u\n", isp->isp_name,
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
fast_post_handle));
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_CTIO_DONE:
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
/* Should only occur when Fast Posting Set for 2100s */
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
PRINTF("%s: CTIO done\n", isp->isp_name);
break;
case ASYNC_LIP_OCCURRED:
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
((fcparam *) isp->isp_param)->isp_lipseq =
ISP_READ(isp, OUTMAILBOX1);
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
((fcparam *) isp->isp_param)->isp_fwstate = FW_CONFIG_WAIT;
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
((fcparam *) isp->isp_param)->isp_loopstate = LOOP_LIP_RCVD;
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
isp->isp_sendmarker = 1;
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_mark_getpdb_all(isp);
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
IDPRINTF(1, ("%s: LIP occurred\n", isp->isp_name));
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
#ifdef ISP_TARGET_MODE
isp_target_async(isp, bus, mbox);
#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_LOOP_UP:
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
isp->isp_sendmarker = 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
((fcparam *) isp->isp_param)->isp_fwstate = FW_CONFIG_WAIT;
((fcparam *) isp->isp_param)->isp_loopstate = LOOP_LIP_RCVD;
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_mark_getpdb_all(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
isp_async(isp, ISPASYNC_LOOP_UP, NULL);
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
#ifdef ISP_TARGET_MODE
isp_target_async(isp, bus, mbox);
#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_LOOP_DOWN:
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_sendmarker = 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
((fcparam *) isp->isp_param)->isp_fwstate = FW_CONFIG_WAIT;
((fcparam *) isp->isp_param)->isp_loopstate = LOOP_NIL;
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_mark_getpdb_all(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
isp_async(isp, ISPASYNC_LOOP_DOWN, NULL);
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
#ifdef ISP_TARGET_MODE
isp_target_async(isp, bus, mbox);
#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_LOOP_RESET:
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
isp->isp_sendmarker = 1 << bus;
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
((fcparam *) isp->isp_param)->isp_fwstate = FW_CONFIG_WAIT;
((fcparam *) isp->isp_param)->isp_loopstate = LOOP_NIL;
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_mark_getpdb_all(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
PRINTF("%s: Loop RESET\n", isp->isp_name);
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
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
isp_target_async(isp, bus, mbox);
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_PDB_CHANGED:
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
isp->isp_sendmarker = 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
((fcparam *) isp->isp_param)->isp_loopstate = LOOP_PDB_RCVD;
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
isp_mark_getpdb_all(isp);
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
IDPRINTF(2, ("%s: Port Database Changed\n", isp->isp_name));
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_CHANGE_NOTIFY:
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
isp_mark_getpdb_all(isp);
/*
* Not correct, but it will force us to rescan the loop.
*/
((fcparam *) isp->isp_param)->isp_loopstate = LOOP_PDB_RCVD;
isp_async(isp, ISPASYNC_CHANGE_NOTIFY, NULL);
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;
default:
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
PRINTF("%s: unknown async code 0x%x\n", isp->isp_name, 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;
}
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
return (fast_post_handle);
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
isp_handle_other_response(isp, sp, optrp)
struct ispsoftc *isp;
ispstatusreq_t *sp;
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
u_int16_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
{
switch (sp->req_header.rqs_entry_type) {
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:
#ifdef ISP_TARGET_MODE
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
return (isp_target_notify(isp, sp, optrp));
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
/* 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
#endif
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:
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
PRINTF("%s: unhandled response type 0x%x\n", isp->isp_name,
sp->req_header.rqs_entry_type);
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);
}
}
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
isp_parse_status(isp, sp, xs)
struct ispsoftc *isp;
ispstatusreq_t *sp;
ISP_SCSI_XFER_T *xs;
{
switch (sp->req_completion_status) {
case RQCS_COMPLETE:
XS_SETERR(xs, HBA_NOERROR);
return;
case RQCS_INCOMPLETE:
if ((sp->req_state_flags & RQSF_GOT_TARGET) == 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
IDPRINTF(3, ("%s: Selection Timeout for target %d\n",
isp->isp_name, XS_TGT(xs)));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
XS_SETERR(xs, HBA_SELTIMEOUT);
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
PRINTF("%s: command incomplete for target %d lun %d, state "
"0x%x\n", isp->isp_name, XS_TGT(xs), XS_LUN(xs),
sp->req_state_flags);
break;
case RQCS_DMA_ERROR:
PRINTF("%s: DMA error for command on target %d, lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
case RQCS_TRANSPORT_ERROR:
PRINTF("%s: transport error\n", isp->isp_name);
isp_prtstst(sp);
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 RQCS_RESET_OCCURRED:
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
IDPRINTF(2, ("%s: bus %d reset destroyed command for target %d "
"lun %d\n", isp->isp_name, XS_CHANNEL(xs), XS_TGT(xs),
XS_LUN(xs)));
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
/*
* XXX: Get port number for bus
*/
isp->isp_sendmarker = 3;
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_BUSRESET);
return;
case RQCS_ABORTED:
PRINTF("%s: command aborted for target %d lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
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
/*
* XXX: Get port number for bus
*/
isp->isp_sendmarker = 3;
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_ABORTED);
return;
case RQCS_TIMEOUT:
IDPRINTF(2, ("%s: command timed out for target %d lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs)));
XS_SETERR(xs, HBA_CMDTIMEOUT);
return;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
case RQCS_DATA_OVERRUN:
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)) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
XS_RESID(xs) = sp->req_resid;
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
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:
PRINTF("%s: command overrun for command on target %d, lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_STATUS_OVERRUN:
PRINTF("%s: status overrun for command on target %d, lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_BAD_MESSAGE:
PRINTF("%s: message not COMMAND COMPLETE after status on "
"target %d, lun %d\n", isp->isp_name, XS_TGT(xs),
XS_LUN(xs));
break;
case RQCS_NO_MESSAGE_OUT:
PRINTF("%s: No MESSAGE OUT phase after selection on "
"target %d, lun %d\n", isp->isp_name, XS_TGT(xs),
XS_LUN(xs));
break;
case RQCS_EXT_ID_FAILED:
PRINTF("%s: EXTENDED IDENTIFY failed on target %d, lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_IDE_MSG_FAILED:
PRINTF("%s: target %d lun %d rejected INITIATOR DETECTED "
"ERROR message\n", isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_ABORT_MSG_FAILED:
PRINTF("%s: target %d lun %d rejected ABORT message\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_REJECT_MSG_FAILED:
PRINTF("%s: target %d lun %d rejected MESSAGE REJECT message\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_NOP_MSG_FAILED:
PRINTF("%s: target %d lun %d rejected NOP message\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_PARITY_ERROR_MSG_FAILED:
PRINTF("%s: target %d lun %d rejected MESSAGE PARITY ERROR "
"message\n", isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_DEVICE_RESET_MSG_FAILED:
PRINTF("%s: target %d lun %d rejected BUS DEVICE RESET "
"message\n", isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_ID_MSG_FAILED:
PRINTF("%s: target %d lun %d rejected IDENTIFY "
"message\n", isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_UNEXP_BUS_FREE:
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
PRINTF("%s: target %d lun %d had an unexpected bus free\n",
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_name, XS_TGT(xs), XS_LUN(xs));
break;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
case RQCS_DATA_UNDERRUN:
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)) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
XS_RESID(xs) = sp->req_resid;
/* an UNDERRUN is not a botch ??? */
}
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
case RQCS_XACT_ERR1:
PRINTF("%s: HBA attempted queued transaction with disconnect "
"not set for target %d lun %d\n", isp->isp_name, XS_TGT(xs),
XS_LUN(xs));
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:
PRINTF("%s: HBA attempted queued transaction to target "
"routine %d on target %d\n", isp->isp_name, XS_LUN(xs),
XS_TGT(xs));
break;
case RQCS_XACT_ERR3:
PRINTF("%s: HBA attempted queued transaction for target %d lun "
"%d when queueing disabled\n", isp->isp_name, XS_TGT(xs),
XS_LUN(xs));
break;
case RQCS_BAD_ENTRY:
PRINTF("%s: invalid IOCB entry type detected\n", isp->isp_name);
break;
case RQCS_QUEUE_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
IDPRINTF(3, ("%s: internal queues full for target %d lun %d "
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
"status 0x%x\n", isp->isp_name, XS_TGT(xs), XS_LUN(xs),
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_STS(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
/*
* If QFULL or some other status byte is set, then this
* isn't an error, per se.
