d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
a7cfd0bd3b
freebsd-skq/sys/dev/isp/isp_library.c

3345 lines
124 KiB
C
Raw Normal View History

Start each of the license/copyright comments with /*-, minor shuffle of lines
2005-01-06 01:43:34 +00:00
/*-
sys/dev: further adoption of SPDX licensing ID tags. Mainly focus on files that use BSD 2-Clause license, however the tool I was using misidentified many licenses so this was mostly a manual - error prone - task. The Software Package Data Exchange (SPDX) group provides a specification to make it easier for automated tools to detect and summarize well known opensource licenses. We are gradually adopting the specification, noting that the tags are considered only advisory and do not, in any way, superceed or replace the license texts.
2017-11-27 14:52:40 +00:00
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
* Copyright (c) 2009-2017 Alexander Motin <mav@FreeBSD.org>
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* Copyright (c) 1997-2009 by Matthew Jacob
Fix some stupid copyright mistakes that have been there for quite some time.
2007-03-10 02:39:54 +00:00
* All rights reserved.
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
*
Fix some stupid copyright mistakes that have been there for quite some time.
2007-03-10 02:39:54 +00:00
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
*
Fix some stupid copyright mistakes that have been there for quite some time.
2007-03-10 02:39:54 +00:00
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
*
Fix some stupid copyright mistakes that have been there for quite some time.
2007-03-10 02:39:54 +00:00
* THIS SOFTWARE IS PROVIDED BY AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
*
Add a file where inline functions for the Qlogic isp cards can go. This reduces duplication in all the platform specific header files.
1999-10-17 18:15:05 +00:00
*/
Some rearrangement of headers to minimize diffs with outside of FreeBSD repository and to clean up the license header so as to not pollute the license with file function. Zero all mailbox structures prior to use (just in case). Change the outgoing mailbox count for INIT_FIRMWARE to be correct.
2006-07-16 20:11:50 +00:00
/*
* Qlogic Host Adapter Internal Library Functions
*/
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
#ifdef __NetBSD__
Fix some stupid copyright mistakes that have been there for quite some time.
2007-03-10 02:39:54 +00:00
#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD$");
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
#include <dev/ic/isp_netbsd.h>
#endif
#ifdef __FreeBSD__
Some rearrangement of headers to minimize diffs with outside of FreeBSD repository and to clean up the license header so as to not pollute the license with file function. Zero all mailbox structures prior to use (just in case). Change the outgoing mailbox count for INIT_FIRMWARE to be correct.
2006-07-16 20:11:50 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
#include <dev/isp/isp_freebsd.h>
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
#endif
#ifdef __OpenBSD__
#include <dev/ic/isp_openbsd.h>
#endif
#ifdef __linux__
#include "isp_linux.h"
#endif
#ifdef __svr4__
#include "isp_solaris.h"
#endif
'Support' for ISP SBus cards. This code does not imply that SBus cards work yet. They hang for me. But I can't netboot the latest snapshot on my ultra1e, and things hang at bus_setup_intr time. Since I'm offline for a while, I thought I'd toss this in in case somebody else who has a bit better luck wants to fart around with it. Please try and wait until I get back to check things in.
2002-07-11 03:25:04 +00:00
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
const char *isp_class3_roles[4] = {
"None", "Target", "Initiator", "Target/Initiator"
};
/*
* Command shipping- finish off first queue entry and do dma mapping and additional segments as needed.
*
* Called with the first queue entry at least partially filled out.
*/
int
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_send_cmd(ispsoftc_t *isp, void *fqe, void *segp, uint32_t nsegs, uint32_t totalcnt, isp_ddir_t ddir, ispds64_t *ecmd)
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
{
uint8_t storage[QENTRY_LEN];
Unify initiator and target DMA setup and command sending. The code is so alike that it is pointless to keep it separate. MFC after: 2 weeks
2017-03-24 14:44:03 +00:00
uint8_t type, nqe, need64;
uint32_t seg, seglim, nxt, nxtnxt, ddf;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ispds_t *dsp = NULL;
ispds64_t *dsp64 = NULL;
void *qe0, *qe1;
qe0 = isp_getrqentry(isp);
if (qe0 == NULL) {
return (CMD_EAGAIN);
}
nxt = ISP_NXT_QENTRY(isp->isp_reqidx, RQUEST_QUEUE_LEN(isp));
type = ((isphdr_t *)fqe)->rqs_entry_type;
nqe = 1;
/*
* If we have no data to transmit, just copy the first IOCB and start it up.
*/
if (ddir == ISP_NOXFR) {
if (type == RQSTYPE_T2RQS || type == RQSTYPE_T3RQS) {
ddf = CT2_NO_DATA;
} else {
ddf = 0;
}
goto copy_and_sync;
}
Unify initiator and target DMA setup and command sending. The code is so alike that it is pointless to keep it separate. MFC after: 2 weeks
2017-03-24 14:44:03 +00:00
need64 = 0;
for (seg = 0; seg < nsegs; seg++)
need64 |= XS_NEED_DMA64_SEG(segp, seg);
if (need64) {
if (type == RQSTYPE_T2RQS)
((isphdr_t *)fqe)->rqs_entry_type = type = RQSTYPE_T3RQS;
else if (type == RQSTYPE_REQUEST)
((isphdr_t *)fqe)->rqs_entry_type = type = RQSTYPE_A64;
else if (type == RQSTYPE_CTIO2)
((isphdr_t *)fqe)->rqs_entry_type = type = RQSTYPE_CTIO3;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
Unify initiator and target DMA setup and command sending. The code is so alike that it is pointless to keep it separate. MFC after: 2 weeks
2017-03-24 14:44:03 +00:00
* First figure out how many pieces of data to transfer, what
* kind and how many we can put into the first queue entry.
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
*/
switch (type) {
case RQSTYPE_REQUEST:
ddf = (ddir == ISP_TO_DEVICE)? REQFLAG_DATA_OUT : REQFLAG_DATA_IN;
dsp = ((ispreq_t *)fqe)->req_dataseg;
seglim = ISP_RQDSEG;
break;
case RQSTYPE_CMDONLY:
ddf = (ddir == ISP_TO_DEVICE)? REQFLAG_DATA_OUT : REQFLAG_DATA_IN;
seglim = 0;
break;
case RQSTYPE_T2RQS:
ddf = (ddir == ISP_TO_DEVICE)? REQFLAG_DATA_OUT : REQFLAG_DATA_IN;
dsp = ((ispreqt2_t *)fqe)->req_dataseg;
seglim = ISP_RQDSEG_T2;
break;
case RQSTYPE_A64:
ddf = (ddir == ISP_TO_DEVICE)? REQFLAG_DATA_OUT : REQFLAG_DATA_IN;
dsp64 = ((ispreqt3_t *)fqe)->req_dataseg;
seglim = ISP_RQDSEG_T3;
break;
case RQSTYPE_T3RQS:
ddf = (ddir == ISP_TO_DEVICE)? REQFLAG_DATA_OUT : REQFLAG_DATA_IN;
dsp64 = ((ispreqt3_t *)fqe)->req_dataseg;
seglim = ISP_RQDSEG_T3;
break;
case RQSTYPE_T7RQS:
ddf = (ddir == ISP_TO_DEVICE)? FCP_CMND_DATA_WRITE : FCP_CMND_DATA_READ;
dsp64 = &((ispreqt7_t *)fqe)->req_dataseg;
seglim = 1;
break;
Unify initiator and target DMA setup and command sending. The code is so alike that it is pointless to keep it separate. MFC after: 2 weeks
2017-03-24 14:44:03 +00:00
#ifdef ISP_TARGET_MODE
case RQSTYPE_CTIO2:
dsp = ((ct2_entry_t *)fqe)->rsp.m0.u.ct_dataseg;
seglim = ISP_RQDSEG_T2;
break;
case RQSTYPE_CTIO3:
dsp64 = ((ct2_entry_t *)fqe)->rsp.m0.u.ct_dataseg64;
seglim = ISP_RQDSEG_T3;
break;
case RQSTYPE_CTIO7:
dsp64 = &((ct7_entry_t *)fqe)->rsp.m0.ds;
seglim = 1;
break;
#endif
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
default:
return (CMD_COMPLETE);
}
Unify initiator and target DMA setup and command sending. The code is so alike that it is pointless to keep it separate. MFC after: 2 weeks
2017-03-24 14:44:03 +00:00
if (seglim > nsegs)
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
seglim = nsegs;
Unify initiator and target DMA setup and command sending. The code is so alike that it is pointless to keep it separate. MFC after: 2 weeks
2017-03-24 14:44:03 +00:00
seg = 0;
while (seg < seglim) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (dsp64) {
XS_GET_DMA64_SEG(dsp64++, segp, seg++);
} else {
XS_GET_DMA_SEG(dsp++, segp, seg++);
}
}
/*
* Second, start building additional continuation segments as needed.
*/
while (seg < nsegs) {
nxtnxt = ISP_NXT_QENTRY(nxt, RQUEST_QUEUE_LEN(isp));
if (nxtnxt == isp->isp_reqodx) {
Fix I/O freezes in some cases, caused by r257916. Delaying isp_reqodx update, we should be ready to update it every time we read it. Otherwise requests using several indexes may be requeued ndefinitely without ever updating the variable. MFC after: 3 days
2014-02-05 16:22:02 +00:00
isp->isp_reqodx = ISP_READ(isp, isp->isp_rqstoutrp);
if (nxtnxt == isp->isp_reqodx)
return (CMD_EAGAIN);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
ISP_MEMZERO(storage, QENTRY_LEN);
qe1 = ISP_QUEUE_ENTRY(isp->isp_rquest, nxt);
nxt = nxtnxt;
if (dsp64) {
ispcontreq64_t *crq = (ispcontreq64_t *) storage;
seglim = ISP_CDSEG64;
crq->req_header.rqs_entry_type = RQSTYPE_A64_CONT;
crq->req_header.rqs_entry_count = 1;
dsp64 = crq->req_dataseg;
} else {
ispcontreq_t *crq = (ispcontreq_t *) storage;
seglim = ISP_CDSEG;
crq->req_header.rqs_entry_type = RQSTYPE_DATASEG;
crq->req_header.rqs_entry_count = 1;
dsp = crq->req_dataseg;
}
Unify initiator and target DMA setup and command sending. The code is so alike that it is pointless to keep it separate. MFC after: 2 weeks
2017-03-24 14:44:03 +00:00
seglim += seg;
if (seglim > nsegs)
seglim = nsegs;
while (seg < seglim) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (dsp64) {
XS_GET_DMA64_SEG(dsp64++, segp, seg++);
} else {
XS_GET_DMA_SEG(dsp++, segp, seg++);
}
}
if (dsp64) {
isp_put_cont64_req(isp, (ispcontreq64_t *)storage, qe1);
} else {
isp_put_cont_req(isp, (ispcontreq_t *)storage, qe1);
}
if (isp->isp_dblev & ISP_LOGDEBUG1) {
Polish debugging IOCB dumping. Add few more missing cases, unify byte order. MFC after: 1 month
2016-04-11 10:48:26 +00:00
isp_print_bytes(isp, "additional queue entry",
QENTRY_LEN, qe1);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
nqe++;
}
copy_and_sync:
((isphdr_t *)fqe)->rqs_entry_count = nqe;
switch (type) {
case RQSTYPE_REQUEST:
((ispreq_t *)fqe)->req_flags |= ddf;
Unify initiator and target DMA setup and command sending. The code is so alike that it is pointless to keep it separate. MFC after: 2 weeks
2017-03-24 14:44:03 +00:00
/* This is historical and not clear whether really needed. */
if (nsegs == 0)
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
nsegs = 1;
((ispreq_t *)fqe)->req_seg_count = nsegs;
isp_put_request(isp, fqe, qe0);
break;
case RQSTYPE_CMDONLY:
((ispreq_t *)fqe)->req_flags |= ddf;
Unify initiator and target DMA setup and command sending. The code is so alike that it is pointless to keep it separate. MFC after: 2 weeks
2017-03-24 14:44:03 +00:00
/* This is historical and not clear whether really needed. */
if (nsegs == 0)
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
nsegs = 1;
((ispextreq_t *)fqe)->req_seg_count = nsegs;
isp_put_extended_request(isp, fqe, qe0);
break;
case RQSTYPE_T2RQS:
((ispreqt2_t *)fqe)->req_flags |= ddf;
((ispreqt2_t *)fqe)->req_seg_count = nsegs;
((ispreqt2_t *)fqe)->req_totalcnt = totalcnt;
if (ISP_CAP_2KLOGIN(isp)) {
isp_put_request_t2e(isp, fqe, qe0);
} else {
isp_put_request_t2(isp, fqe, qe0);
}
break;
case RQSTYPE_A64:
case RQSTYPE_T3RQS:
((ispreqt3_t *)fqe)->req_flags |= ddf;
((ispreqt3_t *)fqe)->req_seg_count = nsegs;
((ispreqt3_t *)fqe)->req_totalcnt = totalcnt;
if (ISP_CAP_2KLOGIN(isp)) {
isp_put_request_t3e(isp, fqe, qe0);
} else {
isp_put_request_t3(isp, fqe, qe0);
}
break;
case RQSTYPE_T7RQS:
Unify initiator and target DMA setup and command sending. The code is so alike that it is pointless to keep it separate. MFC after: 2 weeks
2017-03-24 14:44:03 +00:00
((ispreqt7_t *)fqe)->req_alen_datadir = ddf;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
((ispreqt7_t *)fqe)->req_seg_count = nsegs;
((ispreqt7_t *)fqe)->req_dl = totalcnt;
isp_put_request_t7(isp, fqe, qe0);
break;
Unify initiator and target DMA setup and command sending. The code is so alike that it is pointless to keep it separate. MFC after: 2 weeks
2017-03-24 14:44:03 +00:00
#ifdef ISP_TARGET_MODE
case RQSTYPE_CTIO2:
case RQSTYPE_CTIO3:
if (((ct2_entry_t *)fqe)->ct_flags & CT2_FLAG_MODE2) {
((ct2_entry_t *)fqe)->ct_seg_count = 1;
} else {
((ct2_entry_t *)fqe)->ct_seg_count = nsegs;
}
if (ISP_CAP_2KLOGIN(isp)) {
isp_put_ctio2e(isp, fqe, qe0);
} else {
isp_put_ctio2(isp, fqe, qe0);
}
break;
case RQSTYPE_CTIO7:
if (((ct7_entry_t *)fqe)->ct_flags & CT7_FLAG_MODE2) {
((ct7_entry_t *)fqe)->ct_seg_count = 1;
} else {
((ct7_entry_t *)fqe)->ct_seg_count = nsegs;
}
isp_put_ctio7(isp, fqe, qe0);
break;
#endif
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
default:
return (CMD_COMPLETE);
}
if (isp->isp_dblev & ISP_LOGDEBUG1) {
Polish debugging IOCB dumping. Add few more missing cases, unify byte order. MFC after: 1 month
2016-04-11 10:48:26 +00:00
isp_print_bytes(isp, "first queue entry", QENTRY_LEN, qe0);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
ISP_ADD_REQUEST(isp, nxt);
return (CMD_QUEUED);
}
Unify handles allocation for initiator and target IOCBs. I am not sure why this was split long ago, but I see no reason for it. At this point this unification just slightly reduces memory usage, but as next step I plan to reuse shared handle space for other IOCB types.
2015-12-25 13:03:18 +00:00
uint32_t
isp_allocate_handle(ispsoftc_t *isp, void *xs, int type)
Add a file where inline functions for the Qlogic isp cards can go. This reduces duplication in all the platform specific header files.
1999-10-17 18:15:05 +00:00
{
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
isp_hdl_t *hdp;
Add a file where inline functions for the Qlogic isp cards can go. This reduces duplication in all the platform specific header files.
1999-10-17 18:15:05 +00:00
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
hdp = isp->isp_xffree;
Unify handles allocation for initiator and target IOCBs. I am not sure why this was split long ago, but I see no reason for it. At this point this unification just slightly reduces memory usage, but as next step I plan to reuse shared handle space for other IOCB types.
2015-12-25 13:03:18 +00:00
if (hdp == NULL)
return (ISP_HANDLE_FREE);
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
isp->isp_xffree = hdp->cmd;
hdp->cmd = xs;
hdp->handle = (hdp - isp->isp_xflist);
Unify handles allocation for initiator and target IOCBs. I am not sure why this was split long ago, but I see no reason for it. At this point this unification just slightly reduces memory usage, but as next step I plan to reuse shared handle space for other IOCB types.
2015-12-25 13:03:18 +00:00
hdp->handle |= (type << ISP_HANDLE_USAGE_SHIFT);
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
hdp->handle |= (isp->isp_seqno++ << ISP_HANDLE_SEQ_SHIFT);
Unify handles allocation for initiator and target IOCBs. I am not sure why this was split long ago, but I see no reason for it. At this point this unification just slightly reduces memory usage, but as next step I plan to reuse shared handle space for other IOCB types.
2015-12-25 13:03:18 +00:00
return (hdp->handle);
Add a file where inline functions for the Qlogic isp cards can go. This reduces duplication in all the platform specific header files.
1999-10-17 18:15:05 +00:00
}
Unify handles allocation for initiator and target IOCBs. I am not sure why this was split long ago, but I see no reason for it. At this point this unification just slightly reduces memory usage, but as next step I plan to reuse shared handle space for other IOCB types.
2015-12-25 13:03:18 +00:00
void *
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_find_xs(ispsoftc_t *isp, uint32_t handle)
Add a file where inline functions for the Qlogic isp cards can go. This reduces duplication in all the platform specific header files.
1999-10-17 18:15:05 +00:00
{
Unify handles allocation for initiator and target IOCBs. I am not sure why this was split long ago, but I see no reason for it. At this point this unification just slightly reduces memory usage, but as next step I plan to reuse shared handle space for other IOCB types.
2015-12-25 13:03:18 +00:00
if (!ISP_VALID_HANDLE(isp, handle)) {
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
isp_prt(isp, ISP_LOGERR, "%s: bad handle 0x%x", __func__, handle);
Add a file where inline functions for the Qlogic isp cards can go. This reduces duplication in all the platform specific header files.
1999-10-17 18:15:05 +00:00
return (NULL);
}
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
return (isp->isp_xflist[(handle & ISP_HANDLE_CMD_MASK)].cmd);
Add a file where inline functions for the Qlogic isp cards can go. This reduces duplication in all the platform specific header files.
1999-10-17 18:15:05 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
uint32_t
Unify handles allocation for initiator and target IOCBs. I am not sure why this was split long ago, but I see no reason for it. At this point this unification just slightly reduces memory usage, but as next step I plan to reuse shared handle space for other IOCB types.
2015-12-25 13:03:18 +00:00
isp_find_handle(ispsoftc_t *isp, void *xs)
Add a file where inline functions for the Qlogic isp cards can go. This reduces duplication in all the platform specific header files.
1999-10-17 18:15:05 +00:00
{
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
uint32_t i, foundhdl = ISP_HANDLE_FREE;
Add a file where inline functions for the Qlogic isp cards can go. This reduces duplication in all the platform specific header files.
1999-10-17 18:15:05 +00:00
if (xs != NULL) {
for (i = 0; i < isp->isp_maxcmds; i++) {
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
if (isp->isp_xflist[i].cmd != xs) {
continue;
Add a file where inline functions for the Qlogic isp cards can go. This reduces duplication in all the platform specific header files.
1999-10-17 18:15:05 +00:00
}
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
foundhdl = isp->isp_xflist[i].handle;
break;
Add a file where inline functions for the Qlogic isp cards can go. This reduces duplication in all the platform specific header files.
1999-10-17 18:15:05 +00:00
}
}
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
return (foundhdl);
Add a file where inline functions for the Qlogic isp cards can go. This reduces duplication in all the platform specific header files.
1999-10-17 18:15:05 +00:00
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_destroy_handle(ispsoftc_t *isp, uint32_t handle)
Add a file where inline functions for the Qlogic isp cards can go. This reduces duplication in all the platform specific header files.
1999-10-17 18:15:05 +00:00
{
Unify handles allocation for initiator and target IOCBs. I am not sure why this was split long ago, but I see no reason for it. At this point this unification just slightly reduces memory usage, but as next step I plan to reuse shared handle space for other IOCB types.
2015-12-25 13:03:18 +00:00
if (!ISP_VALID_HANDLE(isp, handle)) {
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
isp_prt(isp, ISP_LOGERR, "%s: bad handle 0x%x", __func__, handle);
} else {
isp->isp_xflist[(handle & ISP_HANDLE_CMD_MASK)].handle = ISP_HANDLE_FREE;
isp->isp_xflist[(handle & ISP_HANDLE_CMD_MASK)].cmd = isp->isp_xffree;
isp->isp_xffree = &isp->isp_xflist[(handle & ISP_HANDLE_CMD_MASK)];
Store the target handles in a separate list from normal commands. Add a CTIO fast post routine to handle CTIO completions. Submitted by: mjacob
2004-05-24 07:02:25 +00:00
}
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
* Make sure we have space to put something on the request queue.
* Return a pointer to that entry if we do. A side effect of this
* function is to update the output index. The input index
* stays the same.
*/
void *
isp_getrqentry(ispsoftc_t *isp)
{
Save one more register read per command by not reading rqstoutrp register every time. The purpose of that register is unlikely output queue overflow detection, so read it only when its last known (and probably stale now) value signals overflow. This reduces CPU load and lock congestion and rises bottleneck in CTL while doing target mode via two 8Gbps ports from 100K to 120K IOPS.
2013-11-10 13:37:44 +00:00
uint32_t next;
next = ISP_NXT_QENTRY(isp->isp_reqidx, RQUEST_QUEUE_LEN(isp));
if (next == isp->isp_reqodx) {
isp->isp_reqodx = ISP_READ(isp, isp->isp_rqstoutrp);
if (next == isp->isp_reqodx)
return (NULL);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
return (ISP_QUEUE_ENTRY(isp->isp_rquest, isp->isp_reqidx));
Add a isp_getrqentry inline function (that is, find the next available request queue entry if any are left).
1999-11-21 03:14:04 +00:00
}
add isp_print_qentry inline function
2000-01-03 22:14:24 +00:00
Remove isp_prtstst (now in case statement in isp.c). Remove isp2100_fw_statename as an INLINE (now a function in isp.c). Remove isp2100_pdb_statename (unused). Redo all ISP_SCSI_XFER_T as XS_T types. Change all RQUEST_QUEUE_LEN/RESULT_QUEUE_LEN macros to take a parameter. Add isp_print_bytes function.
2000-08-01 06:26:04 +00:00
#define TBA (4 * (((QENTRY_LEN >> 2) * 3) + 1) + 1)
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_print_qentry(ispsoftc_t *isp, const char *msg, int idx, void *arg)
add isp_print_qentry inline function
2000-01-03 22:14:24 +00:00
{
Remove isp_prtstst (now in case statement in isp.c). Remove isp2100_fw_statename as an INLINE (now a function in isp.c). Remove isp2100_pdb_statename (unused). Redo all ISP_SCSI_XFER_T as XS_T types. Change all RQUEST_QUEUE_LEN/RESULT_QUEUE_LEN macros to take a parameter. Add isp_print_bytes function.
2000-08-01 06:26:04 +00:00
char buf[TBA];
add isp_print_qentry inline function
2000-01-03 22:14:24 +00:00
int amt, i, j;
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
uint8_t *ptr = arg;
Remove isp_prtstst (now in case statement in isp.c). Remove isp2100_fw_statename as an INLINE (now a function in isp.c). Remove isp2100_pdb_statename (unused). Redo all ISP_SCSI_XFER_T as XS_T types. Change all RQUEST_QUEUE_LEN/RESULT_QUEUE_LEN macros to take a parameter. Add isp_print_bytes function.
2000-08-01 06:26:04 +00:00
Fix isp_print_qentry to print all four lines- it's been broken for months.
2001-03-02 04:48:41 +00:00
isp_prt(isp, ISP_LOGALL, "%s index %d=>", msg, idx);
Remove isp_prtstst (now in case statement in isp.c). Remove isp2100_fw_statename as an INLINE (now a function in isp.c). Remove isp2100_pdb_statename (unused). Redo all ISP_SCSI_XFER_T as XS_T types. Change all RQUEST_QUEUE_LEN/RESULT_QUEUE_LEN macros to take a parameter. Add isp_print_bytes function.
2000-08-01 06:26:04 +00:00
for (buf[0] = 0, amt = i = 0; i < 4; i++) {
buf[0] = 0;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, TBA, " ");
add isp_print_qentry inline function
2000-01-03 22:14:24 +00:00
for (j = 0; j < (QENTRY_LEN >> 2); j++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, TBA, "%s %02x", buf, ptr[amt++] & 0xff);
Remove isp_prtstst (now in case statement in isp.c). Remove isp2100_fw_statename as an INLINE (now a function in isp.c). Remove isp2100_pdb_statename (unused). Redo all ISP_SCSI_XFER_T as XS_T types. Change all RQUEST_QUEUE_LEN/RESULT_QUEUE_LEN macros to take a parameter. Add isp_print_bytes function.
2000-08-01 06:26:04 +00:00
}
Don't pass a buffer directly as a printflike format string. Found by: clang MFC after: 1 month
2010-06-10 19:38:07 +00:00
isp_prt(isp, ISP_LOGALL, "%s", buf);
Remove isp_prtstst (now in case statement in isp.c). Remove isp2100_fw_statename as an INLINE (now a function in isp.c). Remove isp2100_pdb_statename (unused). Redo all ISP_SCSI_XFER_T as XS_T types. Change all RQUEST_QUEUE_LEN/RESULT_QUEUE_LEN macros to take a parameter. Add isp_print_bytes function.
2000-08-01 06:26:04 +00:00
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
MFP4: a) Some constification from NetBSD (gcc 4.1.2) b) Split default param fetching/setting into scsi and fibre functions and retry the fibre fetch more than once. MFC after: 1 week
2007-03-22 23:38:32 +00:00
isp_print_bytes(ispsoftc_t *isp, const char *msg, int amt, void *arg)
Remove isp_prtstst (now in case statement in isp.c). Remove isp2100_fw_statename as an INLINE (now a function in isp.c). Remove isp2100_pdb_statename (unused). Redo all ISP_SCSI_XFER_T as XS_T types. Change all RQUEST_QUEUE_LEN/RESULT_QUEUE_LEN macros to take a parameter. Add isp_print_bytes function.
2000-08-01 06:26:04 +00:00
{
char buf[128];
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
uint8_t *ptr = arg;
Remove isp_prtstst (now in case statement in isp.c). Remove isp2100_fw_statename as an INLINE (now a function in isp.c). Remove isp2100_pdb_statename (unused). Redo all ISP_SCSI_XFER_T as XS_T types. Change all RQUEST_QUEUE_LEN/RESULT_QUEUE_LEN macros to take a parameter. Add isp_print_bytes function.
2000-08-01 06:26:04 +00:00
int off;
various fixes
2000-08-27 23:38:44 +00:00
if (msg)
isp_prt(isp, ISP_LOGALL, "%s:", msg);
Remove isp_prtstst (now in case statement in isp.c). Remove isp2100_fw_statename as an INLINE (now a function in isp.c). Remove isp2100_pdb_statename (unused). Redo all ISP_SCSI_XFER_T as XS_T types. Change all RQUEST_QUEUE_LEN/RESULT_QUEUE_LEN macros to take a parameter. Add isp_print_bytes function.
2000-08-01 06:26:04 +00:00
off = 0;
various fixes
2000-08-27 23:38:44 +00:00
buf[0] = 0;
Remove isp_prtstst (now in case statement in isp.c). Remove isp2100_fw_statename as an INLINE (now a function in isp.c). Remove isp2100_pdb_statename (unused). Redo all ISP_SCSI_XFER_T as XS_T types. Change all RQUEST_QUEUE_LEN/RESULT_QUEUE_LEN macros to take a parameter. Add isp_print_bytes function.
2000-08-01 06:26:04 +00:00
while (off < amt) {
int j, to;
to = off;
for (j = 0; j < 16; j++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_SNPRINTF(buf, 128, "%s %02x", buf, ptr[off++] & 0xff);
if (off == amt) {
Remove isp_prtstst (now in case statement in isp.c). Remove isp2100_fw_statename as an INLINE (now a function in isp.c). Remove isp2100_pdb_statename (unused). Redo all ISP_SCSI_XFER_T as XS_T types. Change all RQUEST_QUEUE_LEN/RESULT_QUEUE_LEN macros to take a parameter. Add isp_print_bytes function.
2000-08-01 06:26:04 +00:00
break;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
add isp_print_qentry inline function
2000-01-03 22:14:24 +00:00
}
various fixes
2000-08-27 23:38:44 +00:00
isp_prt(isp, ISP_LOGALL, "0x%08x:%s", to, buf);
buf[0] = 0;
add isp_print_qentry inline function
2000-01-03 22:14:24 +00:00
}
}
Add isp_fc_runstate function- this function's purpose is to, in stages, and depending on role, make sure link is up, scan the fabric (if we're connected to a fabric), scan the local loop (if appropriate), merge the results into the local port database then, check once again to make sure we have f/w at FW_READY state and the the loopstate is LOOP_READY.
2001-02-11 03:52:04 +00:00
/*
* Do the common path to try and ensure that link is up, we've scanned
* the fabric (if we're on a fabric), and that we've synchronized this
* all with our own database and done the appropriate logins.
*
* We repeatedly check for firmware state and loop state after each
* action because things may have changed while we were doing this.
* Any failure or change of state causes us to return a nonzero value.
*
* We assume we enter here with any locks held.
*/
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
int
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_fc_runstate(ispsoftc_t *isp, int chan, int tval)
Add isp_fc_runstate function- this function's purpose is to, in stages, and depending on role, make sure link is up, scan the fabric (if we're connected to a fabric), scan the local loop (if appropriate), merge the results into the local port database then, check once again to make sure we have f/w at FW_READY state and the the loopstate is LOOP_READY.
2001-02-11 03:52:04 +00:00
{
One more round of port scanner rewrite. - Make scan aborted by event restart immediately and infinitely. - Improve handling of some loop events from firmware. - Remove loop down timer, adding its functionality to scanner thread. - Some more unification and simplification.
