Aren
/
NetBSD
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
NetBSD/sys/dev/ic/isp.c

4353 lines
116 KiB
C
Raw Normal View History

I was misinformed. I cannot get away from specifying tags for FC. Some devices are happy w/o them- some are unhappy (IBM drives).
1999-10-28 06:50:34 +04:00
/* $NetBSD: isp.c,v 1.42 1999/10/28 02:50:34 mjacob Exp $ */
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
/*
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
* Copyright (C) 1997, 1998, 1999 National Aeronautics & Space Administration
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
* notice, this list of conditions and the following disclaimer.
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. The name of the author may not be used to endorse or promote products
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
* derived from this software without specific prior written permission
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
*
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR 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.
*/
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
/*
* Machine and OS Independent (well, as best as possible)
* code for the Qlogic ISP SCSI adapters.
* Matthew Jacob <mjacob@nas.nasa.gov>
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
*/
NetBSD RCS ID tweaks, a few comment block tweaks. Also, make copyright notices consistent (per Matt Jacob).
1997-03-13 00:06:41 +03:00
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
/*
* Inspiration and ideas about this driver are from Erik Moe's Linux driver
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
* (qlogicisp.c) and Dave Miller's SBus version of same (qlogicisp.c). Some
* ideas dredged from the Solaris driver.
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
*/
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
/*
* Include header file appropriate for platform we're building on.
*/
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
#ifdef __NetBSD__
#include <dev/ic/isp_netbsd.h>
#endif
#ifdef __FreeBSD__
#include <dev/isp/isp_freebsd.h>
#endif
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
#ifdef __OpenBSD__
#include <dev/ic/isp_openbsd.h>
#endif
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
#ifdef __linux__
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
#include "isp_linux.h"
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
#endif
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
/*
* General defines
*/
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
#define MBOX_DELAY_COUNT 1000000 / 100
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
/*
* Local static data
*/
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
/*
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
* Local function prototypes.
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
*/
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
static int isp_parse_async __P((struct ispsoftc *, int));
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
static int isp_handle_other_response
__P((struct ispsoftc *, ispstatusreq_t *, u_int8_t *));
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
static void isp_parse_status
__P((struct ispsoftc *, ispstatusreq_t *, ISP_SCSI_XFER_T *));
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
static void isp_fastpost_complete __P((struct ispsoftc *, u_int32_t));
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
static void isp_scsi_init __P((struct ispsoftc *));
static void isp_scsi_channel_init __P((struct ispsoftc *, int));
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
static void isp_fibre_init __P((struct ispsoftc *));
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
static void isp_mark_getpdb_all __P((struct ispsoftc *));
static int isp_getpdb __P((struct ispsoftc *, int, isp_pdb_t *));
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
static u_int64_t isp_get_portname __P((struct ispsoftc *, int, int));
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
static int isp_fclink_test __P((struct ispsoftc *, int));
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
static int isp_same_lportdb __P((struct lportdb *, struct lportdb *));
static int isp_pdb_sync __P((struct ispsoftc *, int));
#ifdef ISP2100_FABRIC
static int isp_scan_fabric __P((struct ispsoftc *));
#endif
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
static void isp_fw_state __P((struct ispsoftc *));
static void isp_dumpregs __P((struct ispsoftc *, const char *));
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
static void isp_mboxcmd __P((struct ispsoftc *, mbreg_t *));
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
static void isp_update __P((struct ispsoftc *));
static void isp_update_bus __P((struct ispsoftc *, int));
static void isp_setdfltparm __P((struct ispsoftc *, int));
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
static int isp_read_nvram __P((struct ispsoftc *));
static void isp_rdnvram_word __P((struct ispsoftc *, int, u_int16_t *));
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
/*
* Reset Hardware.
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
*
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
* Hit the chip over the head, download new f/w and set it running.
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
*
Locking is now in the outer framework for isp_init/isp_reset- this allows us to call isp_reset/isp_init internally with impunity. Rename isp_phoenix to isp_restart and make it global. Clean it up a tad. If we get an ASYNC_SYSTEM_ERROR code in isp_intr, call isp_restart and return- the f/w is toasted at this point (usually), so we have to bring things back to a known state. In isp_mboxcmd, when we don't see the HOST_INT bit go clear, try and find out if the isp is trying to tell us something and try again. This may avoid a potential deadlock where the previous mailbox command hasn't been cleared by the ISP. In isp_init don't try and get device parameters if we already have them- this typically doesn't work if we're in the middle of an isp_restart.
1998-07-19 01:10:16 +04:00
* Locking done elsewhere.
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
*/
void
isp_reset(isp)
struct ispsoftc *isp;
{
mbreg_t mbs;
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
int loops, i, dodnld = 1;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
char *revname;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
isp->isp_state = ISP_NILSTATE;
A) Attempt to get clock rate from ISP chip prior to bashing it over the head in reset. If none there, try and get from the bus/platform specific code. If a nonzero value for either, set the clock rate. This is why the PCI card versions weren't working- they need to be set at 60MHZ, rather than the default 40MHZ (which worked fine for the internal ISP chips on the Alpha 8X00). B) If a isp_poll returns failure (command never completed) to the caller and no error is set in the xs struct, set XS_SELTIMEOUT. And then call... C) Added isp_lostcmd function to try and ask the ISP chip about it's current state as well as the state of commands for a particular target/lun. This is going along to try and figure out why the very first command to the ISP always seems to get swallowed up.
1997-04-05 06:53:22 +04:00
/*
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
* Basic types (SCSI, FibreChannel and PCI or SBus)
* have been set in the MD code. We figure out more
* here.
A) Attempt to get clock rate from ISP chip prior to bashing it over the head in reset. If none there, try and get from the bus/platform specific code. If a nonzero value for either, set the clock rate. This is why the PCI card versions weren't working- they need to be set at 60MHZ, rather than the default 40MHZ (which worked fine for the internal ISP chips on the Alpha 8X00). B) If a isp_poll returns failure (command never completed) to the caller and no error is set in the xs struct, set XS_SELTIMEOUT. And then call... C) Added isp_lostcmd function to try and ask the ISP chip about it's current state as well as the state of commands for a particular target/lun. This is going along to try and figure out why the very first command to the ISP always seems to get swallowed up.
1997-04-05 06:53:22 +04:00
*/
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
isp->isp_dblev = DFLT_DBLEVEL;
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
/*
* After we've fired this chip up, zero out the conf1 register
* for SCSI adapters and other settings for the 2100.
*/
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
/*
* Get the current running firmware revision out of the
* chip before we hit it over the head (if this is our
* first time through). Note that we store this as the
* 'ROM' firmware revision- which it may not be. In any
* case, we don't really use this yet, but we may in
* the future.
*/
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
if (isp->isp_used == 0) {
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
/*
* Just in case it was paused...
*/
+ Some basic deadchip detection. + Enable FIFO bursts, but also detect bogus 1040A with busted FIFO. + Use new MEMZERO crossplatform define. + Handle RQCS_QUEUE_FULL status case and let upper layer parse SCSI ststus byte if nonzero (should be 0x28- Queue Full status) + Fold ISP_NVRAM_FIFO_THRESHOLD_128 into isp_fifo_threshold tag.
1999-01-10 08:07:53 +03:00
ISP_WRITE(isp, HCCR, HCCR_CMD_RELEASE);
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
mbs.param[0] = MBOX_ABOUT_FIRMWARE;
isp_mboxcmd(isp, &mbs);
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
/*
* If this fails, it probably means we're running
* an old prom, if anything at all...
*/
if (mbs.param[0] == MBOX_COMMAND_COMPLETE) {
isp->isp_romfw_rev[0] = mbs.param[1];
isp->isp_romfw_rev[1] = mbs.param[2];
isp->isp_romfw_rev[2] = mbs.param[3];
+ Some basic deadchip detection. + Enable FIFO bursts, but also detect bogus 1040A with busted FIFO. + Use new MEMZERO crossplatform define. + Handle RQCS_QUEUE_FULL status case and let upper layer parse SCSI ststus byte if nonzero (should be 0x28- Queue Full status) + Fold ISP_NVRAM_FIFO_THRESHOLD_128 into isp_fifo_threshold tag.
1999-01-10 08:07:53 +03:00
}
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
isp->isp_used = 1;
+ Some basic deadchip detection. + Enable FIFO bursts, but also detect bogus 1040A with busted FIFO. + Use new MEMZERO crossplatform define. + Handle RQCS_QUEUE_FULL status case and let upper layer parse SCSI ststus byte if nonzero (should be 0x28- Queue Full status) + Fold ISP_NVRAM_FIFO_THRESHOLD_128 into isp_fifo_threshold tag.
1999-01-10 08:07:53 +03:00
}
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
DISABLE_INTS(isp);
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
/*
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
* Put the board into PAUSE mode.
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
*/
ISP_WRITE(isp, HCCR, HCCR_CMD_PAUSE);
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
if (IS_FC(isp)) {
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
revname = "2X00";
switch (isp->isp_type) {
case ISP_HA_FC_2100:
revname[1] = '1';
break;
case ISP_HA_FC_2200:
revname[1] = '2';
/*
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
* Resident firmware for the 2200 appears
* to have SCCLUN enabled.
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
*/
#ifndef ISP2100_SCCLUN
if (isp->isp_mdvec->dv_fwlen == 0) {
PRINTF("%s: WARNING- using resident f/w without"
" SCCLUN support defined\n", isp->isp_name);
}
#endif
break;
default:
break;
}
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
} else if (IS_12X0(isp)) {
revname = "12X0";
isp->isp_clock = 60;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
} else if (IS_1080(isp)) {
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
u_int16_t l;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
sdparam *sdp = isp->isp_param;
revname = "1080";
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp->isp_clock = 100;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
l = ISP_READ(isp, SXP_PINS_DIFF) & ISP1080_MODE_MASK;
switch (l) {
case ISP1080_LVD_MODE:
sdp->isp_lvdmode = 1;
PRINTF("%s: LVD Mode\n", isp->isp_name);
break;
case ISP1080_HVD_MODE:
sdp->isp_diffmode = 1;
PRINTF("%s: Differential Mode\n", isp->isp_name);
break;
case ISP1080_SE_MODE:
sdp->isp_ultramode = 1;
PRINTF("%s: Single-Ended Mode\n", isp->isp_name);
break;
default:
/*
* Hmm. Up in a wierd mode. This means all SCSI I/O
* buffer lines are tristated, so we're in a lot of
* trouble if we don't set things up right.
*/
PRINTF("%s: Illegal Mode 0x%x\n", isp->isp_name, l);
break;
}
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
} else {
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
sdparam *sdp = isp->isp_param;
+ Some basic deadchip detection. + Enable FIFO bursts, but also detect bogus 1040A with busted FIFO. + Use new MEMZERO crossplatform define. + Handle RQCS_QUEUE_FULL status case and let upper layer parse SCSI ststus byte if nonzero (should be 0x28- Queue Full status) + Fold ISP_NVRAM_FIFO_THRESHOLD_128 into isp_fifo_threshold tag.
1999-01-10 08:07:53 +03:00
i = ISP_READ(isp, BIU_CONF0) & BIU_CONF0_HW_MASK;
switch (i) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
default:
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
PRINTF("%s: unknown chip rev. 0x%x- assuming a 1020\n",
+ Some basic deadchip detection. + Enable FIFO bursts, but also detect bogus 1040A with busted FIFO. + Use new MEMZERO crossplatform define. + Handle RQCS_QUEUE_FULL status case and let upper layer parse SCSI ststus byte if nonzero (should be 0x28- Queue Full status) + Fold ISP_NVRAM_FIFO_THRESHOLD_128 into isp_fifo_threshold tag.
1999-01-10 08:07:53 +03:00
isp->isp_name, i);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
/* FALLTHROUGH */
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
case 1:
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
revname = "1020";
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp->isp_type = ISP_HA_SCSI_1020;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp->isp_clock = 40;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
break;
there is an (unimportant) difference between 1020 && 1020A
1998-05-06 02:10:36 +04:00
case 2:
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
/*
* Some 1020A chips are Ultra Capable, but don't
* run the clock rate up for that unless told to
* do so by the Ultra Capable bits being set.
*/
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
revname = "1020A";
there is an (unimportant) difference between 1020 && 1020A
1998-05-06 02:10:36 +04:00
isp->isp_type = ISP_HA_SCSI_1020A;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp->isp_clock = 40;
there is an (unimportant) difference between 1020 && 1020A
1998-05-06 02:10:36 +04:00
break;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
case 3:
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
revname = "1040";
isp->isp_type = ISP_HA_SCSI_1040;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp->isp_clock = 60;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
break;
case 4:
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
revname = "1040A";
isp->isp_type = ISP_HA_SCSI_1040A;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp->isp_clock = 60;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
break;
case 5:
revname = "1040B";
isp->isp_type = ISP_HA_SCSI_1040B;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp->isp_clock = 60;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
break;
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
case 6:
revname = "1040C(?)";
isp->isp_type = ISP_HA_SCSI_1040C;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp->isp_clock = 60;
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
break;
Hmm- seriously funny and sad bug: you need to directly establish the clock rate for this board on Alpha/PCI systems. Under x86/PCI, the board f/w will correctly tell you "I'm running at 60Mhz", so the code that preserved that across a board reset (which would drop the chip back to 40Mhz) worked fine. On the 8200, the chip was saying "I'm 40Mhz"- which wasn't true. This turned out to be okay as long as you didn't have any FAST or UltraFast targets- In fact, setting the chip to 40Mhz allowed you to run up to 8Mhz SCSI. Unfortunately you die bigtime on the devices that go faster than that. The fix here is to only use what the chip tells you the clock rate is in the cases you don't really know (sbus is the only case where this could be different, although with 66Mhz PCI coming up, this may change).