*/
if (XS_STS(xs) != 0) {
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
break;
case RQCS_PHASE_SKIPPED:
PRINTF("%s: SCSI phase skipped (e.g., COMMAND COMPLETE w/o "
"STATUS phase) for target %d lun %d\n", isp->isp_name,
XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_ARQS_FAILED:
PRINTF("%s: Auto Request Sense failed for target %d lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
XS_SETERR(xs, HBA_ARQFAIL);
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_WIDE_FAILED:
PRINTF("%s: Wide Negotiation failed for target %d lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
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_SCSI(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
sdparam *sdp = isp->isp_param;
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 += XS_CHANNEL(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
sdp->isp_devparam[XS_TGT(xs)].dev_flags &= ~DPARM_WIDE;
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_devparam[XS_TGT(xs)].dev_update = 1;
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_update = XS_CHANNEL(xs)+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
}
XS_SETERR(xs, HBA_NOERROR);
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_SYNCXFER_FAILED:
PRINTF("%s: SDTR Message failed for target %d lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
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_SCSI(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
sdparam *sdp = isp->isp_param;
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 += XS_CHANNEL(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
sdp->isp_devparam[XS_TGT(xs)].dev_flags &= ~DPARM_SYNC;
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_devparam[XS_TGT(xs)].dev_update = 1;
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_update = XS_CHANNEL(xs)+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;
case RQCS_LVD_BUSERR:
PRINTF("%s: Bad LVD Bus condition while talking to target %d "
"lun %d\n", isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
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
*/
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
IDPRINTF(3, ("%s: Port Unavailable for target %d\n",
isp->isp_name, XS_TGT(xs)));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
XS_SETERR(xs, HBA_SELTIMEOUT);
return;
case RQCS_PORT_LOGGED_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
/*
* It was there (maybe)- treat as a selection timeout.
*/
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
IDPRINTF(2, ("%s: port logout for target %d\n",
isp->isp_name, XS_TGT(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
XS_SETERR(xs, HBA_SELTIMEOUT);
return;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
case RQCS_PORT_CHANGED:
PRINTF("%s: port changed for target %d\n",
isp->isp_name, XS_TGT(xs));
break;
case RQCS_PORT_BUSY:
PRINTF("%s: port busy for target %d\n",
isp->isp_name, XS_TGT(xs));
XS_SETERR(xs, HBA_TGTBSY);
return;
default:
PRINTF("%s: comp status %x\n", isp->isp_name,
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
sp->req_completion_status);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
}
XS_SETERR(xs, HBA_BOTCH);
}
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
isp_fastpost_complete(isp, fph)
struct ispsoftc *isp;
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
u_int32_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
{
ISP_SCSI_XFER_T *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
if (fph < 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
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) {
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
PRINTF("%s: command for fast posting handle 0x%x not found\n",
isp->isp_name, 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.
*/
XS_RESID(xs) = 0;
XS_STS(xs) = 0;
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
}
XS_CMD_DONE(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
if (isp->isp_nactive)
isp->isp_nactive--;
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
#define HINIB(x) ((x) >> 0x4)
#define LONIB(x) ((x) & 0xf)
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
#define MAKNIB(a, b) (((a) << 4) | (b))
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
static u_int8_t mbpcnt[] = {
MAKNIB(1, 1), /* 0x00: MBOX_NO_OP */
MAKNIB(5, 5), /* 0x01: MBOX_LOAD_RAM */
MAKNIB(2, 0), /* 0x02: MBOX_EXEC_FIRMWARE */
MAKNIB(5, 5), /* 0x03: MBOX_DUMP_RAM */
MAKNIB(3, 3), /* 0x04: MBOX_WRITE_RAM_WORD */
MAKNIB(2, 3), /* 0x05: MBOX_READ_RAM_WORD */
MAKNIB(6, 6), /* 0x06: MBOX_MAILBOX_REG_TEST */
MAKNIB(2, 3), /* 0x07: MBOX_VERIFY_CHECKSUM */
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
MAKNIB(1, 4), /* 0x08: MBOX_ABOUT_FIRMWARE */
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
MAKNIB(0, 0), /* 0x09: */
MAKNIB(0, 0), /* 0x0a: */
MAKNIB(0, 0), /* 0x0b: */
MAKNIB(0, 0), /* 0x0c: */
MAKNIB(0, 0), /* 0x0d: */
MAKNIB(1, 2), /* 0x0e: MBOX_CHECK_FIRMWARE */
MAKNIB(0, 0), /* 0x0f: */
MAKNIB(5, 5), /* 0x10: MBOX_INIT_REQ_QUEUE */
MAKNIB(6, 6), /* 0x11: MBOX_INIT_RES_QUEUE */
MAKNIB(4, 4), /* 0x12: MBOX_EXECUTE_IOCB */
MAKNIB(2, 2), /* 0x13: MBOX_WAKE_UP */
MAKNIB(1, 6), /* 0x14: MBOX_STOP_FIRMWARE */
MAKNIB(4, 4), /* 0x15: MBOX_ABORT */
MAKNIB(2, 2), /* 0x16: MBOX_ABORT_DEVICE */
MAKNIB(3, 3), /* 0x17: MBOX_ABORT_TARGET */
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
MAKNIB(3, 1), /* 0x18: MBOX_BUS_RESET */
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
MAKNIB(2, 3), /* 0x19: MBOX_STOP_QUEUE */
MAKNIB(2, 3), /* 0x1a: MBOX_START_QUEUE */
MAKNIB(2, 3), /* 0x1b: MBOX_SINGLE_STEP_QUEUE */
MAKNIB(2, 3), /* 0x1c: MBOX_ABORT_QUEUE */
MAKNIB(2, 4), /* 0x1d: MBOX_GET_DEV_QUEUE_STATUS */
MAKNIB(0, 0), /* 0x1e: */
MAKNIB(1, 3), /* 0x1f: MBOX_GET_FIRMWARE_STATUS */
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
MAKNIB(1, 4), /* 0x20: MBOX_GET_INIT_SCSI_ID, MBOX_GET_LOOP_ID */
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
MAKNIB(1, 3), /* 0x21: MBOX_GET_SELECT_TIMEOUT */
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
MAKNIB(1, 3), /* 0x22: MBOX_GET_RETRY_COUNT */
MAKNIB(1, 2), /* 0x23: MBOX_GET_TAG_AGE_LIMIT */
MAKNIB(1, 2), /* 0x24: MBOX_GET_CLOCK_RATE */
MAKNIB(1, 2), /* 0x25: MBOX_GET_ACT_NEG_STATE */
MAKNIB(1, 2), /* 0x26: MBOX_GET_ASYNC_DATA_SETUP_TIME */
MAKNIB(1, 3), /* 0x27: MBOX_GET_PCI_PARAMS */
MAKNIB(2, 4), /* 0x28: MBOX_GET_TARGET_PARAMS */
MAKNIB(2, 4), /* 0x29: MBOX_GET_DEV_QUEUE_PARAMS */
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
MAKNIB(1, 2), /* 0x2a: MBOX_GET_RESET_DELAY_PARAMS */
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
MAKNIB(0, 0), /* 0x2b: */
MAKNIB(0, 0), /* 0x2c: */
MAKNIB(0, 0), /* 0x2d: */
MAKNIB(0, 0), /* 0x2e: */
MAKNIB(0, 0), /* 0x2f: */
MAKNIB(2, 2), /* 0x30: MBOX_SET_INIT_SCSI_ID */
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
MAKNIB(2, 3), /* 0x31: MBOX_SET_SELECT_TIMEOUT */
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
MAKNIB(3, 3), /* 0x32: MBOX_SET_RETRY_COUNT */
MAKNIB(2, 2), /* 0x33: MBOX_SET_TAG_AGE_LIMIT */
MAKNIB(2, 2), /* 0x34: MBOX_SET_CLOCK_RATE */
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
MAKNIB(2, 2), /* 0x35: MBOX_SET_ACT_NEG_STATE */
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