2015-11-26 13:04:58 +00:00
fcparam *fcp = FCPARAM(isp, chan);
int res;
Add isp_fc_runstate function- this function's purpose is to, in stages, and depending on role, make sure link is up, scan the fabric (if we're connected to a fabric), scan the local loop (if appropriate), merge the results into the local port database then, check once again to make sure we have f/w at FW_READY state and the the loopstate is LOOP_READY.
2001-02-11 03:52:04 +00:00
One more round of port scanner rewrite. - Make scan aborted by event restart immediately and infinitely. - Improve handling of some loop events from firmware. - Remove loop down timer, adding its functionality to scanner thread. - Some more unification and simplification.
2015-11-26 13:04:58 +00:00
again:
if (fcp->role == ISP_ROLE_NONE)
Unify and cleanup FC ports scan.
2015-11-17 16:33:46 +00:00
return (-1);
One more round of port scanner rewrite. - Make scan aborted by event restart immediately and infinitely. - Improve handling of some loop events from firmware. - Remove loop down timer, adding its functionality to scanner thread. - Some more unification and simplification.
2015-11-26 13:04:58 +00:00
res = isp_control(isp, ISPCTL_FCLINK_TEST, chan, tval);
if (res > 0)
goto again;
if (res < 0)
return (fcp->isp_loopstate);
res = isp_control(isp, ISPCTL_SCAN_LOOP, chan);
if (res > 0)
goto again;
if (res < 0)
return (fcp->isp_loopstate);
res = isp_control(isp, ISPCTL_SCAN_FABRIC, chan);
if (res > 0)
goto again;
if (res < 0)
return (fcp->isp_loopstate);
res = isp_control(isp, ISPCTL_PDB_SYNC, chan);
if (res > 0)
goto again;
return (fcp->isp_loopstate);
Add isp_fc_runstate function- this function's purpose is to, in stages, and depending on role, make sure link is up, scan the fabric (if we're connected to a fabric), scan the local loop (if appropriate), merge the results into the local port database then, check once again to make sure we have f/w at FW_READY state and the the loopstate is LOOP_READY.
2001-02-11 03:52:04 +00:00
}
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
/*
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
* Fibre Channel Support routines
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
*/
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_dump_portdb(ispsoftc_t *isp, int chan)
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
{
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcparam *fcp = FCPARAM(isp, chan);
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
int i;
for (i = 0; i < MAX_FC_TARG; i++) {
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
char buf1[64], buf2[64];
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
const char *dbs[8] = {
"NIL ",
"PROB",
"DEAD",
"CHGD",
"NEW ",
"PVLD",
"ZOMB",
"VLD "
};
fcportdb_t *lp = &fcp->portdb[i];
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
if (lp->state == FC_PORTDB_STATE_NIL) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
continue;
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
isp_gen_role_str(buf1, sizeof (buf1), lp->prli_word3);
isp_gen_role_str(buf2, sizeof (buf2), lp->new_prli_word3);
Completely remove broken now autologin port flag. Firmware automatically logs in only to local loop ports, and those ports can be easily identified without extra flag by zero domain and area IDs. MFC after: 1 week
2016-05-17 13:18:57 +00:00
isp_prt(isp, ISP_LOGALL, "Chan %d [%d]: hdl 0x%x %s %s 0x%06x =>%s 0x%06x; WWNN 0x%08x%08x WWPN 0x%08x%08x",
chan, i, lp->handle, dbs[lp->state], buf1, lp->portid, buf2, lp->new_portid,
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
(uint32_t) (lp->node_wwn >> 32), (uint32_t) (lp->node_wwn), (uint32_t) (lp->port_wwn >> 32), (uint32_t) (lp->port_wwn));
}
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
void
isp_gen_role_str(char *buf, size_t len, uint16_t p3)
{
int nd = 0;
buf[0] = '(';
buf[1] = 0;
if (p3 & PRLI_WD3_ENHANCED_DISCOVERY) {
nd++;
strlcat(buf, "EDisc", len);
}
if (p3 & PRLI_WD3_REC_SUPPORT) {
if (nd++) {
strlcat(buf, ",", len);
}
strlcat(buf, "REC", len);
}
if (p3 & PRLI_WD3_TASK_RETRY_IDENTIFICATION_REQUESTED) {
if (nd++) {
strlcat(buf, ",", len);
}
strlcat(buf, "RetryID", len);
}
if (p3 & PRLI_WD3_RETRY) {
if (nd++) {
strlcat(buf, ",", len);
}
strlcat(buf, "Retry", len);
}
if (p3 & PRLI_WD3_CONFIRMED_COMPLETION_ALLOWED) {
if (nd++) {
strlcat(buf, ",", len);
}
strlcat(buf, "CNFRM", len);
}
if (p3 & PRLI_WD3_DATA_OVERLAY_ALLOWED) {
if (nd++) {
strlcat(buf, ",", len);
}
strlcat(buf, "DOver", len);
}
if (p3 & PRLI_WD3_INITIATOR_FUNCTION) {
if (nd++) {
strlcat(buf, ",", len);
}
strlcat(buf, "INI", len);
}
if (p3 & PRLI_WD3_TARGET_FUNCTION) {
if (nd++) {
strlcat(buf, ",", len);
}
strlcat(buf, "TGT", len);
}
if (p3 & PRLI_WD3_READ_FCP_XFER_RDY_DISABLED) {
if (nd++) {
strlcat(buf, ",", len);
}
strlcat(buf, "RdXfrDis", len);
}
if (p3 & PRLI_WD3_WRITE_FCP_XFER_RDY_DISABLED) {
if (nd++) {
strlcat(buf, ",", len);
}
strlcat(buf, "XfrDis", len);
}
strlcat(buf, ")", len);
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
const char *
isp_fc_fw_statename(int state)
{
switch (state) {
case FW_CONFIG_WAIT: return "Config Wait";
Unify and cleanup FC ports scan.
2015-11-17 16:33:46 +00:00
case FW_WAIT_LINK: return "Wait Link";
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
case FW_WAIT_LOGIN: return "Wait Login";
case FW_READY: return "Ready";
case FW_LOSS_OF_SYNC: return "Loss Of Sync";
case FW_ERROR: return "Error";
case FW_REINIT: return "Re-Init";
case FW_NON_PART: return "Nonparticipating";
default: return "?????";
}
}
const char *
isp_fc_loop_statename(int state)
{
switch (state) {
case LOOP_NIL: return "NIL";
One more round of port scanner rewrite. - Make scan aborted by event restart immediately and infinitely. - Improve handling of some loop events from firmware. - Remove loop down timer, adding its functionality to scanner thread. - Some more unification and simplification.
2015-11-26 13:04:58 +00:00
case LOOP_HAVE_LINK: return "Have Link";
Extract virtual port address from RQSTYPE_RPT_ID_ACQ. This should close the race between request arriving on new target mode virtual port and its scanner thread finally fetch its address for request routing.
2016-04-14 20:49:01 +00:00
case LOOP_HAVE_ADDR: return "Have Address";
Unify and cleanup FC ports scan.
2015-11-17 16:33:46 +00:00
case LOOP_TESTING_LINK: return "Testing Link";
case LOOP_LTEST_DONE: return "Link Test Done";
case LOOP_SCANNING_LOOP: return "Scanning Loop";
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
case LOOP_LSCAN_DONE: return "Loop Scan Done";
case LOOP_SCANNING_FABRIC: return "Scanning Fabric";
case LOOP_FSCAN_DONE: return "Fabric Scan Done";
case LOOP_SYNCING_PDB: return "Syncing PDB";
case LOOP_READY: return "Ready";
default: return "?????";
}
}
const char *
isp_fc_toponame(fcparam *fcp)
{
Extract virtual port address from RQSTYPE_RPT_ID_ACQ. This should close the race between request arriving on new target mode virtual port and its scanner thread finally fetch its address for request routing.
2016-04-14 20:49:01 +00:00
if (fcp->isp_loopstate < LOOP_HAVE_ADDR) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
return "Unavailable";
}
switch (fcp->isp_topo) {
Another round of port scanner rewrite. This change simplifies and unifies port adding/updating for loop and fabric scanners. It also fixes problems with scanning restarts due to concurrent port databases changes. It also fixes many cosmetic issues.
2015-11-19 17:43:47 +00:00
case TOPO_NL_PORT: return "Private Loop (NL_Port)";
case TOPO_FL_PORT: return "Public Loop (FL_Port)";
case TOPO_N_PORT: return "Point-to-Point (N_Port)";
case TOPO_F_PORT: return "Fabric (F_Port)";
case TOPO_PTP_STUB: return "Point-to-Point (no response)";
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
default: return "?????";
}
}
void
isp_clear_commands(ispsoftc_t *isp)
{
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
uint32_t tmp;
isp_hdl_t *hdp;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
#ifdef ISP_TARGET_MODE
isp_notify_t notify;
#endif
for (tmp = 0; isp->isp_xflist && tmp < isp->isp_maxcmds; tmp++) {
Redo how commands handles are created and managed and implement sequence numbers and handle types in rational way. This will better protect from (unwittingly) dealing with stale handles/commands. Fix the watchdog timeout code to better protect itself from mistakes. If we run an abort on a putatively timed out command, the command may in fact get completed, so check to make sure the command we're timing it out is still around. If the abort succeeds, btw, the command should get returned via a different path.
2010-02-03 21:09:32 +00:00
hdp = &isp->isp_xflist[tmp];
Make virtual ports control asynchronous. Before this change virtual ports control IOCBs were executed synchronously via Execute IOCB mailbox command. It required exclusive use of scratch space of driver and mailbox registers of the hardware. Because of that shared resources use this code could not really sleep, having to spin for completion, blocking any other operation. This change introduces new asynchronous design, sending the IOCBs directly on request queue and gracefully waiting for their return on response queue. Returned IOCBs are identified with unified handle space from r292725.
2015-12-26 04:26:32 +00:00
switch (ISP_H2HT(hdp->handle)) {
case ISP_HANDLE_INITIATOR: {
Unify handles allocation for initiator and target IOCBs. I am not sure why this was split long ago, but I see no reason for it. At this point this unification just slightly reduces memory usage, but as next step I plan to reuse shared handle space for other IOCB types.
2015-12-25 13:03:18 +00:00
XS_T *xs = hdp->cmd;
if (XS_XFRLEN(xs)) {
ISP_DMAFREE(isp, xs, hdp->handle);
XS_SET_RESID(xs, XS_XFRLEN(xs));
} else {
XS_SET_RESID(xs, 0);
}
Make virtual ports control asynchronous. Before this change virtual ports control IOCBs were executed synchronously via Execute IOCB mailbox command. It required exclusive use of scratch space of driver and mailbox registers of the hardware. Because of that shared resources use this code could not really sleep, having to spin for completion, blocking any other operation. This change introduces new asynchronous design, sending the IOCBs directly on request queue and gracefully waiting for their return on response queue. Returned IOCBs are identified with unified handle space from r292725.
2015-12-26 04:26:32 +00:00
isp_destroy_handle(isp, hdp->handle);
Unify handles allocation for initiator and target IOCBs. I am not sure why this was split long ago, but I see no reason for it. At this point this unification just slightly reduces memory usage, but as next step I plan to reuse shared handle space for other IOCB types.
2015-12-25 13:03:18 +00:00
XS_SETERR(xs, HBA_BUSRESET);
isp_done(xs);
Make virtual ports control asynchronous. Before this change virtual ports control IOCBs were executed synchronously via Execute IOCB mailbox command. It required exclusive use of scratch space of driver and mailbox registers of the hardware. Because of that shared resources use this code could not really sleep, having to spin for completion, blocking any other operation. This change introduces new asynchronous design, sending the IOCBs directly on request queue and gracefully waiting for their return on response queue. Returned IOCBs are identified with unified handle space from r292725.
2015-12-26 04:26:32 +00:00
break;
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
#ifdef ISP_TARGET_MODE
Make virtual ports control asynchronous. Before this change virtual ports control IOCBs were executed synchronously via Execute IOCB mailbox command. It required exclusive use of scratch space of driver and mailbox registers of the hardware. Because of that shared resources use this code could not really sleep, having to spin for completion, blocking any other operation. This change introduces new asynchronous design, sending the IOCBs directly on request queue and gracefully waiting for their return on response queue. Returned IOCBs are identified with unified handle space from r292725.
2015-12-26 04:26:32 +00:00
case ISP_HANDLE_TARGET: {
Unify handles allocation for initiator and target IOCBs. I am not sure why this was split long ago, but I see no reason for it. At this point this unification just slightly reduces memory usage, but as next step I plan to reuse shared handle space for other IOCB types.
2015-12-25 13:03:18 +00:00
uint8_t local[QENTRY_LEN];
ISP_DMAFREE(isp, hdp->cmd, hdp->handle);
ISP_MEMZERO(local, QENTRY_LEN);
if (IS_24XX(isp)) {
ct7_entry_t *ctio = (ct7_entry_t *) local;
ctio->ct_syshandle = hdp->handle;
ctio->ct_nphdl = CT_HBA_RESET;
ctio->ct_header.rqs_entry_type = RQSTYPE_CTIO7;
} else {
ct2_entry_t *ctio = (ct2_entry_t *) local;
ctio->ct_syshandle = hdp->handle;
ctio->ct_status = CT_HBA_RESET;
ctio->ct_header.rqs_entry_type = RQSTYPE_CTIO2;
}
isp_async(isp, ISPASYNC_TARGET_ACTION, local);
Make virtual ports control asynchronous. Before this change virtual ports control IOCBs were executed synchronously via Execute IOCB mailbox command. It required exclusive use of scratch space of driver and mailbox registers of the hardware. Because of that shared resources use this code could not really sleep, having to spin for completion, blocking any other operation. This change introduces new asynchronous design, sending the IOCBs directly on request queue and gracefully waiting for their return on response queue. Returned IOCBs are identified with unified handle space from r292725.
2015-12-26 04:26:32 +00:00
break;
}
Unify handles allocation for initiator and target IOCBs. I am not sure why this was split long ago, but I see no reason for it. At this point this unification just slightly reduces memory usage, but as next step I plan to reuse shared handle space for other IOCB types.
2015-12-25 13:03:18 +00:00
#endif
Make virtual ports control asynchronous. Before this change virtual ports control IOCBs were executed synchronously via Execute IOCB mailbox command. It required exclusive use of scratch space of driver and mailbox registers of the hardware. Because of that shared resources use this code could not really sleep, having to spin for completion, blocking any other operation. This change introduces new asynchronous design, sending the IOCBs directly on request queue and gracefully waiting for their return on response queue. Returned IOCBs are identified with unified handle space from r292725.
2015-12-26 04:26:32 +00:00
case ISP_HANDLE_CTRL:
wakeup(hdp->cmd);
isp_destroy_handle(isp, hdp->handle);
break;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
}
Unify handles allocation for initiator and target IOCBs. I am not sure why this was split long ago, but I see no reason for it. At this point this unification just slightly reduces memory usage, but as next step I plan to reuse shared handle space for other IOCB types.
2015-12-25 13:03:18 +00:00
#ifdef ISP_TARGET_MODE
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
for (tmp = 0; tmp < isp->isp_nchan; tmp++) {
ISP_MEMZERO(&notify, sizeof (isp_notify_t));
notify.nt_ncode = NT_HBA_RESET;
notify.nt_hba = isp;
notify.nt_wwn = INI_ANY;
notify.nt_nphdl = NIL_HANDLE;
notify.nt_sid = PORT_ANY;
notify.nt_did = PORT_ANY;
notify.nt_tgt = TGT_ANY;
notify.nt_channel = tmp;
notify.nt_lun = LUN_ANY;
notify.nt_tagval = TAG_ANY;
isp_async(isp, ISPASYNC_TARGET_NOTIFY, &notify);
}
#endif
Push things closer to path failover by implementing loop down and gone device timers and zombie state entries. There are tunables that can be used to select a number of parameters. loop_down_limit - how long to wait for loop to come back up before declaring all devices dead (default 300 seconds) gone_device_time- how long to wait for a device that has appeared to leave the loop or fabric to reappear (default 30 seconds) Internal tunables include (which should be externalized): quick_boot_time- how long to wait when booting for loop to come up change_is_bad- whether or not to accept devices with the same WWNN/WWPN that reappear at a different PortID as being the 'same' device. Keen students of some of the subtle issues here will ask how one can keep devices from being re-accepted at all (the answer is to set a gone_device_time to zero- that effectively would be the same thing).
2006-11-14 08:45:48 +00:00
}
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
/*
* Functions to move stuff to a form that the QLogic RISC engine understands
* and functions to move stuff back to a form the processor understands.
*
* Each platform is required to provide the 8, 16 and 32 bit
* swizzle and unswizzle macros (ISP_IOX{PUT|GET}_{8,16,32})
*
* The assumption is that swizzling and unswizzling is mostly done 'in place'
* (with a few exceptions for efficiency).
*/
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
#define ISP_IS_SBUS(isp) (ISP_SBUS_SUPPORTED && (isp)->isp_bustype == ISP_BT_SBUS)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
#define ASIZE(x) (sizeof (x) / sizeof (x[0]))
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
/*
* Swizzle/Copy Functions
*/
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_hdr(ispsoftc_t *isp, isphdr_t *hpsrc, isphdr_t *hpdst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
if (ISP_IS_SBUS(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_8(isp, hpsrc->rqs_entry_type, &hpdst->rqs_entry_count);
ISP_IOXPUT_8(isp, hpsrc->rqs_entry_count, &hpdst->rqs_entry_type);
ISP_IOXPUT_8(isp, hpsrc->rqs_seqno, &hpdst->rqs_flags);
ISP_IOXPUT_8(isp, hpsrc->rqs_flags, &hpdst->rqs_seqno);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_8(isp, hpsrc->rqs_entry_type, &hpdst->rqs_entry_type);
ISP_IOXPUT_8(isp, hpsrc->rqs_entry_count, &hpdst->rqs_entry_count);
ISP_IOXPUT_8(isp, hpsrc->rqs_seqno, &hpdst->rqs_seqno);
ISP_IOXPUT_8(isp, hpsrc->rqs_flags, &hpdst->rqs_flags);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_hdr(ispsoftc_t *isp, isphdr_t *hpsrc, isphdr_t *hpdst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
if (ISP_IS_SBUS(isp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &hpsrc->rqs_entry_type, hpdst->rqs_entry_count);
ISP_IOXGET_8(isp, &hpsrc->rqs_entry_count, hpdst->rqs_entry_type);
ISP_IOXGET_8(isp, &hpsrc->rqs_seqno, hpdst->rqs_flags);
ISP_IOXGET_8(isp, &hpsrc->rqs_flags, hpdst->rqs_seqno);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &hpsrc->rqs_entry_type, hpdst->rqs_entry_type);
ISP_IOXGET_8(isp, &hpsrc->rqs_entry_count, hpdst->rqs_entry_count);
ISP_IOXGET_8(isp, &hpsrc->rqs_seqno, hpdst->rqs_seqno);
ISP_IOXGET_8(isp, &hpsrc->rqs_flags, hpdst->rqs_flags);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
int
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
isp_get_response_type(ispsoftc_t *isp, isphdr_t *hp)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
uint8_t type;
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
if (ISP_IS_SBUS(isp)) {
ISP_IOXGET_8(isp, &hp->rqs_entry_count, type);
} else {
ISP_IOXGET_8(isp, &hp->rqs_entry_type, type);
}
return ((int)type);
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
isp_put_request(ispsoftc_t *isp, ispreq_t *rqsrc, ispreq_t *rqdst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
int i;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_hdr(isp, &rqsrc->req_header, &rqdst->req_header);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
ISP_IOXPUT_32(isp, rqsrc->req_handle, &rqdst->req_handle);
if (ISP_IS_SBUS(isp)) {
ISP_IOXPUT_8(isp, rqsrc->req_lun_trn, &rqdst->req_target);
ISP_IOXPUT_8(isp, rqsrc->req_target, &rqdst->req_lun_trn);
} else {
ISP_IOXPUT_8(isp, rqsrc->req_lun_trn, &rqdst->req_lun_trn);
ISP_IOXPUT_8(isp, rqsrc->req_target, &rqdst->req_target);
}
ISP_IOXPUT_16(isp, rqsrc->req_cdblen, &rqdst->req_cdblen);
ISP_IOXPUT_16(isp, rqsrc->req_flags, &rqdst->req_flags);
ISP_IOXPUT_16(isp, rqsrc->req_time, &rqdst->req_time);
ISP_IOXPUT_16(isp, rqsrc->req_seg_count, &rqdst->req_seg_count);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
for (i = 0; i < ASIZE(rqsrc->req_cdb); i++) {
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
ISP_IOXPUT_8(isp, rqsrc->req_cdb[i], &rqdst->req_cdb[i]);
}
for (i = 0; i < ISP_RQDSEG; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, rqsrc->req_dataseg[i].ds_base, &rqdst->req_dataseg[i].ds_base);
ISP_IOXPUT_32(isp, rqsrc->req_dataseg[i].ds_count, &rqdst->req_dataseg[i].ds_count);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_marker(ispsoftc_t *isp, isp_marker_t *src, isp_marker_t *dst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
int i;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_hdr(isp, &src->mrk_header, &dst->mrk_header);
ISP_IOXPUT_32(isp, src->mrk_handle, &dst->mrk_handle);
if (ISP_IS_SBUS(isp)) {
ISP_IOXPUT_8(isp, src->mrk_reserved0, &dst->mrk_target);
ISP_IOXPUT_8(isp, src->mrk_target, &dst->mrk_reserved0);
} else {
ISP_IOXPUT_8(isp, src->mrk_reserved0, &dst->mrk_reserved0);
ISP_IOXPUT_8(isp, src->mrk_target, &dst->mrk_target);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->mrk_modifier, &dst->mrk_modifier);
ISP_IOXPUT_16(isp, src->mrk_flags, &dst->mrk_flags);
ISP_IOXPUT_16(isp, src->mrk_lun, &dst->mrk_lun);
for (i = 0; i < ASIZE(src->mrk_reserved1); i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_8(isp, src->mrk_reserved1[i], &dst->mrk_reserved1[i]);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_put_marker_24xx(ispsoftc_t *isp, isp_marker_24xx_t *src, isp_marker_24xx_t *dst)
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
{
int i;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_hdr(isp, &src->mrk_header, &dst->mrk_header);
ISP_IOXPUT_32(isp, src->mrk_handle, &dst->mrk_handle);
ISP_IOXPUT_16(isp, src->mrk_nphdl, &dst->mrk_nphdl);
ISP_IOXPUT_8(isp, src->mrk_modifier, &dst->mrk_modifier);
ISP_IOXPUT_8(isp, src->mrk_reserved0, &dst->mrk_reserved0);
ISP_IOXPUT_8(isp, src->mrk_reserved1, &dst->mrk_reserved1);
ISP_IOXPUT_8(isp, src->mrk_vphdl, &dst->mrk_vphdl);
ISP_IOXPUT_8(isp, src->mrk_reserved2, &dst->mrk_reserved2);
for (i = 0; i < ASIZE(src->mrk_lun); i++) {
ISP_IOXPUT_8(isp, src->mrk_lun[i], &dst->mrk_lun[i]);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
for (i = 0; i < ASIZE(src->mrk_reserved3); i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_8(isp, src->mrk_reserved3[i], &dst->mrk_reserved3[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
}
void
isp_put_request_t2(ispsoftc_t *isp, ispreqt2_t *src, ispreqt2_t *dst)
{
int i;
isp_put_hdr(isp, &src->req_header, &dst->req_header);
ISP_IOXPUT_32(isp, src->req_handle, &dst->req_handle);
ISP_IOXPUT_8(isp, src->req_lun_trn, &dst->req_lun_trn);
ISP_IOXPUT_8(isp, src->req_target, &dst->req_target);
ISP_IOXPUT_16(isp, src->req_scclun, &dst->req_scclun);
ISP_IOXPUT_16(isp, src->req_flags, &dst->req_flags);
Fix target mode support for Qlogic 2200 FC adapters. Now target mode works for all supported FC adapters except ancient 2100, which is not tested.
2015-11-23 15:49:50 +00:00
ISP_IOXPUT_8(isp, src->req_crn, &dst->req_crn);
ISP_IOXPUT_8(isp, src->req_reserved, &dst->req_reserved);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->req_time, &dst->req_time);
ISP_IOXPUT_16(isp, src->req_seg_count, &dst->req_seg_count);
for (i = 0; i < ASIZE(src->req_cdb); i++) {
ISP_IOXPUT_8(isp, src->req_cdb[i], &dst->req_cdb[i]);
}
ISP_IOXPUT_32(isp, src->req_totalcnt, &dst->req_totalcnt);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
for (i = 0; i < ISP_RQDSEG_T2; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->req_dataseg[i].ds_base, &dst->req_dataseg[i].ds_base);
ISP_IOXPUT_32(isp, src->req_dataseg[i].ds_count, &dst->req_dataseg[i].ds_count);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_request_t2e(ispsoftc_t *isp, ispreqt2e_t *src, ispreqt2e_t *dst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
int i;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_hdr(isp, &src->req_header, &dst->req_header);
ISP_IOXPUT_32(isp, src->req_handle, &dst->req_handle);
ISP_IOXPUT_16(isp, src->req_target, &dst->req_target);
ISP_IOXPUT_16(isp, src->req_scclun, &dst->req_scclun);
ISP_IOXPUT_16(isp, src->req_flags, &dst->req_flags);
Fix target mode support for Qlogic 2200 FC adapters. Now target mode works for all supported FC adapters except ancient 2100, which is not tested.
2015-11-23 15:49:50 +00:00
ISP_IOXPUT_8(isp, src->req_crn, &dst->req_crn);
ISP_IOXPUT_8(isp, src->req_reserved, &dst->req_reserved);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->req_time, &dst->req_time);
ISP_IOXPUT_16(isp, src->req_seg_count, &dst->req_seg_count);
for (i = 0; i < ASIZE(src->req_cdb); i++) {
ISP_IOXPUT_8(isp, src->req_cdb[i], &dst->req_cdb[i]);
}
ISP_IOXPUT_32(isp, src->req_totalcnt, &dst->req_totalcnt);
for (i = 0; i < ISP_RQDSEG_T2; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->req_dataseg[i].ds_base, &dst->req_dataseg[i].ds_base);
ISP_IOXPUT_32(isp, src->req_dataseg[i].ds_count, &dst->req_dataseg[i].ds_count);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
void
isp_put_request_t3(ispsoftc_t *isp, ispreqt3_t *src, ispreqt3_t *dst)
{
int i;
isp_put_hdr(isp, &src->req_header, &dst->req_header);
ISP_IOXPUT_32(isp, src->req_handle, &dst->req_handle);
ISP_IOXPUT_8(isp, src->req_lun_trn, &dst->req_lun_trn);
ISP_IOXPUT_8(isp, src->req_target, &dst->req_target);
ISP_IOXPUT_16(isp, src->req_scclun, &dst->req_scclun);
ISP_IOXPUT_16(isp, src->req_flags, &dst->req_flags);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ISP_IOXPUT_8(isp, src->req_crn, &dst->req_crn);
ISP_IOXPUT_8(isp, src->req_reserved, &dst->req_reserved);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->req_time, &dst->req_time);
ISP_IOXPUT_16(isp, src->req_seg_count, &dst->req_seg_count);
for (i = 0; i < ASIZE(src->req_cdb); i++) {
ISP_IOXPUT_8(isp, src->req_cdb[i], &dst->req_cdb[i]);
}
ISP_IOXPUT_32(isp, src->req_totalcnt, &dst->req_totalcnt);
Fix buglet in A64 CTIO3 structure copy.