1997-06-22 23:57:06 +04:00
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
/*
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
* Now, while we're at it, gather info about ultra
* and/or differential mode.
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
*/
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
if (ISP_READ(isp, SXP_PINS_DIFF) & SXP_PINS_DIFF_MODE) {
PRINTF("%s: Differential Mode\n", isp->isp_name);
sdp->isp_diffmode = 1;
+ Some basic deadchip detection. + Enable FIFO bursts, but also detect bogus 1040A with busted FIFO. + Use new MEMZERO crossplatform define. + Handle RQCS_QUEUE_FULL status case and let upper layer parse SCSI ststus byte if nonzero (should be 0x28- Queue Full status) + Fold ISP_NVRAM_FIFO_THRESHOLD_128 into isp_fifo_threshold tag.
1999-01-10 08:07:53 +03:00
} else {
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
sdp->isp_diffmode = 0;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
i = ISP_READ(isp, RISC_PSR);
if (isp->isp_bustype == ISP_BT_SBUS) {
i &= RISC_PSR_SBUS_ULTRA;
} else {
i &= RISC_PSR_PCI_ULTRA;
}
if (i != 0) {
PRINTF("%s: Ultra Mode Capable\n", isp->isp_name);
sdp->isp_ultramode = 1;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
/*
* If we're in Ultra Mode, we have to be 60Mhz clock-
* even for the SBus version.
*/
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp->isp_clock = 60;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
} else {
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
sdp->isp_ultramode = 0;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
/*
* Clock is known. Gronk.
*/
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
/*
* Machine dependent clock (if set) overrides
* our generic determinations.
*/
if (isp->isp_mdvec->dv_clock) {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
if (isp->isp_mdvec->dv_clock < isp->isp_clock) {
isp->isp_clock = isp->isp_mdvec->dv_clock;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
}
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
A) Attempt to get clock rate from ISP chip prior to bashing it over the head in reset. If none there, try and get from the bus/platform specific code. If a nonzero value for either, set the clock rate. This is why the PCI card versions weren't working- they need to be set at 60MHZ, rather than the default 40MHZ (which worked fine for the internal ISP chips on the Alpha 8X00). B) If a isp_poll returns failure (command never completed) to the caller and no error is set in the xs struct, set XS_SELTIMEOUT. And then call... C) Added isp_lostcmd function to try and ask the ISP chip about it's current state as well as the state of commands for a particular target/lun. This is going along to try and figure out why the very first command to the ISP always seems to get swallowed up.
1997-04-05 06:53:22 +04:00
}
Hmm- seriously funny and sad bug: you need to directly establish the clock rate for this board on Alpha/PCI systems. Under x86/PCI, the board f/w will correctly tell you "I'm running at 60Mhz", so the code that preserved that across a board reset (which would drop the chip back to 40Mhz) worked fine. On the 8200, the chip was saying "I'm 40Mhz"- which wasn't true. This turned out to be okay as long as you didn't have any FAST or UltraFast targets- In fact, setting the chip to 40Mhz allowed you to run up to 8Mhz SCSI. Unfortunately you die bigtime on the devices that go faster than that. The fix here is to only use what the chip tells you the clock rate is in the cases you don't really know (sbus is the only case where this could be different, although with 66Mhz PCI coming up, this may change).
1997-06-22 23:57:06 +04:00
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
/*
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
* Do MD specific pre initialization
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
*/
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
ISP_RESET0(isp);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
again:
+ Some basic deadchip detection. + Enable FIFO bursts, but also detect bogus 1040A with busted FIFO. + Use new MEMZERO crossplatform define. + Handle RQCS_QUEUE_FULL status case and let upper layer parse SCSI ststus byte if nonzero (should be 0x28- Queue Full status) + Fold ISP_NVRAM_FIFO_THRESHOLD_128 into isp_fifo_threshold tag.
1999-01-10 08:07:53 +03:00
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
/*
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
* Hit the chip over the head with hammer,
* and give the ISP a chance to recover.
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
*/
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
if (IS_SCSI(isp)) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
ISP_WRITE(isp, BIU_ICR, BIU_ICR_SOFT_RESET);
/*
* A slight delay...
*/
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
SYS_DELAY(100);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
#if 0
PRINTF("%s: mbox0-5: 0x%x 0x%x 0x%x 0x%x 0x%x 0x%x\n",
isp->isp_name, ISP_READ(isp, OUTMAILBOX0),
ISP_READ(isp, OUTMAILBOX1), ISP_READ(isp, OUTMAILBOX2),
ISP_READ(isp, OUTMAILBOX3), ISP_READ(isp, OUTMAILBOX4),
ISP_READ(isp, OUTMAILBOX5));
#endif
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
/*
* Clear data && control DMA engines.
*/
ISP_WRITE(isp, CDMA_CONTROL,
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
DMA_CNTRL_CLEAR_CHAN | DMA_CNTRL_RESET_INT);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
ISP_WRITE(isp, DDMA_CONTROL,
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
DMA_CNTRL_CLEAR_CHAN | DMA_CNTRL_RESET_INT);
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
} else {
ISP_WRITE(isp, BIU2100_CSR, BIU2100_SOFT_RESET);
/*
* A slight delay...
*/
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
SYS_DELAY(100);
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
/*
* Clear data && control DMA engines.
*/
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
ISP_WRITE(isp, CDMA2100_CONTROL,
DMA_CNTRL2100_CLEAR_CHAN | DMA_CNTRL2100_RESET_INT);
ISP_WRITE(isp, TDMA2100_CONTROL,
DMA_CNTRL2100_CLEAR_CHAN | DMA_CNTRL2100_RESET_INT);
ISP_WRITE(isp, RDMA2100_CONTROL,
DMA_CNTRL2100_CLEAR_CHAN | DMA_CNTRL2100_RESET_INT);
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
/*
* Wait for ISP to be ready to go...
*/
loops = MBOX_DELAY_COUNT;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
for (;;) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (IS_SCSI(isp)) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
if (!(ISP_READ(isp, BIU_ICR) & BIU_ICR_SOFT_RESET))
break;
} else {
if (!(ISP_READ(isp, BIU2100_CSR) & BIU2100_SOFT_RESET))
break;
}
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
SYS_DELAY(100);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (--loops < 0) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_dumpregs(isp, "chip reset timed out");
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
return;
}
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
/*
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
* After we've fired this chip up, zero out the conf1 register
* for SCSI adapters and other settings for the 2100.
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
*/
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
if (IS_SCSI(isp)) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
ISP_WRITE(isp, BIU_CONF1, 0);
} else {
ISP_WRITE(isp, BIU2100_CSR, 0);
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
/*
* Reset RISC Processor
*/
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
ISP_WRITE(isp, HCCR, HCCR_CMD_RESET);
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
SYS_DELAY(100);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
+ Some basic deadchip detection. + Enable FIFO bursts, but also detect bogus 1040A with busted FIFO. + Use new MEMZERO crossplatform define. + Handle RQCS_QUEUE_FULL status case and let upper layer parse SCSI ststus byte if nonzero (should be 0x28- Queue Full status) + Fold ISP_NVRAM_FIFO_THRESHOLD_128 into isp_fifo_threshold tag.
1999-01-10 08:07:53 +03:00
/*
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
* Establish some initial burst rate stuff.
+ Some basic deadchip detection. + Enable FIFO bursts, but also detect bogus 1040A with busted FIFO. + Use new MEMZERO crossplatform define. + Handle RQCS_QUEUE_FULL status case and let upper layer parse SCSI ststus byte if nonzero (should be 0x28- Queue Full status) + Fold ISP_NVRAM_FIFO_THRESHOLD_128 into isp_fifo_threshold tag.
1999-01-10 08:07:53 +03:00
* (only for the 1XX0 boards). This really should
* be done later after fetching from NVRAM.
*/
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
if (IS_SCSI(isp)) {
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
u_int16_t tmp = isp->isp_mdvec->dv_conf1;
+ Some basic deadchip detection. + Enable FIFO bursts, but also detect bogus 1040A with busted FIFO. + Use new MEMZERO crossplatform define. + Handle RQCS_QUEUE_FULL status case and let upper layer parse SCSI ststus byte if nonzero (should be 0x28- Queue Full status) + Fold ISP_NVRAM_FIFO_THRESHOLD_128 into isp_fifo_threshold tag.
1999-01-10 08:07:53 +03:00
/*
* Busted FIFO. Turn off all but burst enables.
*/
if (isp->isp_type == ISP_HA_SCSI_1040A) {
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
tmp &= BIU_BURST_ENABLE;
+ Some basic deadchip detection. + Enable FIFO bursts, but also detect bogus 1040A with busted FIFO. + Use new MEMZERO crossplatform define. + Handle RQCS_QUEUE_FULL status case and let upper layer parse SCSI ststus byte if nonzero (should be 0x28- Queue Full status) + Fold ISP_NVRAM_FIFO_THRESHOLD_128 into isp_fifo_threshold tag.
1999-01-10 08:07:53 +03:00
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
ISP_SETBITS(isp, BIU_CONF1, tmp);
if (tmp & BIU_BURST_ENABLE) {
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
ISP_SETBITS(isp, CDMA_CONF, DMA_ENABLE_BURST);
ISP_SETBITS(isp, DDMA_CONF, DMA_ENABLE_BURST);
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
#ifdef PTI_CARDS
if (((sdparam *) isp->isp_param)->isp_ultramode) {
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
while (ISP_READ(isp, RISC_MTR) != 0x1313) {
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
ISP_WRITE(isp, RISC_MTR, 0x1313);
ISP_WRITE(isp, HCCR, HCCR_CMD_STEP);
}
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
} else {
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
ISP_WRITE(isp, RISC_MTR, 0x1212);
}
/*
* PTI specific register
*/
ISP_WRITE(isp, RISC_EMB, DUAL_BANK)
#else
ISP_WRITE(isp, RISC_MTR, 0x1212);
#endif
} else {
ISP_WRITE(isp, RISC_MTR2100, 0x1212);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
ISP_WRITE(isp, HCCR, HCCR_CMD_RELEASE); /* release paused processor */
/*
* Do MD specific post initialization
*/
ISP_RESET1(isp);
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
/*
* Wait for everything to finish firing up...
*/
loops = MBOX_DELAY_COUNT;
while (ISP_READ(isp, OUTMAILBOX0) == MBOX_BUSY) {
SYS_DELAY(100);
if (--loops < 0) {
PRINTF("%s: MBOX_BUSY never cleared on reset\n",
isp->isp_name);
return;
}
}
/*
* Up until this point we've done everything by just reading or
* setting registers. From this point on we rely on at least *some*
* kind of firmware running in the card.
*/
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
/*
* Do some sanity checking.
*/
mbs.param[0] = MBOX_NO_OP;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_mboxcmd(isp, &mbs);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_dumpregs(isp, "NOP test failed");
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
return;
}
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
if (IS_SCSI(isp)) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
mbs.param[0] = MBOX_MAILBOX_REG_TEST;
mbs.param[1] = 0xdead;
mbs.param[2] = 0xbeef;
mbs.param[3] = 0xffff;
mbs.param[4] = 0x1111;
mbs.param[5] = 0xa5a5;
isp_mboxcmd(isp, &mbs);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_dumpregs(isp,
"Mailbox Register test didn't complete");
return;
}
if (mbs.param[1] != 0xdead || mbs.param[2] != 0xbeef ||
mbs.param[3] != 0xffff || mbs.param[4] != 0x1111 ||
mbs.param[5] != 0xa5a5) {
isp_dumpregs(isp, "Register Test Failed");
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
return;
}
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
/*
Fix for port-alpha/4903- always download f/w unless config flags say no or we have no firmware to download.
1998-01-28 22:09:24 +03:00
* Download new Firmware, unless requested not to do so.
* This is made slightly trickier in some cases where the
* firmware of the ROM revision is newer than the revision
* compiled into the driver. So, where we used to compare
* versions of our f/w and the ROM f/w, now we just see
* whether we have f/w at all and whether a config flag
* has disabled our download.
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
*/
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
if ((isp->isp_mdvec->dv_fwlen == 0) ||
(isp->isp_confopts & ISP_CFG_NORELOAD)) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
dodnld = 0;
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
if (dodnld && isp->isp_mdvec->dv_fwlen) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
for (i = 0; i < isp->isp_mdvec->dv_fwlen; i++) {
mbs.param[0] = MBOX_WRITE_RAM_WORD;
mbs.param[1] = isp->isp_mdvec->dv_codeorg + i;
mbs.param[2] = isp->isp_mdvec->dv_ispfw[i];
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
PRINTF("%s: F/W download failed at word %d\n",
isp->isp_name, i);
dodnld = 0;
goto again;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
/*
* Verify that it downloaded correctly.
*/
mbs.param[0] = MBOX_VERIFY_CHECKSUM;
mbs.param[1] = isp->isp_mdvec->dv_codeorg;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
isp_dumpregs(isp, "ram checksum failure");
return;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
} else {
update some debugging stuff- also turn off possible ULTRA mode for 1020s
1998-03-23 01:02:29 +03:00
IDPRINTF(3, ("%s: skipping f/w download\n", isp->isp_name));
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
/*
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
* Now start it rolling.
*
* If we didn't actually download f/w,
* we still need to (re)start it.