MAKNIB(2, 2), /* 0x36: MBOX_SET_ASYNC_DATA_SETUP_TIME */
MAKNIB(3, 3), /* 0x37: MBOX_SET_PCI_CONTROL_PARAMS */
MAKNIB(4, 4), /* 0x38: MBOX_SET_TARGET_PARAMS */
MAKNIB(4, 4), /* 0x39: MBOX_SET_DEV_QUEUE_PARAMS */
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
MAKNIB(1, 2), /* 0x3a: MBOX_SET_RESET_DELAY_PARAMS */
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
MAKNIB(0, 0), /* 0x3b: */
MAKNIB(0, 0), /* 0x3c: */
MAKNIB(0, 0), /* 0x3d: */
MAKNIB(0, 0), /* 0x3e: */
MAKNIB(0, 0), /* 0x3f: */
MAKNIB(1, 2), /* 0x40: MBOX_RETURN_BIOS_BLOCK_ADDR */
MAKNIB(6, 1), /* 0x41: MBOX_WRITE_FOUR_RAM_WORDS */
MAKNIB(2, 3), /* 0x42: MBOX_EXEC_BIOS_IOCB */
MAKNIB(0, 0), /* 0x43: */
MAKNIB(0, 0), /* 0x44: */
MAKNIB(0, 0), /* 0x45: */
MAKNIB(0, 0), /* 0x46: */
MAKNIB(0, 0), /* 0x47: */
MAKNIB(0, 0), /* 0x48: */
MAKNIB(0, 0), /* 0x49: */
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
MAKNIB(2, 1), /* 0x4a: MBOX_SET_FIRMWARE_FEATURES */
MAKNIB(1, 2), /* 0x4b: MBOX_GET_FIRMWARE_FEATURES */
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
MAKNIB(0, 0), /* 0x4c: */
MAKNIB(0, 0), /* 0x4d: */
MAKNIB(0, 0), /* 0x4e: */
MAKNIB(0, 0), /* 0x4f: */
MAKNIB(0, 0), /* 0x50: */
MAKNIB(0, 0), /* 0x51: */
MAKNIB(0, 0), /* 0x52: */
MAKNIB(0, 0), /* 0x53: */
MAKNIB(8, 0), /* 0x54: MBOX_EXEC_COMMAND_IOCB_A64 */
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
MAKNIB(2, 1), /* 0x55: MBOX_ENABLE_TARGET_MODE */
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
MAKNIB(0, 0), /* 0x56: */
MAKNIB(0, 0), /* 0x57: */
MAKNIB(0, 0), /* 0x58: */
MAKNIB(0, 0), /* 0x59: */
MAKNIB(0, 0), /* 0x5a: */
MAKNIB(0, 0), /* 0x5b: */
MAKNIB(0, 0), /* 0x5c: */
MAKNIB(0, 0), /* 0x5d: */
MAKNIB(0, 0), /* 0x5e: */
MAKNIB(0, 0), /* 0x5f: */
MAKNIB(8, 6), /* 0x60: MBOX_INIT_FIRMWARE */
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
MAKNIB(0, 0), /* 0x61: */
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
MAKNIB(2, 1), /* 0x62: MBOX_INIT_LIP */
MAKNIB(8, 1), /* 0x63: MBOX_GET_FC_AL_POSITION_MAP */
MAKNIB(8, 1), /* 0x64: MBOX_GET_PORT_DB */
MAKNIB(3, 1), /* 0x65: MBOX_CLEAR_ACA */
MAKNIB(3, 1), /* 0x66: MBOX_TARGET_RESET */
MAKNIB(3, 1), /* 0x67: MBOX_CLEAR_TASK_SET */
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
MAKNIB(3, 1), /* 0x68: MBOX_ABORT_TASK_SET */
MAKNIB(1, 2), /* 0x69: MBOX_GET_FW_STATE */
MAKNIB(2, 8), /* 0x6a: MBOX_GET_PORT_NAME */
MAKNIB(8, 1), /* 0x6b: MBOX_GET_LINK_STATUS */
MAKNIB(4, 4), /* 0x6c: MBOX_INIT_LIP_RESET */
MAKNIB(0, 0), /* 0x6d: */
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
MAKNIB(8, 2), /* 0x6e: MBOX_SEND_SNS */
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
MAKNIB(4, 3), /* 0x6f: MBOX_FABRIC_LOGIN */
MAKNIB(2, 1), /* 0x70: MBOX_SEND_CHANGE_REQUEST */
MAKNIB(2, 1), /* 0x71: MBOX_FABRIC_LOGOUT */
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
MAKNIB(4, 1) /* 0x72: MBOX_INIT_LIP_LOGIN */
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
};
#define NMBCOM (sizeof (mbpcnt) / sizeof (mbpcnt[0]))
static void
isp_mboxcmd(isp, mbp)
struct ispsoftc *isp;
mbreg_t *mbp;
{
int outparam, inparam;
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 loops, dld = 0;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
u_int8_t opcode;
if (mbp->param[0] == ISP2100_SET_PCI_PARAM) {
opcode = mbp->param[0] = MBOX_SET_PCI_PARAMETERS;
inparam = 4;
outparam = 4;
goto command_known;
} else if (mbp->param[0] > NMBCOM) {
PRINTF("%s: bad command %x\n", isp->isp_name, mbp->param[0]);
return;
}
opcode = mbp->param[0];
inparam = HINIB(mbpcnt[mbp->param[0]]);
outparam = LONIB(mbpcnt[mbp->param[0]]);
if (inparam == 0 && outparam == 0) {
PRINTF("%s: no parameters for %x\n", isp->isp_name,
mbp->param[0]);
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
/*
* Check for variants
*/
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 ISP2100_SCCLUN
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)) {
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
switch (mbp->param[0]) {
case MBOX_ABORT:
inparam = 7;
break;
case MBOX_ABORT_DEVICE:
case MBOX_START_QUEUE:
case MBOX_STOP_QUEUE:
case MBOX_SINGLE_STEP_QUEUE:
case MBOX_ABORT_QUEUE:
case MBOX_GET_DEV_QUEUE_STATUS:
inparam = 3;
break;
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
case MBOX_BUS_RESET:
inparam = 2;
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
default:
break;
}
}
#endif
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
command_known:
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
/*
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
* Set semaphore on mailbox registers to win any races to acquire them.
*/
ISP_WRITE(isp, BIU_SEMA, 1);
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
/*
* Qlogic Errata for the ISP2100 says that there is a necessary
* debounce between between writing the semaphore register
* and reading a mailbox register. I believe we're okay 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
/*
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
* Make sure we can send some words.
* Check to see if there's an async mbox event pending.
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
loops = MBOX_DELAY_COUNT;
while ((ISP_READ(isp, HCCR) & HCCR_HOST_INT) != 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
if (ISP_READ(isp, BIU_SEMA) & 1) {
int fph;
u_int16_t mbox = ISP_READ(isp, OUTMAILBOX0);
/*
* We have a pending MBOX async event.
*/
if (mbox & 0x8000) {
fph = isp_parse_async(isp, (int) mbox);
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
IDPRINTF(5, ("%s: line %d, fph %d\n",
isp->isp_name, __LINE__, 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
ISP_WRITE(isp, BIU_SEMA, 0);
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
if (fph < 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
return;
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
} else if (fph > 0) {
isp_fastpost_complete(isp, fph);
}
SYS_DELAY(100);
goto command_known;
}
/*
* We have a pending MBOX completion? Might be
* from a previous command. We can't (sometimes)
* just clear HOST INTERRUPT, so we'll just silently
* eat this here.
*/
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 (mbox & 0x4000) {
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
IDPRINTF(5, ("%s: line %d, mbox 0x%x\n",
isp->isp_name, __LINE__, 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
ISP_WRITE(isp, BIU_SEMA, 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
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
SYS_DELAY(100);
goto command_known;
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
}
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
SYS_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
if (--loops < 0) {
if (dld++ > 10) {
PRINTF("%s: isp_mboxcmd could not get command "
"started\n", isp->isp_name);
return;
}
ISP_WRITE(isp, BIU_SEMA, 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
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
goto command_known;
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
/*
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
* Write input parameters.
*
* Special case some of the setups for the dual port SCSI cards.
* XXX Eventually will be fixed by converting register write/read
* XXX counts to bitmasks.