2002-09-23 05:00:37 +00:00
for (i = 0; i < ISP_RQDSEG_T3; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->req_dataseg[i].ds_base, &dst->req_dataseg[i].ds_base);
ISP_IOXPUT_32(isp, src->req_dataseg[i].ds_basehi, &dst->req_dataseg[i].ds_basehi);
ISP_IOXPUT_32(isp, src->req_dataseg[i].ds_count, &dst->req_dataseg[i].ds_count);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_request_t3e(ispsoftc_t *isp, ispreqt3e_t *src, ispreqt3e_t *dst)
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
{
int i;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_hdr(isp, &src->req_header, &dst->req_header);
ISP_IOXPUT_32(isp, src->req_handle, &dst->req_handle);
ISP_IOXPUT_16(isp, src->req_target, &dst->req_target);
ISP_IOXPUT_16(isp, src->req_scclun, &dst->req_scclun);
ISP_IOXPUT_16(isp, src->req_flags, &dst->req_flags);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ISP_IOXPUT_8(isp, src->req_crn, &dst->req_crn);
ISP_IOXPUT_8(isp, src->req_reserved, &dst->req_reserved);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->req_time, &dst->req_time);
ISP_IOXPUT_16(isp, src->req_seg_count, &dst->req_seg_count);
for (i = 0; i < ASIZE(src->req_cdb); i++) {
ISP_IOXPUT_8(isp, src->req_cdb[i], &dst->req_cdb[i]);
}
ISP_IOXPUT_32(isp, src->req_totalcnt, &dst->req_totalcnt);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
for (i = 0; i < ISP_RQDSEG_T3; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->req_dataseg[i].ds_base, &dst->req_dataseg[i].ds_base);
ISP_IOXPUT_32(isp, src->req_dataseg[i].ds_basehi, &dst->req_dataseg[i].ds_basehi);
ISP_IOXPUT_32(isp, src->req_dataseg[i].ds_count, &dst->req_dataseg[i].ds_count);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_extended_request(ispsoftc_t *isp, ispextreq_t *src, ispextreq_t *dst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
int i;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_hdr(isp, &src->req_header, &dst->req_header);
ISP_IOXPUT_32(isp, src->req_handle, &dst->req_handle);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
if (ISP_IS_SBUS(isp)) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_8(isp, src->req_lun_trn, &dst->req_target);
ISP_IOXPUT_8(isp, src->req_target, &dst->req_lun_trn);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
} else {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_8(isp, src->req_lun_trn, &dst->req_lun_trn);
ISP_IOXPUT_8(isp, src->req_target, &dst->req_target);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->req_cdblen, &dst->req_cdblen);
ISP_IOXPUT_16(isp, src->req_flags, &dst->req_flags);
ISP_IOXPUT_16(isp, src->req_time, &dst->req_time);
ISP_IOXPUT_16(isp, src->req_seg_count, &dst->req_seg_count);
for (i = 0; i < ASIZE(src->req_cdb); i++) {
ISP_IOXPUT_8(isp, src->req_cdb[i], &dst->req_cdb[i]);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_request_t7(ispsoftc_t *isp, ispreqt7_t *src, ispreqt7_t *dst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
int i;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
uint32_t *a, *b;
isp_put_hdr(isp, &src->req_header, &dst->req_header);
ISP_IOXPUT_32(isp, src->req_handle, &dst->req_handle);
ISP_IOXPUT_16(isp, src->req_nphdl, &dst->req_nphdl);
ISP_IOXPUT_16(isp, src->req_time, &dst->req_time);
ISP_IOXPUT_16(isp, src->req_seg_count, &dst->req_seg_count);
ISP_IOXPUT_16(isp, src->req_reserved, &dst->req_reserved);
a = (uint32_t *) src->req_lun;
b = (uint32_t *) dst->req_lun;
for (i = 0; i < (ASIZE(src->req_lun) >> 2); i++ ) {
Recover from some major omissions/problems with the 24XX port. First, we were never correctly checking for a 24XX Status Type 0 response- that cased us to fall through to evaluate status for commands as if this were a 2100/2200/2300 Status Type 0 response. This is *close*, but not quite the same. This has been reported to be apparent with some wierd lun configuration problems with some arrays. It became glaringly apparent on sparc64 where none of the correct byte swap things were done. Fixing this omission then caused a whole universe shifting debug cycle of endian issues for the 2400. The manual for 24XX f/w turns out to be wrong about the endianness of a couple of entities. The lun and cdb fields for the type 7 request are *not* unconditionally big endian- they happen to be opposite of whatever the endian of the current machine type is. Same with the sense data for the 24XX type 0 response. While we're at it investigate and resolve some NVRAM endian issues. Approved by: re (ken) MFC after: 3 days
2007-07-02 20:08:20 +00:00
*b++ = ISP_SWAP32(isp, *a++);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
ISP_IOXPUT_8(isp, src->req_alen_datadir, &dst->req_alen_datadir);
ISP_IOXPUT_8(isp, src->req_task_management, &dst->req_task_management);
ISP_IOXPUT_8(isp, src->req_task_attribute, &dst->req_task_attribute);
ISP_IOXPUT_8(isp, src->req_crn, &dst->req_crn);
a = (uint32_t *) src->req_cdb;
b = (uint32_t *) dst->req_cdb;
Recover from some major omissions/problems with the 24XX port. First, we were never correctly checking for a 24XX Status Type 0 response- that cased us to fall through to evaluate status for commands as if this were a 2100/2200/2300 Status Type 0 response. This is *close*, but not quite the same. This has been reported to be apparent with some wierd lun configuration problems with some arrays. It became glaringly apparent on sparc64 where none of the correct byte swap things were done. Fixing this omission then caused a whole universe shifting debug cycle of endian issues for the 2400. The manual for 24XX f/w turns out to be wrong about the endianness of a couple of entities. The lun and cdb fields for the type 7 request are *not* unconditionally big endian- they happen to be opposite of whatever the endian of the current machine type is. Same with the sense data for the 24XX type 0 response. While we're at it investigate and resolve some NVRAM endian issues. Approved by: re (ken) MFC after: 3 days
2007-07-02 20:08:20 +00:00
for (i = 0; i < (ASIZE(src->req_cdb) >> 2); i++) {
*b++ = ISP_SWAP32(isp, *a++);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
ISP_IOXPUT_32(isp, src->req_dl, &dst->req_dl);
ISP_IOXPUT_16(isp, src->req_tidlo, &dst->req_tidlo);
ISP_IOXPUT_8(isp, src->req_tidhi, &dst->req_tidhi);
ISP_IOXPUT_8(isp, src->req_vpidx, &dst->req_vpidx);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->req_dataseg.ds_base, &dst->req_dataseg.ds_base);
ISP_IOXPUT_32(isp, src->req_dataseg.ds_basehi, &dst->req_dataseg.ds_basehi);
ISP_IOXPUT_32(isp, src->req_dataseg.ds_count, &dst->req_dataseg.ds_count);
}
void
isp_put_24xx_tmf(ispsoftc_t *isp, isp24xx_tmf_t *src, isp24xx_tmf_t *dst)
{
int i;
uint32_t *a, *b;
isp_put_hdr(isp, &src->tmf_header, &dst->tmf_header);
ISP_IOXPUT_32(isp, src->tmf_handle, &dst->tmf_handle);
ISP_IOXPUT_16(isp, src->tmf_nphdl, &dst->tmf_nphdl);
ISP_IOXPUT_16(isp, src->tmf_delay, &dst->tmf_delay);
ISP_IOXPUT_16(isp, src->tmf_timeout, &dst->tmf_timeout);
for (i = 0; i < ASIZE(src->tmf_reserved0); i++) {
ISP_IOXPUT_8(isp, src->tmf_reserved0[i], &dst->tmf_reserved0[i]);
}
a = (uint32_t *) src->tmf_lun;
b = (uint32_t *) dst->tmf_lun;
for (i = 0; i < (ASIZE(src->tmf_lun) >> 2); i++ ) {
*b++ = ISP_SWAP32(isp, *a++);
}
ISP_IOXPUT_32(isp, src->tmf_flags, &dst->tmf_flags);
for (i = 0; i < ASIZE(src->tmf_reserved1); i++) {
ISP_IOXPUT_8(isp, src->tmf_reserved1[i], &dst->tmf_reserved1[i]);
}
ISP_IOXPUT_16(isp, src->tmf_tidlo, &dst->tmf_tidlo);
ISP_IOXPUT_8(isp, src->tmf_tidhi, &dst->tmf_tidhi);
ISP_IOXPUT_8(isp, src->tmf_vpidx, &dst->tmf_vpidx);
for (i = 0; i < ASIZE(src->tmf_reserved2); i++) {
ISP_IOXPUT_8(isp, src->tmf_reserved2[i], &dst->tmf_reserved2[i]);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
void
isp_put_24xx_abrt(ispsoftc_t *isp, isp24xx_abrt_t *src, isp24xx_abrt_t *dst)
{
int i;
isp_put_hdr(isp, &src->abrt_header, &dst->abrt_header);
ISP_IOXPUT_32(isp, src->abrt_handle, &dst->abrt_handle);
ISP_IOXPUT_16(isp, src->abrt_nphdl, &dst->abrt_nphdl);
ISP_IOXPUT_16(isp, src->abrt_options, &dst->abrt_options);
ISP_IOXPUT_32(isp, src->abrt_cmd_handle, &dst->abrt_cmd_handle);
Add new field to Abort IOCB.
2015-10-24 19:38:06 +00:00
ISP_IOXPUT_16(isp, src->abrt_queue_number, &dst->abrt_queue_number);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
for (i = 0; i < ASIZE(src->abrt_reserved); i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_8(isp, src->abrt_reserved[i], &dst->abrt_reserved[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
ISP_IOXPUT_16(isp, src->abrt_tidlo, &dst->abrt_tidlo);
ISP_IOXPUT_8(isp, src->abrt_tidhi, &dst->abrt_tidhi);
ISP_IOXPUT_8(isp, src->abrt_vpidx, &dst->abrt_vpidx);
for (i = 0; i < ASIZE(src->abrt_reserved1); i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_8(isp, src->abrt_reserved1[i], &dst->abrt_reserved1[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
}
void
isp_put_cont_req(ispsoftc_t *isp, ispcontreq_t *src, ispcontreq_t *dst)
{
int i;
isp_put_hdr(isp, &src->req_header, &dst->req_header);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
for (i = 0; i < ISP_CDSEG; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->req_dataseg[i].ds_base, &dst->req_dataseg[i].ds_base);
ISP_IOXPUT_32(isp, src->req_dataseg[i].ds_count, &dst->req_dataseg[i].ds_count);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_cont64_req(ispsoftc_t *isp, ispcontreq64_t *src, ispcontreq64_t *dst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
int i;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_hdr(isp, &src->req_header, &dst->req_header);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
for (i = 0; i < ISP_CDSEG64; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->req_dataseg[i].ds_base, &dst->req_dataseg[i].ds_base);
ISP_IOXPUT_32(isp, src->req_dataseg[i].ds_basehi, &dst->req_dataseg[i].ds_basehi);
ISP_IOXPUT_32(isp, src->req_dataseg[i].ds_count, &dst->req_dataseg[i].ds_count);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_response(ispsoftc_t *isp, ispstatusreq_t *src, ispstatusreq_t *dst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
int i;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_hdr(isp, &src->req_header, &dst->req_header);
ISP_IOXGET_32(isp, &src->req_handle, dst->req_handle);
ISP_IOXGET_16(isp, &src->req_scsi_status, dst->req_scsi_status);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_16(isp, &src->req_completion_status, dst->req_completion_status);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_16(isp, &src->req_state_flags, dst->req_state_flags);
ISP_IOXGET_16(isp, &src->req_status_flags, dst->req_status_flags);
ISP_IOXGET_16(isp, &src->req_time, dst->req_time);
ISP_IOXGET_16(isp, &src->req_sense_len, dst->req_sense_len);
ISP_IOXGET_32(isp, &src->req_resid, dst->req_resid);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
for (i = 0; i < sizeof (src->req_response); i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->req_response[i], dst->req_response[i]);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
for (i = 0; i < sizeof (src->req_sense_data); i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->req_sense_data[i], dst->req_sense_data[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
void
isp_get_cont_response(ispsoftc_t *isp, ispstatus_cont_t *src, ispstatus_cont_t *dst)
{
int i;
isp_get_hdr(isp, &src->req_header, &dst->req_header);
if (IS_24XX(isp)) {
uint32_t *a, *b;
a = (uint32_t *) src->req_sense_data;
b = (uint32_t *) dst->req_sense_data;
for (i = 0; i < (sizeof (src->req_sense_data) / sizeof (uint32_t)); i++) {
ISP_IOZGET_32(isp, a++, *b++);
}
} else {
for (i = 0; i < sizeof (src->req_sense_data); i++) {
ISP_IOXGET_8(isp, &src->req_sense_data[i], dst->req_sense_data[i]);
}
}
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_get_24xx_response(ispsoftc_t *isp, isp24xx_statusreq_t *src, isp24xx_statusreq_t *dst)
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
{
int i;
Recover from some major omissions/problems with the 24XX port. First, we were never correctly checking for a 24XX Status Type 0 response- that cased us to fall through to evaluate status for commands as if this were a 2100/2200/2300 Status Type 0 response. This is *close*, but not quite the same. This has been reported to be apparent with some wierd lun configuration problems with some arrays. It became glaringly apparent on sparc64 where none of the correct byte swap things were done. Fixing this omission then caused a whole universe shifting debug cycle of endian issues for the 2400. The manual for 24XX f/w turns out to be wrong about the endianness of a couple of entities. The lun and cdb fields for the type 7 request are *not* unconditionally big endian- they happen to be opposite of whatever the endian of the current machine type is. Same with the sense data for the 24XX type 0 response. While we're at it investigate and resolve some NVRAM endian issues. Approved by: re (ken) MFC after: 3 days
2007-07-02 20:08:20 +00:00
uint32_t *s, *d;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_hdr(isp, &src->req_header, &dst->req_header);
ISP_IOXGET_32(isp, &src->req_handle, dst->req_handle);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_16(isp, &src->req_completion_status, dst->req_completion_status);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_16(isp, &src->req_oxid, dst->req_oxid);
ISP_IOXGET_32(isp, &src->req_resid, dst->req_resid);
ISP_IOXGET_16(isp, &src->req_reserved0, dst->req_reserved0);
ISP_IOXGET_16(isp, &src->req_state_flags, dst->req_state_flags);
Minor additions to Status Type 0 IOCB.
2015-10-23 21:30:18 +00:00
ISP_IOXGET_16(isp, &src->req_retry_delay, dst->req_retry_delay);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_16(isp, &src->req_scsi_status, dst->req_scsi_status);
ISP_IOXGET_32(isp, &src->req_fcp_residual, dst->req_fcp_residual);
ISP_IOXGET_32(isp, &src->req_sense_len, dst->req_sense_len);
ISP_IOXGET_32(isp, &src->req_response_len, dst->req_response_len);
Recover from some major omissions/problems with the 24XX port. First, we were never correctly checking for a 24XX Status Type 0 response- that cased us to fall through to evaluate status for commands as if this were a 2100/2200/2300 Status Type 0 response. This is *close*, but not quite the same. This has been reported to be apparent with some wierd lun configuration problems with some arrays. It became glaringly apparent on sparc64 where none of the correct byte swap things were done. Fixing this omission then caused a whole universe shifting debug cycle of endian issues for the 2400. The manual for 24XX f/w turns out to be wrong about the endianness of a couple of entities. The lun and cdb fields for the type 7 request are *not* unconditionally big endian- they happen to be opposite of whatever the endian of the current machine type is. Same with the sense data for the 24XX type 0 response. While we're at it investigate and resolve some NVRAM endian issues. Approved by: re (ken) MFC after: 3 days
2007-07-02 20:08:20 +00:00
s = (uint32_t *)src->req_rsp_sense;
d = (uint32_t *)dst->req_rsp_sense;
for (i = 0; i < (ASIZE(src->req_rsp_sense) >> 2); i++) {
d[i] = ISP_SWAP32(isp, s[i]);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_24xx_abrt(ispsoftc_t *isp, isp24xx_abrt_t *src, isp24xx_abrt_t *dst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
int i;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_hdr(isp, &src->abrt_header, &dst->abrt_header);
ISP_IOXGET_32(isp, &src->abrt_handle, dst->abrt_handle);
ISP_IOXGET_16(isp, &src->abrt_nphdl, dst->abrt_nphdl);
ISP_IOXGET_16(isp, &src->abrt_options, dst->abrt_options);
ISP_IOXGET_32(isp, &src->abrt_cmd_handle, dst->abrt_cmd_handle);
Add new field to Abort IOCB.
2015-10-24 19:38:06 +00:00
ISP_IOXGET_16(isp, &src->abrt_queue_number, dst->abrt_queue_number);
Recover from some major omissions/problems with the 24XX port. First, we were never correctly checking for a 24XX Status Type 0 response- that cased us to fall through to evaluate status for commands as if this were a 2100/2200/2300 Status Type 0 response. This is *close*, but not quite the same. This has been reported to be apparent with some wierd lun configuration problems with some arrays. It became glaringly apparent on sparc64 where none of the correct byte swap things were done. Fixing this omission then caused a whole universe shifting debug cycle of endian issues for the 2400. The manual for 24XX f/w turns out to be wrong about the endianness of a couple of entities. The lun and cdb fields for the type 7 request are *not* unconditionally big endian- they happen to be opposite of whatever the endian of the current machine type is. Same with the sense data for the 24XX type 0 response. While we're at it investigate and resolve some NVRAM endian issues. Approved by: re (ken) MFC after: 3 days
2007-07-02 20:08:20 +00:00
for (i = 0; i < ASIZE(src->abrt_reserved); i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->abrt_reserved[i], dst->abrt_reserved[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
ISP_IOXGET_16(isp, &src->abrt_tidlo, dst->abrt_tidlo);
ISP_IOXGET_8(isp, &src->abrt_tidhi, dst->abrt_tidhi);
ISP_IOXGET_8(isp, &src->abrt_vpidx, dst->abrt_vpidx);
Recover from some major omissions/problems with the 24XX port. First, we were never correctly checking for a 24XX Status Type 0 response- that cased us to fall through to evaluate status for commands as if this were a 2100/2200/2300 Status Type 0 response. This is *close*, but not quite the same. This has been reported to be apparent with some wierd lun configuration problems with some arrays. It became glaringly apparent on sparc64 where none of the correct byte swap things were done. Fixing this omission then caused a whole universe shifting debug cycle of endian issues for the 2400. The manual for 24XX f/w turns out to be wrong about the endianness of a couple of entities. The lun and cdb fields for the type 7 request are *not* unconditionally big endian- they happen to be opposite of whatever the endian of the current machine type is. Same with the sense data for the 24XX type 0 response. While we're at it investigate and resolve some NVRAM endian issues. Approved by: re (ken) MFC after: 3 days
2007-07-02 20:08:20 +00:00
for (i = 0; i < ASIZE(src->abrt_reserved1); i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->abrt_reserved1[i], dst->abrt_reserved1[i]);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
void
isp_get_rio1(ispsoftc_t *isp, isp_rio1_t *r1src, isp_rio1_t *r1dst)
{
const int lim = sizeof (r1dst->req_handles) / sizeof (r1dst->req_handles[0]);
int i;
isp_get_hdr(isp, &r1src->req_header, &r1dst->req_header);
if (r1dst->req_header.rqs_seqno > lim) {
r1dst->req_header.rqs_seqno = lim;
}
for (i = 0; i < r1dst->req_header.rqs_seqno; i++) {
ISP_IOXGET_32(isp, &r1src->req_handles[i], r1dst->req_handles[i]);
}
while (i < lim) {
r1dst->req_handles[i++] = 0;
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
isp_get_rio2(ispsoftc_t *isp, isp_rio2_t *r2src, isp_rio2_t *r2dst)
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
{
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
const int lim = sizeof (r2dst->req_handles) / sizeof (r2dst->req_handles[0]);
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
int i;
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_hdr(isp, &r2src->req_header, &r2dst->req_header);
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
if (r2dst->req_header.rqs_seqno > lim) {
r2dst->req_header.rqs_seqno = lim;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
for (i = 0; i < r2dst->req_header.rqs_seqno; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_16(isp, &r2src->req_handles[i], r2dst->req_handles[i]);
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
}
Revamp the pieces of some of the stuff I forgot to do when shifting to 32 bit handles. The RIO (reduced interrupt operation) and fast posting for the parallel SCSI cards were all 16 bit handles. Furthermore, target mode parallel SCSI only can have 16 bit handles. Use part of a supplied patch to switch over to using 32 bit handles. Be a bit more conservative here and only do this for parallel SCSI for the 12160 (Ultra3) cards. There were a lot of marginal Ultra2 cards, and, frankly, few are findable now for testing. Fix the target handle routine to only do 16 bit handles for parallel SCSI cards. This is okay because the upper sixteen bits of the new 32 bit handles is a sequence number to help protect against duplicate completions. This would be very unlikely to happen with parallel SCSI target mode, and wasn't present before, so we're no worse off than we used to be. While we're at it, finally split the async mailbox completion handlers into FC and parallel SCSI functions. This makes it much cleaner and easier to figure out what is or isn't a legal async mailbox completion code for different card classes. PR: kern/144250 Submitted partially by: Charles D MFC after: 1 week
2010-02-27 05:41:23 +00:00
while (i < lim) {
Implement REDUCED INTERRUPT OPERATION usage form FC cards- this allows the firmware to delay completion of commands so that it can attempt to batch a bunch of completions at once- either returning 16 bit handles in mailbox registers, or in a resposne queue entry that has a whole wad of 16 bit handles. Distinguish between 2300 and 2312 chipsets- if only because the revisions on the chips have different meanings. Add more instrumentation plus ISP_GET_STATS and ISP_CLR_STATS ioctls. Run up the maximum number of response queue entities we'll look at per interrupt. If we haven't set HBA role yet, always return success from isp_fc_runstate. MFC after: 2 weeks
2002-01-03 20:43:22 +00:00
r2dst->req_handles[i++] = 0;
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_icb(ispsoftc_t *isp, isp_icb_t *src, isp_icb_t *dst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
int i;
Fix some incorrectly swapped fields in an ICB. Access a PCI register with correct width. Obtained from: Dmitry Valeryevich Trikoz
2005-05-11 03:00:50 +00:00
if (ISP_IS_SBUS(isp)) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_8(isp, src->icb_version, &dst->icb_reserved0);
ISP_IOXPUT_8(isp, src->icb_reserved0, &dst->icb_version);
Fix some incorrectly swapped fields in an ICB. Access a PCI register with correct width. Obtained from: Dmitry Valeryevich Trikoz
2005-05-11 03:00:50 +00:00
} else {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_8(isp, src->icb_version, &dst->icb_version);
ISP_IOXPUT_8(isp, src->icb_reserved0, &dst->icb_reserved0);
Fix some incorrectly swapped fields in an ICB. Access a PCI register with correct width. Obtained from: Dmitry Valeryevich Trikoz
2005-05-11 03:00:50 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->icb_fwoptions, &dst->icb_fwoptions);
ISP_IOXPUT_16(isp, src->icb_maxfrmlen, &dst->icb_maxfrmlen);
ISP_IOXPUT_16(isp, src->icb_maxalloc, &dst->icb_maxalloc);
ISP_IOXPUT_16(isp, src->icb_execthrottle, &dst->icb_execthrottle);
Fix some incorrectly swapped fields in an ICB. Access a PCI register with correct width. Obtained from: Dmitry Valeryevich Trikoz
2005-05-11 03:00:50 +00:00
if (ISP_IS_SBUS(isp)) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_8(isp, src->icb_retry_count, &dst->icb_retry_delay);
ISP_IOXPUT_8(isp, src->icb_retry_delay, &dst->icb_retry_count);
Fix some incorrectly swapped fields in an ICB. Access a PCI register with correct width. Obtained from: Dmitry Valeryevich Trikoz
2005-05-11 03:00:50 +00:00
} else {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_8(isp, src->icb_retry_count, &dst->icb_retry_count);
ISP_IOXPUT_8(isp, src->icb_retry_delay, &dst->icb_retry_delay);
Fix some incorrectly swapped fields in an ICB. Access a PCI register with correct width. Obtained from: Dmitry Valeryevich Trikoz
2005-05-11 03:00:50 +00:00
}
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
for (i = 0; i < 8; i++) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_8(isp, src->icb_portname[i], &dst->icb_portname[i]);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->icb_hardaddr, &dst->icb_hardaddr);
Fix some incorrectly swapped fields in an ICB. Access a PCI register with correct width. Obtained from: Dmitry Valeryevich Trikoz
2005-05-11 03:00:50 +00:00
if (ISP_IS_SBUS(isp)) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_8(isp, src->icb_iqdevtype, &dst->icb_logintime);
ISP_IOXPUT_8(isp, src->icb_logintime, &dst->icb_iqdevtype);
Fix some incorrectly swapped fields in an ICB. Access a PCI register with correct width. Obtained from: Dmitry Valeryevich Trikoz
2005-05-11 03:00:50 +00:00
} else {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_8(isp, src->icb_iqdevtype, &dst->icb_iqdevtype);
ISP_IOXPUT_8(isp, src->icb_logintime, &dst->icb_logintime);
Fix some incorrectly swapped fields in an ICB. Access a PCI register with correct width. Obtained from: Dmitry Valeryevich Trikoz
2005-05-11 03:00:50 +00:00
}
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
for (i = 0; i < 8; i++) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_8(isp, src->icb_nodename[i], &dst->icb_nodename[i]);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->icb_rqstout, &dst->icb_rqstout);
ISP_IOXPUT_16(isp, src->icb_rspnsin, &dst->icb_rspnsin);
ISP_IOXPUT_16(isp, src->icb_rqstqlen, &dst->icb_rqstqlen);
ISP_IOXPUT_16(isp, src->icb_rsltqlen, &dst->icb_rsltqlen);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
for (i = 0; i < 4; i++) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->icb_rqstaddr[i], &dst->icb_rqstaddr[i]);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
for (i = 0; i < 4; i++) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->icb_respaddr[i], &dst->icb_respaddr[i]);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->icb_lunenables, &dst->icb_lunenables);
Fix some incorrectly swapped fields in an ICB. Access a PCI register with correct width. Obtained from: Dmitry Valeryevich Trikoz
2005-05-11 03:00:50 +00:00
if (ISP_IS_SBUS(isp)) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_8(isp, src->icb_ccnt, &dst->icb_icnt);
ISP_IOXPUT_8(isp, src->icb_icnt, &dst->icb_ccnt);
Fix some incorrectly swapped fields in an ICB. Access a PCI register with correct width. Obtained from: Dmitry Valeryevich Trikoz
2005-05-11 03:00:50 +00:00
} else {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_8(isp, src->icb_ccnt, &dst->icb_ccnt);
ISP_IOXPUT_8(isp, src->icb_icnt, &dst->icb_icnt);
Fix some incorrectly swapped fields in an ICB. Access a PCI register with correct width. Obtained from: Dmitry Valeryevich Trikoz
2005-05-11 03:00:50 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->icb_lunetimeout, &dst->icb_lunetimeout);
ISP_IOXPUT_16(isp, src->icb_reserved1, &dst->icb_reserved1);
ISP_IOXPUT_16(isp, src->icb_xfwoptions, &dst->icb_xfwoptions);
Fix some incorrectly swapped fields in an ICB. Access a PCI register with correct width. Obtained from: Dmitry Valeryevich Trikoz
2005-05-11 03:00:50 +00:00
if (ISP_IS_SBUS(isp)) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_8(isp, src->icb_racctimer, &dst->icb_idelaytimer);
ISP_IOXPUT_8(isp, src->icb_idelaytimer, &dst->icb_racctimer);
Fix some incorrectly swapped fields in an ICB. Access a PCI register with correct width. Obtained from: Dmitry Valeryevich Trikoz
2005-05-11 03:00:50 +00:00
} else {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_8(isp, src->icb_racctimer, &dst->icb_racctimer);
ISP_IOXPUT_8(isp, src->icb_idelaytimer, &dst->icb_idelaytimer);
}
ISP_IOXPUT_16(isp, src->icb_zfwoptions, &dst->icb_zfwoptions);
}
void
isp_put_icb_2400(ispsoftc_t *isp, isp_icb_2400_t *src, isp_icb_2400_t *dst)
{
int i;
ISP_IOXPUT_16(isp, src->icb_version, &dst->icb_version);
ISP_IOXPUT_16(isp, src->icb_reserved0, &dst->icb_reserved0);
ISP_IOXPUT_16(isp, src->icb_maxfrmlen, &dst->icb_maxfrmlen);
ISP_IOXPUT_16(isp, src->icb_execthrottle, &dst->icb_execthrottle);
ISP_IOXPUT_16(isp, src->icb_xchgcnt, &dst->icb_xchgcnt);
ISP_IOXPUT_16(isp, src->icb_hardaddr, &dst->icb_hardaddr);
for (i = 0; i < 8; i++) {
ISP_IOXPUT_8(isp, src->icb_portname[i], &dst->icb_portname[i]);
}
for (i = 0; i < 8; i++) {
ISP_IOXPUT_8(isp, src->icb_nodename[i], &dst->icb_nodename[i]);
}
ISP_IOXPUT_16(isp, src->icb_rspnsin, &dst->icb_rspnsin);
ISP_IOXPUT_16(isp, src->icb_rqstout, &dst->icb_rqstout);
ISP_IOXPUT_16(isp, src->icb_retry_count, &dst->icb_retry_count);
ISP_IOXPUT_16(isp, src->icb_priout, &dst->icb_priout);
ISP_IOXPUT_16(isp, src->icb_rsltqlen, &dst->icb_rsltqlen);
ISP_IOXPUT_16(isp, src->icb_rqstqlen, &dst->icb_rqstqlen);
ISP_IOXPUT_16(isp, src->icb_ldn_nols, &dst->icb_ldn_nols);
ISP_IOXPUT_16(isp, src->icb_prqstqlen, &dst->icb_prqstqlen);
for (i = 0; i < 4; i++) {
ISP_IOXPUT_16(isp, src->icb_rqstaddr[i], &dst->icb_rqstaddr[i]);
}
for (i = 0; i < 4; i++) {
ISP_IOXPUT_16(isp, src->icb_respaddr[i], &dst->icb_respaddr[i]);
}
for (i = 0; i < 4; i++) {
ISP_IOXPUT_16(isp, src->icb_priaddr[i], &dst->icb_priaddr[i]);
}
Update isp_put_icb_2400() for new structure fields.
2015-12-04 01:28:48 +00:00
ISP_IOXPUT_16(isp, src->icb_msixresp, &dst->icb_msixresp);
ISP_IOXPUT_16(isp, src->icb_msixatio, &dst->icb_msixatio);
for (i = 0; i < 2; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_16(isp, src->icb_reserved1[i], &dst->icb_reserved1[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
ISP_IOXPUT_16(isp, src->icb_atio_in, &dst->icb_atio_in);
ISP_IOXPUT_16(isp, src->icb_atioqlen, &dst->icb_atioqlen);
for (i = 0; i < 4; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_16(isp, src->icb_atioqaddr[i], &dst->icb_atioqaddr[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
ISP_IOXPUT_16(isp, src->icb_idelaytimer, &dst->icb_idelaytimer);
ISP_IOXPUT_16(isp, src->icb_logintime, &dst->icb_logintime);
ISP_IOXPUT_32(isp, src->icb_fwoptions1, &dst->icb_fwoptions1);
ISP_IOXPUT_32(isp, src->icb_fwoptions2, &dst->icb_fwoptions2);
ISP_IOXPUT_32(isp, src->icb_fwoptions3, &dst->icb_fwoptions3);
Update isp_put_icb_2400() for new structure fields.
2015-12-04 01:28:48 +00:00
ISP_IOXPUT_16(isp, src->icb_qos, &dst->icb_qos);
for (i = 0; i < 3; i++)
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_16(isp, src->icb_reserved2[i], &dst->icb_reserved2[i]);
Update isp_put_icb_2400() for new structure fields.