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
*/
mbs.param[0] = MBOX_EXEC_FIRMWARE;
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
if (isp->isp_mdvec->dv_codeorg)
mbs.param[1] = isp->isp_mdvec->dv_codeorg;
else
mbs.param[1] = 0x1000;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_mboxcmd(isp, &mbs);
A) Attempt to get clock rate from ISP chip prior to bashing it over the head in reset. If none there, try and get from the bus/platform specific code. If a nonzero value for either, set the clock rate. This is why the PCI card versions weren't working- they need to be set at 60MHZ, rather than the default 40MHZ (which worked fine for the internal ISP chips on the Alpha 8X00). B) If a isp_poll returns failure (command never completed) to the caller and no error is set in the xs struct, set XS_SELTIMEOUT. And then call... C) Added isp_lostcmd function to try and ask the ISP chip about it's current state as well as the state of commands for a particular target/lun. This is going along to try and figure out why the very first command to the ISP always seems to get swallowed up.
1997-04-05 06:53:22 +04:00
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (IS_SCSI(isp)) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
/*
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
* Set CLOCK RATE, but only if asked to.
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
*/
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
if (isp->isp_clock) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
mbs.param[0] = MBOX_SET_CLOCK_RATE;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
mbs.param[1] = isp->isp_clock;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
PRINTF("failed to set clockrate (0x%x)\n",
mbs.param[0]);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
/* but continue */
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
A) Attempt to get clock rate from ISP chip prior to bashing it over the head in reset. If none there, try and get from the bus/platform specific code. If a nonzero value for either, set the clock rate. This is why the PCI card versions weren't working- they need to be set at 60MHZ, rather than the default 40MHZ (which worked fine for the internal ISP chips on the Alpha 8X00). B) If a isp_poll returns failure (command never completed) to the caller and no error is set in the xs struct, set XS_SELTIMEOUT. And then call... C) Added isp_lostcmd function to try and ask the ISP chip about it's current state as well as the state of commands for a particular target/lun. This is going along to try and figure out why the very first command to the ISP always seems to get swallowed up.
1997-04-05 06:53:22 +04:00
}
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
mbs.param[0] = MBOX_ABOUT_FIRMWARE;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_mboxcmd(isp, &mbs);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
PRINTF("could not get f/w started (0x%x)\n", mbs.param[0]);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
return;
}
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
CFGPRINTF("%s: Board Revision %s, %s F/W Revision %d.%d.%d\n",
isp->isp_name, revname, dodnld? "loaded" : "resident",
mbs.param[1], mbs.param[2], mbs.param[3]);
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
if (IS_FC(isp)) {
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
if (ISP_READ(isp, BIU2100_CSR) & BIU2100_PCI64) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
CFGPRINTF("%s: in 64-Bit PCI slot\n", isp->isp_name);
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
}
}
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
isp->isp_fwrev[0] = mbs.param[1];
isp->isp_fwrev[1] = mbs.param[2];
isp->isp_fwrev[2] = mbs.param[3];
if (isp->isp_romfw_rev[0] || isp->isp_romfw_rev[1] ||
isp->isp_romfw_rev[2]) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
CFGPRINTF("%s: Last F/W revision was %d.%d.%d\n", isp->isp_name,
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
isp->isp_romfw_rev[0], isp->isp_romfw_rev[1],
isp->isp_romfw_rev[2]);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
}
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
mbs.param[0] = MBOX_GET_FIRMWARE_STATUS;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
PRINTF("%s: could not GET FIRMWARE STATUS\n", isp->isp_name);
return;
}
isp->isp_maxcmds = mbs.param[2];
CFGPRINTF("%s: %d max I/O commands supported\n",
isp->isp_name, mbs.param[2]);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_fw_state(isp);
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
/*
* Set up DMA for the request and result mailboxes.
*/
if (ISP_MBOXDMASETUP(isp) != 0) {
PRINTF("%s: can't setup dma mailboxes\n", isp->isp_name);
return;
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
isp->isp_state = ISP_RESETSTATE;
}
/*
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
* Initialize Parameters of Hardware to a known state.
Locking is now in the outer framework for isp_init/isp_reset- this allows us to call isp_reset/isp_init internally with impunity. Rename isp_phoenix to isp_restart and make it global. Clean it up a tad. If we get an ASYNC_SYSTEM_ERROR code in isp_intr, call isp_restart and return- the f/w is toasted at this point (usually), so we have to bring things back to a known state. In isp_mboxcmd, when we don't see the HOST_INT bit go clear, try and find out if the isp is trying to tell us something and try again. This may avoid a potential deadlock where the previous mailbox command hasn't been cleared by the ISP. In isp_init don't try and get device parameters if we already have them- this typically doesn't work if we're in the middle of an isp_restart.
1998-07-19 01:10:16 +04:00
*
* Locks are held before coming here.
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
*/
Locking is now in the outer framework for isp_init/isp_reset- this allows us to call isp_reset/isp_init internally with impunity. Rename isp_phoenix to isp_restart and make it global. Clean it up a tad. If we get an ASYNC_SYSTEM_ERROR code in isp_intr, call isp_restart and return- the f/w is toasted at this point (usually), so we have to bring things back to a known state. In isp_mboxcmd, when we don't see the HOST_INT bit go clear, try and find out if the isp is trying to tell us something and try again. This may avoid a potential deadlock where the previous mailbox command hasn't been cleared by the ISP. In isp_init don't try and get device parameters if we already have them- this typically doesn't work if we're in the middle of an isp_restart.
1998-07-19 01:10:16 +04:00
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
void
isp_init(isp)
struct ispsoftc *isp;
{
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
/*
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
* Must do this first to get defaults established.
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
*/
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp_setdfltparm(isp, 0);
if (IS_12X0(isp)) {
isp_setdfltparm(isp, 1);
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
}
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
if (IS_FC(isp)) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_fibre_init(isp);
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
} else {
isp_scsi_init(isp);
}
}
static void
isp_scsi_init(isp)
struct ispsoftc *isp;
{
sdparam *sdp_chan0, *sdp_chan1;
mbreg_t mbs;
sdp_chan0 = isp->isp_param;
sdp_chan1 = sdp_chan0;
if (IS_12X0(isp)) {
sdp_chan1++;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
/* First do overall per-card settings. */
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
/*
* If we have fast memory timing enabled, turn it on.
*/
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
if (isp->isp_fast_mttr) {
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
ISP_WRITE(isp, RISC_MTR, 0x1313);
}
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
/*
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
* Set Retry Delay and Count.
* You set both channels at the same time.
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
*/
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
mbs.param[0] = MBOX_SET_RETRY_COUNT;
mbs.param[1] = sdp_chan0->isp_retry_count;
mbs.param[2] = sdp_chan0->isp_retry_delay;
mbs.param[6] = sdp_chan1->isp_retry_count;
mbs.param[7] = sdp_chan1->isp_retry_delay;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_mboxcmd(isp, &mbs);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
PRINTF("%s: failed to set retry count and retry delay\n",
isp->isp_name);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
return;
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
/*
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
* Set ASYNC DATA SETUP time. This is very important.
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
*/
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
mbs.param[0] = MBOX_SET_ASYNC_DATA_SETUP_TIME;
mbs.param[1] = sdp_chan0->isp_async_data_setup;
mbs.param[2] = sdp_chan1->isp_async_data_setup;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_mboxcmd(isp, &mbs);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
PRINTF("%s: failed to set asynchronous data setup time\n",
isp->isp_name);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
return;
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
/*
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
* Set ACTIVE Negation State.
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
*/
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
mbs.param[0] = MBOX_SET_ACT_NEG_STATE;
mbs.param[1] =
(sdp_chan0->isp_req_ack_active_neg << 4) |
(sdp_chan0->isp_data_line_active_neg << 5);
mbs.param[2] =
(sdp_chan1->isp_req_ack_active_neg << 4) |
(sdp_chan1->isp_data_line_active_neg << 5);
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
PRINTF("%s: failed to set active negation state "
"(%d,%d),(%d,%d)\n", isp->isp_name,
sdp_chan0->isp_req_ack_active_neg,
sdp_chan0->isp_data_line_active_neg,
sdp_chan1->isp_req_ack_active_neg,
sdp_chan1->isp_data_line_active_neg);
/*
* But don't return.
*/
}
/*
* Set the Tag Aging limit
*/
mbs.param[0] = MBOX_SET_TAG_AGE_LIMIT;
mbs.param[1] = sdp_chan0->isp_tag_aging;
mbs.param[2] = sdp_chan1->isp_tag_aging;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_mboxcmd(isp, &mbs);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
PRINTF("%s: failed to set tag age limit (%d,%d)\n",
isp->isp_name, sdp_chan0->isp_tag_aging,
sdp_chan1->isp_tag_aging);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
return;
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
/*
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
* Set selection timeout.
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
*/
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
mbs.param[0] = MBOX_SET_SELECT_TIMEOUT;
mbs.param[1] = sdp_chan0->isp_selection_timeout;
mbs.param[2] = sdp_chan1->isp_selection_timeout;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_mboxcmd(isp, &mbs);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
PRINTF("%s: failed to set selection timeout\n", isp->isp_name);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
return;
}
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
/* now do per-channel settings */
isp_scsi_channel_init(isp, 0);
if (IS_12X0(isp))
isp_scsi_channel_init(isp, 1);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
/*
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
* Now enable request/response queues
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
*/
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
mbs.param[0] = MBOX_INIT_RES_QUEUE;
mbs.param[1] = RESULT_QUEUE_LEN;
mbs.param[2] = DMA_MSW(isp->isp_result_dma);
mbs.param[3] = DMA_LSW(isp->isp_result_dma);
mbs.param[4] = 0;
mbs.param[5] = 0;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
PRINTF("%s: set of response queue failed\n", isp->isp_name);
return;
}
isp->isp_residx = 0;
mbs.param[0] = MBOX_INIT_REQ_QUEUE;
mbs.param[1] = RQUEST_QUEUE_LEN;
mbs.param[2] = DMA_MSW(isp->isp_rquest_dma);
mbs.param[3] = DMA_LSW(isp->isp_rquest_dma);
mbs.param[4] = 0;
mbs.param[5] = 0;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_mboxcmd(isp, &mbs);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
PRINTF("%s: set of request queue failed\n", isp->isp_name);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
return;
}
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp->isp_reqidx = isp->isp_reqodx = 0;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
/*
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
* Turn on Fast Posting, LVD transitions
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
*/
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
if (IS_1080(isp) ||
ISP_FW_REVX(isp->isp_fwrev) >= ISP_FW_REV(7, 55, 0)) {
mbs.param[0] = MBOX_SET_FW_FEATURES;
#ifndef ISP_NO_FASTPOST_SCSI
mbs.param[1] |= FW_FEATURE_FAST_POST;
#else
mbs.param[1] = 0;
#endif
if (IS_1080(isp))
mbs.param[1] |= FW_FEATURE_LVD_NOTIFY;
if (mbs.param[1] != 0) {
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
PRINTF("%s: unable enable FW features\n",
isp->isp_name);
}
}
}
/*
* Let the outer layers decide whether to issue a SCSI bus reset.
*/
isp->isp_state = ISP_INITSTATE;
}
static void
isp_scsi_channel_init(isp, channel)
struct ispsoftc *isp;
int channel;
{
sdparam *sdp;
mbreg_t mbs;
int tgt;
sdp = isp->isp_param;
sdp += channel;
/*
* Set (possibly new) Initiator ID.
*/
mbs.param[0] = MBOX_SET_INIT_SCSI_ID;
mbs.param[1] = (channel << 7) | sdp->isp_initiator_id;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_mboxcmd(isp, &mbs);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
PRINTF("%s: cannot set initiator id on bus %d to %d\n",
isp->isp_name, channel, sdp->isp_initiator_id);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
return;
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
/*
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
* Set current per-target parameters to a safe minimum.
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
*/
for (tgt = 0; tgt < MAX_TARGETS; tgt++) {
int maxlun, lun;
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
u_int16_t sdf;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
if (sdp->isp_devparam[tgt].dev_enable == 0) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
IDPRINTF(1, ("%s: skipping target %d bus %d settings\n",
isp->isp_name, tgt, channel));
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
continue;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
/*
* If we're in LVD mode, then we pretty much should
* only disable tagged queuing.
*/
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
if (IS_1080(isp) && sdp->isp_lvdmode) {
sdf = DPARM_DEFAULT & ~DPARM_TQING;
} else {
sdf = DPARM_SAFE_DFLT;
/*
* It is not quite clear when this changed over so that
* we could force narrow and async, so assume >= 7.55.
*/
if (ISP_FW_REVX(isp->isp_fwrev) >=
ISP_FW_REV(7, 55, 0)) {
sdf |= DPARM_NARROW | DPARM_ASYNC;
}
update some debugging stuff- also turn off possible ULTRA mode for 1020s
1998-03-23 01:02:29 +03:00
}
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
mbs.param[0] = MBOX_SET_TARGET_PARAMS;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
mbs.param[1] = (tgt << 8) | (channel << 15);
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
mbs.param[2] = sdf;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
mbs.param[3] =
(sdp->isp_devparam[tgt].sync_offset << 8) |
(sdp->isp_devparam[tgt].sync_period);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_mboxcmd(isp, &mbs);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
sdf = DPARM_SAFE_DFLT;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
mbs.param[0] = MBOX_SET_TARGET_PARAMS;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
mbs.param[1] = (tgt << 8) | (channel << 15);
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
mbs.param[2] = sdf;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
mbs.param[3] =
(sdp->isp_devparam[tgt].sync_offset << 8) |
(sdp->isp_devparam[tgt].sync_period);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
PRINTF("%s: failed even to set defaults for "
"target %d\n", isp->isp_name, tgt);
continue;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
#if 0
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
/*
* We don't update dev_flags with what we've set
* because that's not the ultimate goal setting.
* If we succeed with the command, we *do* update
* cur_dflags by getting target parameters.
*/
mbs.param[0] = MBOX_GET_TARGET_PARAMS;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
mbs.param[1] = (tgt << 8) | (channel << 15);
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
/*
* Urrr.... We'll set cur_dflags to DPARM_SAFE_DFLT so
* we don't try and do tags if tags aren't enabled.