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 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
switch (opcode) {
case MBOX_GET_RETRY_COUNT:
case MBOX_SET_RETRY_COUNT:
ISP_WRITE(isp, INMAILBOX7, mbp->param[7]);
mbp->param[7] = 0;
ISP_WRITE(isp, INMAILBOX6, mbp->param[6]);
mbp->param[6] = 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;
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
case MBOX_SET_ASYNC_DATA_SETUP_TIME:
case MBOX_SET_ACT_NEG_STATE:
case MBOX_SET_TAG_AGE_LIMIT:
case MBOX_SET_SELECT_TIMEOUT:
ISP_WRITE(isp, INMAILBOX2, mbp->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
mbp->param[2] = 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;
}
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
switch (inparam) {
case 8: ISP_WRITE(isp, INMAILBOX7, mbp->param[7]); mbp->param[7] = 0;
case 7: ISP_WRITE(isp, INMAILBOX6, mbp->param[6]); mbp->param[6] = 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 6:
/*
* The Qlogic 2100 cannot have registers 4 and 5 written to
* after initialization or BAD THINGS HAPPEN (tm).
*/
if (IS_SCSI(isp) || mbp->param[0] == MBOX_INIT_FIRMWARE)
ISP_WRITE(isp, INMAILBOX5, mbp->param[5]);
mbp->param[5] = 0;
case 5:
if (IS_SCSI(isp) || mbp->param[0] == MBOX_INIT_FIRMWARE)
ISP_WRITE(isp, INMAILBOX4, mbp->param[4]);
mbp->param[4] = 0;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
case 4: ISP_WRITE(isp, INMAILBOX3, mbp->param[3]); mbp->param[3] = 0;
case 3: ISP_WRITE(isp, INMAILBOX2, mbp->param[2]); mbp->param[2] = 0;
case 2: ISP_WRITE(isp, INMAILBOX1, mbp->param[1]); mbp->param[1] = 0;
case 1: ISP_WRITE(isp, INMAILBOX0, mbp->param[0]); mbp->param[0] = 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
* Clear RISC int condition.
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
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_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
* Clear semaphore on mailbox registers so that the Qlogic
* may update outgoing registers.
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
ISP_WRITE(isp, BIU_SEMA, 0);
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.
*/
ISP_WRITE(isp, HCCR, HCCR_CMD_SET_HOST_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
/*
* Wait until HOST INT has gone away (meaning that the Qlogic
* has picked up the mailbox command. Wait a long time.
*/
loops = MBOX_DELAY_COUNT * 5;
while ((ISP_READ(isp, HCCR) & HCCR_CMD_CLEAR_RISC_INT) != 0) {
SYS_DELAY(100);
if (--loops < 0) {
PRINTF("%s: isp_mboxcmd timeout #2\n", isp->isp_name);
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
* While the Semaphore registers isn't set, wait for the Qlogic
* to process the mailbox command. Again- wait a long time.
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
loops = MBOX_DELAY_COUNT * 5;
while ((ISP_READ(isp, BIU_SEMA) & 1) == 0) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
SYS_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
/*
* Wierd- I've seen the case where the semaphore register
* isn't getting set- sort of a violation of the protocol..
*/
if (ISP_READ(isp, OUTMAILBOX0) & 0x4000)
break;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
if (--loops < 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
PRINTF("%s: isp_mboxcmd timeout #3\n", isp->isp_name);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
return;
}
}
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
/*
* Make sure that the MBOX_BUSY has gone away
*/
loops = MBOX_DELAY_COUNT;
for (;;) {
u_int16_t mbox = ISP_READ(isp, OUTMAILBOX0);
if (mbox == MBOX_BUSY) {
if (--loops < 0) {
PRINTF("%s: isp_mboxcmd timeout #4\n",
isp->isp_name);
return;
}
SYS_DELAY(100);
continue;
}
/*
* We have a pending MBOX async event.
*/
if (mbox & 0x8000) {
int fph = isp_parse_async(isp, (int) mbox);
ISP_WRITE(isp, BIU_SEMA, 0);
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
if (fph < 0) {
return;
} else if (fph > 0) {
isp_fastpost_complete(isp, fph);
}
SYS_DELAY(100);
continue;
}
break;
}
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
* Pick up output parameters. Special case some of the readbacks
* for the dual port SCSI cards.
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
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
switch (opcode) {
case MBOX_GET_RETRY_COUNT:
case MBOX_SET_RETRY_COUNT:
mbp->param[7] = ISP_READ(isp, OUTMAILBOX7);
mbp->param[6] = ISP_READ(isp, OUTMAILBOX6);
break;
case MBOX_GET_TAG_AGE_LIMIT:
case MBOX_SET_TAG_AGE_LIMIT:
case MBOX_GET_ACT_NEG_STATE:
case MBOX_SET_ACT_NEG_STATE:
case MBOX_SET_ASYNC_DATA_SETUP_TIME:
case MBOX_GET_ASYNC_DATA_SETUP_TIME:
case MBOX_GET_RESET_DELAY_PARAMS:
case MBOX_SET_RESET_DELAY_PARAMS:
mbp->param[2] = ISP_READ(isp, OUTMAILBOX2);
break;
}
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
switch (outparam) {
case 8: mbp->param[7] = ISP_READ(isp, OUTMAILBOX7);
case 7: mbp->param[6] = ISP_READ(isp, OUTMAILBOX6);
case 6: mbp->param[5] = ISP_READ(isp, OUTMAILBOX5);
case 5: mbp->param[4] = ISP_READ(isp, OUTMAILBOX4);
case 4: mbp->param[3] = ISP_READ(isp, OUTMAILBOX3);
case 3: mbp->param[2] = ISP_READ(isp, OUTMAILBOX2);
case 2: mbp->param[1] = ISP_READ(isp, OUTMAILBOX1);
case 1: mbp->param[0] = ISP_READ(isp, OUTMAILBOX0);
}
/*
* Clear RISC int.
*/
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
/*
* Release semaphore on mailbox registers
*/
ISP_WRITE(isp, BIU_SEMA, 0);
/*
* Just to be chatty here...
*/
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:
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
IDPRINTF(2, ("%s: mbox cmd %x failed with INVALID_COMMAND\n",
isp->isp_name, opcode));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
case MBOX_HOST_INTERFACE_ERROR:
PRINTF("%s: mbox cmd %x failed with HOST_INTERFACE_ERROR\n",
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_name, opcode);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
case MBOX_TEST_FAILED:
PRINTF("%s: mbox cmd %x failed with TEST_FAILED\n",
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_name, opcode);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
case MBOX_COMMAND_ERROR:
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 (opcode != MBOX_ABOUT_FIRMWARE)
PRINTF("%s: mbox cmd %x failed with COMMAND_ERROR\n",
isp->isp_name, opcode);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
break;
case MBOX_COMMAND_PARAM_ERROR:
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 (opcode) {
case MBOX_GET_PORT_DB:
case MBOX_GET_PORT_NAME:
case MBOX_GET_DEV_QUEUE_PARAMS:
break;
default:
PRINTF("%s: mbox cmd %x failed with "
"COMMAND_PARAM_ERROR\n", isp->isp_name, opcode);
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +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
/*
* Be silent about these...
*/
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
case ASYNC_PDB_CHANGED:
((fcparam *) isp->isp_param)->isp_loopstate = LOOP_PDB_RCVD;
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
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
case ASYNC_LIP_OCCURRED:
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
((fcparam *) isp->isp_param)->isp_lipseq = mbp->param[1];
/* FALLTHROUGH */
case ASYNC_LOOP_UP:
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
((fcparam *) isp->isp_param)->isp_fwstate = FW_CONFIG_WAIT;
((fcparam *) isp->isp_param)->isp_loopstate = LOOP_LIP_RCVD;
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
case ASYNC_LOOP_DOWN:
case ASYNC_LOOP_RESET:
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
((fcparam *) isp->isp_param)->isp_fwstate = FW_CONFIG_WAIT;
((fcparam *) isp->isp_param)->isp_loopstate = LOOP_NIL;
/* FALLTHROUGH */
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 ASYNC_CHANGE_NOTIFY:
break;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
default:
/*
* The expected return of EXEC_FIRMWARE is zero.