2015-12-04 01:28:48 +00:00
for (i = 0; i < 3; i++)
ISP_IOXPUT_16(isp, src->icb_enodemac[i], &dst->icb_enodemac[i]);
ISP_IOXPUT_16(isp, src->icb_disctime, &dst->icb_disctime);
for (i = 0; i < 4; i++)
ISP_IOXPUT_16(isp, src->icb_reserved3[i], &dst->icb_reserved3[i]);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
void
isp_put_icb_2400_vpinfo(ispsoftc_t *isp, isp_icb_2400_vpinfo_t *src, isp_icb_2400_vpinfo_t *dst)
{
ISP_IOXPUT_16(isp, src->vp_count, &dst->vp_count);
ISP_IOXPUT_16(isp, src->vp_global_options, &dst->vp_global_options);
}
void
isp_put_vp_port_info(ispsoftc_t *isp, vp_port_info_t *src, vp_port_info_t *dst)
{
int i;
ISP_IOXPUT_16(isp, src->vp_port_status, &dst->vp_port_status);
ISP_IOXPUT_8(isp, src->vp_port_options, &dst->vp_port_options);
ISP_IOXPUT_8(isp, src->vp_port_loopid, &dst->vp_port_loopid);
for (i = 0; i < 8; i++) {
ISP_IOXPUT_8(isp, src->vp_port_portname[i], &dst->vp_port_portname[i]);
}
for (i = 0; i < 8; i++) {
ISP_IOXPUT_8(isp, src->vp_port_nodename[i], &dst->vp_port_nodename[i]);
}
/* we never *put* portid_lo/portid_hi */
}
void
isp_get_vp_port_info(ispsoftc_t *isp, vp_port_info_t *src, vp_port_info_t *dst)
{
int i;
ISP_IOXGET_16(isp, &src->vp_port_status, dst->vp_port_status);
ISP_IOXGET_8(isp, &src->vp_port_options, dst->vp_port_options);
ISP_IOXGET_8(isp, &src->vp_port_loopid, dst->vp_port_loopid);
for (i = 0; i < ASIZE(src->vp_port_portname); i++) {
ISP_IOXGET_8(isp, &src->vp_port_portname[i], dst->vp_port_portname[i]);
}
for (i = 0; i < ASIZE(src->vp_port_nodename); i++) {
ISP_IOXGET_8(isp, &src->vp_port_nodename[i], dst->vp_port_nodename[i]);
}
ISP_IOXGET_16(isp, &src->vp_port_portid_lo, dst->vp_port_portid_lo);
ISP_IOXGET_16(isp, &src->vp_port_portid_hi, dst->vp_port_portid_hi);
}
void
isp_put_vp_ctrl_info(ispsoftc_t *isp, vp_ctrl_info_t *src, vp_ctrl_info_t *dst)
{
int i;
isp_put_hdr(isp, &src->vp_ctrl_hdr, &dst->vp_ctrl_hdr);
ISP_IOXPUT_32(isp, src->vp_ctrl_handle, &dst->vp_ctrl_handle);
ISP_IOXPUT_16(isp, src->vp_ctrl_index_fail, &dst->vp_ctrl_index_fail);
ISP_IOXPUT_16(isp, src->vp_ctrl_status, &dst->vp_ctrl_status);
ISP_IOXPUT_16(isp, src->vp_ctrl_command, &dst->vp_ctrl_command);
ISP_IOXPUT_16(isp, src->vp_ctrl_vp_count, &dst->vp_ctrl_vp_count);
for (i = 0; i < ASIZE(src->vp_ctrl_idmap); i++) {
ISP_IOXPUT_16(isp, src->vp_ctrl_idmap[i], &dst->vp_ctrl_idmap[i]);
}
for (i = 0; i < ASIZE(src->vp_ctrl_reserved); i++) {
Reimplement enable and implement disable of virtual ports. Now on 24xx and above chips it is really possible to simulate several virtual FC ports with single physical one. For example, it allows to configure several targets in ctl.conf, assign each of them to separate virtual port, and let user to control access to them with switch zoning. I still doubt that all problems are solved there, but at now it passes at least basic tests.
2015-10-26 18:14:15 +00:00
ISP_IOXPUT_16(isp, src->vp_ctrl_reserved[i], &dst->vp_ctrl_reserved[i]);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Reimplement enable and implement disable of virtual ports. Now on 24xx and above chips it is really possible to simulate several virtual FC ports with single physical one. For example, it allows to configure several targets in ctl.conf, assign each of them to separate virtual port, and let user to control access to them with switch zoning. I still doubt that all problems are solved there, but at now it passes at least basic tests.
2015-10-26 18:14:15 +00:00
ISP_IOXPUT_16(isp, src->vp_ctrl_fcf_index, &dst->vp_ctrl_fcf_index);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
void
isp_get_vp_ctrl_info(ispsoftc_t *isp, vp_ctrl_info_t *src, vp_ctrl_info_t *dst)
{
int i;
isp_get_hdr(isp, &src->vp_ctrl_hdr, &dst->vp_ctrl_hdr);
ISP_IOXGET_32(isp, &src->vp_ctrl_handle, dst->vp_ctrl_handle);
ISP_IOXGET_16(isp, &src->vp_ctrl_index_fail, dst->vp_ctrl_index_fail);
ISP_IOXGET_16(isp, &src->vp_ctrl_status, dst->vp_ctrl_status);
ISP_IOXGET_16(isp, &src->vp_ctrl_command, dst->vp_ctrl_command);
ISP_IOXGET_16(isp, &src->vp_ctrl_vp_count, dst->vp_ctrl_vp_count);
for (i = 0; i < ASIZE(src->vp_ctrl_idmap); i++) {
ISP_IOXGET_16(isp, &src->vp_ctrl_idmap[i], dst->vp_ctrl_idmap[i]);
}
for (i = 0; i < ASIZE(src->vp_ctrl_reserved); i++) {
Reimplement enable and implement disable of virtual ports. Now on 24xx and above chips it is really possible to simulate several virtual FC ports with single physical one. For example, it allows to configure several targets in ctl.conf, assign each of them to separate virtual port, and let user to control access to them with switch zoning. I still doubt that all problems are solved there, but at now it passes at least basic tests.
2015-10-26 18:14:15 +00:00
ISP_IOXGET_16(isp, &src->vp_ctrl_reserved[i], dst->vp_ctrl_reserved[i]);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Reimplement enable and implement disable of virtual ports. Now on 24xx and above chips it is really possible to simulate several virtual FC ports with single physical one. For example, it allows to configure several targets in ctl.conf, assign each of them to separate virtual port, and let user to control access to them with switch zoning. I still doubt that all problems are solved there, but at now it passes at least basic tests.
2015-10-26 18:14:15 +00:00
ISP_IOXGET_16(isp, &src->vp_ctrl_fcf_index, dst->vp_ctrl_fcf_index);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
void
isp_put_vp_modify(ispsoftc_t *isp, vp_modify_t *src, vp_modify_t *dst)
{
int i, j;
isp_put_hdr(isp, &src->vp_mod_hdr, &dst->vp_mod_hdr);
ISP_IOXPUT_32(isp, src->vp_mod_hdl, &dst->vp_mod_hdl);
ISP_IOXPUT_16(isp, src->vp_mod_reserved0, &dst->vp_mod_reserved0);
ISP_IOXPUT_16(isp, src->vp_mod_status, &dst->vp_mod_status);
ISP_IOXPUT_8(isp, src->vp_mod_cmd, &dst->vp_mod_cmd);
ISP_IOXPUT_8(isp, src->vp_mod_cnt, &dst->vp_mod_cnt);
ISP_IOXPUT_8(isp, src->vp_mod_idx0, &dst->vp_mod_idx0);
ISP_IOXPUT_8(isp, src->vp_mod_idx1, &dst->vp_mod_idx1);
for (i = 0; i < ASIZE(src->vp_mod_ports); i++) {
ISP_IOXPUT_8(isp, src->vp_mod_ports[i].options, &dst->vp_mod_ports[i].options);
ISP_IOXPUT_8(isp, src->vp_mod_ports[i].loopid, &dst->vp_mod_ports[i].loopid);
ISP_IOXPUT_16(isp, src->vp_mod_ports[i].reserved1, &dst->vp_mod_ports[i].reserved1);
for (j = 0; j < ASIZE(src->vp_mod_ports[i].wwpn); j++) {
ISP_IOXPUT_8(isp, src->vp_mod_ports[i].wwpn[j], &dst->vp_mod_ports[i].wwpn[j]);
}
for (j = 0; j < ASIZE(src->vp_mod_ports[i].wwnn); j++) {
ISP_IOXPUT_8(isp, src->vp_mod_ports[i].wwnn[j], &dst->vp_mod_ports[i].wwnn[j]);
}
}
for (i = 0; i < ASIZE(src->vp_mod_reserved2); i++) {
ISP_IOXPUT_8(isp, src->vp_mod_reserved2[i], &dst->vp_mod_reserved2[i]);
}
}
void
isp_get_vp_modify(ispsoftc_t *isp, vp_modify_t *src, vp_modify_t *dst)
{
int i, j;
isp_get_hdr(isp, &src->vp_mod_hdr, &dst->vp_mod_hdr);
ISP_IOXGET_32(isp, &src->vp_mod_hdl, dst->vp_mod_hdl);
ISP_IOXGET_16(isp, &src->vp_mod_reserved0, dst->vp_mod_reserved0);
ISP_IOXGET_16(isp, &src->vp_mod_status, dst->vp_mod_status);
ISP_IOXGET_8(isp, &src->vp_mod_cmd, dst->vp_mod_cmd);
ISP_IOXGET_8(isp, &src->vp_mod_cnt, dst->vp_mod_cnt);
ISP_IOXGET_8(isp, &src->vp_mod_idx0, dst->vp_mod_idx0);
ISP_IOXGET_8(isp, &src->vp_mod_idx1, dst->vp_mod_idx1);
for (i = 0; i < ASIZE(src->vp_mod_ports); i++) {
ISP_IOXGET_8(isp, &src->vp_mod_ports[i].options, dst->vp_mod_ports[i].options);
ISP_IOXGET_8(isp, &src->vp_mod_ports[i].loopid, dst->vp_mod_ports[i].loopid);
ISP_IOXGET_16(isp, &src->vp_mod_ports[i].reserved1, dst->vp_mod_ports[i].reserved1);
for (j = 0; j < ASIZE(src->vp_mod_ports[i].wwpn); j++) {
ISP_IOXGET_8(isp, &src->vp_mod_ports[i].wwpn[j], dst->vp_mod_ports[i].wwpn[j]);
}
for (j = 0; j < ASIZE(src->vp_mod_ports[i].wwnn); j++) {
ISP_IOXGET_8(isp, &src->vp_mod_ports[i].wwnn[j], dst->vp_mod_ports[i].wwnn[j]);
}
}
for (i = 0; i < ASIZE(src->vp_mod_reserved2); i++) {
ISP_IOXGET_8(isp, &src->vp_mod_reserved2[i], dst->vp_mod_reserved2[i]);
Fix some incorrectly swapped fields in an ICB. Access a PCI register with correct width. Obtained from: Dmitry Valeryevich Trikoz
2005-05-11 03:00:50 +00:00
}
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_pdb_21xx(ispsoftc_t *isp, isp_pdb_21xx_t *src, isp_pdb_21xx_t *dst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
int i;
ISP_IOXGET_16(isp, &src->pdb_options, dst->pdb_options);
ISP_IOXGET_8(isp, &src->pdb_mstate, dst->pdb_mstate);
ISP_IOXGET_8(isp, &src->pdb_sstate, dst->pdb_sstate);
for (i = 0; i < 4; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->pdb_hardaddr_bits[i], dst->pdb_hardaddr_bits[i]);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
for (i = 0; i < 4; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->pdb_portid_bits[i], dst->pdb_portid_bits[i]);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
for (i = 0; i < 8; i++) {
ISP_IOXGET_8(isp, &src->pdb_nodename[i], dst->pdb_nodename[i]);
}
for (i = 0; i < 8; i++) {
ISP_IOXGET_8(isp, &src->pdb_portname[i], dst->pdb_portname[i]);
}
ISP_IOXGET_16(isp, &src->pdb_execthrottle, dst->pdb_execthrottle);
ISP_IOXGET_16(isp, &src->pdb_exec_count, dst->pdb_exec_count);
ISP_IOXGET_8(isp, &src->pdb_retry_count, dst->pdb_retry_count);
ISP_IOXGET_8(isp, &src->pdb_retry_delay, dst->pdb_retry_delay);
ISP_IOXGET_16(isp, &src->pdb_resalloc, dst->pdb_resalloc);
ISP_IOXGET_16(isp, &src->pdb_curalloc, dst->pdb_curalloc);
ISP_IOXGET_16(isp, &src->pdb_qhead, dst->pdb_qhead);
ISP_IOXGET_16(isp, &src->pdb_qtail, dst->pdb_qtail);
ISP_IOXGET_16(isp, &src->pdb_tl_next, dst->pdb_tl_next);
ISP_IOXGET_16(isp, &src->pdb_tl_last, dst->pdb_tl_last);
ISP_IOXGET_16(isp, &src->pdb_features, dst->pdb_features);
ISP_IOXGET_16(isp, &src->pdb_pconcurrnt, dst->pdb_pconcurrnt);
ISP_IOXGET_16(isp, &src->pdb_roi, dst->pdb_roi);
ISP_IOXGET_8(isp, &src->pdb_target, dst->pdb_target);
ISP_IOXGET_8(isp, &src->pdb_initiator, dst->pdb_initiator);
ISP_IOXGET_16(isp, &src->pdb_rdsiz, dst->pdb_rdsiz);
ISP_IOXGET_16(isp, &src->pdb_ncseq, dst->pdb_ncseq);
ISP_IOXGET_16(isp, &src->pdb_noseq, dst->pdb_noseq);
ISP_IOXGET_16(isp, &src->pdb_labrtflg, dst->pdb_labrtflg);
ISP_IOXGET_16(isp, &src->pdb_lstopflg, dst->pdb_lstopflg);
ISP_IOXGET_16(isp, &src->pdb_sqhead, dst->pdb_sqhead);
ISP_IOXGET_16(isp, &src->pdb_sqtail, dst->pdb_sqtail);
ISP_IOXGET_16(isp, &src->pdb_ptimer, dst->pdb_ptimer);
ISP_IOXGET_16(isp, &src->pdb_nxt_seqid, dst->pdb_nxt_seqid);
ISP_IOXGET_16(isp, &src->pdb_fcount, dst->pdb_fcount);
ISP_IOXGET_16(isp, &src->pdb_prli_len, dst->pdb_prli_len);
ISP_IOXGET_16(isp, &src->pdb_prli_svc0, dst->pdb_prli_svc0);
ISP_IOXGET_16(isp, &src->pdb_prli_svc3, dst->pdb_prli_svc3);
ISP_IOXGET_16(isp, &src->pdb_loopid, dst->pdb_loopid);
ISP_IOXGET_16(isp, &src->pdb_il_ptr, dst->pdb_il_ptr);
ISP_IOXGET_16(isp, &src->pdb_sl_ptr, dst->pdb_sl_ptr);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
void
isp_get_pdb_24xx(ispsoftc_t *isp, isp_pdb_24xx_t *src, isp_pdb_24xx_t *dst)
{
int i;
ISP_IOXGET_16(isp, &src->pdb_flags, dst->pdb_flags);
ISP_IOXGET_8(isp, &src->pdb_curstate, dst->pdb_curstate);
ISP_IOXGET_8(isp, &src->pdb_laststate, dst->pdb_laststate);
for (i = 0; i < 4; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->pdb_hardaddr_bits[i], dst->pdb_hardaddr_bits[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
for (i = 0; i < 4; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->pdb_portid_bits[i], dst->pdb_portid_bits[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
ISP_IOXGET_16(isp, &src->pdb_retry_timer, dst->pdb_retry_timer);
ISP_IOXGET_16(isp, &src->pdb_handle, dst->pdb_handle);
ISP_IOXGET_16(isp, &src->pdb_rcv_dsize, dst->pdb_rcv_dsize);
ISP_IOXGET_16(isp, &src->pdb_reserved0, dst->pdb_reserved0);
ISP_IOXGET_16(isp, &src->pdb_prli_svc0, dst->pdb_prli_svc0);
ISP_IOXGET_16(isp, &src->pdb_prli_svc3, dst->pdb_prli_svc3);
for (i = 0; i < 8; i++) {
ISP_IOXGET_8(isp, &src->pdb_nodename[i], dst->pdb_nodename[i]);
}
for (i = 0; i < 8; i++) {
ISP_IOXGET_8(isp, &src->pdb_portname[i], dst->pdb_portname[i]);
}
for (i = 0; i < 24; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->pdb_reserved1[i], dst->pdb_reserved1[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
}
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
Improve/fix loop scanning routine. For the most of chips (except anscient ones) port handlers have no relation to port IDs. In such situation old code scanning first 125 handlers was quite naive. Instead of doing that, send to chip single request to get full list of port handlers available on specific virtual port and scan only them. Old code had problems with case of several virtual ports enabled, when port handlers allocated from global address space could easily go above 125. This change was successfully tested on 23xx, 24xx and 25xx chips in loop mode with 4 virtual initiator ports, each seing 50 virtual target ports.
2015-10-28 12:34:59 +00:00
void
isp_get_pnhle_21xx(ispsoftc_t *isp, isp_pnhle_21xx_t *src, isp_pnhle_21xx_t *dst)
{
ISP_IOXGET_16(isp, &src->pnhle_port_id_lo, dst->pnhle_port_id_lo);
ISP_IOXGET_16(isp, &src->pnhle_port_id_hi_handle, dst->pnhle_port_id_hi_handle);
}
void
isp_get_pnhle_23xx(ispsoftc_t *isp, isp_pnhle_23xx_t *src, isp_pnhle_23xx_t *dst)
{
ISP_IOXGET_16(isp, &src->pnhle_port_id_lo, dst->pnhle_port_id_lo);
ISP_IOXGET_16(isp, &src->pnhle_port_id_hi, dst->pnhle_port_id_hi);
ISP_IOXGET_16(isp, &src->pnhle_handle, dst->pnhle_handle);
}
void
isp_get_pnhle_24xx(ispsoftc_t *isp, isp_pnhle_24xx_t *src, isp_pnhle_24xx_t *dst)
{
ISP_IOXGET_16(isp, &src->pnhle_port_id_lo, dst->pnhle_port_id_lo);
ISP_IOXGET_16(isp, &src->pnhle_port_id_hi, dst->pnhle_port_id_hi);
ISP_IOXGET_16(isp, &src->pnhle_handle, dst->pnhle_handle);
ISP_IOXGET_16(isp, &src->pnhle_reserved, dst->pnhle_reserved);
}
void
isp_get_pnnle(ispsoftc_t *isp, isp_pnnle_t *src, isp_pnnle_t *dst)
{
int i;
for (i = 0; i < 8; i++)
ISP_IOXGET_8(isp, &src->pnnle_name[i], dst->pnnle_name[i]);
ISP_IOXGET_16(isp, &src->pnnle_handle, dst->pnnle_handle);
ISP_IOXGET_16(isp, &src->pnnle_reserved, dst->pnnle_reserved);
}
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
/*
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
* PLOGI/LOGO IOCB canonicalization
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
*/
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_plogx(ispsoftc_t *isp, isp_plogx_t *src, isp_plogx_t *dst)
{
int i;
isp_get_hdr(isp, &src->plogx_header, &dst->plogx_header);
ISP_IOXGET_32(isp, &src->plogx_handle, dst->plogx_handle);
ISP_IOXGET_16(isp, &src->plogx_status, dst->plogx_status);
ISP_IOXGET_16(isp, &src->plogx_nphdl, dst->plogx_nphdl);
ISP_IOXGET_16(isp, &src->plogx_flags, dst->plogx_flags);
ISP_IOXGET_16(isp, &src->plogx_vphdl, dst->plogx_vphdl);
ISP_IOXGET_16(isp, &src->plogx_portlo, dst->plogx_portlo);
ISP_IOXGET_16(isp, &src->plogx_rspsz_porthi, dst->plogx_rspsz_porthi);
for (i = 0; i < 11; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_16(isp, &src->plogx_ioparm[i].lo16, dst->plogx_ioparm[i].lo16);
ISP_IOXGET_16(isp, &src->plogx_ioparm[i].hi16, dst->plogx_ioparm[i].hi16);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
}
void
isp_put_plogx(ispsoftc_t *isp, isp_plogx_t *src, isp_plogx_t *dst)
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
{
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
int i;
isp_put_hdr(isp, &src->plogx_header, &dst->plogx_header);
ISP_IOXPUT_32(isp, src->plogx_handle, &dst->plogx_handle);
ISP_IOXPUT_16(isp, src->plogx_status, &dst->plogx_status);
ISP_IOXPUT_16(isp, src->plogx_nphdl, &dst->plogx_nphdl);
ISP_IOXPUT_16(isp, src->plogx_flags, &dst->plogx_flags);
ISP_IOXPUT_16(isp, src->plogx_vphdl, &dst->plogx_vphdl);
ISP_IOXPUT_16(isp, src->plogx_portlo, &dst->plogx_portlo);
ISP_IOXPUT_16(isp, src->plogx_rspsz_porthi, &dst->plogx_rspsz_porthi);
for (i = 0; i < 11; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_16(isp, src->plogx_ioparm[i].lo16, &dst->plogx_ioparm[i].lo16);
ISP_IOXPUT_16(isp, src->plogx_ioparm[i].hi16, &dst->plogx_ioparm[i].hi16);
}
}
/*
* Report ID canonicalization
*/
void
isp_get_ridacq(ispsoftc_t *isp, isp_ridacq_t *src, isp_ridacq_t *dst)
{
int i;
isp_get_hdr(isp, &src->ridacq_hdr, &dst->ridacq_hdr);
ISP_IOXGET_32(isp, &src->ridacq_handle, dst->ridacq_handle);
Add real initial support for RQSTYPE_RPT_ID_ACQ.
2015-11-17 13:02:44 +00:00
ISP_IOXGET_8(isp, &src->ridacq_vp_acquired, dst->ridacq_vp_acquired);
ISP_IOXGET_8(isp, &src->ridacq_vp_setup, dst->ridacq_vp_setup);
ISP_IOXGET_8(isp, &src->ridacq_vp_index, dst->ridacq_vp_index);
ISP_IOXGET_8(isp, &src->ridacq_vp_status, dst->ridacq_vp_status);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_16(isp, &src->ridacq_vp_port_lo, dst->ridacq_vp_port_lo);
ISP_IOXGET_8(isp, &src->ridacq_vp_port_hi, dst->ridacq_vp_port_hi);
ISP_IOXGET_8(isp, &src->ridacq_format, dst->ridacq_format);
for (i = 0; i < sizeof (src->ridacq_map) / sizeof (src->ridacq_map[0]); i++) {
ISP_IOXGET_16(isp, &src->ridacq_map[i], dst->ridacq_map[i]);
}
for (i = 0; i < sizeof (src->ridacq_reserved1) / sizeof (src->ridacq_reserved1[0]); i++) {
ISP_IOXGET_16(isp, &src->ridacq_reserved1[i], dst->ridacq_reserved1[i]);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
* CT Passthru canonicalization
*/
void
isp_get_ct_pt(ispsoftc_t *isp, isp_ct_pt_t *src, isp_ct_pt_t *dst)
{
int i;
isp_get_hdr(isp, &src->ctp_header, &dst->ctp_header);
ISP_IOXGET_32(isp, &src->ctp_handle, dst->ctp_handle);
ISP_IOXGET_16(isp, &src->ctp_status, dst->ctp_status);
ISP_IOXGET_16(isp, &src->ctp_nphdl, dst->ctp_nphdl);
ISP_IOXGET_16(isp, &src->ctp_cmd_cnt, dst->ctp_cmd_cnt);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->ctp_vpidx, dst->ctp_vpidx);
ISP_IOXGET_8(isp, &src->ctp_reserved0, dst->ctp_reserved0);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_16(isp, &src->ctp_time, dst->ctp_time);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_16(isp, &src->ctp_reserved1, dst->ctp_reserved1);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_16(isp, &src->ctp_rsp_cnt, dst->ctp_rsp_cnt);
for (i = 0; i < 5; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_16(isp, &src->ctp_reserved2[i], dst->ctp_reserved2[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
ISP_IOXGET_32(isp, &src->ctp_rsp_bcnt, dst->ctp_rsp_bcnt);
ISP_IOXGET_32(isp, &src->ctp_cmd_bcnt, dst->ctp_cmd_bcnt);
for (i = 0; i < 2; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_32(isp, &src->ctp_dataseg[i].ds_base, dst->ctp_dataseg[i].ds_base);
ISP_IOXGET_32(isp, &src->ctp_dataseg[i].ds_basehi, dst->ctp_dataseg[i].ds_basehi);
ISP_IOXGET_32(isp, &src->ctp_dataseg[i].ds_count, dst->ctp_dataseg[i].ds_count);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
}
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
void
isp_get_ms(ispsoftc_t *isp, isp_ms_t *src, isp_ms_t *dst)
{
int i;
isp_get_hdr(isp, &src->ms_header, &dst->ms_header);
ISP_IOXGET_32(isp, &src->ms_handle, dst->ms_handle);
ISP_IOXGET_16(isp, &src->ms_nphdl, dst->ms_nphdl);
ISP_IOXGET_16(isp, &src->ms_status, dst->ms_status);
ISP_IOXGET_16(isp, &src->ms_flags, dst->ms_flags);
ISP_IOXGET_16(isp, &src->ms_reserved1, dst->ms_reserved1);
ISP_IOXGET_16(isp, &src->ms_time, dst->ms_time);
ISP_IOXGET_16(isp, &src->ms_cmd_cnt, dst->ms_cmd_cnt);
ISP_IOXGET_16(isp, &src->ms_tot_cnt, dst->ms_tot_cnt);
ISP_IOXGET_8(isp, &src->ms_type, dst->ms_type);
ISP_IOXGET_8(isp, &src->ms_r_ctl, dst->ms_r_ctl);
ISP_IOXGET_16(isp, &src->ms_rxid, dst->ms_rxid);
ISP_IOXGET_16(isp, &src->ms_reserved2, dst->ms_reserved2);
ISP_IOXGET_32(isp, &src->ms_rsp_bcnt, dst->ms_rsp_bcnt);
ISP_IOXGET_32(isp, &src->ms_cmd_bcnt, dst->ms_cmd_bcnt);
for (i = 0; i < 2; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_32(isp, &src->ms_dataseg[i].ds_base, dst->ms_dataseg[i].ds_base);
ISP_IOXGET_32(isp, &src->ms_dataseg[i].ds_basehi, dst->ms_dataseg[i].ds_basehi);
ISP_IOXGET_32(isp, &src->ms_dataseg[i].ds_count, dst->ms_dataseg[i].ds_count);
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
}
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
void
isp_put_ct_pt(ispsoftc_t *isp, isp_ct_pt_t *src, isp_ct_pt_t *dst)
{
int i;
isp_put_hdr(isp, &src->ctp_header, &dst->ctp_header);
ISP_IOXPUT_32(isp, src->ctp_handle, &dst->ctp_handle);
ISP_IOXPUT_16(isp, src->ctp_status, &dst->ctp_status);
ISP_IOXPUT_16(isp, src->ctp_nphdl, &dst->ctp_nphdl);
ISP_IOXPUT_16(isp, src->ctp_cmd_cnt, &dst->ctp_cmd_cnt);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_8(isp, src->ctp_vpidx, &dst->ctp_vpidx);
ISP_IOXPUT_8(isp, src->ctp_reserved0, &dst->ctp_reserved0);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->ctp_time, &dst->ctp_time);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_16(isp, src->ctp_reserved1, &dst->ctp_reserved1);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->ctp_rsp_cnt, &dst->ctp_rsp_cnt);
for (i = 0; i < 5; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_16(isp, src->ctp_reserved2[i], &dst->ctp_reserved2[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
ISP_IOXPUT_32(isp, src->ctp_rsp_bcnt, &dst->ctp_rsp_bcnt);
ISP_IOXPUT_32(isp, src->ctp_cmd_bcnt, &dst->ctp_cmd_bcnt);
for (i = 0; i < 2; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->ctp_dataseg[i].ds_base, &dst->ctp_dataseg[i].ds_base);
ISP_IOXPUT_32(isp, src->ctp_dataseg[i].ds_basehi, &dst->ctp_dataseg[i].ds_basehi);
ISP_IOXPUT_32(isp, src->ctp_dataseg[i].ds_count, &dst->ctp_dataseg[i].ds_count);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
}
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
void
isp_put_ms(ispsoftc_t *isp, isp_ms_t *src, isp_ms_t *dst)
{
int i;
isp_put_hdr(isp, &src->ms_header, &dst->ms_header);
ISP_IOXPUT_32(isp, src->ms_handle, &dst->ms_handle);
ISP_IOXPUT_16(isp, src->ms_nphdl, &dst->ms_nphdl);
ISP_IOXPUT_16(isp, src->ms_status, &dst->ms_status);
ISP_IOXPUT_16(isp, src->ms_flags, &dst->ms_flags);
ISP_IOXPUT_16(isp, src->ms_reserved1, &dst->ms_reserved1);
ISP_IOXPUT_16(isp, src->ms_time, &dst->ms_time);
ISP_IOXPUT_16(isp, src->ms_cmd_cnt, &dst->ms_cmd_cnt);
ISP_IOXPUT_16(isp, src->ms_tot_cnt, &dst->ms_tot_cnt);
ISP_IOXPUT_8(isp, src->ms_type, &dst->ms_type);
ISP_IOXPUT_8(isp, src->ms_r_ctl, &dst->ms_r_ctl);
ISP_IOXPUT_16(isp, src->ms_rxid, &dst->ms_rxid);
ISP_IOXPUT_16(isp, src->ms_reserved2, &dst->ms_reserved2);
ISP_IOXPUT_32(isp, src->ms_rsp_bcnt, &dst->ms_rsp_bcnt);
ISP_IOXPUT_32(isp, src->ms_cmd_bcnt, &dst->ms_cmd_bcnt);
for (i = 0; i < 2; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->ms_dataseg[i].ds_base, &dst->ms_dataseg[i].ds_base);
ISP_IOXPUT_32(isp, src->ms_dataseg[i].ds_basehi, &dst->ms_dataseg[i].ds_basehi);
ISP_IOXPUT_32(isp, src->ms_dataseg[i].ds_count, &dst->ms_dataseg[i].ds_count);
Make the SAN login/logout stuff more common between different chipsets and provied an isp_control entry point so that the outer layers can do PLOGI/LOGO explicitly. Add MS IOCB support. This completes the cycle for base support for SMI-S.