*/
sdp->isp_devparam[tgt].cur_dflags = DPARM_SAFE_DFLT;
} else {
sdp->isp_devparam[tgt].cur_dflags = mbs.param[2];
sdp->isp_devparam[tgt].cur_offset = mbs.param[3] >> 8;
sdp->isp_devparam[tgt].cur_period = mbs.param[3] & 0xff;
}
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
IDPRINTF(3, ("%s: set flags 0x%x got 0x%x back for target %d\n",
isp->isp_name, sdf, mbs.param[2], tgt));
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
#else
/*
* We don't update any information because we need to run
* at least one command per target to cause a new state
* to be latched.
*/
#endif
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
/*
* Ensure that we don't believe tagged queuing is enabled yet.
* It turns out that sometimes the ISP just ignores our
* attempts to set parameters for devices that it hasn't
* seen yet.
*/
sdp->isp_devparam[tgt].cur_dflags &= ~DPARM_TQING;
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
if (ISP_FW_REVX(isp->isp_fwrev) >= ISP_FW_REV(7, 55, 0))
maxlun = 32;
else
maxlun = 8;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
for (lun = 0; lun < maxlun; lun++) {
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
mbs.param[0] = MBOX_SET_DEV_QUEUE_PARAMS;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
mbs.param[1] = (channel << 15) | (tgt << 8) | lun;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
mbs.param[2] = sdp->isp_max_queue_depth;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
mbs.param[3] = sdp->isp_devparam[tgt].exc_throttle;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_mboxcmd(isp, &mbs);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
PRINTF("%s: failed to set device queue "
"parameters for target %d, lun %d\n",
isp->isp_name, tgt, lun);
break;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Locking is now in the outer framework for isp_init/isp_reset- this allows us to call isp_reset/isp_init internally with impunity. Rename isp_phoenix to isp_restart and make it global. Clean it up a tad. If we get an ASYNC_SYSTEM_ERROR code in isp_intr, call isp_restart and return- the f/w is toasted at this point (usually), so we have to bring things back to a known state. In isp_mboxcmd, when we don't see the HOST_INT bit go clear, try and find out if the isp is trying to tell us something and try again. This may avoid a potential deadlock where the previous mailbox command hasn't been cleared by the ISP. In isp_init don't try and get device parameters if we already have them- this typically doesn't work if we're in the middle of an isp_restart.
1998-07-19 01:10:16 +04:00
/*
* Fibre Channel specific initialization.
*
* Locks are held before coming here.
*/
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
static void
isp_fibre_init(isp)
struct ispsoftc *isp;
{
fcparam *fcp;
isp_icb_t *icbp;
mbreg_t mbs;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
int loopid;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
fcp = isp->isp_param;
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
/*
* For systems that don't have BIOS methods for which
* we can easily change the NVRAM based loopid, we'll
* override that here. Note that when we initialize
* the firmware we may get back a different loopid than
* we asked for anyway. XXX This is probably not the
* best way to figure this out XXX
*/
#ifndef __i386__
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
loopid = DEFAULT_LOOPID(isp);
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
#else
loopid = fcp->isp_loopid;
#endif
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
icbp = (isp_icb_t *) fcp->isp_scratch;
+ Some basic deadchip detection. + Enable FIFO bursts, but also detect bogus 1040A with busted FIFO. + Use new MEMZERO crossplatform define. + Handle RQCS_QUEUE_FULL status case and let upper layer parse SCSI ststus byte if nonzero (should be 0x28- Queue Full status) + Fold ISP_NVRAM_FIFO_THRESHOLD_128 into isp_fifo_threshold tag.
1999-01-10 08:07:53 +03:00
MEMZERO(icbp, sizeof (*icbp));
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
icbp->icb_version = ICB_VERSION1;
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
#ifdef ISP_TARGET_MODE
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
fcp->isp_fwoptions = ICBOPT_TGT_ENABLE;
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
#else
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
fcp->isp_fwoptions = 0;
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
#endif
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
fcp->isp_fwoptions |= ICBOPT_FAIRNESS;
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
fcp->isp_fwoptions |= ICBOPT_PDBCHANGE_AE;
fcp->isp_fwoptions |= ICBOPT_HARD_ADDRESS;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
/*
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
* We have to use FULL LOGIN even though it resets the loop too much
* because otherwise port database entries don't get updated after
* a LIP- this is a known f/w bug.
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
*/
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (ISP_FW_REVX(isp->isp_fwrev) < ISP_FW_REV(1, 17, 0)) {
fcp->isp_fwoptions |= ICBOPT_FULL_LOGIN;
}
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
#ifndef ISP_NO_FASTPOST_FC
fcp->isp_fwoptions |= ICBOPT_FAST_POST;
#endif
if (isp->isp_confopts & ISP_CFG_FULL_DUPLEX)
fcp->isp_fwoptions |= ICBOPT_FULL_DUPLEX;
/*
* We don't set ICBOPT_PORTNAME because we want our
* Node Name && Port Names to be distinct.
*/
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
icbp->icb_fwoptions = fcp->isp_fwoptions;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
icbp->icb_maxfrmlen = fcp->isp_maxfrmlen;
if (icbp->icb_maxfrmlen < ICB_MIN_FRMLEN ||
icbp->icb_maxfrmlen > ICB_MAX_FRMLEN) {
PRINTF("%s: bad frame length (%d) from NVRAM- using %d\n",
isp->isp_name, fcp->isp_maxfrmlen, ICB_DFLT_FRMLEN);
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
icbp->icb_maxfrmlen = ICB_DFLT_FRMLEN;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
}
icbp->icb_maxalloc = fcp->isp_maxalloc;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
if (icbp->icb_maxalloc < 1) {
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
PRINTF("%s: bad maximum allocation (%d)- using 16\n",
isp->isp_name, fcp->isp_maxalloc);
icbp->icb_maxalloc = 16;
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
icbp->icb_execthrottle = fcp->isp_execthrottle;
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
if (icbp->icb_execthrottle < 1) {
PRINTF("%s: bad execution throttle of %d- using 16\n",
isp->isp_name, fcp->isp_execthrottle);
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
icbp->icb_execthrottle = ICB_DFLT_THROTTLE;
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
icbp->icb_retry_delay = fcp->isp_retry_delay;
icbp->icb_retry_count = fcp->isp_retry_count;
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
icbp->icb_hardaddr = loopid;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
icbp->icb_logintime = 30; /* 30 second login timeout */
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
if (fcp->isp_nodewwn) {
u_int64_t pn;
MAKE_NODE_NAME_FROM_WWN(icbp->icb_nodename, fcp->isp_nodewwn);
if (fcp->isp_portwwn) {
pn = fcp->isp_portwwn;
} else {
pn = fcp->isp_nodewwn |
(((u_int64_t)(isp->isp_unit+1)) << 56);
}
/*
* If the top nibble is 2, we can construct a port name
* from the node name by setting a nonzero instance in
* bits 56..59. Otherwise, we need to make it identical
* to Node name...
*/
if ((fcp->isp_nodewwn >> 60) == 2) {
MAKE_NODE_NAME_FROM_WWN(icbp->icb_portname, pn);
} else {
MAKE_NODE_NAME_FROM_WWN(icbp->icb_portname,
fcp->isp_nodewwn);
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
}
} else {
fcp->isp_fwoptions &= ~(ICBOPT_USE_PORTNAME|ICBOPT_FULL_LOGIN);
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
}
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
icbp->icb_rqstqlen = RQUEST_QUEUE_LEN;
icbp->icb_rsltqlen = RESULT_QUEUE_LEN;
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
icbp->icb_rqstaddr[RQRSP_ADDR0015] = DMA_LSW(isp->isp_rquest_dma);
icbp->icb_rqstaddr[RQRSP_ADDR1631] = DMA_MSW(isp->isp_rquest_dma);
icbp->icb_respaddr[RQRSP_ADDR0015] = DMA_LSW(isp->isp_result_dma);
icbp->icb_respaddr[RQRSP_ADDR1631] = DMA_MSW(isp->isp_result_dma);
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
ISP_SWIZZLE_ICB(isp, icbp);
/*
* Do this *before* initializing the firmware.
*/
isp_mark_getpdb_all(isp);
fcp->isp_fwstate = FW_CONFIG_WAIT;
fcp->isp_loopstate = LOOP_NIL;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
MemoryBarrier();
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
for (;;) {
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
mbs.param[0] = MBOX_INIT_FIRMWARE;
mbs.param[1] = 0;
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
mbs.param[2] = DMA_MSW(fcp->isp_scdma);
mbs.param[3] = DMA_LSW(fcp->isp_scdma);
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
mbs.param[4] = 0;
mbs.param[5] = 0;
mbs.param[6] = 0;
mbs.param[7] = 0;
isp_mboxcmd(isp, &mbs);
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
PRINTF("%s: INIT FIRMWARE failed (code 0x%x)\n",
isp->isp_name, mbs.param[0]);
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (mbs.param[0] & 0x8000) {
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
SYS_DELAY(1000);
continue;
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
}
return;
}
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
break;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
isp->isp_reqidx = isp->isp_reqodx = 0;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp->isp_residx = 0;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
isp->isp_sendmarker = 1;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
/*
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
* Whatever happens, we're now committed to being here.
*/
isp->isp_state = ISP_INITSTATE;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
#ifdef ISP_TARGET_MODE
if (isp_modify_lun(isp, 0, 1, 1)) {
PRINTF("%s: failed to enable target mode\n", isp->isp_name);
}
#endif
}
/*
* Fibre Channel Support- get the port database for the id.
*
* Locks are held before coming here. Return 0 if success,
* else failure.
*/
static void
isp_mark_getpdb_all(isp)
struct ispsoftc *isp;
{
fcparam *fcp = (fcparam *) isp->isp_param;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
int i;
for (i = 0; i < MAX_FC_TARG; i++) {
fcp->portdb[i].valid = 0;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
}
}
static int
isp_getpdb(isp, id, pdbp)
struct ispsoftc *isp;
int id;
isp_pdb_t *pdbp;
{
fcparam *fcp = (fcparam *) isp->isp_param;
mbreg_t mbs;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
mbs.param[0] = MBOX_GET_PORT_DB;
mbs.param[1] = id << 8;
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
mbs.param[2] = DMA_MSW(fcp->isp_scdma);
mbs.param[3] = DMA_LSW(fcp->isp_scdma);
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
/*
* Unneeded. For the 2100, except for initializing f/w, registers
* 4/5 have to not be written to.
* mbs.param[4] = 0;
* mbs.param[5] = 0;
*
*/
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
mbs.param[6] = 0;
mbs.param[7] = 0;
isp_mboxcmd(isp, &mbs);
switch (mbs.param[0]) {
case MBOX_COMMAND_COMPLETE:
MemoryBarrier();
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
ISP_UNSWIZZLE_AND_COPY_PDBP(isp, pdbp, fcp->isp_scratch);
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
break;
case MBOX_HOST_INTERFACE_ERROR:
PRINTF("%s: DMA error getting port database\n", isp->isp_name);
return (-1);
case MBOX_COMMAND_PARAM_ERROR:
/* Not Logged In */
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
IDPRINTF(3, ("%s: Param Error on Get Port Database for id %d\n",
isp->isp_name, id));
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
return (-1);
default:
PRINTF("%s: error 0x%x getting port database for ID %d\n",
isp->isp_name, mbs.param[0], id);
return (-1);
}
return (0);
}
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
static u_int64_t
isp_get_portname(isp, loopid, nodename)
struct ispsoftc *isp;
int loopid;
int nodename;
{
u_int64_t wwn = 0;
mbreg_t mbs;
mbs.param[0] = MBOX_GET_PORT_NAME;
mbs.param[1] = loopid << 8;
if (nodename)
mbs.param[1] |= 1;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] == MBOX_COMMAND_COMPLETE) {
wwn =
(((u_int64_t)(mbs.param[2] & 0xff)) << 56) |
(((u_int64_t)(mbs.param[2] >> 8)) << 48) |
(((u_int64_t)(mbs.param[3] & 0xff)) << 40) |
(((u_int64_t)(mbs.param[3] >> 8)) << 32) |
(((u_int64_t)(mbs.param[6] & 0xff)) << 24) |
(((u_int64_t)(mbs.param[6] >> 8)) << 16) |
(((u_int64_t)(mbs.param[7] & 0xff)) << 8) |
(((u_int64_t)(mbs.param[7] >> 8)));
}
return (wwn);
}
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
/*
* Make sure we have good FC link and know our Loop ID.
*/
static int
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
isp_fclink_test(isp, waitdelay)
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
struct ispsoftc *isp;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
int waitdelay;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
{
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
static char *toponames[] = {
"Private Loop",
"FL Port",
"N-Port to N-Port",
"F Port"
};
char *tname;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
mbreg_t mbs;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
int count, topo = -1;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
u_int8_t lwfs;
fcparam *fcp;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
#if defined(ISP2100_FABRIC)
isp_pdb_t pdb;
#endif
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
fcp = isp->isp_param;
/*
* Wait up to N microseconds for F/W to go to a ready state.
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
*/
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
lwfs = FW_CONFIG_WAIT;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
for (count = 0; count < waitdelay; count += 100) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_fw_state(isp);
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
if (lwfs != fcp->isp_fwstate) {
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
PRINTF("%s: Firmware State %s -> %s\n",
isp->isp_name, isp2100_fw_statename((int)lwfs),
isp2100_fw_statename((int)fcp->isp_fwstate));
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
lwfs = fcp->isp_fwstate;
}
if (fcp->isp_fwstate == FW_READY) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
break;
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
}
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
SYS_DELAY(100); /* wait 100 microseconds */
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
/*
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
* If we haven't gone to 'ready' state, return.