*/
if ((opcode == MBOX_EXEC_FIRMWARE && mbp->param[0] != 0) ||
(opcode != MBOX_EXEC_FIRMWARE)) {
PRINTF("%s: mbox cmd %x failed with error %x\n",
isp->isp_name, opcode, mbp->param[0]);
}
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
void
isp_lostcmd(isp, xs)
struct ispsoftc *isp;
ISP_SCSI_XFER_T *xs;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
{
mbreg_t mbs;
mbs.param[0] = MBOX_GET_FIRMWARE_STATUS;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
isp_dumpregs(isp, "couldn't GET FIRMWARE STATUS");
return;
}
if (mbs.param[1]) {
PRINTF("%s: %d commands on completion queue\n",
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
isp->isp_name, mbs.param[1]);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
}
if (XS_NULL(xs))
return;
mbs.param[0] = MBOX_GET_DEV_QUEUE_STATUS;
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] = (XS_TGT(xs) << 8) | XS_LUN(xs); /* XXX: WHICH BUS? */
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
isp_dumpregs(isp, "couldn't GET DEVICE QUEUE STATUS");
return;
}
PRINTF("%s: lost command for target %d lun %d, %d active of %d, "
"Queue State: %x\n", isp->isp_name, XS_TGT(xs),
XS_LUN(xs), mbs.param[2], mbs.param[3], mbs.param[1]);
isp_dumpregs(isp, "lost command");
/*
* XXX: Need to try and do something to recover.
*/
}
static void
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_dumpregs(isp, msg)
struct ispsoftc *isp;
const char *msg;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
{
PRINTF("%s: %s\n", isp->isp_name, msg);
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 support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
PRINTF(" biu_conf1=%x", ISP_READ(isp, BIU_CONF1));
else
PRINTF(" biu_csr=%x", ISP_READ(isp, BIU2100_CSR));
PRINTF(" biu_icr=%x biu_isr=%x biu_sema=%x ", ISP_READ(isp, BIU_ICR),
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
ISP_READ(isp, BIU_ISR), ISP_READ(isp, BIU_SEMA));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
PRINTF("risc_hccr=%x\n", ISP_READ(isp, HCCR));
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
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 support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
ISP_WRITE(isp, HCCR, HCCR_CMD_PAUSE);
PRINTF(" cdma_conf=%x cdma_sts=%x cdma_fifostat=%x\n",
ISP_READ(isp, CDMA_CONF), ISP_READ(isp, CDMA_STATUS),
ISP_READ(isp, CDMA_FIFO_STS));
PRINTF(" ddma_conf=%x ddma_sts=%x ddma_fifostat=%x\n",
ISP_READ(isp, DDMA_CONF), ISP_READ(isp, DDMA_STATUS),
ISP_READ(isp, DDMA_FIFO_STS));
PRINTF(" sxp_int=%x sxp_gross=%x sxp(scsi_ctrl)=%x\n",
ISP_READ(isp, SXP_INTERRUPT),
ISP_READ(isp, SXP_GROSS_ERR),
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
ISP_READ(isp, SXP_PINS_CTRL));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
ISP_WRITE(isp, HCCR, HCCR_CMD_RELEASE);
}
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
PRINTF(" mbox regs: %x %x %x %x %x\n",
ISP_READ(isp, OUTMAILBOX0), ISP_READ(isp, OUTMAILBOX1),
ISP_READ(isp, OUTMAILBOX2), ISP_READ(isp, OUTMAILBOX3),
ISP_READ(isp, OUTMAILBOX4));
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
ISP_DUMPREGS(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
static void
isp_fw_state(isp)
struct ispsoftc *isp;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
{
mbreg_t mbs;
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)) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
int once = 0;
fcparam *fcp = isp->isp_param;
again:
mbs.param[0] = MBOX_GET_FW_STATE;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
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
IDPRINTF(1, ("%s: isp_fw_state 0x%x\n", isp->isp_name,
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
mbs.param[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
switch (mbs.param[0]) {
case ASYNC_PDB_CHANGED:
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 (once++ < 10) {
goto again;
}
fcp->isp_fwstate = FW_CONFIG_WAIT;
fcp->isp_loopstate = LOOP_PDB_RCVD;
goto again;
case ASYNC_LIP_OCCURRED:
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
fcp->isp_lipseq = mbs.param[1];
/* FALLTHROUGH */
case ASYNC_LOOP_UP:
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
fcp->isp_fwstate = FW_CONFIG_WAIT;
fcp->isp_loopstate = LOOP_LIP_RCVD;
if (once++ < 10) {
goto again;
}
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
case ASYNC_LOOP_RESET:
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
case ASYNC_LOOP_DOWN:
fcp->isp_fwstate = FW_CONFIG_WAIT;
fcp->isp_loopstate = LOOP_NIL;
/* FALLTHROUGH */
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 ASYNC_CHANGE_NOTIFY:
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 (once++ < 10) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
goto again;
}
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
break;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +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
PRINTF("%s: GET FIRMWARE STATE failed (0x%x)\n",
isp->isp_name, mbs.param[0]);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
return;
}
fcp->isp_fwstate = mbs.param[1];
}
}
static void
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_update(isp)
struct ispsoftc *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
{
int bus;
for (bus = 0; isp->isp_update != 0; bus++) {
if (isp->isp_update & (1 << bus)) {
isp_update_bus(isp, bus);
isp->isp_update ^= (1 << bus);
}
}
}
static void
isp_update_bus(isp, bus)
struct ispsoftc *isp;
int bus;
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)) {
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
sdp = isp->isp_param;
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 += 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
for (tgt = 0; tgt < MAX_TARGETS; tgt++) {
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
u_int16_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) {
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
IDPRINTF(1, ("%s: skipping target %d bus %d update\n",
isp->isp_name, tgt, 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
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
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,
* take what's in dev_flags and try and set the device
* toward that. Otherwise, if we're just refreshing the
* current device state, get the current parameters.
*/
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 (sdp->isp_devparam[tgt].dev_update) {
mbs.param[0] = MBOX_SET_TARGET_PARAMS;
mbs.param[2] = sdp->isp_devparam[tgt].dev_flags;
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
/*
* Insist that PARITY must be enabled if SYNC
* is enabled.
*/
if (mbs.param[2] & DPARM_SYNC) {
mbs.param[2] |= DPARM_PARITY;
}
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[3] =
(sdp->isp_devparam[tgt].sync_offset << 8) |
(sdp->isp_devparam[tgt].sync_period);
sdp->isp_devparam[tgt].dev_update = 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
/*
* A command completion later that has
* RQSTF_NEGOTIATION set will cause
* the dev_refresh/announce cycle.
*
* 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.
*/
sdp->isp_devparam[tgt].cur_dflags &= ~DPARM_TQING;
sdp->isp_devparam[tgt].cur_dflags |=
(sdp->isp_devparam[tgt].dev_flags & DPARM_TQING);
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->isp_devparam[tgt].dev_refresh = 1;
be a bit more chatty about some speed negotiations
1999-05-12 18:56:55 +00:00
IDPRINTF(3, ("%s: bus %d set tgt %d flags 0x%x off 0x%x"
" period 0x%x\n", isp->isp_name, bus, 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 if (sdp->isp_devparam[tgt].dev_refresh) {
mbs.param[0] = MBOX_GET_TARGET_PARAMS;
sdp->isp_devparam[tgt].dev_refresh = 0;
get = 1;
} 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;
}
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] = (bus << 15) | (tgt << 8) ;
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_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
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
PRINTF("%s: failed to %cet SCSI parameters for "
"target %d\n", isp->isp_name, (get)? 'g' : 's',
tgt);
continue;
}
if (get == 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
isp->isp_sendmarker |= (1 << bus);
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
continue;
}
flags = mbs.param[2];
period = mbs.param[3] & 0xff;
offset = mbs.param[3] >> 8;
sdp->isp_devparam[tgt].cur_dflags = flags;
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
sdp->isp_devparam[tgt].cur_period = period;
sdp->isp_devparam[tgt].cur_offset = offset;
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
get = (bus << 16) | tgt;
(void) isp_async(isp, ISPASYNC_NEW_TGT_PARAMS, &get);
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
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 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_setdfltparm(isp, channel)
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
struct ispsoftc *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
int channel;
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;
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
mbreg_t mbs;
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
sdparam *sdp;
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_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
fcparam *fcp = (fcparam *) isp->isp_param;
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
fcp += channel;
if (fcp->isp_gotdparms) {
return;
}
fcp->isp_gotdparms = 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
fcp->isp_maxfrmlen = ICB_DFLT_FRMLEN;
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
fcp->isp_maxalloc = ICB_DFLT_ALLOC;
fcp->isp_execthrottle = ICB_DFLT_THROTTLE;
fcp->isp_retry_delay = ICB_DFLT_RDELAY;
fcp->isp_retry_count = ICB_DFLT_RCOUNT;
/* Platform specific.... */
fcp->isp_loopid = DEFAULT_LOOPID(isp);
fcp->isp_nodewwn = DEFAULT_WWN(isp);
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
fcp->isp_portwwn = 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
/*
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
* Now try and read NVRAM
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
if ((isp->isp_confopts & (ISP_CFG_NONVRAM|ISP_CFG_OWNWWN)) ||
(isp_read_nvram(isp))) {
PRINTF("%s: using Node WWN 0x%08x%08x\n",
isp->isp_name, (u_int32_t)(fcp->isp_nodewwn >> 32),
(u_int32_t)(fcp->isp_nodewwn & 0xffffffff));
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
}
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;
}
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
sdp = (sdparam *) isp->isp_param;
sdp += channel;
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
/*
* Been there, done that, got the T-shirt...