2006-11-18 03:53:16 +00:00
}
}
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
/*
* Generic SNS request - not particularly useful since the per-command data
* isn't always 16 bit words.
*/
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
a) clean up some declaration stuff (i.e., make more modern with respect to getting rid u_int for uint and so on). b) Turn back on 64 bit DAC support. Cheeze it a bit in that we have two DMA callback functions- one when we have bus_addr_t > 4 bits in width and the other which should be normal. Even Cheezier in that we turn off setting up DMA maps to be BUS_SPACE_MAXADDR if we're in ISP_TARGET_MODE. More work on this in a week or so. c) Tested under amd64 and 1MB DFLTPHYS, sparc64, i386 (PAE, but insufficient memory to really test > 4GB). LINT check under amd64. MFC after: 1 month
2006-02-15 00:31:48 +00:00
isp_put_sns_request(ispsoftc_t *isp, sns_screq_t *src, sns_screq_t *dst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
int i, nw = (int) src->snscb_sblen;
ISP_IOXPUT_16(isp, src->snscb_rblen, &dst->snscb_rblen);
for (i = 0; i < 4; i++) {
ISP_IOXPUT_16(isp, src->snscb_addr[i], &dst->snscb_addr[i]);
}
ISP_IOXPUT_16(isp, src->snscb_sblen, &dst->snscb_sblen);
for (i = 0; i < nw; i++) {
ISP_IOXPUT_16(isp, src->snscb_data[i], &dst->snscb_data[i]);
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_put_gid_ft_request(ispsoftc_t *isp, sns_gid_ft_req_t *src, sns_gid_ft_req_t *dst)
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
{
ISP_IOXPUT_16(isp, src->snscb_rblen, &dst->snscb_rblen);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->snscb_reserved0, &dst->snscb_reserved0);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
ISP_IOXPUT_16(isp, src->snscb_addr[0], &dst->snscb_addr[0]);
ISP_IOXPUT_16(isp, src->snscb_addr[1], &dst->snscb_addr[1]);
ISP_IOXPUT_16(isp, src->snscb_addr[2], &dst->snscb_addr[2]);
ISP_IOXPUT_16(isp, src->snscb_addr[3], &dst->snscb_addr[3]);
ISP_IOXPUT_16(isp, src->snscb_sblen, &dst->snscb_sblen);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->snscb_reserved1, &dst->snscb_reserved1);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
ISP_IOXPUT_16(isp, src->snscb_cmd, &dst->snscb_cmd);
ISP_IOXPUT_16(isp, src->snscb_mword_div_2, &dst->snscb_mword_div_2);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_32(isp, src->snscb_reserved3, &dst->snscb_reserved3);
Send 32 bytes out for fc4_types... Interestingly enough the Solaris/Sparc version worked fine, but Linux/Sparc && FreeBSD/Sparc choked. MFC after: 1 week
2002-04-05 01:40:05 +00:00
ISP_IOXPUT_32(isp, src->snscb_fc4_type, &dst->snscb_fc4_type);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
isp_put_gid_pt_request(ispsoftc_t *isp, sns_gid_pt_req_t *src, sns_gid_pt_req_t *dst)
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
{
ISP_IOXPUT_16(isp, src->snscb_rblen, &dst->snscb_rblen);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->snscb_reserved0, &dst->snscb_reserved0);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
ISP_IOXPUT_16(isp, src->snscb_addr[0], &dst->snscb_addr[0]);
ISP_IOXPUT_16(isp, src->snscb_addr[1], &dst->snscb_addr[1]);
ISP_IOXPUT_16(isp, src->snscb_addr[2], &dst->snscb_addr[2]);
ISP_IOXPUT_16(isp, src->snscb_addr[3], &dst->snscb_addr[3]);
ISP_IOXPUT_16(isp, src->snscb_sblen, &dst->snscb_sblen);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->snscb_reserved1, &dst->snscb_reserved1);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
ISP_IOXPUT_16(isp, src->snscb_cmd, &dst->snscb_cmd);
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOXPUT_16(isp, src->snscb_mword_div_2, &dst->snscb_mword_div_2);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_32(isp, src->snscb_reserved3, &dst->snscb_reserved3);
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOXPUT_8(isp, src->snscb_port_type, &dst->snscb_port_type);
ISP_IOXPUT_8(isp, src->snscb_domain, &dst->snscb_domain);
ISP_IOXPUT_8(isp, src->snscb_area, &dst->snscb_area);
ISP_IOXPUT_8(isp, src->snscb_flags, &dst->snscb_flags);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
isp_put_gxx_id_request(ispsoftc_t *isp, sns_gxx_id_req_t *src, sns_gxx_id_req_t *dst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOXPUT_16(isp, src->snscb_rblen, &dst->snscb_rblen);
ISP_IOXPUT_16(isp, src->snscb_reserved0, &dst->snscb_reserved0);
ISP_IOXPUT_16(isp, src->snscb_addr[0], &dst->snscb_addr[0]);
ISP_IOXPUT_16(isp, src->snscb_addr[1], &dst->snscb_addr[1]);
ISP_IOXPUT_16(isp, src->snscb_addr[2], &dst->snscb_addr[2]);
ISP_IOXPUT_16(isp, src->snscb_addr[3], &dst->snscb_addr[3]);
ISP_IOXPUT_16(isp, src->snscb_sblen, &dst->snscb_sblen);
ISP_IOXPUT_16(isp, src->snscb_reserved1, &dst->snscb_reserved1);
ISP_IOXPUT_16(isp, src->snscb_cmd, &dst->snscb_cmd);
ISP_IOXPUT_16(isp, src->snscb_mword_div_2, &dst->snscb_mword_div_2);
ISP_IOXPUT_32(isp, src->snscb_reserved3, &dst->snscb_reserved3);
ISP_IOXPUT_32(isp, src->snscb_portid, &dst->snscb_portid);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
isp_get_gid_xx_response(ispsoftc_t *isp, sns_gid_xx_rsp_t *src, sns_gid_xx_rsp_t *dst, int nwords)
Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. This was found on linux/PPC. This mandated creating another inline: isp_get_gan_response.
2001-12-11 21:58:04 +00:00
{
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
int i, j;
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
isp_get_ct_hdr(isp, &src->snscb_cthdr, &dst->snscb_cthdr);
for (i = 0; i < nwords; i++) {
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_8(isp, &src->snscb_ports[i].control,
dst->snscb_ports[i].control);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
for (j = 0; j < 3; j++) {
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_8(isp, &src->snscb_ports[i].portid[j],
dst->snscb_ports[i].portid[j]);
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
}
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
if (dst->snscb_ports[i].control & 0x80)
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
break;
Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. This was found on linux/PPC. This mandated creating another inline: isp_get_gan_response.
2001-12-11 21:58:04 +00:00
}
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_get_gxn_id_response(ispsoftc_t *isp, sns_gxn_id_rsp_t *src, sns_gxn_id_rsp_t *dst)
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
{
int i;
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
isp_get_ct_hdr(isp, &src->snscb_cthdr, &dst->snscb_cthdr);
for (i = 0; i < 8; i++)
ISP_IOZGET_8(isp, &src->snscb_wwn[i], dst->snscb_wwn[i]);
}
void
isp_get_gft_id_response(ispsoftc_t *isp, sns_gft_id_rsp_t *src, sns_gft_id_rsp_t *dst)
{
int i;
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
isp_get_ct_hdr(isp, &src->snscb_cthdr, &dst->snscb_cthdr);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
for (i = 0; i < 8; i++) {
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_32(isp, &src->snscb_fc4_types[i],
dst->snscb_fc4_types[i]);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_get_gff_id_response(ispsoftc_t *isp, sns_gff_id_rsp_t *src, sns_gff_id_rsp_t *dst)
Scale back # of luns supported for SCC to 16384- oops- top 3 bits are a lun address modifier of sorts. Only an HP XP-512 seems to have cared. Fix a few misplaced pointers for the new fabric goop, which has been demonstrated to work on newer Brocades and McData switches now. Put in commented out code which would run GFF_ID if the QLogic f/w allowed it. Don't whine about not being able to find a handle for a command if it was a command aborted (by us).
2002-04-16 19:55:35 +00:00
{
int i;
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
Scale back # of luns supported for SCC to 16384- oops- top 3 bits are a lun address modifier of sorts. Only an HP XP-512 seems to have cared. Fix a few misplaced pointers for the new fabric goop, which has been demonstrated to work on newer Brocades and McData switches now. Put in commented out code which would run GFF_ID if the QLogic f/w allowed it. Don't whine about not being able to find a handle for a command if it was a command aborted (by us).
2002-04-16 19:55:35 +00:00
isp_get_ct_hdr(isp, &src->snscb_cthdr, &dst->snscb_cthdr);
for (i = 0; i < 32; i++) {
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_32(isp, &src->snscb_fc4_features[i],
dst->snscb_fc4_features[i]);
Scale back # of luns supported for SCC to 16384- oops- top 3 bits are a lun address modifier of sorts. Only an HP XP-512 seems to have cared. Fix a few misplaced pointers for the new fabric goop, which has been demonstrated to work on newer Brocades and McData switches now. Put in commented out code which would run GFF_ID if the QLogic f/w allowed it. Don't whine about not being able to find a handle for a command if it was a command aborted (by us).
2002-04-16 19:55:35 +00:00
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_get_ga_nxt_response(ispsoftc_t *isp, sns_ga_nxt_rsp_t *src, sns_ga_nxt_rsp_t *dst)
Fix bus dma segment count to be based off of MAXPHYS, not BUS_SPACE_MAXSIZE. Grumble. I've seen better documented architectures out of Redmond. Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list. MFC after: 1 week
2002-04-04 23:46:01 +00:00
{
int i;
isp_get_ct_hdr(isp, &src->snscb_cthdr, &dst->snscb_cthdr);
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_8(isp, &src->snscb_port_type, dst->snscb_port_type);
Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. This was found on linux/PPC. This mandated creating another inline: isp_get_gan_response.
2001-12-11 21:58:04 +00:00
for (i = 0; i < 3; i++) {
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_8(isp, &src->snscb_port_id[i], dst->snscb_port_id[i]);
Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. This was found on linux/PPC. This mandated creating another inline: isp_get_gan_response.
2001-12-11 21:58:04 +00:00
}
for (i = 0; i < 8; i++) {
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_8(isp, &src->snscb_portname[i], dst->snscb_portname[i]);
Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. This was found on linux/PPC. This mandated creating another inline: isp_get_gan_response.
2001-12-11 21:58:04 +00:00
}
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_8(isp, &src->snscb_pnlen, dst->snscb_pnlen);
Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. This was found on linux/PPC. This mandated creating another inline: isp_get_gan_response.
2001-12-11 21:58:04 +00:00
for (i = 0; i < 255; i++) {
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_8(isp, &src->snscb_pname[i], dst->snscb_pname[i]);
Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. This was found on linux/PPC. This mandated creating another inline: isp_get_gan_response.
2001-12-11 21:58:04 +00:00
}
for (i = 0; i < 8; i++) {
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_8(isp, &src->snscb_nodename[i], dst->snscb_nodename[i]);
Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. This was found on linux/PPC. This mandated creating another inline: isp_get_gan_response.
2001-12-11 21:58:04 +00:00
}
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_8(isp, &src->snscb_nnlen, dst->snscb_nnlen);
Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. This was found on linux/PPC. This mandated creating another inline: isp_get_gan_response.
2001-12-11 21:58:04 +00:00
for (i = 0; i < 255; i++) {
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_8(isp, &src->snscb_nname[i], dst->snscb_nname[i]);
Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. This was found on linux/PPC. This mandated creating another inline: isp_get_gan_response.
2001-12-11 21:58:04 +00:00
}
for (i = 0; i < 8; i++) {
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_8(isp, &src->snscb_ipassoc[i], dst->snscb_ipassoc[i]);
Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. This was found on linux/PPC. This mandated creating another inline: isp_get_gan_response.
2001-12-11 21:58:04 +00:00
}
for (i = 0; i < 16; i++) {
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_8(isp, &src->snscb_ipaddr[i], dst->snscb_ipaddr[i]);
Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. This was found on linux/PPC. This mandated creating another inline: isp_get_gan_response.
2001-12-11 21:58:04 +00:00
}
for (i = 0; i < 4; i++) {
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_8(isp, &src->snscb_svc_class[i], dst->snscb_svc_class[i]);
Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. This was found on linux/PPC. This mandated creating another inline: isp_get_gan_response.
2001-12-11 21:58:04 +00:00
}
for (i = 0; i < 32; i++) {
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_8(isp, &src->snscb_fc4_types[i], dst->snscb_fc4_types[i]);
Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. This was found on linux/PPC. This mandated creating another inline: isp_get_gan_response.
2001-12-11 21:58:04 +00:00
}
for (i = 0; i < 8; i++) {
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_8(isp, &src->snscb_fpname[i], dst->snscb_fpname[i]);
Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. This was found on linux/PPC. This mandated creating another inline: isp_get_gan_response.
2001-12-11 21:58:04 +00:00
}
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_8(isp, &src->snscb_reserved, dst->snscb_reserved);
Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. This was found on linux/PPC. This mandated creating another inline: isp_get_gan_response.
2001-12-11 21:58:04 +00:00
for (i = 0; i < 3; i++) {
Switch fabric scans from GID_FT to GID_PT+GFF_ID/GFT_ID. Instead of using GID_FT SNS request to get list of registered FCP ports, use GID_PT to get list of all Nx_Ports, and then use GFF_ID and/or GFT_ID requests to find whether they are FCP and target capable. The problem with old approach is that GID_FT does not report ports without FC-4 type registered. In particular it was impossible to boot OS from FreeBSD FC target using QLogic FC BIOS, since one does not register FC-4 type even on new cards and so ignored by old code as incompatible. As a side bonus this allows initiator to skip pointless logins to other initiators by fetching that information from SNS instead. In case some switches do not implement GFF_ID/GFT_ID correctly, add sysctls to disable that functionality. I handled broken GFF_ID of my Brocade 200E, but there may be other switches with different bugs. Linux also uses GID_PT, but GFF_ID is disabled by default there, and GFT_ID is not supported. Sponsored by: iXsystems, Inc.
2017-07-03 15:56:45 +00:00
ISP_IOZGET_8(isp, &src->snscb_hardaddr[i], dst->snscb_hardaddr[i]);
Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. This was found on linux/PPC. This mandated creating another inline: isp_get_gan_response.
2001-12-11 21:58:04 +00:00
}
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
void
isp_get_els(ispsoftc_t *isp, els_t *src, els_t *dst)
{
int i;
isp_get_hdr(isp, &src->els_hdr, &dst->els_hdr);
ISP_IOXGET_32(isp, &src->els_handle, dst->els_handle);
ISP_IOXGET_16(isp, &src->els_status, dst->els_status);
ISP_IOXGET_16(isp, &src->els_nphdl, dst->els_nphdl);
ISP_IOXGET_16(isp, &src->els_xmit_dsd_count, dst->els_xmit_dsd_count);
ISP_IOXGET_8(isp, &src->els_vphdl, dst->els_vphdl);
ISP_IOXGET_8(isp, &src->els_sof, dst->els_sof);
ISP_IOXGET_32(isp, &src->els_rxid, dst->els_rxid);
ISP_IOXGET_16(isp, &src->els_recv_dsd_count, dst->els_recv_dsd_count);
ISP_IOXGET_8(isp, &src->els_opcode, dst->els_opcode);
ISP_IOXGET_8(isp, &src->els_reserved2, dst->els_reserved1);
ISP_IOXGET_8(isp, &src->els_did_lo, dst->els_did_lo);
ISP_IOXGET_8(isp, &src->els_did_mid, dst->els_did_mid);
ISP_IOXGET_8(isp, &src->els_did_hi, dst->els_did_hi);
ISP_IOXGET_8(isp, &src->els_reserved2, dst->els_reserved2);
ISP_IOXGET_16(isp, &src->els_reserved3, dst->els_reserved3);
ISP_IOXGET_16(isp, &src->els_ctl_flags, dst->els_ctl_flags);
ISP_IOXGET_32(isp, &src->els_bytecnt, dst->els_bytecnt);
ISP_IOXGET_32(isp, &src->els_subcode1, dst->els_subcode1);
ISP_IOXGET_32(isp, &src->els_subcode2, dst->els_subcode2);
for (i = 0; i < 20; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->els_reserved4[i], dst->els_reserved4[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
}
void
isp_put_els(ispsoftc_t *isp, els_t *src, els_t *dst)
{
isp_put_hdr(isp, &src->els_hdr, &dst->els_hdr);
ISP_IOXPUT_32(isp, src->els_handle, &dst->els_handle);
ISP_IOXPUT_16(isp, src->els_status, &dst->els_status);
ISP_IOXPUT_16(isp, src->els_nphdl, &dst->els_nphdl);
ISP_IOXPUT_16(isp, src->els_xmit_dsd_count, &dst->els_xmit_dsd_count);
ISP_IOXPUT_8(isp, src->els_vphdl, &dst->els_vphdl);
ISP_IOXPUT_8(isp, src->els_sof, &dst->els_sof);
ISP_IOXPUT_32(isp, src->els_rxid, &dst->els_rxid);
ISP_IOXPUT_16(isp, src->els_recv_dsd_count, &dst->els_recv_dsd_count);
ISP_IOXPUT_8(isp, src->els_opcode, &dst->els_opcode);
ISP_IOXPUT_8(isp, src->els_reserved2, &dst->els_reserved1);
ISP_IOXPUT_8(isp, src->els_did_lo, &dst->els_did_lo);
ISP_IOXPUT_8(isp, src->els_did_mid, &dst->els_did_mid);
ISP_IOXPUT_8(isp, src->els_did_hi, &dst->els_did_hi);
ISP_IOXPUT_8(isp, src->els_reserved2, &dst->els_reserved2);
ISP_IOXPUT_16(isp, src->els_reserved3, &dst->els_reserved3);
ISP_IOXPUT_16(isp, src->els_ctl_flags, &dst->els_ctl_flags);
ISP_IOXPUT_32(isp, src->els_recv_bytecnt, &dst->els_recv_bytecnt);
ISP_IOXPUT_32(isp, src->els_xmit_bytecnt, &dst->els_xmit_bytecnt);
ISP_IOXPUT_32(isp, src->els_xmit_dsd_length, &dst->els_xmit_dsd_length);
ISP_IOXPUT_16(isp, src->els_xmit_dsd_a1500, &dst->els_xmit_dsd_a1500);
ISP_IOXPUT_16(isp, src->els_xmit_dsd_a3116, &dst->els_xmit_dsd_a3116);
ISP_IOXPUT_16(isp, src->els_xmit_dsd_a4732, &dst->els_xmit_dsd_a4732);
ISP_IOXPUT_16(isp, src->els_xmit_dsd_a6348, &dst->els_xmit_dsd_a6348);
ISP_IOXPUT_32(isp, src->els_recv_dsd_length, &dst->els_recv_dsd_length);
ISP_IOXPUT_16(isp, src->els_recv_dsd_a1500, &dst->els_recv_dsd_a1500);
ISP_IOXPUT_16(isp, src->els_recv_dsd_a3116, &dst->els_recv_dsd_a3116);
ISP_IOXPUT_16(isp, src->els_recv_dsd_a4732, &dst->els_recv_dsd_a4732);
ISP_IOXPUT_16(isp, src->els_recv_dsd_a6348, &dst->els_recv_dsd_a6348);
}
/*
* FC Structure Canonicalization
*/
void
isp_get_fc_hdr(ispsoftc_t *isp, fc_hdr_t *src, fc_hdr_t *dst)
{
ISP_IOZGET_8(isp, &src->r_ctl, dst->r_ctl);
ISP_IOZGET_8(isp, &src->d_id[0], dst->d_id[0]);
ISP_IOZGET_8(isp, &src->d_id[1], dst->d_id[1]);
ISP_IOZGET_8(isp, &src->d_id[2], dst->d_id[2]);
ISP_IOZGET_8(isp, &src->cs_ctl, dst->cs_ctl);
ISP_IOZGET_8(isp, &src->s_id[0], dst->s_id[0]);
ISP_IOZGET_8(isp, &src->s_id[1], dst->s_id[1]);
ISP_IOZGET_8(isp, &src->s_id[2], dst->s_id[2]);
ISP_IOZGET_8(isp, &src->type, dst->type);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOZGET_8(isp, &src->f_ctl[0], dst->f_ctl[0]);
ISP_IOZGET_8(isp, &src->f_ctl[1], dst->f_ctl[1]);
ISP_IOZGET_8(isp, &src->f_ctl[2], dst->f_ctl[2]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOZGET_8(isp, &src->seq_id, dst->seq_id);
ISP_IOZGET_8(isp, &src->df_ctl, dst->df_ctl);
ISP_IOZGET_16(isp, &src->seq_cnt, dst->seq_cnt);
ISP_IOZGET_16(isp, &src->ox_id, dst->ox_id);
ISP_IOZGET_16(isp, &src->rx_id, dst->rx_id);
ISP_IOZGET_32(isp, &src->parameter, dst->parameter);
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
void
isp_put_fc_hdr(ispsoftc_t *isp, fc_hdr_t *src, fc_hdr_t *dst)
{
ISP_IOZPUT_8(isp, src->r_ctl, &dst->r_ctl);
ISP_IOZPUT_8(isp, src->d_id[0], &dst->d_id[0]);
ISP_IOZPUT_8(isp, src->d_id[1], &dst->d_id[1]);
ISP_IOZPUT_8(isp, src->d_id[2], &dst->d_id[2]);
ISP_IOZPUT_8(isp, src->cs_ctl, &dst->cs_ctl);
ISP_IOZPUT_8(isp, src->s_id[0], &dst->s_id[0]);
ISP_IOZPUT_8(isp, src->s_id[1], &dst->s_id[1]);
ISP_IOZPUT_8(isp, src->s_id[2], &dst->s_id[2]);
ISP_IOZPUT_8(isp, src->type, &dst->type);
ISP_IOZPUT_8(isp, src->f_ctl[0], &dst->f_ctl[0]);
ISP_IOZPUT_8(isp, src->f_ctl[1], &dst->f_ctl[1]);
ISP_IOZPUT_8(isp, src->f_ctl[2], &dst->f_ctl[2]);
ISP_IOZPUT_8(isp, src->seq_id, &dst->seq_id);
ISP_IOZPUT_8(isp, src->df_ctl, &dst->df_ctl);
ISP_IOZPUT_16(isp, src->seq_cnt, &dst->seq_cnt);
ISP_IOZPUT_16(isp, src->ox_id, &dst->ox_id);
ISP_IOZPUT_16(isp, src->rx_id, &dst->rx_id);
ISP_IOZPUT_32(isp, src->parameter, &dst->parameter);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
void
isp_get_fcp_cmnd_iu(ispsoftc_t *isp, fcp_cmnd_iu_t *src, fcp_cmnd_iu_t *dst)
{
int i;
for (i = 0; i < 8; i++) {
ISP_IOZGET_8(isp, &src->fcp_cmnd_lun[i], dst->fcp_cmnd_lun[i]);
}
ISP_IOZGET_8(isp, &src->fcp_cmnd_crn, dst->fcp_cmnd_crn);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOZGET_8(isp, &src->fcp_cmnd_task_attribute, dst->fcp_cmnd_task_attribute);
ISP_IOZGET_8(isp, &src->fcp_cmnd_task_management, dst->fcp_cmnd_task_management);
ISP_IOZGET_8(isp, &src->fcp_cmnd_alen_datadir, dst->fcp_cmnd_alen_datadir);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
for (i = 0; i < 16; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOZGET_8(isp, &src->cdb_dl.sf.fcp_cmnd_cdb[i], dst->cdb_dl.sf.fcp_cmnd_cdb[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOZGET_32(isp, &src->cdb_dl.sf.fcp_cmnd_dl, dst->cdb_dl.sf.fcp_cmnd_dl);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
void
isp_put_rft_id(ispsoftc_t *isp, rft_id_t *src, rft_id_t *dst)
{
int i;
isp_put_ct_hdr(isp, &src->rftid_hdr, &dst->rftid_hdr);
ISP_IOZPUT_8(isp, src->rftid_reserved, &dst->rftid_reserved);
for (i = 0; i < 3; i++) {
ISP_IOZPUT_8(isp, src->rftid_portid[i], &dst->rftid_portid[i]);
}
for (i = 0; i < 8; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOZPUT_32(isp, src->rftid_fc4types[i], &dst->rftid_fc4types[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
}
Register symbolic port/node names in FC name server. This is cosmetics that simplifies identification of new ports on FC switch. It would be good to use target name from CTL here instead of hostname, but it is not passed here through CAM now. MFC after: 2 weeks
2016-04-09 14:50:47 +00:00
void
isp_put_rspn_id(ispsoftc_t *isp, rspn_id_t *src, rspn_id_t *dst)
{
/* int i;*/
isp_put_ct_hdr(isp, &src->rspnid_hdr, &dst->rspnid_hdr);
ISP_IOZPUT_8(isp, src->rspnid_reserved, &dst->rspnid_reserved);
ISP_IOZPUT_8(isp, src->rspnid_length, &dst->rspnid_length);
/* for (i = 0; i < src->rspnid_length; i++)
ISP_IOZPUT_8(isp, src->rspnid_name[i], &dst->rspnid_name[i]);*/
}
Register our FC4 Features in SNS.
2015-11-17 19:57:49 +00:00
void
isp_put_rff_id(ispsoftc_t *isp, rff_id_t *src, rff_id_t *dst)
{
int i;
isp_put_ct_hdr(isp, &src->rffid_hdr, &dst->rffid_hdr);
ISP_IOZPUT_8(isp, src->rffid_reserved, &dst->rffid_reserved);
for (i = 0; i < 3; i++)
ISP_IOZPUT_8(isp, src->rffid_portid[i], &dst->rffid_portid[i]);
ISP_IOZPUT_16(isp, src->rffid_reserved2, &dst->rffid_reserved2);
ISP_IOZPUT_8(isp, src->rffid_fc4features, &dst->rffid_fc4features);
ISP_IOZPUT_8(isp, src->rffid_fc4type, &dst->rffid_fc4type);
}
Register symbolic port/node names in FC name server. This is cosmetics that simplifies identification of new ports on FC switch. It would be good to use target name from CTL here instead of hostname, but it is not passed here through CAM now. MFC after: 2 weeks
2016-04-09 14:50:47 +00:00
void
isp_put_rsnn_nn(ispsoftc_t *isp, rsnn_nn_t *src, rsnn_nn_t *dst)
{
int i;
isp_put_ct_hdr(isp, &src->rsnnnn_hdr, &dst->rsnnnn_hdr);
for (i = 0; i < 8; i++)
ISP_IOZPUT_8(isp, src->rsnnnn_nodename[i], &dst->rsnnnn_nodename[i]);
ISP_IOZPUT_8(isp, src->rsnnnn_length, &dst->rsnnnn_length);
/* for (i = 0; i < src->rsnnnn_length; i++)
ISP_IOZPUT_8(isp, src->rsnnnn_name[i], &dst->rsnnnn_name[i]);*/
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
void
isp_get_ct_hdr(ispsoftc_t *isp, ct_hdr_t *src, ct_hdr_t *dst)
{
ISP_IOZGET_8(isp, &src->ct_revision, dst->ct_revision);
ISP_IOZGET_8(isp, &src->ct_in_id[0], dst->ct_in_id[0]);
ISP_IOZGET_8(isp, &src->ct_in_id[1], dst->ct_in_id[1]);
ISP_IOZGET_8(isp, &src->ct_in_id[2], dst->ct_in_id[2]);
ISP_IOZGET_8(isp, &src->ct_fcs_type, dst->ct_fcs_type);
ISP_IOZGET_8(isp, &src->ct_fcs_subtype, dst->ct_fcs_subtype);
ISP_IOZGET_8(isp, &src->ct_options, dst->ct_options);
ISP_IOZGET_8(isp, &src->ct_reserved0, dst->ct_reserved0);
ISP_IOZGET_16(isp, &src->ct_cmd_resp, dst->ct_cmd_resp);
ISP_IOZGET_16(isp, &src->ct_bcnt_resid, dst->ct_bcnt_resid);
ISP_IOZGET_8(isp, &src->ct_reserved1, dst->ct_reserved1);
ISP_IOZGET_8(isp, &src->ct_reason, dst->ct_reason);
ISP_IOZGET_8(isp, &src->ct_explanation, dst->ct_explanation);
ISP_IOZGET_8(isp, &src->ct_vunique, dst->ct_vunique);
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
void
isp_put_ct_hdr(ispsoftc_t *isp, ct_hdr_t *src, ct_hdr_t *dst)
{
ISP_IOZPUT_8(isp, src->ct_revision, &dst->ct_revision);
ISP_IOZPUT_8(isp, src->ct_in_id[0], &dst->ct_in_id[0]);
ISP_IOZPUT_8(isp, src->ct_in_id[1], &dst->ct_in_id[1]);
ISP_IOZPUT_8(isp, src->ct_in_id[2], &dst->ct_in_id[2]);
ISP_IOZPUT_8(isp, src->ct_fcs_type, &dst->ct_fcs_type);
ISP_IOZPUT_8(isp, src->ct_fcs_subtype, &dst->ct_fcs_subtype);
ISP_IOZPUT_8(isp, src->ct_options, &dst->ct_options);
ISP_IOZPUT_8(isp, src->ct_reserved0, &dst->ct_reserved0);
ISP_IOZPUT_16(isp, src->ct_cmd_resp, &dst->ct_cmd_resp);
ISP_IOZPUT_16(isp, src->ct_bcnt_resid, &dst->ct_bcnt_resid);
ISP_IOZPUT_8(isp, src->ct_reserved1, &dst->ct_reserved1);
ISP_IOZPUT_8(isp, src->ct_reason, &dst->ct_reason);
ISP_IOZPUT_8(isp, src->ct_explanation, &dst->ct_explanation);
ISP_IOZPUT_8(isp, src->ct_vunique, &dst->ct_vunique);
}
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
void
isp_put_fcp_rsp_iu(ispsoftc_t *isp, fcp_rsp_iu_t *src, fcp_rsp_iu_t *dst)
{
int i;
for (i = 0; i < ((sizeof (src->fcp_rsp_reserved))/(sizeof (src->fcp_rsp_reserved[0]))); i++) {
ISP_IOZPUT_8(isp, src->fcp_rsp_reserved[i], &dst->fcp_rsp_reserved[i]);
}
ISP_IOZPUT_16(isp, src->fcp_rsp_status_qualifier, &dst->fcp_rsp_status_qualifier);
ISP_IOZPUT_8(isp, src->fcp_rsp_bits, &dst->fcp_rsp_bits);
ISP_IOZPUT_8(isp, src->fcp_rsp_scsi_status, &dst->fcp_rsp_scsi_status);
ISP_IOZPUT_32(isp, src->fcp_rsp_resid, &dst->fcp_rsp_resid);
ISP_IOZPUT_32(isp, src->fcp_rsp_snslen, &dst->fcp_rsp_snslen);
ISP_IOZPUT_32(isp, src->fcp_rsp_rsplen, &dst->fcp_rsp_rsplen);
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
/*
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
* Find port database entries
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
*/
int
Another round of port scanner rewrite. This change simplifies and unifies port adding/updating for loop and fabric scanners. It also fixes problems with scanning restarts due to concurrent port databases changes. It also fixes many cosmetic issues.