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
*/
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
if (fcp->isp_fwstate != FW_READY) {
return (-1);
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
}
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
/*
* Get our Loop ID (if possible). We really need to have it.
*/
mbs.param[0] = MBOX_GET_LOOP_ID;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
PRINTF("%s: GET LOOP ID failed\n", isp->isp_name);
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
return (-1);
}
fcp->isp_loopid = mbs.param[1];
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (isp->isp_type == ISP_HA_FC_2100) {
if (ISP_FW_REVX(isp->isp_fwrev) >= ISP_FW_REV(2, 0, 14)) {
topo = (int) mbs.param[6];
}
} else if (isp->isp_type == ISP_HA_FC_2100) {
if (ISP_FW_REVX(isp->isp_fwrev) >= ISP_FW_REV(1, 17, 26)) {
topo = (int) mbs.param[6];
}
}
if (topo < 0 || topo > 3)
tname = "unknown";
else
tname = toponames[topo];
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
/*
* If we're not on a fabric, the low 8 bits will be our AL_PA.
* If we're on a fabric, the low 8 bits will still be our AL_PA.
*/
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
fcp->isp_alpa = mbs.param[2];
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
#if defined(ISP2100_FABRIC)
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
fcp->isp_onfabric = 0;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
if (isp_getpdb(isp, FL_PORT_ID, &pdb) == 0) {
fcp->isp_portid = mbs.param[2] | (((int)mbs.param[3]) << 16);
fcp->isp_onfabric = 1;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
CFGPRINTF("%s: Loop ID %d, AL_PA 0x%x, Port ID 0x%x Loop State "
"0x%x topology %s\n", isp->isp_name, fcp->isp_loopid,
fcp->isp_alpa, fcp->isp_portid, fcp->isp_loopstate, tname);
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
/*
* Make sure we're logged out of all fabric devices.
*/
for (count = FC_SNS_ID+1; count < MAX_FC_TARG; count++) {
struct lportdb *lp = &fcp->portdb[count];
if (lp->valid == 0 || lp->fabdev == 0)
continue;
PRINTF("%s: logging out target %d at Loop ID %d "
"(port id 0x%x)\n", isp->isp_name, count,
lp->loopid, lp->portid);
mbs.param[0] = MBOX_FABRIC_LOGOUT;
mbs.param[1] = lp->loopid << 8;
mbs.param[2] = 0;
mbs.param[3] = 0;
isp_mboxcmd(isp, &mbs);
}
} else
#endif
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
CFGPRINTF("%s: Loop ID %d, ALPA 0x%x Loop State 0x%x topology %s\n",
isp->isp_name, fcp->isp_loopid, fcp->isp_alpa, fcp->isp_loopstate,
tname);
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
fcp->loop_seen_once = 1;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
return (0);
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
}
/*
* Compare two local port db entities and return 1 if they're the same, else 0.
*/
static int
isp_same_lportdb(a, b)
struct lportdb *a, *b;
{
/*
* We decide two lports are the same if they have non-zero and
* identical port WWNs and identical loop IDs.
*/
if (a->port_wwn == 0 || a->port_wwn != b->port_wwn ||
a->loopid != b->loopid) {
return (0);
} else {
return (1);
}
}
/*
* Synchronize our soft copy of the port database with what the f/w thinks
* (with a view toward possibly for a specific target....)
*/
static int
isp_pdb_sync(isp, target)
struct ispsoftc *isp;
int target;
{
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
struct lportdb *lp, *tport;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
fcparam *fcp = isp->isp_param;
isp_pdb_t pdb;
int loopid, lim;
#ifdef ISP2100_FABRIC
/*
* XXX: If we do this *after* building up our local port database,
* XXX: the commands simply don't work.
*/
/*
* (Re)discover all fabric devices
*/
if (fcp->isp_onfabric)
(void) isp_scan_fabric(isp);
#endif
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
/*
* Run through the local loop ports and get port database info
* for each loop ID.
*
* There's a somewhat unexplained situation where the f/w passes back
* the wrong database entity- if that happens, just restart (up to
* FL_PORT_ID times).
*/
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
tport = fcp->tport;
MEMZERO((void *) tport, sizeof (tport));
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
for (lim = loopid = 0; loopid < FL_PORT_ID; loopid++) {
/*
* make sure the temp port database is clean...
*/
lp = &tport[loopid];
lp->node_wwn = isp_get_portname(isp, loopid, 1);
if (lp->node_wwn == 0)
continue;
lp->port_wwn = isp_get_portname(isp, loopid, 0);
if (lp->port_wwn == 0) {
lp->node_wwn = 0;
continue;
}
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
/*
* Get an entry....
*/
if (isp_getpdb(isp, loopid, &pdb) != 0) {
continue;
}
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
/*
* If the returned database element doesn't match what we
* asked for, restart the process entirely (up to a point...).
*/
if (pdb.pdb_loopid != loopid) {
IDPRINTF(0, ("%s: wankage (%d != %d)\n",
isp->isp_name, pdb.pdb_loopid, loopid));
loopid = 0;
if (lim++ < FL_PORT_ID) {
continue;
}
PRINTF("%s: giving up on synchronizing the port "
"database\n", isp->isp_name);
return (-1);
}
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
/*
* Save the pertinent info locally.
*/
lp->node_wwn =
(((u_int64_t)pdb.pdb_nodename[0]) << 56) |
(((u_int64_t)pdb.pdb_nodename[1]) << 48) |
(((u_int64_t)pdb.pdb_nodename[2]) << 40) |
(((u_int64_t)pdb.pdb_nodename[3]) << 32) |
(((u_int64_t)pdb.pdb_nodename[4]) << 24) |
(((u_int64_t)pdb.pdb_nodename[5]) << 16) |
(((u_int64_t)pdb.pdb_nodename[6]) << 8) |
(((u_int64_t)pdb.pdb_nodename[7]));
lp->port_wwn =
(((u_int64_t)pdb.pdb_portname[0]) << 56) |
(((u_int64_t)pdb.pdb_portname[1]) << 48) |
(((u_int64_t)pdb.pdb_portname[2]) << 40) |
(((u_int64_t)pdb.pdb_portname[3]) << 32) |
(((u_int64_t)pdb.pdb_portname[4]) << 24) |
(((u_int64_t)pdb.pdb_portname[5]) << 16) |
(((u_int64_t)pdb.pdb_portname[6]) << 8) |
(((u_int64_t)pdb.pdb_portname[7]));
lp->roles =
(pdb.pdb_prli_svc3 & SVC3_ROLE_MASK) >> SVC3_ROLE_SHIFT;
lp->portid = BITS2WORD(pdb.pdb_portid_bits);
lp->loopid = pdb.pdb_loopid;
/*
* Do a quick check to see whether this matches the saved port
* database for the same loopid. We do this here to save
* searching later (if possible). Note that this fails over
* time as things shuffle on the loop- we get the current
* loop state (where loop id as an index matches loop id in
* use) and then compare it to our saved database which
* never shifts.
*/
if (isp_same_lportdb(lp, &fcp->portdb[target])) {
lp->valid = 1;
}
}
/*
* If we get this far, we've settled our differences with the f/w
* and we can say that the loop state is ready.
*/
fcp->isp_loopstate = LOOP_READY;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
/*
* Mark all of the permanent local loop database entries as invalid.
*/
for (loopid = 0; loopid < FL_PORT_ID; loopid++) {
fcp->portdb[loopid].valid = 0;
}
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
/*
* Now merge our local copy of the port database into our saved copy.
* Notify the outer layers of new devices arriving.
*/
for (loopid = 0; loopid < FL_PORT_ID; loopid++) {
int i;
/*
* If we don't have a non-zero Port WWN, we're not here.
*/
if (tport[loopid].port_wwn == 0) {
continue;
}
/*
* If we've already marked our tmp copy as valid,
* this means that we've decided that it's the
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
* same as our saved data base. This didn't include
* the 'valid' marking so we have set that here.
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
*/
if (tport[loopid].valid) {
fcp->portdb[loopid].valid = 1;
continue;
}
/*
* For the purposes of deciding whether this is the
* 'same' device or not, we only search for an identical
* Port WWN. Node WWNs may or may not be the same as
* the Port WWN, and there may be multiple different
* Port WWNs with the same Node WWN. It would be chaos
* to have multiple identical Port WWNs, so we don't
* allow that.
*/
for (i = 0; i < FL_PORT_ID; i++) {
int j;
if (fcp->portdb[i].port_wwn == 0)
continue;
if (fcp->portdb[i].port_wwn != tport[loopid].port_wwn)
continue;
/*
* We found this WWN elsewhere- it's changed
* loopids then. We don't change it's actual
* position in our cached port database- we
* just change the actual loop ID we'd use.
*/
if (fcp->portdb[i].loopid != loopid) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
PRINTF("%s: Target ID %d Loop 0x%x (Port 0x%x) "
"=> Loop 0x%x (Port 0x%x) \n",
isp->isp_name, i, fcp->portdb[i].loopid,
fcp->portdb[i].portid, loopid,
tport[loopid].portid);
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
}
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
fcp->portdb[i].portid = tport[loopid].portid;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
fcp->portdb[i].loopid = loopid;
fcp->portdb[i].valid = 1;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
/*
* XXX: Should we also propagate roles in case they
* XXX: changed?
*/
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
/*
* Now make sure this Port WWN doesn't exist elsewhere
* in the port database.
*/
for (j = i+1; j < FL_PORT_ID; j++) {
if (fcp->portdb[i].port_wwn !=
fcp->portdb[j].port_wwn) {
continue;
}
PRINTF("%s: Target ID %d Duplicates Target ID "
"%d- killing off both\n",
isp->isp_name, j, i);
/*
* Invalidate the 'old' *and* 'new' ones.
* This is really harsh and not quite right,
* but if this happens, we really don't know
* who is what at this point.
*/
fcp->portdb[i].valid = 0;
fcp->portdb[j].valid = 0;
}
break;
}
/*
* If we didn't traverse the entire port database,
* then we found (and remapped) an existing entry.
* No need to notify anyone- go for the next one.
*/
if (i < FL_PORT_ID) {
continue;
}
/*
* We've not found this Port WWN anywhere. It's a new entry.
* See if we can leave it where it is (with target == loopid).
*/
if (fcp->portdb[loopid].port_wwn != 0) {
for (lim = 0; lim < FL_PORT_ID; lim++) {
if (fcp->portdb[lim].port_wwn == 0)
break;
}
/* "Cannot Happen" */
if (lim == FL_PORT_ID) {
PRINTF("%s: remap overflow?\n", isp->isp_name);
continue;
}
i = lim;
} else {
i = loopid;
}
/*
* NB: The actual loopid we use here is loopid- we may
* in fact be at a completely different index (target).
*/
fcp->portdb[i].loopid = loopid;
fcp->portdb[i].port_wwn = tport[loopid].port_wwn;
fcp->portdb[i].node_wwn = tport[loopid].node_wwn;
fcp->portdb[i].roles = tport[loopid].roles;
fcp->portdb[i].portid = tport[loopid].portid;
fcp->portdb[i].valid = 1;
/*
* Tell the outside world we've arrived.
*/
(void) isp_async(isp, ISPASYNC_PDB_CHANGED, &i);
}
/*
* Now find all previously used targets that are now invalid and
* notify the outer layers that they're gone.
*/
for (lp = fcp->portdb; lp < &fcp->portdb[FL_PORT_ID]; lp++) {
if (lp->valid || lp->port_wwn == 0)
continue;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
/*
* Tell the outside world we've gone away;
*/
loopid = lp - fcp->portdb;
(void) isp_async(isp, ISPASYNC_PDB_CHANGED, &loopid);
MEMZERO((void *) lp, sizeof (*lp));
}
#ifdef ISP2100_FABRIC
/*
* Now log in any fabric devices
*/
for (lp = &fcp->portdb[FC_SNS_ID+1];
lp < &fcp->portdb[MAX_FC_TARG]; lp++) {
mbreg_t mbs;
/*
* Nothing here?
*/
if (lp->port_wwn == 0)
continue;
/*
* Don't try to log into yourself.
*/
if (lp->portid == fcp->isp_portid)
continue;
/*
* Force a logout.
*/
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
lp->loopid = loopid = lp - fcp->portdb;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
mbs.param[0] = MBOX_FABRIC_LOGOUT;
mbs.param[1] = lp->loopid << 8;
mbs.param[2] = 0;
mbs.param[3] = 0;
isp_mboxcmd(isp, &mbs);
/*
* And log in....
*/
mbs.param[0] = MBOX_FABRIC_LOGIN;
mbs.param[1] = lp->loopid << 8;
mbs.param[2] = lp->portid >> 16;
mbs.param[3] = lp->portid & 0xffff;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] == MBOX_COMMAND_COMPLETE) {
lp->valid = 1;
lp->fabdev = 1;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (isp_getpdb(isp, loopid, &pdb) != 0) {
/*
* Be kind...