*/
if (sdp->isp_gotdparms) {
return;
}
sdp->isp_gotdparms = 1;
/*
* If we've not been told to avoid reading NVRAM, try and read it.
* If we're successful reading it, we can return since NVRAM will
* tell us the right thing to do. Otherwise, establish some reasonable
* defaults.
*/
if ((isp->isp_confopts & ISP_CFG_NONVRAM) == 0) {
if (isp_read_nvram(isp) == 0) {
return;
}
}
/*
* Now try and see whether we have specific values for them.
*/
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 ((isp->isp_confopts & ISP_CFG_NONVRAM) == 0) {
mbs.param[0] = MBOX_GET_ACT_NEG_STATE;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
IDPRINTF(2, ("could not GET ACT NEG STATE\n"));
sdp->isp_req_ack_active_neg = 1;
sdp->isp_data_line_active_neg = 1;
} else {
sdp->isp_req_ack_active_neg =
(mbs.param[1+channel] >> 4) & 0x1;
sdp->isp_data_line_active_neg =
(mbs.param[1+channel] >> 5) & 0x1;
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
} else {
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
sdp->isp_req_ack_active_neg = 1;
sdp->isp_data_line_active_neg = 1;
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
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
IDPRINTF(3, ("%s: defaulting bus %d REQ/ACK Active Negation is %d\n",
isp->isp_name, channel, sdp->isp_req_ack_active_neg));
IDPRINTF(3, ("%s: defaulting bus %d DATA Active Negation is %d\n",
isp->isp_name, channel, sdp->isp_data_line_active_neg));
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!)
* state (dev_flags). Then try and get the current status from
* 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++) {
sdp->isp_devparam[tgt].cur_offset = 0;
sdp->isp_devparam[tgt].cur_period = 0;
sdp->isp_devparam[tgt].dev_flags = DPARM_DEFAULT;
sdp->isp_devparam[tgt].cur_dflags = 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
if (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 && 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
sdp->isp_devparam[tgt].sync_offset =
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_10M_SYNCPARMS >> 8;
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
sdp->isp_devparam[tgt].sync_period =
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_10M_SYNCPARMS & 0xff;
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
} else if (IS_ULTRA2(isp)) {
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
sdp->isp_devparam[tgt].sync_offset =
ISP_40M_SYNCPARMS >> 8;
sdp->isp_devparam[tgt].sync_period =
ISP_40M_SYNCPARMS & 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
} else {
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
sdp->isp_devparam[tgt].sync_offset =
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_20M_SYNCPARMS >> 8;
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
sdp->isp_devparam[tgt].sync_period =
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_20M_SYNCPARMS & 0xff;
}
/*
* Don't get current target parameters if we've been
* told not to use NVRAM- it's really the same thing.
*/
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 ((isp->isp_confopts & ISP_CFG_NONVRAM) == 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
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
mbs.param[0] = MBOX_GET_TARGET_PARAMS;
mbs.param[1] = tgt << 8;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
continue;
}
sdp->isp_devparam[tgt].cur_dflags = mbs.param[2];
sdp->isp_devparam[tgt].dev_flags = mbs.param[2];
sdp->isp_devparam[tgt].cur_period = mbs.param[3] & 0xff;
sdp->isp_devparam[tgt].cur_offset = mbs.param[3] >> 8;
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
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
/*
* The maximum period we can really see
* here is 100 (decimal), or 400 ns.
* For some unknown reason we sometimes
* get back wildass numbers from the
* boot device's parameters (alpha only).
*/
if ((mbs.param[3] & 0xff) <= 0x64) {
sdp->isp_devparam[tgt].sync_period =
mbs.param[3] & 0xff;
sdp->isp_devparam[tgt].sync_offset =
mbs.param[3] >> 8;
}
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 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
/*
* It is not safe to run Ultra Mode with a clock < 60.
*/
if (((isp->isp_clock && isp->isp_clock < 60) ||
(isp->isp_type < ISP_HA_SCSI_1020A)) &&
(sdp->isp_devparam[tgt].sync_period <=
(ISP_20M_SYNCPARMS & 0xff))) {
sdp->isp_devparam[tgt].sync_offset =
ISP_10M_SYNCPARMS >> 8;
sdp->isp_devparam[tgt].sync_period =
ISP_10M_SYNCPARMS & 0xff;
}
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
IDPRINTF(3, ("%s: bus %d tgt %d flags %x offset %x period %x\n",
isp->isp_name, channel, tgt,
sdp->isp_devparam[tgt].dev_flags,
sdp->isp_devparam[tgt].sync_period,
sdp->isp_devparam[tgt].sync_period));
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
* Establish default some more default parameters.
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
*/
sdp->isp_cmd_dma_burst_enable = 1;
sdp->isp_data_dma_burst_enabl = 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
sdp->isp_fifo_threshold = 0;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
sdp->isp_initiator_id = 7;
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
/* XXXX This is probably based upon clock XXXX */
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_type >= ISP_HA_SCSI_1040) {
sdp->isp_async_data_setup = 9;
} else {
sdp->isp_async_data_setup = 6;
}
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
sdp->isp_selection_timeout = 250;
temp fix for internal queue overflow problem
1999-04-14 17:37:36 +00:00
sdp->isp_max_queue_depth = MAXISPREQUEST;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
sdp->isp_tag_aging = 8;
sdp->isp_bus_reset_delay = 3;
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->isp_retry_count = 2;
sdp->isp_retry_delay = 2;
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
for (tgt = 0; tgt < MAX_TARGETS; tgt++) {
sdp->isp_devparam[tgt].exc_throttle = 16;
sdp->isp_devparam[tgt].dev_enable = 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
/*
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
*
* Locks held prior to coming here.
*/
void
isp_restart(isp)
struct ispsoftc *isp;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +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
ISP_SCSI_XFER_T *xs;
u_int32_t handle;
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
#if 0
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_gotdparms = 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
#endif
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
isp_reset(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
if (isp->isp_state == ISP_RESETSTATE) {
isp_init(isp);
if (isp->isp_state == ISP_INITSTATE) {
isp->isp_state = ISP_RUNSTATE;
}
}
if (isp->isp_state != ISP_RUNSTATE) {
PRINTF("%s: isp_restart cannot restart ISP\n", isp->isp_name);
}
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 = 0;
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +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
for (handle = 1; handle <= (int) isp->isp_maxcmds; handle++) {
xs = isp_find_xs(isp, handle);
if (xs == NULL) {
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
continue;
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
}
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, handle);
if (XS_XFRLEN(xs)) {
ISP_DMAFREE(isp, xs, handle);
XS_RESID(xs) = XS_XFRLEN(xs);
} else {
XS_RESID(xs) = 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
XS_SETERR(xs, HBA_BUSRESET);
Add support for the Qlogic ISP SCSI && FC/AL Adapters
1998-04-22 17:54:58 +00:00
XS_CMD_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
/*
* NVRAM Routines
*/
static int
isp_read_nvram(isp)
struct ispsoftc *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
static char *tru = "true";
static char *not = "false";
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, amt;
u_int8_t csum, minversion;
union {
u_int8_t _x[ISP2100_NVRAM_SIZE];
u_int16_t _s[ISP2100_NVRAM_SIZE>>1];
} _n;
#define nvram_data _n._x
#define nvram_words _n._s
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)) {
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;
}
/*
* Just read the first two words first to see if we have a valid
* NVRAM to continue reading the rest with.