2015-11-19 17:43:47 +00:00
isp_find_pdb_empty(ispsoftc_t *isp, int chan, fcportdb_t **lptr)
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
{
Another round of port scanner rewrite. This change simplifies and unifies port adding/updating for loop and fabric scanners. It also fixes problems with scanning restarts due to concurrent port databases changes. It also fixes many cosmetic issues.
2015-11-19 17:43:47 +00:00
fcparam *fcp = FCPARAM(isp, chan);
int i;
for (i = 0; i < MAX_FC_TARG; i++) {
fcportdb_t *lp = &fcp->portdb[i];
if (lp->state == FC_PORTDB_STATE_NIL) {
*lptr = lp;
return (1);
}
}
return (0);
}
int
isp_find_pdb_by_wwpn(ispsoftc_t *isp, int chan, uint64_t wwpn, fcportdb_t **lptr)
{
fcparam *fcp = FCPARAM(isp, chan);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
int i;
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
for (i = 0; i < MAX_FC_TARG; i++) {
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
fcportdb_t *lp = &fcp->portdb[i];
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
if (lp->state == FC_PORTDB_STATE_NIL)
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
continue;
Another round of port scanner rewrite. This change simplifies and unifies port adding/updating for loop and fabric scanners. It also fixes problems with scanning restarts due to concurrent port databases changes. It also fixes many cosmetic issues.
2015-11-19 17:43:47 +00:00
if (lp->port_wwn == wwpn) {
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
*lptr = lp;
return (1);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
}
return (0);
}
int
Another round of port scanner rewrite. This change simplifies and unifies port adding/updating for loop and fabric scanners. It also fixes problems with scanning restarts due to concurrent port databases changes. It also fixes many cosmetic issues.
2015-11-19 17:43:47 +00:00
isp_find_pdb_by_handle(ispsoftc_t *isp, int chan, uint16_t handle,
fcportdb_t **lptr)
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
{
Another round of port scanner rewrite. This change simplifies and unifies port adding/updating for loop and fabric scanners. It also fixes problems with scanning restarts due to concurrent port databases changes. It also fixes many cosmetic issues.
2015-11-19 17:43:47 +00:00
fcparam *fcp = FCPARAM(isp, chan);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
int i;
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
for (i = 0; i < MAX_FC_TARG; i++) {
fcportdb_t *lp = &fcp->portdb[i];
if (lp->state == FC_PORTDB_STATE_NIL)
continue;
if (lp->handle == handle) {
*lptr = lp;
return (1);
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
return (0);
}
int
Another round of port scanner rewrite. This change simplifies and unifies port adding/updating for loop and fabric scanners. It also fixes problems with scanning restarts due to concurrent port databases changes. It also fixes many cosmetic issues.
2015-11-19 17:43:47 +00:00
isp_find_pdb_by_portid(ispsoftc_t *isp, int chan, uint32_t portid,
fcportdb_t **lptr)
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
{
Another round of port scanner rewrite. This change simplifies and unifies port adding/updating for loop and fabric scanners. It also fixes problems with scanning restarts due to concurrent port databases changes. It also fixes many cosmetic issues.
2015-11-19 17:43:47 +00:00
fcparam *fcp = FCPARAM(isp, chan);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
int i;
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
for (i = 0; i < MAX_FC_TARG; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcportdb_t *lp = &fcp->portdb[i];
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
if (lp->state == FC_PORTDB_STATE_NIL)
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
continue;
Another round of port scanner rewrite. This change simplifies and unifies port adding/updating for loop and fabric scanners. It also fixes problems with scanning restarts due to concurrent port databases changes. It also fixes many cosmetic issues.
2015-11-19 17:43:47 +00:00
if (lp->portid == portid) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
*lptr = lp;
return (1);
}
}
return (0);
}
Another round of port scanner rewrite. This change simplifies and unifies port adding/updating for loop and fabric scanners. It also fixes problems with scanning restarts due to concurrent port databases changes. It also fixes many cosmetic issues.
2015-11-19 17:43:47 +00:00
#ifdef ISP_TARGET_MODE
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
void
isp_find_chan_by_did(ispsoftc_t *isp, uint32_t did, uint16_t *cp)
{
uint16_t chan;
*cp = ISP_NOCHAN;
for (chan = 0; chan < isp->isp_nchan; chan++) {
fcparam *fcp = FCPARAM(isp, chan);
Unify and cleanup FC ports scan.
2015-11-17 16:33:46 +00:00
if ((fcp->role & ISP_ROLE_TARGET) == 0 ||
Extract virtual port address from RQSTYPE_RPT_ID_ACQ. This should close the race between request arriving on new target mode virtual port and its scanner thread finally fetch its address for request routing.
2016-04-14 20:49:01 +00:00
fcp->isp_loopstate < LOOP_HAVE_ADDR) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
continue;
}
if (fcp->isp_portid == did) {
*cp = chan;
break;
}
}
}
/*
* Add an initiator device to the port database
*/
void
Improve INOTs handling for 24xx and above chips.
2015-10-23 16:54:24 +00:00
isp_add_wwn_entry(ispsoftc_t *isp, int chan, uint64_t wwpn, uint64_t wwnn,
uint16_t nphdl, uint32_t s_id, uint16_t prli_params)
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
{
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
char buf[64];
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcparam *fcp;
fcportdb_t *lp;
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
int i, change;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
fcp = FCPARAM(isp, chan);
if (nphdl >= MAX_NPORT_HANDLE) {
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
isp_prt(isp, ISP_LOGTINFO|ISP_LOGWARN, "Chan %d WWPN 0x%016llx "
"PortID 0x%06x handle 0x%x -- bad handle",
Improve INOTs handling for 24xx and above chips.
2015-10-23 16:54:24 +00:00
chan, (unsigned long long) wwpn, s_id, nphdl);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
return;
}
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
/*
* If valid record for requested handle already exists, update it
* with new parameters. Some cases of update can be suspicious,
* so log them verbosely and dump the whole port database.
*/
Another round of port scanner rewrite. This change simplifies and unifies port adding/updating for loop and fabric scanners. It also fixes problems with scanning restarts due to concurrent port databases changes. It also fixes many cosmetic issues.
2015-11-19 17:43:47 +00:00
if ((VALID_INI(wwpn) && isp_find_pdb_by_wwpn(isp, chan, wwpn, &lp)) ||
Fix target mode with fabric for pre-24xx chips. For those chips we are not receiving login events, adding initiators based on ATIO requests. But there is no port ID in that structure, so in fabric mode we have to explicitly fetch it from firmware to be able to do normal scan after that.
2015-11-21 21:01:00 +00:00
(VALID_PORT(s_id) && isp_find_pdb_by_portid(isp, chan, s_id, &lp))) {
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
change = 0;
lp->new_portid = lp->portid;
lp->new_prli_word3 = lp->prli_word3;
Fix target mode with fabric for pre-24xx chips. For those chips we are not receiving login events, adding initiators based on ATIO requests. But there is no port ID in that structure, so in fabric mode we have to explicitly fetch it from firmware to be able to do normal scan after that.
2015-11-21 21:01:00 +00:00
if (VALID_PORT(s_id) && lp->portid != s_id) {
if (!VALID_PORT(lp->portid)) {
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
isp_prt(isp, ISP_LOGTINFO,
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
"Chan %d WWPN 0x%016llx handle 0x%x "
"gets PortID 0x%06x",
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
chan, (unsigned long long) lp->port_wwn,
nphdl, s_id);
} else {
isp_prt(isp, ISP_LOGTINFO|ISP_LOGWARN,
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
"Chan %d WWPN 0x%016llx handle 0x%x "
"changes PortID 0x%06x to 0x%06x",
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
chan, (unsigned long long) lp->port_wwn,
nphdl, lp->portid, s_id);
if (isp->isp_dblev & (ISP_LOGTINFO|ISP_LOGWARN))
isp_dump_portdb(isp, chan);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
lp->new_portid = s_id;
change++;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Improve INOTs handling for 24xx and above chips.
2015-10-23 16:54:24 +00:00
if (VALID_INI(wwpn) && lp->port_wwn != wwpn) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
if (!VALID_INI(lp->port_wwn)) {
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
isp_prt(isp, ISP_LOGTINFO,
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
"Chan %d PortID 0x%06x handle 0x%x "
Improve INOTs handling for 24xx and above chips.
2015-10-23 16:54:24 +00:00
"gets WWPN 0x%016llxx",
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
chan, lp->portid, nphdl,
Improve INOTs handling for 24xx and above chips.
2015-10-23 16:54:24 +00:00
(unsigned long long) wwpn);
} else if (lp->port_wwn != wwpn) {
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
isp_prt(isp, ISP_LOGTINFO|ISP_LOGWARN,
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
"Chan %d PortID 0x%06x handle 0x%x "
Improve INOTs handling for 24xx and above chips.
2015-10-23 16:54:24 +00:00
"changes WWPN 0x%016llx to 0x%016llx",
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
chan, lp->portid, nphdl,
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
(unsigned long long) lp->port_wwn,
Improve INOTs handling for 24xx and above chips.
2015-10-23 16:54:24 +00:00
(unsigned long long) wwpn);
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
if (isp->isp_dblev & (ISP_LOGTINFO|ISP_LOGWARN))
isp_dump_portdb(isp, chan);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Improve INOTs handling for 24xx and above chips.
2015-10-23 16:54:24 +00:00
lp->port_wwn = wwpn;
change++;
}
if (VALID_INI(wwnn) && lp->node_wwn != wwnn) {
if (!VALID_INI(lp->node_wwn)) {
isp_prt(isp, ISP_LOGTINFO,
"Chan %d PortID 0x%06x handle 0x%x "
"gets WWNN 0x%016llxx",
chan, lp->portid, nphdl,
(unsigned long long) wwnn);
} else if (lp->port_wwn != wwnn) {
isp_prt(isp, ISP_LOGTINFO,
"Chan %d PortID 0x%06x handle 0x%x "
"changes WWNN 0x%016llx to 0x%016llx",
chan, lp->portid, nphdl,
(unsigned long long) lp->node_wwn,
(unsigned long long) wwnn);
}
lp->node_wwn = wwnn;
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
change++;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
if (prli_params != 0 && lp->prli_word3 != prli_params) {
isp_gen_role_str(buf, sizeof (buf), prli_params);
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
isp_prt(isp, ISP_LOGTINFO|ISP_LOGCONFIG,
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
"Chan %d WWPN 0x%016llx PortID 0x%06x "
"handle 0x%x changes PRLI Word 3 %s",
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
chan, (unsigned long long) lp->port_wwn,
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
lp->portid, lp->handle, buf);
lp->new_prli_word3 = prli_params;
change++;
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
}
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
if (lp->handle != nphdl) {
isp_prt(isp, ISP_LOGTINFO|ISP_LOGCONFIG,
"Chan %d WWPN 0x%016llx PortID 0x%06x "
"changes handle 0x%x to 0x%x",
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
chan, (unsigned long long) lp->port_wwn,
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
lp->portid, lp->handle, nphdl);
lp->handle = nphdl;
change++;
}
lp->state = FC_PORTDB_STATE_VALID;
if (change) {
isp_async(isp, ISPASYNC_DEV_CHANGED, chan, lp);
lp->portid = lp->new_portid;
lp->prli_word3 = lp->new_prli_word3;
} else {
isp_prt(isp, ISP_LOGTINFO,
"Chan %d WWPN 0x%016llx PortID 0x%06x "
"handle 0x%x reentered",
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
chan, (unsigned long long) lp->port_wwn,
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
lp->portid, lp->handle);
isp_async(isp, ISPASYNC_DEV_STAYED, chan, lp);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
return;
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
}
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
/* Search for room to insert new record. */
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
for (i = 0; i < MAX_FC_TARG; i++) {
if (fcp->portdb[i].state == FC_PORTDB_STATE_NIL)
break;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
if (i >= MAX_FC_TARG) {
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
isp_prt(isp, ISP_LOGTINFO|ISP_LOGWARN,
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
"Chan %d WWPN 0x%016llx PortID 0x%06x handle 0x%x "
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
"-- no room in port database",
Improve INOTs handling for 24xx and above chips.
2015-10-23 16:54:24 +00:00
chan, (unsigned long long) wwpn, s_id, nphdl);
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
if (isp->isp_dblev & (ISP_LOGTINFO|ISP_LOGWARN))
isp_dump_portdb(isp, chan);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
return;
}
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
/* Insert new record and mark it valid. */
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
lp = &fcp->portdb[i];
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_MEMZERO(lp, sizeof (fcportdb_t));
lp->handle = nphdl;
lp->portid = s_id;
Improve INOTs handling for 24xx and above chips.
2015-10-23 16:54:24 +00:00
lp->port_wwn = wwpn;
lp->node_wwn = wwnn;
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
lp->prli_word3 = (prli_params != 0) ? prli_params : PRLI_WD3_INITIATOR_FUNCTION;
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
lp->state = FC_PORTDB_STATE_VALID;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
Rewrite port database handling for target mode. Previous implementation was too fragile to initiator parameters changes. In case of port role change it could not survive different handle assigned to the same initiator by firmware, even though initiator was logged out. The new implementation should be more resillient to this kind of problems, trying to work in any situation and only warn user about suspisious events. MFC after: 1 week Sponsored by: iXsystems, Inc.
2015-06-21 14:18:54 +00:00
isp_gen_role_str(buf, sizeof (buf), lp->prli_word3);
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
isp_prt(isp, ISP_LOGTINFO, "Chan %d WWPN 0x%016llx "
"PortID 0x%06x handle 0x%x vtgt %d %s added", chan,
Improve INOTs handling for 24xx and above chips.
2015-10-23 16:54:24 +00:00
(unsigned long long) wwpn, s_id, nphdl, i, buf);
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
/* Notify above levels about new port arrival. */
isp_async(isp, ISPASYNC_DEV_ARRIVED, chan, lp);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
/*
* Remove a target device to the port database
*/
void
Improve INOTs handling for 24xx and above chips.
2015-10-23 16:54:24 +00:00
isp_del_wwn_entry(ispsoftc_t *isp, int chan, uint64_t wwpn, uint16_t nphdl, uint32_t s_id)
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
{
fcparam *fcp;
fcportdb_t *lp;
if (nphdl >= MAX_NPORT_HANDLE) {
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
isp_prt(isp, ISP_LOGWARN, "Chan %d WWPN 0x%016llx PortID 0x%06x bad handle 0x%x",
Improve INOTs handling for 24xx and above chips.
2015-10-23 16:54:24 +00:00
chan, (unsigned long long) wwpn, s_id, nphdl);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
return;
}
fcp = FCPARAM(isp, chan);
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
if (isp_find_pdb_by_handle(isp, chan, nphdl, &lp) == 0) {
isp_prt(isp, ISP_LOGWARN, "Chan %d WWPN 0x%016llx PortID 0x%06x handle 0x%x cannot be found to be deleted",
Improve INOTs handling for 24xx and above chips.
2015-10-23 16:54:24 +00:00
chan, (unsigned long long) wwpn, s_id, nphdl);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_dump_portdb(isp, chan);
return;
}
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
isp_prt(isp, ISP_LOGTINFO, "Chan %d WWPN 0x%016llx PortID 0x%06x handle 0x%x vtgt %d deleted",
chan, (unsigned long long) lp->port_wwn, lp->portid, nphdl, FC_PORTDB_TGT(isp, chan, lp));
lp->state = FC_PORTDB_STATE_NIL;
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
/* Notify above levels about gone port. */
isp_async(isp, ISPASYNC_DEV_GONE, chan, lp);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
void
isp_del_all_wwn_entries(ispsoftc_t *isp, int chan)
{
fcparam *fcp;
int i;
if (!IS_FC(isp)) {
return;
}
/*
* Handle iterations over all channels via recursion
*/
if (chan == ISP_NOCHAN) {
for (chan = 0; chan < isp->isp_nchan; chan++) {
isp_del_all_wwn_entries(isp, chan);
}
return;
}
if (chan > isp->isp_nchan) {
return;
}
fcp = FCPARAM(isp, chan);
if (fcp == NULL) {
return;
}
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
for (i = 0; i < MAX_FC_TARG; i++) {
fcportdb_t *lp = &fcp->portdb[i];
if (lp->state != FC_PORTDB_STATE_NIL)
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_del_wwn_entry(isp, chan, lp->port_wwn, lp->handle, lp->portid);
}
}
void
isp_del_wwn_entries(ispsoftc_t *isp, isp_notify_t *mp)
{
fcportdb_t *lp;
/*
* Handle iterations over all channels via recursion
*/
if (mp->nt_channel == ISP_NOCHAN) {
for (mp->nt_channel = 0; mp->nt_channel < isp->isp_nchan; mp->nt_channel++) {
isp_del_wwn_entries(isp, mp);
}
mp->nt_channel = ISP_NOCHAN;
return;
}
/*
* We have an entry which is only partially identified.
*
* It's only known by WWN, N-Port handle, or Port ID.
* We need to find the actual entry so we can delete it.
*/
if (mp->nt_nphdl != NIL_HANDLE) {
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
if (isp_find_pdb_by_handle(isp, mp->nt_channel, mp->nt_nphdl, &lp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_del_wwn_entry(isp, mp->nt_channel, lp->port_wwn, lp->handle, lp->portid);
return;
}
}
Fix target mode with fabric for pre-24xx chips. For those chips we are not receiving login events, adding initiators based on ATIO requests. But there is no port ID in that structure, so in fabric mode we have to explicitly fetch it from firmware to be able to do normal scan after that.
2015-11-21 21:01:00 +00:00
if (VALID_INI(mp->nt_wwn)) {
Another round of port scanner rewrite. This change simplifies and unifies port adding/updating for loop and fabric scanners. It also fixes problems with scanning restarts due to concurrent port databases changes. It also fixes many cosmetic issues.
2015-11-19 17:43:47 +00:00
if (isp_find_pdb_by_wwpn(isp, mp->nt_channel, mp->nt_wwn, &lp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_del_wwn_entry(isp, mp->nt_channel, lp->port_wwn, lp->handle, lp->portid);
return;
}
}
Fix target mode with fabric for pre-24xx chips. For those chips we are not receiving login events, adding initiators based on ATIO requests. But there is no port ID in that structure, so in fabric mode we have to explicitly fetch it from firmware to be able to do normal scan after that.
2015-11-21 21:01:00 +00:00
if (VALID_PORT(mp->nt_sid)) {
Another round of port scanner rewrite. This change simplifies and unifies port adding/updating for loop and fabric scanners. It also fixes problems with scanning restarts due to concurrent port databases changes. It also fixes many cosmetic issues.
2015-11-19 17:43:47 +00:00
if (isp_find_pdb_by_portid(isp, mp->nt_channel, mp->nt_sid, &lp)) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_del_wwn_entry(isp, mp->nt_channel, lp->port_wwn, lp->handle, lp->portid);
return;
}
}
Unify port database use for target and initiator roles. Aside from cleaner and more consistent code, this allows ports to be both target and initiator same time, and easily switch from any role to any. Sponsored by: iXsystems, Inc.
2015-07-13 15:11:05 +00:00
isp_prt(isp, ISP_LOGWARN, "Chan %d unable to find entry to delete WWPN 0x%016jx PortID 0x%06x handle 0x%x",
mp->nt_channel, mp->nt_wwn, mp->nt_sid, mp->nt_nphdl);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_atio2(ispsoftc_t *isp, at2_entry_t *src, at2_entry_t *dst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
int i;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_hdr(isp, &src->at_header, &dst->at_header);
ISP_IOXPUT_32(isp, src->at_reserved, &dst->at_reserved);
ISP_IOXPUT_8(isp, src->at_lun, &dst->at_lun);
ISP_IOXPUT_8(isp, src->at_iid, &dst->at_iid);
ISP_IOXPUT_16(isp, src->at_rxid, &dst->at_rxid);
ISP_IOXPUT_16(isp, src->at_flags, &dst->at_flags);
ISP_IOXPUT_16(isp, src->at_status, &dst->at_status);
ISP_IOXPUT_8(isp, src->at_crn, &dst->at_crn);
ISP_IOXPUT_8(isp, src->at_taskcodes, &dst->at_taskcodes);
ISP_IOXPUT_8(isp, src->at_taskflags, &dst->at_taskflags);
ISP_IOXPUT_8(isp, src->at_execodes, &dst->at_execodes);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
for (i = 0; i < ATIO2_CDBLEN; i++) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_8(isp, src->at_cdb[i], &dst->at_cdb[i]);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_32(isp, src->at_datalen, &dst->at_datalen);
ISP_IOXPUT_16(isp, src->at_scclun, &dst->at_scclun);
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
for (i = 0; i < 4; i++) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->at_wwpn[i], &dst->at_wwpn[i]);
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
}
for (i = 0; i < 6; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_16(isp, src->at_reserved2[i], &dst->at_reserved2[i]);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->at_oxid, &dst->at_oxid);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_atio2e(ispsoftc_t *isp, at2e_entry_t *src, at2e_entry_t *dst)
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
{
int i;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_hdr(isp, &src->at_header, &dst->at_header);
ISP_IOXPUT_32(isp, src->at_reserved, &dst->at_reserved);
ISP_IOXPUT_16(isp, src->at_iid, &dst->at_iid);
ISP_IOXPUT_16(isp, src->at_rxid, &dst->at_rxid);
ISP_IOXPUT_16(isp, src->at_flags, &dst->at_flags);
ISP_IOXPUT_16(isp, src->at_status, &dst->at_status);
ISP_IOXPUT_8(isp, src->at_crn, &dst->at_crn);
ISP_IOXPUT_8(isp, src->at_taskcodes, &dst->at_taskcodes);
ISP_IOXPUT_8(isp, src->at_taskflags, &dst->at_taskflags);
ISP_IOXPUT_8(isp, src->at_execodes, &dst->at_execodes);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
for (i = 0; i < ATIO2_CDBLEN; i++) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_8(isp, src->at_cdb[i], &dst->at_cdb[i]);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_32(isp, src->at_datalen, &dst->at_datalen);
ISP_IOXPUT_16(isp, src->at_scclun, &dst->at_scclun);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
for (i = 0; i < 4; i++) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->at_wwpn[i], &dst->at_wwpn[i]);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
for (i = 0; i < 6; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_16(isp, src->at_reserved2[i], &dst->at_reserved2[i]);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->at_oxid, &dst->at_oxid);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_atio2(ispsoftc_t *isp, at2_entry_t *src, at2_entry_t *dst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
int i;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_hdr(isp, &src->at_header, &dst->at_header);
ISP_IOXGET_32(isp, &src->at_reserved, dst->at_reserved);
ISP_IOXGET_8(isp, &src->at_lun, dst->at_lun);
ISP_IOXGET_8(isp, &src->at_iid, dst->at_iid);
ISP_IOXGET_16(isp, &src->at_rxid, dst->at_rxid);
ISP_IOXGET_16(isp, &src->at_flags, dst->at_flags);
ISP_IOXGET_16(isp, &src->at_status, dst->at_status);
ISP_IOXGET_8(isp, &src->at_crn, dst->at_crn);
ISP_IOXGET_8(isp, &src->at_taskcodes, dst->at_taskcodes);
ISP_IOXGET_8(isp, &src->at_taskflags, dst->at_taskflags);
ISP_IOXGET_8(isp, &src->at_execodes, dst->at_execodes);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
for (i = 0; i < ATIO2_CDBLEN; i++) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_8(isp, &src->at_cdb[i], dst->at_cdb[i]);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_32(isp, &src->at_datalen, dst->at_datalen);
ISP_IOXGET_16(isp, &src->at_scclun, dst->at_scclun);
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
for (i = 0; i < 4; i++) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_16(isp, &src->at_wwpn[i], dst->at_wwpn[i]);
+ A variety of 23XX changes: disable MWI on 2300 based on function code, set an 'isp_port' for the 2312- it's a separate instance, but the NVRAM is shared, and the second port's NVRAM is at offset 256. + Enable RIO operation for LVD SCSI cards. This makes a *big* difference as even under reasonable load we get batched completions of about 30 commands at a time on, say, an ISP1080. + Do 'continuation' mailbox commands- this allows us to specify a work area within the softc and 'continue' repeated mailbox commands. This is more or less on an ad hoc basis and is currently only used for firmware loading (which f/w now loads substantially faster becuase the calling thread is only woken when all the f/w words are loaded- not for each one of the 40000 f/w words that gets loaded). + If we're about to return from isp_intr with a 'bogus interrupt' indication, and we're not a 23XX card, check to see whether the semaphore register is currently *2* (not *1* as it should be) and whether there's an async completion sitting in outgoing mailbox0. This seems to capture cases of lost fast posting and RIO interrupts that the 12160 && 1080 have been known to pump out under extreme load (extreme, as in > 250 active commands). + FC_SCRATCH_ACQUIRE/FC_SCRATCH_RELEASE macros. + Endian correct swizzle/unswizzle of an ATIO2 that has a WWPN in it. MFC after: 1 week
2002-02-04 21:04:25 +00:00
}
for (i = 0; i < 6; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_16(isp, &src->at_reserved2[i], dst->at_reserved2[i]);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_16(isp, &src->at_oxid, dst->at_oxid);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_atio2e(ispsoftc_t *isp, at2e_entry_t *src, at2e_entry_t *dst)
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
{
int i;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_hdr(isp, &src->at_header, &dst->at_header);
ISP_IOXGET_32(isp, &src->at_reserved, dst->at_reserved);
ISP_IOXGET_16(isp, &src->at_iid, dst->at_iid);
ISP_IOXGET_16(isp, &src->at_rxid, dst->at_rxid);
ISP_IOXGET_16(isp, &src->at_flags, dst->at_flags);
ISP_IOXGET_16(isp, &src->at_status, dst->at_status);
ISP_IOXGET_8(isp, &src->at_crn, dst->at_crn);
ISP_IOXGET_8(isp, &src->at_taskcodes, dst->at_taskcodes);
ISP_IOXGET_8(isp, &src->at_taskflags, dst->at_taskflags);
ISP_IOXGET_8(isp, &src->at_execodes, dst->at_execodes);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
for (i = 0; i < ATIO2_CDBLEN; i++) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_8(isp, &src->at_cdb[i], dst->at_cdb[i]);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_32(isp, &src->at_datalen, dst->at_datalen);
ISP_IOXGET_16(isp, &src->at_scclun, dst->at_scclun);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
for (i = 0; i < 4; i++) {
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_16(isp, &src->at_wwpn[i], dst->at_wwpn[i]);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
for (i = 0; i < 6; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_16(isp, &src->at_reserved2[i], dst->at_reserved2[i]);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_16(isp, &src->at_oxid, dst->at_oxid);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_atio7(ispsoftc_t *isp, at7_entry_t *src, at7_entry_t *dst)
{
ISP_IOXGET_8(isp, &src->at_type, dst->at_type);
ISP_IOXGET_8(isp, &src->at_count, dst->at_count);
ISP_IOXGET_16(isp, &src->at_ta_len, dst->at_ta_len);
ISP_IOXGET_32(isp, &src->at_rxid, dst->at_rxid);
isp_get_fc_hdr(isp, &src->at_hdr, &dst->at_hdr);
isp_get_fcp_cmnd_iu(isp, &src->at_cmnd, &dst->at_cmnd);
}
void
isp_put_ctio2(ispsoftc_t *isp, ct2_entry_t *src, ct2_entry_t *dst)
{
int i;