*/
lp->roles = (SVC3_TGT_ROLE >> SVC3_ROLE_SHIFT);
PRINTF("%s: Faked PortID 0x%x into LoopID %d\n",
isp->isp_name, lp->portid, lp->loopid);
} else if (pdb.pdb_loopid != lp->loopid) {
lp->roles = (SVC3_TGT_ROLE >> SVC3_ROLE_SHIFT);
PRINTF("%s: Wanked PortID 0x%x to LoopID %d\n",
isp->isp_name, lp->portid, lp->loopid);
} else {
lp->roles =
(pdb.pdb_prli_svc3 & SVC3_ROLE_MASK) >>
SVC3_ROLE_SHIFT;
lp->portid = BITS2WORD(pdb.pdb_portid_bits);
lp->loopid = pdb.pdb_loopid;
lp->node_wwn =
(((u_int64_t)pdb.pdb_nodename[0]) << 56) |
(((u_int64_t)pdb.pdb_nodename[1]) << 48) |
(((u_int64_t)pdb.pdb_nodename[2]) << 40) |
(((u_int64_t)pdb.pdb_nodename[3]) << 32) |
(((u_int64_t)pdb.pdb_nodename[4]) << 24) |
(((u_int64_t)pdb.pdb_nodename[5]) << 16) |
(((u_int64_t)pdb.pdb_nodename[6]) << 8) |
(((u_int64_t)pdb.pdb_nodename[7]));
lp->port_wwn =
(((u_int64_t)pdb.pdb_portname[0]) << 56) |
(((u_int64_t)pdb.pdb_portname[1]) << 48) |
(((u_int64_t)pdb.pdb_portname[2]) << 40) |
(((u_int64_t)pdb.pdb_portname[3]) << 32) |
(((u_int64_t)pdb.pdb_portname[4]) << 24) |
(((u_int64_t)pdb.pdb_portname[5]) << 16) |
(((u_int64_t)pdb.pdb_portname[6]) << 8) |
(((u_int64_t)pdb.pdb_portname[7]));
(void) isp_async(isp, ISPASYNC_PDB_CHANGED,
&loopid);
}
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
}
}
#endif
return (0);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
#ifdef ISP2100_FABRIC
static int
isp_scan_fabric(isp)
struct ispsoftc *isp;
{
fcparam *fcp = isp->isp_param;
u_int32_t portid, first_nz_portid;
sns_screq_t *reqp;
sns_scrsp_t *resp;
mbreg_t mbs;
int hicap;
reqp = (sns_screq_t *) fcp->isp_scratch;
resp = (sns_scrsp_t *) (&((char *)fcp->isp_scratch)[0x100]);
first_nz_portid = portid = fcp->isp_portid;
for (hicap = 0; hicap < 1024; hicap++) {
MEMZERO((void *) reqp, SNS_GAN_REQ_SIZE);
reqp->snscb_rblen = SNS_GAN_RESP_SIZE >> 1;
reqp->snscb_addr[RQRSP_ADDR0015] =
DMA_LSW(fcp->isp_scdma + 0x100);
reqp->snscb_addr[RQRSP_ADDR1631] =
DMA_MSW(fcp->isp_scdma + 0x100);
reqp->snscb_sblen = 6;
reqp->snscb_data[0] = SNS_GAN;
reqp->snscb_data[4] = portid & 0xffff;
reqp->snscb_data[5] = (portid >> 16) & 0xff;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
ISP_SWIZZLE_SNS_REQ(isp, reqp);
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
mbs.param[0] = MBOX_SEND_SNS;
mbs.param[1] = SNS_GAN_REQ_SIZE >> 1;
mbs.param[2] = DMA_MSW(fcp->isp_scdma);
mbs.param[3] = DMA_LSW(fcp->isp_scdma);
mbs.param[6] = 0;
mbs.param[7] = 0;
MemoryBarrier();
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
return (-1);
}
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
ISP_UNSWIZZLE_SNS_RSP(isp, resp, SNS_GAN_RESP_SIZE >> 1);
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
portid = (((u_int32_t) resp->snscb_port_id[0]) << 16) |
(((u_int32_t) resp->snscb_port_id[1]) << 8) |
(((u_int32_t) resp->snscb_port_id[2]));
if (isp_async(isp, ISPASYNC_FABRIC_DEV, resp)) {
return (-1);
}
if (first_nz_portid == 0 && portid) {
first_nz_portid = portid;
}
if (first_nz_portid == portid) {
return (0);
}
}
/*
* We either have a broken name server or a huge fabric if we get here.
*/
return (0);
}
#endif
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
/*
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
* Start a command. Locking is assumed done in the caller.
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
*/
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
int32_t
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
ispscsicmd(xs)
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
ISP_SCSI_XFER_T *xs;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
{
struct ispsoftc *isp;
u_int8_t iptr, optr;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
union {
ispreq_t *_reqp;
ispreqt2_t *_t2reqp;
} _u;
#define reqp _u._reqp
#define t2reqp _u._t2reqp
Add a new async notification, and also fix a bug where the queue entry wasn't getting fully zeroed as it should be.
1997-12-05 22:41:49 +03:00
#define UZSIZE max(sizeof (ispreq_t), sizeof (ispreqt2_t))
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
int target, i;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_INITERR(xs);
isp = XS_ISP(xs);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
Locking is now in the outer framework for isp_init/isp_reset- this allows us to call isp_reset/isp_init internally with impunity. Rename isp_phoenix to isp_restart and make it global. Clean it up a tad. If we get an ASYNC_SYSTEM_ERROR code in isp_intr, call isp_restart and return- the f/w is toasted at this point (usually), so we have to bring things back to a known state. In isp_mboxcmd, when we don't see the HOST_INT bit go clear, try and find out if the isp is trying to tell us something and try again. This may avoid a potential deadlock where the previous mailbox command hasn't been cleared by the ISP. In isp_init don't try and get device parameters if we already have them- this typically doesn't work if we're in the middle of an isp_restart.
1998-07-19 01:10:16 +04:00
if (isp->isp_state != ISP_RUNSTATE) {
PRINTF("%s: adapter not ready\n", isp->isp_name);
XS_SETERR(xs, HBA_BOTCH);
return (CMD_COMPLETE);
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
/*
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
* We *could* do the different sequence type that has close
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
* to the whole Queue Entry for the command...
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
*/
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
if (XS_CDBLEN(xs) > (IS_FC(isp) ? 16 : 12) || XS_CDBLEN(xs) == 0) {
PRINTF("%s: unsupported cdb length (%d, CDB[0]=0x%x)\n",
isp->isp_name, XS_CDBLEN(xs), XS_CDBP(xs)[0]);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
XS_SETERR(xs, HBA_BOTCH);
return (CMD_COMPLETE);
}
/*
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
* Check to see whether we have good firmware state still or
* need to refresh our port database for this target.
*/
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
target = XS_TGT(xs);
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
if (IS_FC(isp)) {
fcparam *fcp = isp->isp_param;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
struct lportdb *lp;
#if defined(ISP2100_FABRIC)
if (target >= FL_PORT_ID) {
/*
* If we're not on a Fabric, we can't have a target
* above FL_PORT_ID-1. If we're on a fabric, we
* can't have a target less than FC_SNS_ID+1.
*/
if (fcp->isp_onfabric == 0 || target <= FC_SNS_ID) {
XS_SETERR(xs, HBA_SELTIMEOUT);
return (CMD_COMPLETE);
}
}
#endif
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
/*
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
* Check for f/w being in ready state. If the f/w
* isn't in ready state, then we don't know our
* loop ID and the f/w hasn't completed logging
* into all targets on the loop. If this is the
* case, then bounce the command. We pretend this is
* a SELECTION TIMEOUT error if we've never gone to
* FW_READY state at all- in this case we may not
* be hooked to a loop at all and we shouldn't hang
* the machine for this. Otherwise, defer this command
* until later.
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
*/
if (fcp->isp_fwstate != FW_READY) {
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
if (isp_fclink_test(isp, FC_FW_READY_DELAY)) {
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
XS_SETERR(xs, HBA_SELTIMEOUT);
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
if (fcp->loop_seen_once) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
return (CMD_RQLATER);
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
} else {
return (CMD_COMPLETE);
}
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
}
}
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
/*
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
* If our loop state is such that we haven't yet received
* a "Port Database Changed" notification (after a LIP or
* a Loop Reset or firmware initialization), then defer
* sending commands for a little while.
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
*/
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
if (fcp->isp_loopstate < LOOP_PDB_RCVD) {
XS_SETERR(xs, HBA_SELTIMEOUT);
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
return (CMD_RQLATER);
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
}
/*
* If our loop state is now such that we've just now
* received a Port Database Change notification, then
* we have to go off and (re)synchronize our
*/
if (fcp->isp_loopstate == LOOP_PDB_RCVD) {
if (isp_pdb_sync(isp, target)) {
XS_SETERR(xs, HBA_SELTIMEOUT);
return (CMD_COMPLETE);
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
}
}
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
/*
* Now check whether we should even think about pursuing this.
*/
lp = &fcp->portdb[target];
if (lp->valid == 0) {
XS_SETERR(xs, HBA_SELTIMEOUT);
return (CMD_COMPLETE);
}
if ((lp->roles & (SVC3_TGT_ROLE >> SVC3_ROLE_SHIFT)) == 0) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
IDPRINTF(3, ("%s: target %d is not a target\n",
isp->isp_name, target));
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
XS_SETERR(xs, HBA_SELTIMEOUT);
return (CMD_COMPLETE);
}
/*
* Now turn target into what the actual loop ID is.
*/
target = lp->loopid;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
}
/*
* Next check to see if any HBA or Device
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
* parameters need to be updated.
*/
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
if (isp->isp_update != 0) {
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
isp_update(isp);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
optr = isp->isp_reqodx = ISP_READ(isp, OUTMAILBOX4);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
iptr = isp->isp_reqidx;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
reqp = (ispreq_t *) ISP_QUEUE_ENTRY(isp->isp_rquest, iptr);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
iptr = ISP_NXT_QENTRY(iptr, RQUEST_QUEUE_LEN);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (iptr == optr) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
IDPRINTF(0, ("%s: Request Queue Overflow\n", isp->isp_name));
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_SETERR(xs, HBA_BOTCH);
return (CMD_EAGAIN);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Some formatting cleanup, and for the ISP2100, the flags hadn't being set quite right, but things worked okay anyhow.
1997-09-10 06:18:19 +04:00
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
/*
* Now see if we need to synchronize the ISP with respect to anything.
* We do dual duty here (cough) for synchronizing for busses other
* than which we got here to send a command to.
*/
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (isp->isp_sendmarker) {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
u_int8_t niptr, n = (IS_12X0(isp)? 2: 1);
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
/*
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
* Check ports to send markers for...
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
*/
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
for (i = 0; i < n; i++) {
if ((isp->isp_sendmarker & (1 << i)) == 0) {
continue;
}
MEMZERO((void *) reqp, sizeof (*reqp));
reqp->req_header.rqs_entry_count = 1;
reqp->req_header.rqs_entry_type = RQSTYPE_MARKER;
reqp->req_modifier = SYNC_ALL;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
reqp->req_target = i << 7; /* insert bus number */
ISP_SWIZZLE_REQUEST(isp, reqp);
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
/*
* Unconditionally update the input pointer anyway.
*/
ISP_WRITE(isp, INMAILBOX4, iptr);
isp->isp_reqidx = iptr;
niptr = ISP_NXT_QENTRY(iptr, RQUEST_QUEUE_LEN);
if (niptr == optr) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
IDPRINTF(0, ("%s: Request Queue Overflow+\n",
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp->isp_name));
XS_SETERR(xs, HBA_BOTCH);
return (CMD_EAGAIN);
}
reqp = (ispreq_t *)
ISP_QUEUE_ENTRY(isp->isp_rquest, iptr);
iptr = niptr;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
}
+ Some basic deadchip detection. + Enable FIFO bursts, but also detect bogus 1040A with busted FIFO. + Use new MEMZERO crossplatform define. + Handle RQCS_QUEUE_FULL status case and let upper layer parse SCSI ststus byte if nonzero (should be 0x28- Queue Full status) + Fold ISP_NVRAM_FIFO_THRESHOLD_128 into isp_fifo_threshold tag.
1999-01-10 08:07:53 +03:00
MEMZERO((void *) reqp, UZSIZE);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
reqp->req_header.rqs_entry_count = 1;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (IS_FC(isp)) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
reqp->req_header.rqs_entry_type = RQSTYPE_T2RQS;
} else {
reqp->req_header.rqs_entry_type = RQSTYPE_REQUEST;
}
reqp->req_header.rqs_flags = 0;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
reqp->req_header.rqs_seqno = 0;
if (IS_FC(isp)) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
/*
* See comment in isp_intr
*/
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_RESID(xs) = 0;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
/*
I was misinformed. I cannot get away from specifying tags for FC. Some devices are happy w/o them- some are unhappy (IBM drives).
1999-10-28 06:50:34 +04:00
* Fibre Channel always requires some kind of tag.
* The Qlogic drivers seem be happy not to use a tag,
* but this breaks for some devices (IBM drives).
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
*/
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
if (XS_CANTAG(xs)) {
t2reqp->req_flags = XS_KINDOF_TAG(xs);
I was misinformed. I cannot get away from specifying tags for FC. Some devices are happy w/o them- some are unhappy (IBM drives).
1999-10-28 06:50:34 +04:00
} else {
if (XS_CDBP(xs)[0] == 0x3) /* REQUEST SENSE */
t2reqp->req_flags = REQFLAG_HTAG;
else
t2reqp->req_flags = REQFLAG_OTAG;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
} else {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
sdparam *sdp = (sdparam *)isp->isp_param;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
if ((sdp->isp_devparam[target].cur_dflags & DPARM_TQING) &&
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_CANTAG(xs)) {
reqp->req_flags = XS_KINDOF_TAG(xs);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
}
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
reqp->req_target = target | (XS_CHANNEL(xs) << 7);
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (IS_SCSI(isp)) {
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
reqp->req_lun_trn = XS_LUN(xs);
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
reqp->req_cdblen = XS_CDBLEN(xs);
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
} else {
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
#ifdef ISP2100_SCCLUN
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
t2reqp->req_scclun = XS_LUN(xs);
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
#else
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
t2reqp->req_lun_trn = XS_LUN(xs);
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
#endif
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
MEMCPY(reqp->req_cdb, XS_CDBP(xs), XS_CDBLEN(xs));
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
reqp->req_time = XS_TIME(xs) / 1000;
if (reqp->req_time == 0 && XS_TIME(xs))
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
reqp->req_time = 1;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
/*
* Always give a bit more leeway to commands after a bus reset.