*/
for (i = 0; i < 2; i++) {
isp_rdnvram_word(isp, i, &nvram_words[i]);
}
if (nvram_data[0] != 'I' || nvram_data[1] != 'S' ||
nvram_data[2] != 'P') {
if (isp->isp_bustype != ISP_BT_SBUS) {
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
PRINTF("%s: invalid NVRAM header (%x,%x,%x,%x)\n",
isp->isp_name, nvram_data[0], nvram_data[1],
nvram_data[2], nvram_data[3]);
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);
}
for (i = 2; i < amt>>1; i++) {
isp_rdnvram_word(isp, i, &nvram_words[i]);
}
for (csum = 0, i = 0; i < amt; i++) {
csum += nvram_data[i];
}
if (csum != 0) {
PRINTF("%s: invalid NVRAM checksum\n", isp->isp_name);
return (-1);
}
if (ISP_NVRAM_VERSION(nvram_data) < minversion) {
PRINTF("%s: version %d NVRAM not understood\n", isp->isp_name,
ISP_NVRAM_VERSION(nvram_data));
return (-1);
}
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
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
int bus;
sdparam *sdp = (sdparam *) isp->isp_param;
Add Dual LVD bus (1280) support
1999-12-16 05:42:02 +00:00
for (bus = 0; bus < (IS_DUALBUS(isp)? 2 : 1); bus++, sdp++) {
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->isp_fifo_threshold =
ISP1080_NVRAM_FIFO_THRESHOLD(nvram_data);
sdp->isp_initiator_id =
ISP1080_NVRAM_INITIATOR_ID(nvram_data, bus);
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 =
ISP1080_NVRAM_ASYNC_DATA_SETUP_TIME(nvram_data,
bus);
sdp->isp_req_ack_active_neg =
ISP1080_NVRAM_REQ_ACK_ACTIVE_NEGATION(nvram_data,
bus);
sdp->isp_data_line_active_neg =
ISP1080_NVRAM_DATA_LINE_ACTIVE_NEGATION(nvram_data,
bus);
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);
if (isp->isp_dblev >= 3) {
PRINTF("%s: ISP1080 bus %d NVRAM values:\n",
isp->isp_name, bus);
PRINTF(" Initiator ID = %d\n",
sdp->isp_initiator_id);
PRINTF(" Fifo Threshold = 0x%x\n",
sdp->isp_fifo_threshold);
PRINTF(" Bus Reset Delay = %d\n",
sdp->isp_bus_reset_delay);
PRINTF(" Retry Count = %d\n",
sdp->isp_retry_count);
PRINTF(" Retry Delay = %d\n",
sdp->isp_retry_delay);
PRINTF(" Tag Age Limit = %d\n",
sdp->isp_tag_aging);
PRINTF(" Selection Timeout = %d\n",
sdp->isp_selection_timeout);
PRINTF(" Max Queue Depth = %d\n",
sdp->isp_max_queue_depth);
PRINTF(" Async Data Setup = 0x%x\n",
sdp->isp_async_data_setup);
PRINTF(" REQ/ACK Active Negation = %s\n",
sdp->isp_req_ack_active_neg? tru : not);
PRINTF(" Data Line Active Negation = %s\n",
sdp->isp_data_line_active_neg? tru : not);
PRINTF(" Cmd DMA Burst Enable = %s\n",
sdp->isp_cmd_dma_burst_enable? tru : not);
}
for (i = 0; i < MAX_TARGETS; i++) {
sdp->isp_devparam[i].dev_enable =
ISP1080_NVRAM_TGT_DEVICE_ENABLE(nvram_data, i, bus);
sdp->isp_devparam[i].exc_throttle =
ISP1080_NVRAM_TGT_EXEC_THROTTLE(nvram_data, i, bus);
sdp->isp_devparam[i].sync_offset =
ISP1080_NVRAM_TGT_SYNC_OFFSET(nvram_data, i, bus);
sdp->isp_devparam[i].sync_period =
ISP1080_NVRAM_TGT_SYNC_PERIOD(nvram_data, i, bus);
sdp->isp_devparam[i].dev_flags = 0;
if (ISP1080_NVRAM_TGT_RENEG(nvram_data, i, bus))
sdp->isp_devparam[i].dev_flags |= DPARM_RENEG;
if (ISP1080_NVRAM_TGT_QFRZ(nvram_data, i, bus)) {
PRINTF("%s: not supporting QFRZ option "
"for target %d bus %d\n",
isp->isp_name, i, bus);
}
sdp->isp_devparam[i].dev_flags |= DPARM_ARQ;
if (ISP1080_NVRAM_TGT_ARQ(nvram_data, i, bus) == 0) {
PRINTF("%s: not disabling ARQ option "
"for target %d bus %d\n",
isp->isp_name, i, bus);
}
if (ISP1080_NVRAM_TGT_TQING(nvram_data, i, bus))
sdp->isp_devparam[i].dev_flags |= DPARM_TQING;
if (ISP1080_NVRAM_TGT_SYNC(nvram_data, i, bus))
sdp->isp_devparam[i].dev_flags |= DPARM_SYNC;
if (ISP1080_NVRAM_TGT_WIDE(nvram_data, i, bus))
sdp->isp_devparam[i].dev_flags |= DPARM_WIDE;
if (ISP1080_NVRAM_TGT_PARITY(nvram_data, i, bus))
sdp->isp_devparam[i].dev_flags |= DPARM_PARITY;
if (ISP1080_NVRAM_TGT_DISC(nvram_data, i, bus))
sdp->isp_devparam[i].dev_flags |= DPARM_DISC;
sdp->isp_devparam[i].cur_dflags = 0;
if (isp->isp_dblev >= 3) {
PRINTF(" Target %d: Ena %d Throttle "
"%d Offset %d Period %d Flags "
"0x%x\n", i,
sdp->isp_devparam[i].dev_enable,
sdp->isp_devparam[i].exc_throttle,
sdp->isp_devparam[i].sync_offset,
sdp->isp_devparam[i].sync_period,
sdp->isp_devparam[i].dev_flags);
}
}
}
} else if (IS_SCSI(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
sdparam *sdp = (sdparam *) isp->isp_param;
sdp->isp_fifo_threshold =
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
ISP_NVRAM_FIFO_THRESHOLD(nvram_data) |
(ISP_NVRAM_FIFO_THRESHOLD_128(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
sdp->isp_initiator_id =
ISP_NVRAM_INITIATOR_ID(nvram_data);
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;
}
}
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
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);
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_fast_mttr = ISP_NVRAM_FAST_MTTR_ENABLE(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
if (isp->isp_dblev > 2) {
PRINTF("%s: NVRAM values:\n", isp->isp_name);
PRINTF(" Fifo Threshold = 0x%x\n",
sdp->isp_fifo_threshold);
PRINTF(" Bus Reset Delay = %d\n",
sdp->isp_bus_reset_delay);
PRINTF(" Retry Count = %d\n",
sdp->isp_retry_count);
PRINTF(" Retry Delay = %d\n",
sdp->isp_retry_delay);
PRINTF(" Tag Age Limit = %d\n",
sdp->isp_tag_aging);
PRINTF(" Selection Timeout = %d\n",
sdp->isp_selection_timeout);
PRINTF(" Max Queue Depth = %d\n",
sdp->isp_max_queue_depth);
PRINTF(" Async Data Setup = 0x%x\n",
sdp->isp_async_data_setup);
PRINTF(" REQ/ACK Active Negation = %s\n",
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->isp_req_ack_active_neg? tru : not);
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
PRINTF(" Data Line Active Negation = %s\n",
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->isp_data_line_active_neg? tru : not);
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
PRINTF(" Data DMA Burst Enable = %s\n",
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->isp_data_dma_burst_enabl? tru : not);
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
PRINTF(" Cmd DMA Burst Enable = %s\n",
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->isp_cmd_dma_burst_enable? tru : not);
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
PRINTF(" Fast MTTR = %s\n",
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_fast_mttr? tru : not);
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
}
for (i = 0; i < MAX_TARGETS; 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
sdp->isp_devparam[i].dev_enable =
ISP_NVRAM_TGT_DEVICE_ENABLE(nvram_data, i);
sdp->isp_devparam[i].exc_throttle =
ISP_NVRAM_TGT_EXEC_THROTTLE(nvram_data, i);
sdp->isp_devparam[i].sync_offset =
ISP_NVRAM_TGT_SYNC_OFFSET(nvram_data, i);
sdp->isp_devparam[i].sync_period =
ISP_NVRAM_TGT_SYNC_PERIOD(nvram_data, i);
if (isp->isp_type < ISP_HA_SCSI_1040) {
/*
* If we're not ultra, we can't possibly
* be a shorter period than this.