isp_put_hdr(isp, &src->ct_header, &dst->ct_header);
ISP_IOXPUT_32(isp, src->ct_syshandle, &dst->ct_syshandle);
ISP_IOXPUT_8(isp, src->ct_lun, &dst->ct_lun);
ISP_IOXPUT_8(isp, src->ct_iid, &dst->ct_iid);
ISP_IOXPUT_16(isp, src->ct_rxid, &dst->ct_rxid);
ISP_IOXPUT_16(isp, src->ct_flags, &dst->ct_flags);
ISP_IOXPUT_16(isp, src->ct_timeout, &dst->ct_timeout);
ISP_IOXPUT_16(isp, src->ct_seg_count, &dst->ct_seg_count);
ISP_IOXPUT_32(isp, src->ct_resid, &dst->ct_resid);
ISP_IOXPUT_32(isp, src->ct_reloff, &dst->ct_reloff);
if ((src->ct_flags & CT2_FLAG_MMASK) == CT2_FLAG_MODE0) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->rsp.m0._reserved, &dst->rsp.m0._reserved);
ISP_IOXPUT_16(isp, src->rsp.m0._reserved2, &dst->rsp.m0._reserved2);
ISP_IOXPUT_16(isp, src->rsp.m0.ct_scsi_status, &dst->rsp.m0.ct_scsi_status);
ISP_IOXPUT_32(isp, src->rsp.m0.ct_xfrlen, &dst->rsp.m0.ct_xfrlen);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (src->ct_header.rqs_entry_type == RQSTYPE_CTIO2) {
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
for (i = 0; i < ISP_RQDSEG_T2; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->rsp.m0.u.ct_dataseg[i].ds_base, &dst->rsp.m0.u.ct_dataseg[i].ds_base);
ISP_IOXPUT_32(isp, src->rsp.m0.u.ct_dataseg[i].ds_count, &dst->rsp.m0.u.ct_dataseg[i].ds_count);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else if (src->ct_header.rqs_entry_type == RQSTYPE_CTIO3) {
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
for (i = 0; i < ISP_RQDSEG_T3; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->rsp.m0.u.ct_dataseg64[i].ds_base, &dst->rsp.m0.u.ct_dataseg64[i].ds_base);
ISP_IOXPUT_32(isp, src->rsp.m0.u.ct_dataseg64[i].ds_basehi, &dst->rsp.m0.u.ct_dataseg64[i].ds_basehi);
ISP_IOXPUT_32(isp, src->rsp.m0.u.ct_dataseg64[i].ds_count, &dst->rsp.m0.u.ct_dataseg64[i].ds_count);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else if (src->ct_header.rqs_entry_type == RQSTYPE_CTIO4) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_16(isp, src->rsp.m0.u.ct_dslist.ds_type, &dst->rsp.m0.u.ct_dslist.ds_type); ISP_IOXPUT_32(isp, src->rsp.m0.u.ct_dslist.ds_segment,
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
&dst->rsp.m0.u.ct_dslist.ds_segment);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->rsp.m0.u.ct_dslist.ds_base, &dst->rsp.m0.u.ct_dslist.ds_base);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else if ((src->ct_flags & CT2_FLAG_MMASK) == CT2_FLAG_MODE1) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_16(isp, src->rsp.m1._reserved, &dst->rsp.m1._reserved);
ISP_IOXPUT_16(isp, src->rsp.m1._reserved2, &dst->rsp.m1._reserved2);
ISP_IOXPUT_16(isp, src->rsp.m1.ct_senselen, &dst->rsp.m1.ct_senselen);
ISP_IOXPUT_16(isp, src->rsp.m1.ct_scsi_status, &dst->rsp.m1.ct_scsi_status);
ISP_IOXPUT_16(isp, src->rsp.m1.ct_resplen, &dst->rsp.m1.ct_resplen);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
for (i = 0; i < MAXRESPLEN; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_8(isp, src->rsp.m1.ct_resp[i], &dst->rsp.m1.ct_resp[i]);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->rsp.m2._reserved, &dst->rsp.m2._reserved);
ISP_IOXPUT_16(isp, src->rsp.m2._reserved2, &dst->rsp.m2._reserved2);
ISP_IOXPUT_16(isp, src->rsp.m2._reserved3, &dst->rsp.m2._reserved3);
ISP_IOXPUT_32(isp, src->rsp.m2.ct_datalen, &dst->rsp.m2.ct_datalen);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (src->ct_header.rqs_entry_type == RQSTYPE_CTIO2) {
ISP_IOXPUT_32(isp, src->rsp.m2.u.ct_fcp_rsp_iudata_32.ds_base, &dst->rsp.m2.u.ct_fcp_rsp_iudata_32.ds_base);
ISP_IOXPUT_32(isp, src->rsp.m2.u.ct_fcp_rsp_iudata_32.ds_count, &dst->rsp.m2.u.ct_fcp_rsp_iudata_32.ds_count);
} else {
ISP_IOXPUT_32(isp, src->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_base, &dst->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_base);
ISP_IOXPUT_32(isp, src->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_basehi, &dst->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_basehi);
ISP_IOXPUT_32(isp, src->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_count, &dst->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_count);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
}
void
isp_put_ctio2e(ispsoftc_t *isp, ct2e_entry_t *src, ct2e_entry_t *dst)
{
int i;
isp_put_hdr(isp, &src->ct_header, &dst->ct_header);
ISP_IOXPUT_32(isp, src->ct_syshandle, &dst->ct_syshandle);
ISP_IOXPUT_16(isp, src->ct_iid, &dst->ct_iid);
ISP_IOXPUT_16(isp, src->ct_rxid, &dst->ct_rxid);
ISP_IOXPUT_16(isp, src->ct_flags, &dst->ct_flags);
ISP_IOXPUT_16(isp, src->ct_timeout, &dst->ct_timeout);
ISP_IOXPUT_16(isp, src->ct_seg_count, &dst->ct_seg_count);
ISP_IOXPUT_32(isp, src->ct_resid, &dst->ct_resid);
ISP_IOXPUT_32(isp, src->ct_reloff, &dst->ct_reloff);
if ((src->ct_flags & CT2_FLAG_MMASK) == CT2_FLAG_MODE0) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->rsp.m0._reserved, &dst->rsp.m0._reserved);
ISP_IOXPUT_16(isp, src->rsp.m0._reserved2, &dst->rsp.m0._reserved2);
ISP_IOXPUT_16(isp, src->rsp.m0.ct_scsi_status, &dst->rsp.m0.ct_scsi_status);
ISP_IOXPUT_32(isp, src->rsp.m0.ct_xfrlen, &dst->rsp.m0.ct_xfrlen);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (src->ct_header.rqs_entry_type == RQSTYPE_CTIO2) {
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
for (i = 0; i < ISP_RQDSEG_T2; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->rsp.m0.u.ct_dataseg[i].ds_base, &dst->rsp.m0.u.ct_dataseg[i].ds_base);
ISP_IOXPUT_32(isp, src->rsp.m0.u.ct_dataseg[i].ds_count, &dst->rsp.m0.u.ct_dataseg[i].ds_count);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else if (src->ct_header.rqs_entry_type == RQSTYPE_CTIO3) {
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
for (i = 0; i < ISP_RQDSEG_T3; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->rsp.m0.u.ct_dataseg64[i].ds_base, &dst->rsp.m0.u.ct_dataseg64[i].ds_base);
ISP_IOXPUT_32(isp, src->rsp.m0.u.ct_dataseg64[i].ds_basehi, &dst->rsp.m0.u.ct_dataseg64[i].ds_basehi);
ISP_IOXPUT_32(isp, src->rsp.m0.u.ct_dataseg64[i].ds_count, &dst->rsp.m0.u.ct_dataseg64[i].ds_count);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
} else if (src->ct_header.rqs_entry_type == RQSTYPE_CTIO4) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_16(isp, src->rsp.m0.u.ct_dslist.ds_type, &dst->rsp.m0.u.ct_dslist.ds_type);
ISP_IOXPUT_32(isp, src->rsp.m0.u.ct_dslist.ds_segment, &dst->rsp.m0.u.ct_dslist.ds_segment);
ISP_IOXPUT_32(isp, src->rsp.m0.u.ct_dslist.ds_base, &dst->rsp.m0.u.ct_dslist.ds_base);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
} else if ((src->ct_flags & CT2_FLAG_MMASK) == CT2_FLAG_MODE1) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_16(isp, src->rsp.m1._reserved, &dst->rsp.m1._reserved);
ISP_IOXPUT_16(isp, src->rsp.m1._reserved2, &dst->rsp.m1._reserved2);
ISP_IOXPUT_16(isp, src->rsp.m1.ct_senselen, &dst->rsp.m1.ct_senselen);
ISP_IOXPUT_16(isp, src->rsp.m1.ct_scsi_status, &dst->rsp.m1.ct_scsi_status);
ISP_IOXPUT_16(isp, src->rsp.m1.ct_resplen, &dst->rsp.m1.ct_resplen);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
for (i = 0; i < MAXRESPLEN; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_8(isp, src->rsp.m1.ct_resp[i], &dst->rsp.m1.ct_resp[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->rsp.m2._reserved, &dst->rsp.m2._reserved);
ISP_IOXPUT_16(isp, src->rsp.m2._reserved2, &dst->rsp.m2._reserved2);
ISP_IOXPUT_16(isp, src->rsp.m2._reserved3, &dst->rsp.m2._reserved3);
ISP_IOXPUT_32(isp, src->rsp.m2.ct_datalen, &dst->rsp.m2.ct_datalen);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (src->ct_header.rqs_entry_type == RQSTYPE_CTIO2) {
ISP_IOXPUT_32(isp, src->rsp.m2.u.ct_fcp_rsp_iudata_32.ds_base, &dst->rsp.m2.u.ct_fcp_rsp_iudata_32.ds_base);
ISP_IOXPUT_32(isp, src->rsp.m2.u.ct_fcp_rsp_iudata_32.ds_count, &dst->rsp.m2.u.ct_fcp_rsp_iudata_32.ds_count);
} else {
ISP_IOXPUT_32(isp, src->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_base, &dst->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_base);
ISP_IOXPUT_32(isp, src->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_basehi, &dst->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_basehi);
ISP_IOXPUT_32(isp, src->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_count, &dst->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_count);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
}
void
isp_put_ctio7(ispsoftc_t *isp, ct7_entry_t *src, ct7_entry_t *dst)
{
int i;
isp_put_hdr(isp, &src->ct_header, &dst->ct_header);
ISP_IOXPUT_32(isp, src->ct_syshandle, &dst->ct_syshandle);
ISP_IOXPUT_16(isp, src->ct_nphdl, &dst->ct_nphdl);
ISP_IOXPUT_16(isp, src->ct_timeout, &dst->ct_timeout);
ISP_IOXPUT_16(isp, src->ct_seg_count, &dst->ct_seg_count);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_8(isp, src->ct_vpidx, &dst->ct_vpidx);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_8(isp, src->ct_xflags, &dst->ct_xflags);
ISP_IOXPUT_16(isp, src->ct_iid_lo, &dst->ct_iid_lo);
ISP_IOXPUT_8(isp, src->ct_iid_hi, &dst->ct_iid_hi);
ISP_IOXPUT_8(isp, src->ct_reserved, &dst->ct_reserved);
ISP_IOXPUT_32(isp, src->ct_rxid, &dst->ct_rxid);
ISP_IOXPUT_16(isp, src->ct_senselen, &dst->ct_senselen);
ISP_IOXPUT_16(isp, src->ct_flags, &dst->ct_flags);
ISP_IOXPUT_32(isp, src->ct_resid, &dst->ct_resid);
ISP_IOXPUT_16(isp, src->ct_oxid, &dst->ct_oxid);
ISP_IOXPUT_16(isp, src->ct_scsi_status, &dst->ct_scsi_status);
if ((dst->ct_flags & CT7_FLAG_MMASK) == CT7_FLAG_MODE0) {
ISP_IOXPUT_32(isp, src->rsp.m0.reloff, &dst->rsp.m0.reloff);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->rsp.m0.reserved0, &dst->rsp.m0.reserved0);
ISP_IOXPUT_32(isp, src->rsp.m0.ct_xfrlen, &dst->rsp.m0.ct_xfrlen);
ISP_IOXPUT_32(isp, src->rsp.m0.reserved1, &dst->rsp.m0.reserved1);
ISP_IOXPUT_32(isp, src->rsp.m0.ds.ds_base, &dst->rsp.m0.ds.ds_base);
ISP_IOXPUT_32(isp, src->rsp.m0.ds.ds_basehi, &dst->rsp.m0.ds.ds_basehi);
ISP_IOXPUT_32(isp, src->rsp.m0.ds.ds_count, &dst->rsp.m0.ds.ds_count);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else if ((dst->ct_flags & CT7_FLAG_MMASK) == CT7_FLAG_MODE1) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
uint32_t *a, *b;
ISP_IOXPUT_16(isp, src->rsp.m1.ct_resplen, &dst->rsp.m1.ct_resplen);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->rsp.m1.reserved, &dst->rsp.m1.reserved);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
a = (uint32_t *) src->rsp.m1.ct_resp;
b = (uint32_t *) dst->rsp.m1.ct_resp;
for (i = 0; i < (ASIZE(src->rsp.m1.ct_resp) >> 2); i++) {
*b++ = ISP_SWAP32(isp, *a++);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->rsp.m2.reserved0, &dst->rsp.m2.reserved0);
ISP_IOXPUT_32(isp, src->rsp.m2.reserved1, &dst->rsp.m2.reserved1);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
ISP_IOXPUT_32(isp, src->rsp.m2.ct_datalen, &dst->rsp.m2.ct_datalen);
ISP_IOXPUT_32(isp, src->rsp.m2.reserved2, &dst->rsp.m2.reserved2);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_32(isp, src->rsp.m2.ct_fcp_rsp_iudata.ds_base, &dst->rsp.m2.ct_fcp_rsp_iudata.ds_base);
ISP_IOXPUT_32(isp, src->rsp.m2.ct_fcp_rsp_iudata.ds_basehi, &dst->rsp.m2.ct_fcp_rsp_iudata.ds_basehi);
ISP_IOXPUT_32(isp, src->rsp.m2.ct_fcp_rsp_iudata.ds_count, &dst->rsp.m2.ct_fcp_rsp_iudata.ds_count);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
}
void
isp_get_ctio2(ispsoftc_t *isp, ct2_entry_t *src, ct2_entry_t *dst)
{
int i;
isp_get_hdr(isp, &src->ct_header, &dst->ct_header);
ISP_IOXGET_32(isp, &src->ct_syshandle, dst->ct_syshandle);
ISP_IOXGET_8(isp, &src->ct_lun, dst->ct_lun);
ISP_IOXGET_8(isp, &src->ct_iid, dst->ct_iid);
ISP_IOXGET_16(isp, &src->ct_rxid, dst->ct_rxid);
ISP_IOXGET_16(isp, &src->ct_flags, dst->ct_flags);
ISP_IOXGET_16(isp, &src->ct_status, dst->ct_status);
ISP_IOXGET_16(isp, &src->ct_timeout, dst->ct_timeout);
ISP_IOXGET_16(isp, &src->ct_seg_count, dst->ct_seg_count);
ISP_IOXGET_32(isp, &src->ct_reloff, dst->ct_reloff);
ISP_IOXGET_32(isp, &src->ct_resid, dst->ct_resid);
if ((dst->ct_flags & CT2_FLAG_MMASK) == CT2_FLAG_MODE0) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_32(isp, &src->rsp.m0._reserved, dst->rsp.m0._reserved);
ISP_IOXGET_16(isp, &src->rsp.m0._reserved2, dst->rsp.m0._reserved2);
ISP_IOXGET_16(isp, &src->rsp.m0.ct_scsi_status, dst->rsp.m0.ct_scsi_status);
ISP_IOXGET_32(isp, &src->rsp.m0.ct_xfrlen, dst->rsp.m0.ct_xfrlen);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (dst->ct_header.rqs_entry_type == RQSTYPE_CTIO2) {
for (i = 0; i < ISP_RQDSEG_T2; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_32(isp, &src->rsp.m0.u.ct_dataseg[i].ds_base, dst->rsp.m0.u.ct_dataseg[i].ds_base);
ISP_IOXGET_32(isp, &src->rsp.m0.u.ct_dataseg[i].ds_count, dst->rsp.m0.u.ct_dataseg[i].ds_count);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
} else if (dst->ct_header.rqs_entry_type == RQSTYPE_CTIO3) {
for (i = 0; i < ISP_RQDSEG_T3; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_32(isp, &src->rsp.m0.u.ct_dataseg64[i].ds_base, dst->rsp.m0.u.ct_dataseg64[i].ds_base);
ISP_IOXGET_32(isp, &src->rsp.m0.u.ct_dataseg64[i].ds_basehi, dst->rsp.m0.u.ct_dataseg64[i].ds_basehi);
ISP_IOXGET_32(isp, &src->rsp.m0.u.ct_dataseg64[i].ds_count, dst->rsp.m0.u.ct_dataseg64[i].ds_count);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else if (dst->ct_header.rqs_entry_type == RQSTYPE_CTIO4) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_16(isp, &src->rsp.m0.u.ct_dslist.ds_type, dst->rsp.m0.u.ct_dslist.ds_type);
ISP_IOXGET_32(isp, &src->rsp.m0.u.ct_dslist.ds_segment, dst->rsp.m0.u.ct_dslist.ds_segment);
ISP_IOXGET_32(isp, &src->rsp.m0.u.ct_dslist.ds_base, dst->rsp.m0.u.ct_dslist.ds_base);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else if ((dst->ct_flags & CT2_FLAG_MMASK) == CT2_FLAG_MODE1) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_16(isp, &src->rsp.m1._reserved, dst->rsp.m1._reserved);
ISP_IOXGET_16(isp, &src->rsp.m1._reserved2, dst->rsp.m1._reserved2);
ISP_IOXGET_16(isp, &src->rsp.m1.ct_senselen, dst->rsp.m1.ct_senselen);
ISP_IOXGET_16(isp, &src->rsp.m1.ct_scsi_status, dst->rsp.m1.ct_scsi_status);
ISP_IOXGET_16(isp, &src->rsp.m1.ct_resplen, dst->rsp.m1.ct_resplen);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
for (i = 0; i < MAXRESPLEN; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->rsp.m1.ct_resp[i], dst->rsp.m1.ct_resp[i]);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_32(isp, &src->rsp.m2._reserved, dst->rsp.m2._reserved);
ISP_IOXGET_16(isp, &src->rsp.m2._reserved2, dst->rsp.m2._reserved2);
ISP_IOXGET_16(isp, &src->rsp.m2._reserved3, dst->rsp.m2._reserved3);
ISP_IOXGET_32(isp, &src->rsp.m2.ct_datalen, dst->rsp.m2.ct_datalen);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (src->ct_header.rqs_entry_type == RQSTYPE_CTIO2) {
ISP_IOXGET_32(isp, &src->rsp.m2.u.ct_fcp_rsp_iudata_32.ds_base, dst->rsp.m2.u.ct_fcp_rsp_iudata_32.ds_base);
ISP_IOXGET_32(isp, &src->rsp.m2.u.ct_fcp_rsp_iudata_32.ds_count, dst->rsp.m2.u.ct_fcp_rsp_iudata_32.ds_count);
} else {
ISP_IOXGET_32(isp, &src->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_base, dst->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_base);
ISP_IOXGET_32(isp, &src->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_basehi, dst->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_basehi);
ISP_IOXGET_32(isp, &src->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_count, dst->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_count);
}
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_ctio2e(ispsoftc_t *isp, ct2e_entry_t *src, ct2e_entry_t *dst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
int i;
isp_get_hdr(isp, &src->ct_header, &dst->ct_header);
ISP_IOXGET_32(isp, &src->ct_syshandle, dst->ct_syshandle);
ISP_IOXGET_16(isp, &src->ct_iid, dst->ct_iid);
ISP_IOXGET_16(isp, &src->ct_rxid, dst->ct_rxid);
ISP_IOXGET_16(isp, &src->ct_flags, dst->ct_flags);
ISP_IOXGET_16(isp, &src->ct_status, dst->ct_status);
ISP_IOXGET_16(isp, &src->ct_timeout, dst->ct_timeout);
ISP_IOXGET_16(isp, &src->ct_seg_count, dst->ct_seg_count);
ISP_IOXGET_32(isp, &src->ct_reloff, dst->ct_reloff);
ISP_IOXGET_32(isp, &src->ct_resid, dst->ct_resid);
if ((dst->ct_flags & CT2_FLAG_MMASK) == CT2_FLAG_MODE0) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_32(isp, &src->rsp.m0._reserved, dst->rsp.m0._reserved);
ISP_IOXGET_16(isp, &src->rsp.m0._reserved2, dst->rsp.m0._reserved2);
ISP_IOXGET_16(isp, &src->rsp.m0.ct_scsi_status, dst->rsp.m0.ct_scsi_status);
ISP_IOXGET_32(isp, &src->rsp.m0.ct_xfrlen, dst->rsp.m0.ct_xfrlen);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
if (src->ct_header.rqs_entry_type == RQSTYPE_CTIO2) {
for (i = 0; i < ISP_RQDSEG_T2; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_32(isp, &src->rsp.m0.u.ct_dataseg[i].ds_base, dst->rsp.m0.u.ct_dataseg[i].ds_base);
ISP_IOXGET_32(isp, &src->rsp.m0.u.ct_dataseg[i].ds_count, dst->rsp.m0.u.ct_dataseg[i].ds_count);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
} else if (dst->ct_header.rqs_entry_type == RQSTYPE_CTIO3) {
for (i = 0; i < ISP_RQDSEG_T3; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_32(isp, &src->rsp.m0.u.ct_dataseg64[i].ds_base, dst->rsp.m0.u.ct_dataseg64[i].ds_base);
ISP_IOXGET_32(isp, &src->rsp.m0.u.ct_dataseg64[i].ds_basehi, dst->rsp.m0.u.ct_dataseg64[i].ds_basehi);
ISP_IOXGET_32(isp, &src->rsp.m0.u.ct_dataseg64[i].ds_count, dst->rsp.m0.u.ct_dataseg64[i].ds_count);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
} else if (dst->ct_header.rqs_entry_type == RQSTYPE_CTIO4) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_16(isp, &src->rsp.m0.u.ct_dslist.ds_type, dst->rsp.m0.u.ct_dslist.ds_type);
ISP_IOXGET_32(isp, &src->rsp.m0.u.ct_dslist.ds_segment, dst->rsp.m0.u.ct_dslist.ds_segment);
ISP_IOXGET_32(isp, &src->rsp.m0.u.ct_dslist.ds_base, dst->rsp.m0.u.ct_dslist.ds_base);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
} else if ((dst->ct_flags & CT2_FLAG_MMASK) == CT2_FLAG_MODE1) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_16(isp, &src->rsp.m1._reserved, dst->rsp.m1._reserved);
ISP_IOXGET_16(isp, &src->rsp.m1._reserved2, dst->rsp.m1._reserved2);
ISP_IOXGET_16(isp, &src->rsp.m1.ct_senselen, dst->rsp.m1.ct_senselen);
ISP_IOXGET_16(isp, &src->rsp.m1.ct_scsi_status, dst->rsp.m1.ct_scsi_status);
ISP_IOXGET_16(isp, &src->rsp.m1.ct_resplen, dst->rsp.m1.ct_resplen);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
for (i = 0; i < MAXRESPLEN; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->rsp.m1.ct_resp[i], dst->rsp.m1.ct_resp[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_32(isp, &src->rsp.m2._reserved, dst->rsp.m2._reserved);
ISP_IOXGET_16(isp, &src->rsp.m2._reserved2, dst->rsp.m2._reserved2);
ISP_IOXGET_16(isp, &src->rsp.m2._reserved3, dst->rsp.m2._reserved3);
ISP_IOXGET_32(isp, &src->rsp.m2.ct_datalen, dst->rsp.m2.ct_datalen);
----------- MISC CHANGES Add a new async event- ISP_TARGET_NOTIFY_ACK, that will guarantee eventual delivery of a NOTIFY ACK. This is tons better than just ignoring the return from isp_notify_ack and hoping for the best. Clean up the lower level lun enable code to be a bit more sensible. Fix a botch in isp_endcmd which was messing up the sense data. Fix notify ack for SRR to use a sensible error code in the case of a reject. Clean up and make clear what kind of firmware we've loaded and what capabilities it has. ----------- FULL (252 byte) SENSE DATA In CTIOs for the ISP, there's only a limimted amount of space to load SENSE DATA for associated CHECK CONDITIONS (24 or 26 bytes). This makes it difficult to send full SENSE DATA that can be up to 252 bytes. Implement MODE 2 responses which have us build the FCP Response in system memory which the ISP will put onto the wire directly. On the initiator side, the same problem occurs in that a command status response only has a limited amount of space for SENSE DATA. This data is supplemented by status continuation responses that the ISP pushes onto the response queue after the status response. We now pull them all together so that full sense data can be returned to the periph driver. This is supported on 23XX, 24XX and 25XX cards. This is also preparation for doing >16 byte CDBs. ----------- FC TAPE Implement full FC-TAPE on both initiator and target mode side. This capability is driven by firmware loaded, board type, board NVRAM settings, or hint configuration options to enable or disable. This is supported for 23XX, 24XX and 25XX cards. On the initiator side, we pretty much just have to generate a command reference number for each command we send out. This is FCP-4 compliant in that we do this per ITL nexus to generate the allowed 1 thru 255 CRN. In order to support the target side of FC-TAPE, we now pay attention to more of the PRLI word 3 parameters which will tell us whether an initiator wants confirmed responses. While we're at it, we'll pay attention to the initiator view too and report it. On sending back CTIOs, we will notice whether the initiator wants confirmed responses and we'll set up flags to do so. If a response or data frame is lost the initiator sends us an SRR (Sequence Retransmit Request) ELS which shows up as an SRR notify and all outstanding CTIOs are nuked with SRR Received status. The SRR notify contains the offset that the initiator wants us to restart the data transfer from or to retransmit the response frame. If the ISP driver still has the CCB around for which the data segment or response applies, it will retransmit. However, we typically don't know about a lost data frame until we send the FCP Response and the initiator totes up counters for data moved and notices missing segments. In this case we've already completed the data CCBs already and sent themn back up to the periph driver. Because there's no really clean mechanism yet in CAM to handle this, a hack has been put into place to complete the CTIO CCB with the CAM_MESSAGE_RECV status which will have a MODIFY DATA POINTER extended message in it. The internal ISP target groks this and ctl(8) will be modified to deal with this as well. At any rate, the data is retransmitted and an an FCP response is sent. The whole point here is to successfully complete a command so that you don't have to depend on ULP (SCSI) to have to recover, which in the case of tape is not really possible (hence the name FC-TAPE). Sponsored by: Spectralogic MFC after: 1 month
2012-07-28 20:06:29 +00:00
if (src->ct_header.rqs_entry_type == RQSTYPE_CTIO2) {
ISP_IOXGET_32(isp, &src->rsp.m2.u.ct_fcp_rsp_iudata_32.ds_base, dst->rsp.m2.u.ct_fcp_rsp_iudata_32.ds_base);
ISP_IOXGET_32(isp, &src->rsp.m2.u.ct_fcp_rsp_iudata_32.ds_count, dst->rsp.m2.u.ct_fcp_rsp_iudata_32.ds_count);
} else {
ISP_IOXGET_32(isp, &src->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_base, dst->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_base);
ISP_IOXGET_32(isp, &src->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_basehi, dst->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_basehi);
ISP_IOXGET_32(isp, &src->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_count, dst->rsp.m2.u.ct_fcp_rsp_iudata_64.ds_count);
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_ctio7(ispsoftc_t *isp, ct7_entry_t *src, ct7_entry_t *dst)
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
{
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
int i;
isp_get_hdr(isp, &src->ct_header, &dst->ct_header);
ISP_IOXGET_32(isp, &src->ct_syshandle, dst->ct_syshandle);
ISP_IOXGET_16(isp, &src->ct_nphdl, dst->ct_nphdl);
ISP_IOXGET_16(isp, &src->ct_timeout, dst->ct_timeout);
ISP_IOXGET_16(isp, &src->ct_seg_count, dst->ct_seg_count);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->ct_vpidx, dst->ct_vpidx);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_8(isp, &src->ct_xflags, dst->ct_xflags);
ISP_IOXGET_16(isp, &src->ct_iid_lo, dst->ct_iid_lo);
ISP_IOXGET_8(isp, &src->ct_iid_hi, dst->ct_iid_hi);
ISP_IOXGET_8(isp, &src->ct_reserved, dst->ct_reserved);
ISP_IOXGET_32(isp, &src->ct_rxid, dst->ct_rxid);
ISP_IOXGET_16(isp, &src->ct_senselen, dst->ct_senselen);
ISP_IOXGET_16(isp, &src->ct_flags, dst->ct_flags);
ISP_IOXGET_32(isp, &src->ct_resid, dst->ct_resid);
ISP_IOXGET_16(isp, &src->ct_oxid, dst->ct_oxid);
ISP_IOXGET_16(isp, &src->ct_scsi_status, dst->ct_scsi_status);
if ((dst->ct_flags & CT7_FLAG_MMASK) == CT7_FLAG_MODE0) {
ISP_IOXGET_32(isp, &src->rsp.m0.reloff, dst->rsp.m0.reloff);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_32(isp, &src->rsp.m0.reserved0, dst->rsp.m0.reserved0);
ISP_IOXGET_32(isp, &src->rsp.m0.ct_xfrlen, dst->rsp.m0.ct_xfrlen);
ISP_IOXGET_32(isp, &src->rsp.m0.reserved1, dst->rsp.m0.reserved1);
ISP_IOXGET_32(isp, &src->rsp.m0.ds.ds_base, dst->rsp.m0.ds.ds_base);
ISP_IOXGET_32(isp, &src->rsp.m0.ds.ds_basehi, dst->rsp.m0.ds.ds_basehi);
ISP_IOXGET_32(isp, &src->rsp.m0.ds.ds_count, dst->rsp.m0.ds.ds_count);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
} else if ((dst->ct_flags & CT7_FLAG_MMASK) == CT7_FLAG_MODE1) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
uint32_t *a, *b;
ISP_IOXGET_16(isp, &src->rsp.m1.ct_resplen, dst->rsp.m1.ct_resplen);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_16(isp, &src->rsp.m1.reserved, dst->rsp.m1.reserved);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
a = (uint32_t *) src->rsp.m1.ct_resp;
b = (uint32_t *) dst->rsp.m1.ct_resp;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
for (i = 0; i < MAXRESPLEN_24XX; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->rsp.m1.ct_resp[i], dst->rsp.m1.ct_resp[i]);
}
for (i = 0; i < (ASIZE(src->rsp.m1.ct_resp) >> 2); i++) {
*b++ = ISP_SWAP32(isp, *a++);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
} else {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_32(isp, &src->rsp.m2.reserved0, dst->rsp.m2.reserved0);
ISP_IOXGET_32(isp, &src->rsp.m2.ct_datalen, dst->rsp.m2.ct_datalen);
ISP_IOXGET_32(isp, &src->rsp.m2.reserved1, dst->rsp.m2.reserved1);
ISP_IOXGET_32(isp, &src->rsp.m2.ct_fcp_rsp_iudata.ds_base, dst->rsp.m2.ct_fcp_rsp_iudata.ds_base);
ISP_IOXGET_32(isp, &src->rsp.m2.ct_fcp_rsp_iudata.ds_basehi, dst->rsp.m2.ct_fcp_rsp_iudata.ds_basehi);
ISP_IOXGET_32(isp, &src->rsp.m2.ct_fcp_rsp_iudata.ds_count, dst->rsp.m2.ct_fcp_rsp_iudata.ds_count);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_put_notify_fc(ispsoftc_t *isp, in_fcentry_t *src, in_fcentry_t *dst)
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
{
isp_put_hdr(isp, &src->in_header, &dst->in_header);
ISP_IOXPUT_32(isp, src->in_reserved, &dst->in_reserved);
ISP_IOXPUT_8(isp, src->in_lun, &dst->in_lun);
ISP_IOXPUT_8(isp, src->in_iid, &dst->in_iid);
ISP_IOXPUT_16(isp, src->in_scclun, &dst->in_scclun);
ISP_IOXPUT_32(isp, src->in_reserved2, &dst->in_reserved2);
ISP_IOXPUT_16(isp, src->in_status, &dst->in_status);
ISP_IOXPUT_16(isp, src->in_task_flags, &dst->in_task_flags);
ISP_IOXPUT_16(isp, src->in_seqid, &dst->in_seqid);
}
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_put_notify_fc_e(ispsoftc_t *isp, in_fcentry_e_t *src, in_fcentry_e_t *dst)
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
{
isp_put_hdr(isp, &src->in_header, &dst->in_header);
ISP_IOXPUT_32(isp, src->in_reserved, &dst->in_reserved);
ISP_IOXPUT_16(isp, src->in_iid, &dst->in_iid);
ISP_IOXPUT_16(isp, src->in_scclun, &dst->in_scclun);
ISP_IOXPUT_32(isp, src->in_reserved2, &dst->in_reserved2);
ISP_IOXPUT_16(isp, src->in_status, &dst->in_status);
ISP_IOXPUT_16(isp, src->in_task_flags, &dst->in_task_flags);
ISP_IOXPUT_16(isp, src->in_seqid, &dst->in_seqid);
}
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_put_notify_24xx(ispsoftc_t *isp, in_fcentry_24xx_t *src, in_fcentry_24xx_t *dst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
int i;
isp_put_hdr(isp, &src->in_header, &dst->in_header);
ISP_IOXPUT_32(isp, src->in_reserved, &dst->in_reserved);
ISP_IOXPUT_16(isp, src->in_nphdl, &dst->in_nphdl);
ISP_IOXPUT_16(isp, src->in_reserved1, &dst->in_reserved1);
ISP_IOXPUT_16(isp, src->in_flags, &dst->in_flags);
ISP_IOXPUT_16(isp, src->in_srr_rxid, &dst->in_srr_rxid);
ISP_IOXPUT_16(isp, src->in_status, &dst->in_status);
ISP_IOXPUT_8(isp, src->in_status_subcode, &dst->in_status_subcode);
Improve INOTs handling for 24xx and above chips.