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
* XXX: DOES NOT DISTINGUISH WHICH PORT MAY HAVE BEEN SYNCED
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
*/
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (isp->isp_sendmarker && reqp->req_time < 5) {
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
reqp->req_time = 5;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
}
if (isp_save_xs(isp, xs, &reqp->req_handle)) {
IDPRINTF(2, ("%s: out of xflist pointers\n", isp->isp_name));
XS_SETERR(xs, HBA_BOTCH);
return (CMD_EAGAIN);
}
/*
* Set up DMA and/or do any bus swizzling of the request entry
* so that the Qlogic F/W understands what is being asked of it.
*/
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
i = ISP_DMASETUP(isp, xs, reqp, &iptr, optr);
if (i != CMD_QUEUED) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
isp_destroy_handle(isp, reqp->req_handle);
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
/*
* dmasetup sets actual error in packet, and
* return what we were given to return.
*/
return (i);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_SETERR(xs, HBA_NOERROR);
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
IDPRINTF(5, ("%s(%d.%d.%d): START cmd 0x%x datalen %d\n",
isp->isp_name, XS_CHANNEL(xs), target, XS_LUN(xs),
reqp->req_cdb[0], XS_XFRLEN(xs)));
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
MemoryBarrier();
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
ISP_WRITE(isp, INMAILBOX4, iptr);
isp->isp_reqidx = iptr;
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
isp->isp_nactive++;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
if (isp->isp_sendmarker)
isp->isp_sendmarker = 0;
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
return (CMD_QUEUED);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
#undef reqp
#undef t2reqp
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
/*
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
* isp control
* Locks (ints blocked) assumed held.
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
*/
int
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
isp_control(isp, ctl, arg)
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
struct ispsoftc *isp;
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
ispctl_t ctl;
void *arg;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
{
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
ISP_SCSI_XFER_T *xs;
mbreg_t mbs;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
int bus, tgt;
u_int32_t handle;
clean up isp_poll() so that it works more correctly. (Modeled on the version in the BusLogic driver.)
1997-03-23 03:50:07 +03:00
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
switch (ctl) {
default:
PRINTF("%s: isp_control unknown control op %x\n",
isp->isp_name, ctl);
break;
clean up isp_poll() so that it works more correctly. (Modeled on the version in the BusLogic driver.)
1997-03-23 03:50:07 +03:00
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
case ISPCTL_RESET_BUS:
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
/*
* Issue a bus reset.
*/
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
mbs.param[0] = MBOX_BUS_RESET;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (IS_SCSI(isp)) {
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
mbs.param[1] =
((sdparam *) isp->isp_param)->isp_bus_reset_delay;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
if (mbs.param[1] < 2)
mbs.param[1] = 2;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
bus = *((int *) arg);
mbs.param[2] = bus;
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
} else {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
mbs.param[1] = 10;
remember to initialize mailbox 2 for FC isp bus resets
1999-10-22 21:06:12 +04:00
mbs.param[2] = 0;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
bus = 0;
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
}
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp->isp_sendmarker = 1 << bus;
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
isp_dumpregs(isp, "isp_control SCSI bus reset failed");
break;
}
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
PRINTF("%s: driver initiated bus reset of bus %d\n",
isp->isp_name, bus);
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
return (0);
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
case ISPCTL_RESET_DEV:
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
tgt = (*((int *) arg)) & 0xffff;
bus = (*((int *) arg)) >> 16;
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
mbs.param[0] = MBOX_ABORT_TARGET;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
mbs.param[1] = (tgt << 8) | (bus << 15);
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
mbs.param[2] = 3; /* 'delay', in seconds */
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
PRINTF("%s: isp_control MBOX_RESET_DEV failure (code "
"%x)\n", isp->isp_name, mbs.param[0]);
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
break;
}
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
PRINTF("%s: Target %d on Bus %d Reset Succeeded\n",
isp->isp_name, tgt, bus);
isp->isp_sendmarker = 1 << bus;
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
return (0);
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
case ISPCTL_ABORT_CMD:
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
xs = (ISP_SCSI_XFER_T *) arg;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
handle = isp_find_handle(isp, xs);
if (handle == 0) {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF("%s: isp_control- cannot find command to abort "
"in active list\n", isp->isp_name);
break;
}
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
bus = XS_CHANNEL(xs);
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
mbs.param[0] = MBOX_ABORT;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (IS_FC(isp)) {
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
#ifdef ISP2100_SCCLUN
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
mbs.param[1] = XS_TGT(xs) << 8;
mbs.param[4] = 0;
mbs.param[5] = 0;
mbs.param[6] = XS_LUN(xs);
#else
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
mbs.param[1] = XS_TGT(xs) << 8 | XS_LUN(xs);
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
#endif
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
} else {
mbs.param[1] =
(bus << 15) | (XS_TGT(xs) << 8) | XS_LUN(xs);
}
mbs.param[2] = handle >> 16;
mbs.param[3] = handle & 0xffff;
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
PRINTF("%s: isp_control MBOX_ABORT failure (code %x)\n",
isp->isp_name, mbs.param[0]);
break;
}
PRINTF("%s: command for target %d lun %d was aborted\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
return (0);
turn into a debug message an ABOUT F/W failure. Add ISPCTL_UPDATE_PARAMS isp_control case
1998-09-10 21:10:27 +04:00
case ISPCTL_UPDATE_PARAMS:
isp_update(isp);
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
return (0);
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
case ISPCTL_FCLINK_TEST:
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
return (isp_fclink_test(isp, FC_FW_READY_DELAY));
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
return (-1);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
/*
* Interrupt Service Routine(s).
*
* External (OS) framework has done the appropriate locking,
* and the locking will be held throughout this function.
*/
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
int
isp_intr(arg)
void *arg;
{
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
ISP_SCSI_XFER_T *complist[RESULT_QUEUE_LEN], *xs;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
struct ispsoftc *isp = arg;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
u_int8_t iptr, optr;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
u_int16_t isr, isrb, sema;
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
int i, nlooked = 0, ndone = 0;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
/*
* Well, if we've disabled interrupts, we may get a case where
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
* isr isn't set, but sema is. In any case, debounce isr reads.
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
*/
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
do {
isr = ISP_READ(isp, BIU_ISR);
isrb = ISP_READ(isp, BIU_ISR);
} while (isr != isrb);
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
sema = ISP_READ(isp, BIU_SEMA) & 0x1;
IDPRINTF(5, ("%s: isp_intr isr %x sem %x\n", isp->isp_name, isr, sema));
Some stylistic changes, but also use the new INT_PENDING macro. Start tracking LIP sequence numbers.
1999-10-17 05:38:27 +04:00
if (isr == 0) {
return (0);
}
if (!INT_PENDING(isp, isr)) {
IDPRINTF(4, ("%s: isp_intr isr=%x\n", isp->isp_name, isr));
return (0);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
if (isp->isp_state != ISP_RUNSTATE) {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
IDPRINTF(3, ("%s: interrupt (isr=%x,sema=%x) when not ready\n",
isp->isp_name, isr, sema));
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
ISP_WRITE(isp, INMAILBOX5, ISP_READ(isp, OUTMAILBOX5));
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
ISP_WRITE(isp, BIU_SEMA, 0);
ENABLE_INTS(isp);
return (1);
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
if (sema) {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
u_int16_t mbox = ISP_READ(isp, OUTMAILBOX0);
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
if (mbox & 0x4000) {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
IDPRINTF(3, ("%s: Command Mbox 0x%x\n",
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
isp->isp_name, mbox));
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
} else {
u_int32_t fhandle = isp_parse_async(isp, (int) mbox);
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
IDPRINTF(3, ("%s: Async Mbox 0x%x\n",
isp->isp_name, mbox));
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
if (fhandle > 0) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
isp_fastpost_complete(isp, fhandle);
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
}
}
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
ISP_WRITE(isp, BIU_SEMA, 0);
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
ENABLE_INTS(isp);
return (1);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
/*
* You *must* read OUTMAILBOX5 prior to clearing the RISC interrupt.
*/
Locking is now in the outer framework for isp_init/isp_reset- this allows us to call isp_reset/isp_init internally with impunity. Rename isp_phoenix to isp_restart and make it global. Clean it up a tad. If we get an ASYNC_SYSTEM_ERROR code in isp_intr, call isp_restart and return- the f/w is toasted at this point (usually), so we have to bring things back to a known state. In isp_mboxcmd, when we don't see the HOST_INT bit go clear, try and find out if the isp is trying to tell us something and try again. This may avoid a potential deadlock where the previous mailbox command hasn't been cleared by the ISP. In isp_init don't try and get device parameters if we already have them- this typically doesn't work if we're in the middle of an isp_restart.
1998-07-19 01:10:16 +04:00
optr = isp->isp_residx;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
iptr = ISP_READ(isp, OUTMAILBOX5);
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
if (optr == iptr) {
update some debugging stuff- also turn off possible ULTRA mode for 1020s
1998-03-23 01:02:29 +03:00
IDPRINTF(4, ("why intr? isr %x iptr %x optr %x\n",
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
isr, optr, iptr));
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
while (optr != iptr) {
ispstatusreq_t *sp;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
u_int8_t oop;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
int buddaboom = 0;
- Simplify mailbox DMA setup, allowing the front-ends to allocate the mailbox storage in a way they see fit. - Define macros to index and size the request and result queue entries.
1997-06-08 10:31:52 +04:00
sp = (ispstatusreq_t *) ISP_QUEUE_ENTRY(isp->isp_result, optr);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
oop = optr;
optr = ISP_NXT_QENTRY(optr, RESULT_QUEUE_LEN);
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
nlooked++;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
MemoryBarrier();
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
/*
* Do any appropriate unswizzling of what the Qlogic f/w has
* written into memory so it makes sense to us.
*/
ISP_UNSWIZZLE_RESPONSE(isp, sp);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (sp->req_header.rqs_entry_type != RQSTYPE_RESPONSE) {
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
if (isp_handle_other_response(isp, sp, &optr) == 0) {
ISP_WRITE(isp, INMAILBOX5, optr);
continue;
}
/*
* It really has to be a bounced request just copied
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
* from the request queue to the response queue. If
* not, something bad has happened.
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
*/
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (sp->req_header.rqs_entry_type != RQSTYPE_REQUEST) {
ISP_WRITE(isp, INMAILBOX5, optr);
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
PRINTF("%s: not RESPONSE in RESPONSE Queue "
"(type 0x%x) @ idx %d (next %d)\n",
isp->isp_name,
sp->req_header.rqs_entry_type, oop, optr);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
continue;
}
buddaboom = 1;
}
if (sp->req_header.rqs_flags & 0xf) {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
#define _RQS_OFLAGS \
~(RQSFLAG_CONTINUATION|RQSFLAG_FULL|RQSFLAG_BADHEADER|RQSFLAG_BADPACKET)
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (sp->req_header.rqs_flags & RQSFLAG_CONTINUATION) {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
IDPRINTF(3, ("%s: continuation segment\n",
isp->isp_name));
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
ISP_WRITE(isp, INMAILBOX5, optr);
continue;
}
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
if (sp->req_header.rqs_flags & RQSFLAG_FULL) {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
IDPRINTF(2, ("%s: internal queues full\n",
isp->isp_name));
/*
* We'll synthesize a QUEUE FULL message below.
*/
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
}
if (sp->req_header.rqs_flags & RQSFLAG_BADHEADER) {
PRINTF("%s: bad header\n", isp->isp_name);
buddaboom++;
}
if (sp->req_header.rqs_flags & RQSFLAG_BADPACKET) {
PRINTF("%s: bad request packet\n",
isp->isp_name);
buddaboom++;
}
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
if (sp->req_header.rqs_flags & _RQS_OFLAGS) {
PRINTF("%s: unknown flags in response (0x%x)\n",
isp->isp_name, sp->req_header.rqs_flags);
buddaboom++;
}
#undef _RQS_OFLAGS
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (sp->req_handle > isp->isp_maxcmds || sp->req_handle < 1) {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF("%s: bad request handle %d\n", isp->isp_name,
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
sp->req_handle);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
ISP_WRITE(isp, INMAILBOX5, optr);
continue;
}
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
xs = isp_find_xs(isp, sp->req_handle);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (xs == NULL) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
PRINTF("%s: NULL xs in xflist (handle 0x%x)\n",
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
isp->isp_name, sp->req_handle);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
ISP_WRITE(isp, INMAILBOX5, optr);
continue;
}
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
isp_destroy_handle(isp, sp->req_handle);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (sp->req_status_flags & RQSTF_BUS_RESET) {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp->isp_sendmarker |= (1 << XS_CHANNEL(xs));
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
if (buddaboom) {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_SETERR(xs, HBA_BOTCH);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_STS(xs) = sp->req_scsi_status & 0xff;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
if (IS_SCSI(isp)) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
if (sp->req_state_flags & RQSF_GOT_SENSE) {
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
MEMCPY(XS_SNSP(xs), sp->req_sense_data,
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_SNSLEN(xs));
XS_SNS_IS_VALID(xs);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
/*
* A new synchronous rate was negotiated for this
* target. Mark state such that we'll go look up
* that which has changed later.