*/
if (sdp->isp_devparam[i].sync_period < 0x19) {
sdp->isp_devparam[i].sync_period =
0x19;
}
if (sdp->isp_devparam[i].sync_offset > 0xc) {
sdp->isp_devparam[i].sync_offset =
0x0c;
}
} else {
if (sdp->isp_devparam[i].sync_offset > 0x8) {
sdp->isp_devparam[i].sync_offset = 0x8;
}
}
sdp->isp_devparam[i].dev_flags = 0;
if (ISP_NVRAM_TGT_RENEG(nvram_data, i))
sdp->isp_devparam[i].dev_flags |= DPARM_RENEG;
if (ISP_NVRAM_TGT_QFRZ(nvram_data, i)) {
PRINTF("%s: not supporting QFRZ option for "
"target %d\n", isp->isp_name, i);
}
sdp->isp_devparam[i].dev_flags |= DPARM_ARQ;
if (ISP_NVRAM_TGT_ARQ(nvram_data, i) == 0) {
PRINTF("%s: not disabling ARQ option for "
"target %d\n", isp->isp_name, i);
}
if (ISP_NVRAM_TGT_TQING(nvram_data, i))
sdp->isp_devparam[i].dev_flags |= DPARM_TQING;
if (ISP_NVRAM_TGT_SYNC(nvram_data, i))
sdp->isp_devparam[i].dev_flags |= DPARM_SYNC;
if (ISP_NVRAM_TGT_WIDE(nvram_data, i))
sdp->isp_devparam[i].dev_flags |= DPARM_WIDE;
if (ISP_NVRAM_TGT_PARITY(nvram_data, i))
sdp->isp_devparam[i].dev_flags |= DPARM_PARITY;
if (ISP_NVRAM_TGT_DISC(nvram_data, i))
sdp->isp_devparam[i].dev_flags |= DPARM_DISC;
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
sdp->isp_devparam[i].cur_dflags = 0; /* we don't know */
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 (isp->isp_dblev > 2) {
PRINTF(" Target %d: Enabled %d Throttle %d "
"Offset %d Period %d Flags 0x%x\n", i,
sdp->isp_devparam[i].dev_enable,
sdp->isp_devparam[i].exc_throttle,
sdp->isp_devparam[i].sync_offset,
sdp->isp_devparam[i].sync_period,
sdp->isp_devparam[i].dev_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
}
} else {
fcparam *fcp = (fcparam *) isp->isp_param;
union {
struct {
#if BYTE_ORDER == BIG_ENDIAN
u_int32_t hi32;
u_int32_t lo32;
#else
u_int32_t lo32;
u_int32_t hi32;
#endif
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
} wd;
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
u_int64_t full64;
} wwnstore;
wwnstore.full64 = ISP2100_NVRAM_NODE_NAME(nvram_data);
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
/*
* Broken PTI cards with nothing in the top nibble. Pah.
*/
if ((wwnstore.wd.hi32 >> 28) == 0) {
wwnstore.wd.hi32 |= (2 << 28);
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
CFGPRINTF("%s: (corrected) Adapter WWN 0x%08x%08x\n",
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
isp->isp_name, wwnstore.wd.hi32, wwnstore.wd.lo32);
} else {
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
CFGPRINTF("%s: Adapter WWN 0x%08x%08x\n", isp->isp_name,
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
wwnstore.wd.hi32, wwnstore.wd.lo32);
}
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
fcp->isp_nodewwn = wwnstore.full64;
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
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 the Node WWN has 2 in the top nibble, we can
* authoritatively construct a Port WWN by adding
* our unit number (plus one to make it nonzero) and
* putting it into bits 59..56. If the top nibble isn't
* 2, then we just set them identically.
*/
if ((fcp->isp_nodewwn >> 60) == 2) {
fcp->isp_portwwn = fcp->isp_nodewwn |
(((u_int64_t)(isp->isp_unit+1)) << 56);
} else {
fcp->isp_portwwn = fcp->isp_nodewwn;
}
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
wwnstore.full64 = ISP2100_NVRAM_BOOT_NODE_NAME(nvram_data);
if (wwnstore.full64 != 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
PRINTF("%s: BOOT DEVICE WWN 0x%08x%08x\n",
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
isp->isp_name, wwnstore.wd.hi32, wwnstore.wd.lo32);
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
}
fcp->isp_maxalloc =
ISP2100_NVRAM_MAXIOCBALLOCATION(nvram_data);
fcp->isp_maxfrmlen =
ISP2100_NVRAM_MAXFRAMELENGTH(nvram_data);
fcp->isp_retry_delay =
ISP2100_NVRAM_RETRY_DELAY(nvram_data);
fcp->isp_retry_count =
ISP2100_NVRAM_RETRY_COUNT(nvram_data);
fcp->isp_loopid =
ISP2100_NVRAM_HARDLOOPID(nvram_data);
fcp->isp_execthrottle =
ISP2100_NVRAM_EXECUTION_THROTTLE(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
fcp->isp_fwoptions = ISP2100_NVRAM_OPTIONS(nvram_data);
if (isp->isp_dblev > 2) {
PRINTF("%s: NVRAM values:\n", isp->isp_name);
PRINTF(" Max IOCB Allocation = %d\n",
fcp->isp_maxalloc);
PRINTF(" Max Frame Length = %d\n",
fcp->isp_maxfrmlen);
PRINTF(" Execution Throttle = %d\n",
fcp->isp_execthrottle);
PRINTF(" Retry Count = %d\n",
fcp->isp_retry_count);
PRINTF(" Retry Delay = %d\n",
fcp->isp_retry_delay);
PRINTF(" Hard Loop ID = %d\n",
fcp->isp_loopid);
PRINTF(" Options = 0x%x\n",
fcp->isp_fwoptions);
PRINTF(" HBA Options = 0x%x\n",
ISP2100_NVRAM_HBA_OPTIONS(nvram_data));
}
}
IDPRINTF(3, ("%s: NVRAM is valid\n", isp->isp_name));
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);
}
static void
isp_rdnvram_word(isp, wo, rp)
struct ispsoftc *isp;
int wo;
u_int16_t *rp;
{
int i, cbits;
u_int16_t bit, rqst;
ISP_WRITE(isp, BIU_NVRAM, BIU_NVRAM_SELECT);
SYS_DELAY(2);
ISP_WRITE(isp, BIU_NVRAM, BIU_NVRAM_SELECT|BIU_NVRAM_CLOCK);
SYS_DELAY(2);
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)) {
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
wo &= ((ISP2100_NVRAM_SIZE >> 1) - 1);
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);
SYS_DELAY(2);
ISP_WRITE(isp, BIU_NVRAM, bit | BIU_NVRAM_CLOCK);
SYS_DELAY(2);
ISP_WRITE(isp, BIU_NVRAM, bit);
SYS_DELAY(2);
}
/*
* Now read the result back in (bits come back in MSB format).
*/
*rp = 0;
for (i = 0; i < 16; i++) {
u_int16_t rv;
*rp <<= 1;
ISP_WRITE(isp, BIU_NVRAM, BIU_NVRAM_SELECT|BIU_NVRAM_CLOCK);
SYS_DELAY(2);
rv = ISP_READ(isp, BIU_NVRAM);
if (rv & BIU_NVRAM_DATAIN) {
*rp |= 1;
}
SYS_DELAY(2);
ISP_WRITE(isp, BIU_NVRAM, BIU_NVRAM_SELECT);
SYS_DELAY(2);
}
ISP_WRITE(isp, BIU_NVRAM, 0);
SYS_DELAY(2);
#if BYTE_ORDER == BIG_ENDIAN
*rp = ((*rp >> 8) | ((*rp & 0xff) << 8));
#endif
}
Reference in New Issue Copy Permalink