2015-10-23 16:54:24 +00:00
ISP_IOXPUT_8(isp, src->in_fwhandle, &dst->in_fwhandle);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_32(isp, src->in_rxid, &dst->in_rxid);
ISP_IOXPUT_16(isp, src->in_srr_reloff_hi, &dst->in_srr_reloff_hi);
ISP_IOXPUT_16(isp, src->in_srr_reloff_lo, &dst->in_srr_reloff_lo);
ISP_IOXPUT_16(isp, src->in_srr_iu, &dst->in_srr_iu);
ISP_IOXPUT_16(isp, src->in_srr_oxid, &dst->in_srr_oxid);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_16(isp, src->in_nport_id_hi, &dst->in_nport_id_hi);
ISP_IOXPUT_8(isp, src->in_nport_id_lo, &dst->in_nport_id_lo);
ISP_IOXPUT_8(isp, src->in_reserved3, &dst->in_reserved3);
ISP_IOXPUT_16(isp, src->in_np_handle, &dst->in_np_handle);
for (i = 0; i < ASIZE(src->in_reserved4); i++) {
ISP_IOXPUT_8(isp, src->in_reserved4[i], &dst->in_reserved4[i]);
}
ISP_IOXPUT_8(isp, src->in_reserved5, &dst->in_reserved5);
ISP_IOXPUT_8(isp, src->in_vpidx, &dst->in_vpidx);
ISP_IOXPUT_32(isp, src->in_reserved6, &dst->in_reserved6);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->in_portid_lo, &dst->in_portid_lo);
ISP_IOXPUT_8(isp, src->in_portid_hi, &dst->in_portid_hi);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_8(isp, src->in_reserved7, &dst->in_reserved7);
ISP_IOXPUT_16(isp, src->in_reserved8, &dst->in_reserved8);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_16(isp, src->in_oxid, &dst->in_oxid);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_get_notify_fc(ispsoftc_t *isp, in_fcentry_t *src, in_fcentry_t *dst)
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
{
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_hdr(isp, &src->in_header, &dst->in_header);
ISP_IOXGET_32(isp, &src->in_reserved, dst->in_reserved);
ISP_IOXGET_8(isp, &src->in_lun, dst->in_lun);
ISP_IOXGET_8(isp, &src->in_iid, dst->in_iid);
ISP_IOXGET_16(isp, &src->in_scclun, dst->in_scclun);
ISP_IOXGET_32(isp, &src->in_reserved2, dst->in_reserved2);
ISP_IOXGET_16(isp, &src->in_status, dst->in_status);
ISP_IOXGET_16(isp, &src->in_task_flags, dst->in_task_flags);
ISP_IOXGET_16(isp, &src->in_seqid, dst->in_seqid);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_get_notify_fc_e(ispsoftc_t *isp, in_fcentry_e_t *src, in_fcentry_e_t *dst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_hdr(isp, &src->in_header, &dst->in_header);
ISP_IOXGET_32(isp, &src->in_reserved, dst->in_reserved);
ISP_IOXGET_16(isp, &src->in_iid, dst->in_iid);
ISP_IOXGET_16(isp, &src->in_scclun, dst->in_scclun);
ISP_IOXGET_32(isp, &src->in_reserved2, dst->in_reserved2);
ISP_IOXGET_16(isp, &src->in_status, dst->in_status);
ISP_IOXGET_16(isp, &src->in_task_flags, dst->in_task_flags);
ISP_IOXGET_16(isp, &src->in_seqid, dst->in_seqid);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_get_notify_24xx(ispsoftc_t *isp, in_fcentry_24xx_t *src, in_fcentry_24xx_t *dst)
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
{
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
int i;
isp_get_hdr(isp, &src->in_header, &dst->in_header);
ISP_IOXGET_32(isp, &src->in_reserved, dst->in_reserved);
ISP_IOXGET_16(isp, &src->in_nphdl, dst->in_nphdl);
ISP_IOXGET_16(isp, &src->in_reserved1, dst->in_reserved1);
ISP_IOXGET_16(isp, &src->in_flags, dst->in_flags);
ISP_IOXGET_16(isp, &src->in_srr_rxid, dst->in_srr_rxid);
ISP_IOXGET_16(isp, &src->in_status, dst->in_status);
ISP_IOXGET_8(isp, &src->in_status_subcode, dst->in_status_subcode);
Improve INOTs handling for 24xx and above chips.
2015-10-23 16:54:24 +00:00
ISP_IOXGET_8(isp, &src->in_fwhandle, dst->in_fwhandle);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_32(isp, &src->in_rxid, dst->in_rxid);
ISP_IOXGET_16(isp, &src->in_srr_reloff_hi, dst->in_srr_reloff_hi);
ISP_IOXGET_16(isp, &src->in_srr_reloff_lo, dst->in_srr_reloff_lo);
ISP_IOXGET_16(isp, &src->in_srr_iu, dst->in_srr_iu);
ISP_IOXGET_16(isp, &src->in_srr_oxid, dst->in_srr_oxid);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_16(isp, &src->in_nport_id_hi, dst->in_nport_id_hi);
ISP_IOXGET_8(isp, &src->in_nport_id_lo, dst->in_nport_id_lo);
ISP_IOXGET_8(isp, &src->in_reserved3, dst->in_reserved3);
ISP_IOXGET_16(isp, &src->in_np_handle, dst->in_np_handle);
for (i = 0; i < ASIZE(src->in_reserved4); i++) {
ISP_IOXGET_8(isp, &src->in_reserved4[i], dst->in_reserved4[i]);
}
ISP_IOXGET_8(isp, &src->in_reserved5, dst->in_reserved5);
ISP_IOXGET_8(isp, &src->in_vpidx, dst->in_vpidx);
ISP_IOXGET_32(isp, &src->in_reserved6, dst->in_reserved6);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_16(isp, &src->in_portid_lo, dst->in_portid_lo);
ISP_IOXGET_8(isp, &src->in_portid_hi, dst->in_portid_hi);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->in_reserved7, dst->in_reserved7);
ISP_IOXGET_16(isp, &src->in_reserved8, dst->in_reserved8);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_16(isp, &src->in_oxid, dst->in_oxid);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_put_notify_ack_fc(ispsoftc_t *isp, na_fcentry_t *src, na_fcentry_t *dst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
int i;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_hdr(isp, &src->na_header, &dst->na_header);
ISP_IOXPUT_32(isp, src->na_reserved, &dst->na_reserved);
ISP_IOXPUT_8(isp, src->na_reserved1, &dst->na_reserved1);
ISP_IOXPUT_8(isp, src->na_iid, &dst->na_iid);
ISP_IOXPUT_16(isp, src->na_response, &dst->na_response);
ISP_IOXPUT_16(isp, src->na_flags, &dst->na_flags);
ISP_IOXPUT_16(isp, src->na_reserved2, &dst->na_reserved2);
ISP_IOXPUT_16(isp, src->na_status, &dst->na_status);
ISP_IOXPUT_16(isp, src->na_task_flags, &dst->na_task_flags);
ISP_IOXPUT_16(isp, src->na_seqid, &dst->na_seqid);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
for (i = 0; i < NA2_RSVDLEN; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_16(isp, src->na_reserved3[i], &dst->na_reserved3[i]);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_put_notify_ack_fc_e(ispsoftc_t *isp, na_fcentry_e_t *src, na_fcentry_e_t *dst)
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
{
int i;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_hdr(isp, &src->na_header, &dst->na_header);
ISP_IOXPUT_32(isp, src->na_reserved, &dst->na_reserved);
ISP_IOXPUT_16(isp, src->na_iid, &dst->na_iid);
ISP_IOXPUT_16(isp, src->na_response, &dst->na_response);
ISP_IOXPUT_16(isp, src->na_flags, &dst->na_flags);
ISP_IOXPUT_16(isp, src->na_reserved2, &dst->na_reserved2);
ISP_IOXPUT_16(isp, src->na_status, &dst->na_status);
ISP_IOXPUT_16(isp, src->na_task_flags, &dst->na_task_flags);
ISP_IOXPUT_16(isp, src->na_seqid, &dst->na_seqid);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
for (i = 0; i < NA2_RSVDLEN; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_16(isp, src->na_reserved3[i], &dst->na_reserved3[i]);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_put_notify_24xx_ack(ispsoftc_t *isp, na_fcentry_24xx_t *src, na_fcentry_24xx_t *dst)
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
{
int i;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_put_hdr(isp, &src->na_header, &dst->na_header);
ISP_IOXPUT_32(isp, src->na_handle, &dst->na_handle);
ISP_IOXPUT_16(isp, src->na_nphdl, &dst->na_nphdl);
ISP_IOXPUT_16(isp, src->na_reserved1, &dst->na_reserved1);
ISP_IOXPUT_16(isp, src->na_flags, &dst->na_flags);
ISP_IOXPUT_16(isp, src->na_srr_rxid, &dst->na_srr_rxid);
ISP_IOXPUT_16(isp, src->na_status, &dst->na_status);
ISP_IOXPUT_8(isp, src->na_status_subcode, &dst->na_status_subcode);
Improve INOTs handling for 24xx and above chips.
2015-10-23 16:54:24 +00:00
ISP_IOXPUT_8(isp, src->na_fwhandle, &dst->na_fwhandle);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_32(isp, src->na_rxid, &dst->na_rxid);
ISP_IOXPUT_16(isp, src->na_srr_reloff_hi, &dst->na_srr_reloff_hi);
ISP_IOXPUT_16(isp, src->na_srr_reloff_lo, &dst->na_srr_reloff_lo);
ISP_IOXPUT_16(isp, src->na_srr_iu, &dst->na_srr_iu);
ISP_IOXPUT_16(isp, src->na_srr_flags, &dst->na_srr_flags);
for (i = 0; i < 18; i++) {
ISP_IOXPUT_8(isp, src->na_reserved3[i], &dst->na_reserved3[i]);
}
ISP_IOXPUT_8(isp, src->na_reserved4, &dst->na_reserved4);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_8(isp, src->na_vpidx, &dst->na_vpidx);
ISP_IOXPUT_8(isp, src->na_srr_reject_vunique, &dst->na_srr_reject_vunique);
ISP_IOXPUT_8(isp, src->na_srr_reject_explanation, &dst->na_srr_reject_explanation);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXPUT_8(isp, src->na_srr_reject_code, &dst->na_srr_reject_code);
ISP_IOXPUT_8(isp, src->na_reserved5, &dst->na_reserved5);
for (i = 0; i < 6; i++) {
ISP_IOXPUT_8(isp, src->na_reserved6[i], &dst->na_reserved6[i]);
}
ISP_IOXPUT_16(isp, src->na_oxid, &dst->na_oxid);
}
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_get_notify_ack_fc(ispsoftc_t *isp, na_fcentry_t *src, na_fcentry_t *dst)
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
{
int i;
isp_get_hdr(isp, &src->na_header, &dst->na_header);
ISP_IOXGET_32(isp, &src->na_reserved, dst->na_reserved);
ISP_IOXGET_8(isp, &src->na_reserved1, dst->na_reserved1);
ISP_IOXGET_8(isp, &src->na_iid, dst->na_iid);
ISP_IOXGET_16(isp, &src->na_response, dst->na_response);
ISP_IOXGET_16(isp, &src->na_flags, dst->na_flags);
ISP_IOXGET_16(isp, &src->na_reserved2, dst->na_reserved2);
ISP_IOXGET_16(isp, &src->na_status, dst->na_status);
ISP_IOXGET_16(isp, &src->na_task_flags, dst->na_task_flags);
ISP_IOXGET_16(isp, &src->na_seqid, dst->na_seqid);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
for (i = 0; i < NA2_RSVDLEN; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_16(isp, &src->na_reserved3[i], dst->na_reserved3[i]);
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). It also has to handle the SBus cards (for platforms that have them) which, while on a Big Endian machine, do *not* require *most* of the request/response queue entry fields to be swizzled or unswizzled. One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). MFC after: 2 weeks
2001-12-11 00:18:45 +00:00
}
}
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_get_notify_ack_fc_e(ispsoftc_t *isp, na_fcentry_e_t *src, na_fcentry_e_t *dst)
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
{
int i;
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
isp_get_hdr(isp, &src->na_header, &dst->na_header);
ISP_IOXGET_32(isp, &src->na_reserved, dst->na_reserved);
ISP_IOXGET_16(isp, &src->na_iid, dst->na_iid);
ISP_IOXGET_16(isp, &src->na_response, dst->na_response);
ISP_IOXGET_16(isp, &src->na_flags, dst->na_flags);
ISP_IOXGET_16(isp, &src->na_reserved2, dst->na_reserved2);
ISP_IOXGET_16(isp, &src->na_status, dst->na_status);
ISP_IOXGET_16(isp, &src->na_task_flags, dst->na_task_flags);
ISP_IOXGET_16(isp, &src->na_seqid, dst->na_seqid);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
for (i = 0; i < NA2_RSVDLEN; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_16(isp, &src->na_reserved3[i], dst->na_reserved3[i]);
First of several commits as this driver is dusted off and maybe brought up to date. Principle changes for this reelase is to support 2K Port Login firmware. This allows us to support the 2322 (and 2422 4Gb) cards which only come with the 2K Port Login firmware. The 2322 should now work- but we don't have firmware sets for it in ispfw (as the change to load 2K Port Login f/w hasn't been made- that f/w is so big it has to be loaded in more than one chunk). Other changes are the beginnings of cleaning up some long standing target mode issues. The next changes here will incorporate a lot of bug fixes from others. Finally, some copyright cleanup and attempts to make the parts of the driver that are FreeBSD specific start conforming more to FreeBSD style. MFC after: 1 month
2006-01-23 06:23:37 +00:00
}
}
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
void
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
isp_get_notify_ack_24xx(ispsoftc_t *isp, na_fcentry_24xx_t *src, na_fcentry_24xx_t *dst)
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
{
int i;
isp_get_hdr(isp, &src->na_header, &dst->na_header);
ISP_IOXGET_32(isp, &src->na_handle, dst->na_handle);
ISP_IOXGET_16(isp, &src->na_nphdl, dst->na_nphdl);
ISP_IOXGET_16(isp, &src->na_reserved1, dst->na_reserved1);
ISP_IOXGET_16(isp, &src->na_flags, dst->na_flags);
ISP_IOXGET_16(isp, &src->na_srr_rxid, dst->na_srr_rxid);
ISP_IOXGET_16(isp, &src->na_status, dst->na_status);
ISP_IOXGET_8(isp, &src->na_status_subcode, dst->na_status_subcode);
Improve INOTs handling for 24xx and above chips.
2015-10-23 16:54:24 +00:00
ISP_IOXGET_8(isp, &src->na_fwhandle, dst->na_fwhandle);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_32(isp, &src->na_rxid, dst->na_rxid);
ISP_IOXGET_16(isp, &src->na_srr_reloff_hi, dst->na_srr_reloff_hi);
ISP_IOXGET_16(isp, &src->na_srr_reloff_lo, dst->na_srr_reloff_lo);
ISP_IOXGET_16(isp, &src->na_srr_iu, dst->na_srr_iu);
ISP_IOXGET_16(isp, &src->na_srr_flags, dst->na_srr_flags);
for (i = 0; i < 18; i++) {
ISP_IOXGET_8(isp, &src->na_reserved3[i], dst->na_reserved3[i]);
}
ISP_IOXGET_8(isp, &src->na_reserved4, dst->na_reserved4);
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->na_vpidx, dst->na_vpidx);
ISP_IOXGET_8(isp, &src->na_srr_reject_vunique, dst->na_srr_reject_vunique);
ISP_IOXGET_8(isp, &src->na_srr_reject_explanation, dst->na_srr_reject_explanation);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_8(isp, &src->na_srr_reject_code, dst->na_srr_reject_code);
ISP_IOXGET_8(isp, &src->na_reserved5, dst->na_reserved5);
for (i = 0; i < 6; i++) {
ISP_IOXGET_8(isp, &src->na_reserved6[i], dst->na_reserved6[i]);
}
ISP_IOXGET_16(isp, &src->na_oxid, dst->na_oxid);
}
void
isp_get_abts(ispsoftc_t *isp, abts_t *src, abts_t *dst)
{
int i;
isp_get_hdr(isp, &src->abts_header, &dst->abts_header);
for (i = 0; i < 6; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->abts_reserved0[i], dst->abts_reserved0[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
ISP_IOXGET_16(isp, &src->abts_nphdl, dst->abts_nphdl);
ISP_IOXGET_16(isp, &src->abts_reserved1, dst->abts_reserved1);
ISP_IOXGET_16(isp, &src->abts_sof, dst->abts_sof);
ISP_IOXGET_32(isp, &src->abts_rxid_abts, dst->abts_rxid_abts);
ISP_IOXGET_16(isp, &src->abts_did_lo, dst->abts_did_lo);
ISP_IOXGET_8(isp, &src->abts_did_hi, dst->abts_did_hi);
ISP_IOXGET_8(isp, &src->abts_r_ctl, dst->abts_r_ctl);
ISP_IOXGET_16(isp, &src->abts_sid_lo, dst->abts_sid_lo);
ISP_IOXGET_8(isp, &src->abts_sid_hi, dst->abts_sid_hi);
ISP_IOXGET_8(isp, &src->abts_cs_ctl, dst->abts_cs_ctl);
ISP_IOXGET_16(isp, &src->abts_fs_ctl, dst->abts_fs_ctl);
ISP_IOXGET_8(isp, &src->abts_f_ctl, dst->abts_f_ctl);
ISP_IOXGET_8(isp, &src->abts_type, dst->abts_type);
ISP_IOXGET_16(isp, &src->abts_seq_cnt, dst->abts_seq_cnt);
ISP_IOXGET_8(isp, &src->abts_df_ctl, dst->abts_df_ctl);
ISP_IOXGET_8(isp, &src->abts_seq_id, dst->abts_seq_id);
ISP_IOXGET_16(isp, &src->abts_rx_id, dst->abts_rx_id);
ISP_IOXGET_16(isp, &src->abts_ox_id, dst->abts_ox_id);
ISP_IOXGET_32(isp, &src->abts_param, dst->abts_param);
for (i = 0; i < 16; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->abts_reserved2[i], dst->abts_reserved2[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
ISP_IOXGET_32(isp, &src->abts_rxid_task, dst->abts_rxid_task);
}
void
isp_put_abts_rsp(ispsoftc_t *isp, abts_rsp_t *src, abts_rsp_t *dst)
{
int i;
isp_put_hdr(isp, &src->abts_rsp_header, &dst->abts_rsp_header);
ISP_IOXPUT_32(isp, src->abts_rsp_handle, &dst->abts_rsp_handle);
ISP_IOXPUT_16(isp, src->abts_rsp_status, &dst->abts_rsp_status);
ISP_IOXPUT_16(isp, src->abts_rsp_nphdl, &dst->abts_rsp_nphdl);
ISP_IOXPUT_16(isp, src->abts_rsp_ctl_flags, &dst->abts_rsp_ctl_flags);
ISP_IOXPUT_16(isp, src->abts_rsp_sof, &dst->abts_rsp_sof);
ISP_IOXPUT_32(isp, src->abts_rsp_rxid_abts, &dst->abts_rsp_rxid_abts);
ISP_IOXPUT_16(isp, src->abts_rsp_did_lo, &dst->abts_rsp_did_lo);
ISP_IOXPUT_8(isp, src->abts_rsp_did_hi, &dst->abts_rsp_did_hi);
ISP_IOXPUT_8(isp, src->abts_rsp_r_ctl, &dst->abts_rsp_r_ctl);
ISP_IOXPUT_16(isp, src->abts_rsp_sid_lo, &dst->abts_rsp_sid_lo);
ISP_IOXPUT_8(isp, src->abts_rsp_sid_hi, &dst->abts_rsp_sid_hi);
ISP_IOXPUT_8(isp, src->abts_rsp_cs_ctl, &dst->abts_rsp_cs_ctl);
ISP_IOXPUT_16(isp, src->abts_rsp_f_ctl_lo, &dst->abts_rsp_f_ctl_lo);
ISP_IOXPUT_8(isp, src->abts_rsp_f_ctl_hi, &dst->abts_rsp_f_ctl_hi);
ISP_IOXPUT_8(isp, src->abts_rsp_type, &dst->abts_rsp_type);
ISP_IOXPUT_16(isp, src->abts_rsp_seq_cnt, &dst->abts_rsp_seq_cnt);
ISP_IOXPUT_8(isp, src->abts_rsp_df_ctl, &dst->abts_rsp_df_ctl);
ISP_IOXPUT_8(isp, src->abts_rsp_seq_id, &dst->abts_rsp_seq_id);
ISP_IOXPUT_16(isp, src->abts_rsp_rx_id, &dst->abts_rsp_rx_id);
ISP_IOXPUT_16(isp, src->abts_rsp_ox_id, &dst->abts_rsp_ox_id);
ISP_IOXPUT_32(isp, src->abts_rsp_param, &dst->abts_rsp_param);
if (src->abts_rsp_r_ctl == BA_ACC) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_16(isp, src->abts_rsp_payload.ba_acc.reserved, &dst->abts_rsp_payload.ba_acc.reserved);
ISP_IOXPUT_8(isp, src->abts_rsp_payload.ba_acc.last_seq_id, &dst->abts_rsp_payload.ba_acc.last_seq_id);
ISP_IOXPUT_8(isp, src->abts_rsp_payload.ba_acc.seq_id_valid, &dst->abts_rsp_payload.ba_acc.seq_id_valid);
ISP_IOXPUT_16(isp, src->abts_rsp_payload.ba_acc.aborted_rx_id, &dst->abts_rsp_payload.ba_acc.aborted_rx_id);
ISP_IOXPUT_16(isp, src->abts_rsp_payload.ba_acc.aborted_ox_id, &dst->abts_rsp_payload.ba_acc.aborted_ox_id);
ISP_IOXPUT_16(isp, src->abts_rsp_payload.ba_acc.high_seq_cnt, &dst->abts_rsp_payload.ba_acc.high_seq_cnt);
ISP_IOXPUT_16(isp, src->abts_rsp_payload.ba_acc.low_seq_cnt, &dst->abts_rsp_payload.ba_acc.low_seq_cnt);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
for (i = 0; i < 4; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_16(isp, src->abts_rsp_payload.ba_acc.reserved2[i], &dst->abts_rsp_payload.ba_acc.reserved2[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
} else if (src->abts_rsp_r_ctl == BA_RJT) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_8(isp, src->abts_rsp_payload.ba_rjt.vendor_unique, &dst->abts_rsp_payload.ba_rjt.vendor_unique);
ISP_IOXPUT_8(isp, src->abts_rsp_payload.ba_rjt.explanation, &dst->abts_rsp_payload.ba_rjt.explanation);
ISP_IOXPUT_8(isp, src->abts_rsp_payload.ba_rjt.reason, &dst->abts_rsp_payload.ba_rjt.reason);
ISP_IOXPUT_8(isp, src->abts_rsp_payload.ba_rjt.reserved, &dst->abts_rsp_payload.ba_rjt.reserved);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
for (i = 0; i < 12; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_16(isp, src->abts_rsp_payload.ba_rjt.reserved2[i], &dst->abts_rsp_payload.ba_rjt.reserved2[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
} else {
for (i = 0; i < 16; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXPUT_8(isp, src->abts_rsp_payload.reserved[i], &dst->abts_rsp_payload.reserved[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
}
ISP_IOXPUT_32(isp, src->abts_rsp_rxid_task, &dst->abts_rsp_rxid_task);
}
void
isp_get_abts_rsp(ispsoftc_t *isp, abts_rsp_t *src, abts_rsp_t *dst)
{
int i;
isp_get_hdr(isp, &src->abts_rsp_header, &dst->abts_rsp_header);
ISP_IOXGET_32(isp, &src->abts_rsp_handle, dst->abts_rsp_handle);
ISP_IOXGET_16(isp, &src->abts_rsp_status, dst->abts_rsp_status);
ISP_IOXGET_16(isp, &src->abts_rsp_nphdl, dst->abts_rsp_nphdl);
ISP_IOXGET_16(isp, &src->abts_rsp_ctl_flags, dst->abts_rsp_ctl_flags);
ISP_IOXGET_16(isp, &src->abts_rsp_sof, dst->abts_rsp_sof);
ISP_IOXGET_32(isp, &src->abts_rsp_rxid_abts, dst->abts_rsp_rxid_abts);
ISP_IOXGET_16(isp, &src->abts_rsp_did_lo, dst->abts_rsp_did_lo);
ISP_IOXGET_8(isp, &src->abts_rsp_did_hi, dst->abts_rsp_did_hi);
ISP_IOXGET_8(isp, &src->abts_rsp_r_ctl, dst->abts_rsp_r_ctl);
ISP_IOXGET_16(isp, &src->abts_rsp_sid_lo, dst->abts_rsp_sid_lo);
ISP_IOXGET_8(isp, &src->abts_rsp_sid_hi, dst->abts_rsp_sid_hi);
ISP_IOXGET_8(isp, &src->abts_rsp_cs_ctl, dst->abts_rsp_cs_ctl);
ISP_IOXGET_16(isp, &src->abts_rsp_f_ctl_lo, dst->abts_rsp_f_ctl_lo);
ISP_IOXGET_8(isp, &src->abts_rsp_f_ctl_hi, dst->abts_rsp_f_ctl_hi);
ISP_IOXGET_8(isp, &src->abts_rsp_type, dst->abts_rsp_type);
ISP_IOXGET_16(isp, &src->abts_rsp_seq_cnt, dst->abts_rsp_seq_cnt);
ISP_IOXGET_8(isp, &src->abts_rsp_df_ctl, dst->abts_rsp_df_ctl);
ISP_IOXGET_8(isp, &src->abts_rsp_seq_id, dst->abts_rsp_seq_id);
ISP_IOXGET_16(isp, &src->abts_rsp_rx_id, dst->abts_rsp_rx_id);
ISP_IOXGET_16(isp, &src->abts_rsp_ox_id, dst->abts_rsp_ox_id);
ISP_IOXGET_32(isp, &src->abts_rsp_param, dst->abts_rsp_param);
for (i = 0; i < 8; i++) {
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_8(isp, &src->abts_rsp_payload.rsp.reserved[i], dst->abts_rsp_payload.rsp.reserved[i]);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
}
Add 8Gb support (isp_2500). Fix a fair number of configuration and firmware loading bugs. Target mode support has received some serious attention to make it more usable and stable. Some backward compatible additions to CAM have been made that make target mode async events easier to deal with have also been put into place. Further refinement and better support for NP-IV (N-port Virtualization) is now in place. Code for release prior to RELENG_7 has been stripped away for code clarity. Sponsored by: Copan Systems Reviewed by: scottl, ken, jung-uk kim Approved by: re
2009-08-01 01:04:26 +00:00
ISP_IOXGET_32(isp, &src->abts_rsp_payload.rsp.subcode1, dst->abts_rsp_payload.rsp.subcode1);
ISP_IOXGET_32(isp, &src->abts_rsp_payload.rsp.subcode2, dst->abts_rsp_payload.rsp.subcode2);
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
ISP_IOXGET_32(isp, &src->abts_rsp_rxid_task, dst->abts_rsp_rxid_task);
}
Remove use of inlines and use the functions as a library. Larger code space, possibly performance hit, but more portable. Certainly less questionable use of inlining. Suggested by: des
2006-02-02 21:31:34 +00:00
#endif /* ISP_TARGET_MODE */
Add 4Gb (24XX) support and lay the foundation for a lot of new stuff.
2006-11-02 03:21:32 +00:00
/*
* vim:ts=8:sw=8
*/
Reference in New Issue Copy Permalink