*/
if (sp->req_status_flags & RQSTF_NEGOTIATION) {
sdparam *sdp = isp->isp_param;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
sdp += XS_CHANNEL(xs);
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
sdp->isp_devparam[XS_TGT(xs)].dev_refresh = 1;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp->isp_update |= (1 << XS_CHANNEL(xs));
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
}
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
} else {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
if (XS_STS(xs) == SCSI_CHECK) {
XS_SNS_IS_VALID(xs);
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
MEMCPY(XS_SNSP(xs), sp->req_sense_data,
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_SNSLEN(xs));
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
sp->req_state_flags |= RQSF_GOT_SENSE;
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
if (XS_NOERR(xs) && XS_STS(xs) == SCSI_BUSY) {
XS_SETERR(xs, HBA_TGTBSY);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (sp->req_header.rqs_entry_type == RQSTYPE_RESPONSE) {
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
if (XS_NOERR(xs)) {
if (sp->req_completion_status != RQCS_COMPLETE) {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
isp_parse_status(isp, sp, xs);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
} else {
XS_SETERR(xs, HBA_NOERROR);
}
}
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
} else if (sp->req_header.rqs_entry_type == RQSTYPE_REQUEST) {
if (sp->req_header.rqs_flags & RQSFLAG_FULL) {
/*
* Force Queue Full status.
*/
XS_STS(xs) = SCSI_QFULL;
XS_SETERR(xs, HBA_NOERROR);
} else if (XS_NOERR(xs)) {
XS_SETERR(xs, HBA_BOTCH);
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
} else {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
PRINTF("%s: unhandled respose queue type 0x%x\n",
isp->isp_name, sp->req_header.rqs_entry_type);
+ Some basic deadchip detection. + Enable FIFO bursts, but also detect bogus 1040A with busted FIFO. + Use new MEMZERO crossplatform define. + Handle RQCS_QUEUE_FULL status case and let upper layer parse SCSI ststus byte if nonzero (should be 0x28- Queue Full status) + Fold ISP_NVRAM_FIFO_THRESHOLD_128 into isp_fifo_threshold tag.
1999-01-10 08:07:53 +03:00
if (XS_NOERR(xs)) {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_SETERR(xs, HBA_BOTCH);
+ Some basic deadchip detection. + Enable FIFO bursts, but also detect bogus 1040A with busted FIFO. + Use new MEMZERO crossplatform define. + Handle RQCS_QUEUE_FULL status case and let upper layer parse SCSI ststus byte if nonzero (should be 0x28- Queue Full status) + Fold ISP_NVRAM_FIFO_THRESHOLD_128 into isp_fifo_threshold tag.
1999-01-10 08:07:53 +03:00
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (IS_SCSI(isp)) {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_RESID(xs) = sp->req_resid;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
} else if (sp->req_scsi_status & RQCS_RU) {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_RESID(xs) = sp->req_resid;
update some debugging stuff- also turn off possible ULTRA mode for 1020s
1998-03-23 01:02:29 +03:00
IDPRINTF(4, ("%s: cnt %d rsd %d\n", isp->isp_name,
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_XFRLEN(xs), sp->req_resid));
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
if (XS_XFRLEN(xs)) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
ISP_DMAFREE(isp, xs, sp->req_handle);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
/*
* XXX: If we have a check condition, but no Sense Data,
* XXX: mark it as an error (ARQ failed). We need to
* XXX: to do a more distinct job because there may
* XXX: cases where ARQ is disabled.
*/
if (XS_STS(xs) == SCSI_CHECK && !(XS_IS_SNS_VALID(xs))) {
if (XS_NOERR(xs)) {
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
PRINTF("%s: ARQ failure for target %d lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_SETERR(xs, HBA_ARQFAIL);
}
}
if ((isp->isp_dblev >= 5) ||
(isp->isp_dblev > 2 && !XS_NOERR(xs))) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
PRINTF("%s(%d.%d): FIN dl%d resid%d STS %x",
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
isp->isp_name, XS_TGT(xs), XS_LUN(xs),
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
XS_XFRLEN(xs), XS_RESID(xs), XS_STS(xs));
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
if (sp->req_state_flags & RQSF_GOT_SENSE) {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF(" Skey: %x", XS_SNSKEY(xs));
if (!(XS_IS_SNS_VALID(xs))) {
PRINTF(" BUT NOT SET");
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
PRINTF(" XS_ERR=0x%x\n", (unsigned int) XS_ERR(xs));
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
if (isp->isp_nactive > 0)
isp->isp_nactive--;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
complist[ndone++] = xs; /* defer completion call until later */
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
/*
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
* If we looked at any commands, then it's valid to find out
* what the outpointer is. It also is a trigger to update the
* ISP's notion of what we've seen so far.
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
*/
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
if (nlooked) {
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
ISP_WRITE(isp, INMAILBOX5, optr);
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
isp->isp_reqodx = ISP_READ(isp, OUTMAILBOX4);
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
isp->isp_residx = optr;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
for (i = 0; i < ndone; i++) {
xs = complist[i];
if (xs) {
XS_CMD_DONE(xs);
}
}
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
ENABLE_INTS(isp);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
return (1);
}
/*
* Support routines.
*/
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
static int
isp_parse_async(isp, mbox)
struct ispsoftc *isp;
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
int mbox;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
{
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
u_int32_t fast_post_handle = 0;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
switch (mbox) {
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
case MBOX_COMMAND_COMPLETE: /* sometimes these show up */
break;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
case ASYNC_BUS_RESET:
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
{
int bus;
if (IS_1080(isp) || IS_12X0(isp)) {
bus = ISP_READ(isp, OUTMAILBOX6);
} else {
bus = 0;
}
isp->isp_sendmarker = (1 << bus);
isp_async(isp, ISPASYNC_BUS_RESET, &bus);
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
#ifdef ISP_TARGET_MODE
isp_notify_ack(isp, NULL);
#endif
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
break;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
case ASYNC_SYSTEM_ERROR:
mbox = ISP_READ(isp, OUTMAILBOX1);
PRINTF("%s: Internal FW Error @ RISC Addr 0x%x\n",
isp->isp_name, mbox);
isp_restart(isp);
/* no point continuing after this */
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
return (-1);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
case ASYNC_RQS_XFER_ERR:
PRINTF("%s: Request Queue Transfer Error\n", isp->isp_name);
break;
case ASYNC_RSP_XFER_ERR:
PRINTF("%s: Response Queue Transfer Error\n", isp->isp_name);
break;
case ASYNC_QWAKEUP:
/* don't need to be chatty */
mbox = ISP_READ(isp, OUTMAILBOX4);
break;
case ASYNC_TIMEOUT_RESET:
PRINTF("%s: timeout initiated SCSI bus reset\n", isp->isp_name);
isp->isp_sendmarker = 1;
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
#ifdef ISP_TARGET_MODE
isp_notify_ack(isp, NULL);
#endif
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
break;
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
case ASYNC_DEVICE_RESET:
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
/*
* XXX: WHICH BUS?
*/
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
isp->isp_sendmarker = 1;
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
PRINTF("%s: device reset\n", isp->isp_name);
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
#ifdef ISP_TARGET_MODE
isp_notify_ack(isp, NULL);
#endif
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
break;
case ASYNC_EXTMSG_UNDERRUN:
PRINTF("%s: extended message underrun\n", isp->isp_name);
break;
case ASYNC_SCAM_INT:
PRINTF("%s: SCAM interrupt\n", isp->isp_name);
break;
case ASYNC_HUNG_SCSI:
PRINTF("%s: stalled SCSI Bus after DATA Overrun\n",
isp->isp_name);
/* XXX: Need to issue SCSI reset at this point */
break;
case ASYNC_KILLED_BUS:
PRINTF("%s: SCSI Bus reset after DATA Overrun\n",
isp->isp_name);
break;
case ASYNC_BUS_TRANSIT:
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
/*
* XXX: WHICH BUS?
*/
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
mbox = ISP_READ(isp, OUTMAILBOX2);
switch (mbox & 0x1c00) {
case SXP_PINS_LVD_MODE:
PRINTF("%s: Transition to LVD mode\n", isp->isp_name);
((sdparam *)isp->isp_param)->isp_diffmode = 0;
((sdparam *)isp->isp_param)->isp_ultramode = 0;
((sdparam *)isp->isp_param)->isp_lvdmode = 1;
break;
case SXP_PINS_HVD_MODE:
PRINTF("%s: Transition to Differential mode\n",
isp->isp_name);
((sdparam *)isp->isp_param)->isp_diffmode = 1;
((sdparam *)isp->isp_param)->isp_ultramode = 0;
((sdparam *)isp->isp_param)->isp_lvdmode = 0;
break;
case SXP_PINS_SE_MODE:
PRINTF("%s: Transition to Single Ended mode\n",
isp->isp_name);
((sdparam *)isp->isp_param)->isp_diffmode = 0;
((sdparam *)isp->isp_param)->isp_ultramode = 1;
((sdparam *)isp->isp_param)->isp_lvdmode = 0;
break;
default:
PRINTF("%s: Transition to unknown mode 0x%x\n",
isp->isp_name, mbox);
break;
}
/*
* XXX: Set up to renegotiate again!
*/
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
/* Can only be for a 1080... */
isp->isp_sendmarker = (1 << ISP_READ(isp, OUTMAILBOX6));
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
break;
case ASYNC_CMD_CMPLT:
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
fast_post_handle = (ISP_READ(isp, OUTMAILBOX2) << 16) |
ISP_READ(isp, OUTMAILBOX1);
IDPRINTF(3, ("%s: fast post completion of %u\n", isp->isp_name,
fast_post_handle));
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
break;
case ASYNC_CTIO_DONE:
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
/* Should only occur when Fast Posting Set for 2100s */
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
PRINTF("%s: CTIO done\n", isp->isp_name);
break;
case ASYNC_LIP_OCCURRED:
Some stylistic changes, but also use the new INT_PENDING macro. Start tracking LIP sequence numbers.
1999-10-17 05:38:27 +04:00
((fcparam *) isp->isp_param)->isp_lipseq =
ISP_READ(isp, OUTMAILBOX1);
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
((fcparam *) isp->isp_param)->isp_fwstate = FW_CONFIG_WAIT;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
((fcparam *) isp->isp_param)->isp_loopstate = LOOP_LIP_RCVD;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
isp->isp_sendmarker = 1;
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
isp_mark_getpdb_all(isp);
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
IDPRINTF(1, ("%s: LIP occurred\n", isp->isp_name));
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
break;
case ASYNC_LOOP_UP:
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
isp->isp_sendmarker = 1;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
((fcparam *) isp->isp_param)->isp_fwstate = FW_CONFIG_WAIT;
((fcparam *) isp->isp_param)->isp_loopstate = LOOP_LIP_RCVD;
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
isp_mark_getpdb_all(isp);
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
isp_async(isp, ISPASYNC_LOOP_UP, NULL);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
break;
case ASYNC_LOOP_DOWN:
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
isp->isp_sendmarker = 1;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
((fcparam *) isp->isp_param)->isp_fwstate = FW_CONFIG_WAIT;
((fcparam *) isp->isp_param)->isp_loopstate = LOOP_NIL;
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
isp_mark_getpdb_all(isp);
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
isp_async(isp, ISPASYNC_LOOP_DOWN, NULL);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
break;
case ASYNC_LOOP_RESET:
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
isp->isp_sendmarker = 1;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
((fcparam *) isp->isp_param)->isp_fwstate = FW_CONFIG_WAIT;
((fcparam *) isp->isp_param)->isp_loopstate = LOOP_NIL;
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
isp_mark_getpdb_all(isp);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
PRINTF("%s: Loop RESET\n", isp->isp_name);
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
#ifdef ISP_TARGET_MODE
isp_notify_ack(isp, NULL);
#endif
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
break;
case ASYNC_PDB_CHANGED:
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
isp->isp_sendmarker = 1;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
((fcparam *) isp->isp_param)->isp_loopstate = LOOP_PDB_RCVD;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
isp_mark_getpdb_all(isp);
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
IDPRINTF(2, ("%s: Port Database Changed\n", isp->isp_name));
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
break;
case ASYNC_CHANGE_NOTIFY:
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
isp_mark_getpdb_all(isp);
/*
* Not correct, but it will force us to rescan the loop.
*/
((fcparam *) isp->isp_param)->isp_loopstate = LOOP_PDB_RCVD;
isp_async(isp, ISPASYNC_CHANGE_NOTIFY, NULL);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
break;
default:
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
PRINTF("%s: unknown async code 0x%x\n", isp->isp_name, mbox);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
break;
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
return (fast_post_handle);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
}
static int
isp_handle_other_response(isp, sp, optrp)
struct ispsoftc *isp;
ispstatusreq_t *sp;
u_int8_t *optrp;
{
switch (sp->req_header.rqs_entry_type) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
case RQSTYPE_ATIO:
case RQSTYPE_CTIO0:
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
case RQSTYPE_ENABLE_LUN:
case RQSTYPE_MODIFY_LUN:
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
case RQSTYPE_NOTIFY:
case RQSTYPE_NOTIFY_ACK:
case RQSTYPE_CTIO1:
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
case RQSTYPE_ATIO2:
case RQSTYPE_CTIO2:
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
case RQSTYPE_CTIO3:
#ifdef ISP_TARGET_MODE
return(isp_target_notify(isp, sp, optrp));
#else
/* FALLTHROUGH */
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
#endif
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
case RQSTYPE_REQUEST:
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
default:
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
return (-1);
}
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
static void
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_parse_status(isp, sp, xs)
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
struct ispsoftc *isp;
ispstatusreq_t *sp;
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
ISP_SCSI_XFER_T *xs;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
{
switch (sp->req_completion_status) {
case RQCS_COMPLETE:
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_SETERR(xs, HBA_NOERROR);
return;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
case RQCS_INCOMPLETE:
if ((sp->req_state_flags & RQSF_GOT_TARGET) == 0) {
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
IDPRINTF(3, ("%s: Selection Timeout for target %d\n",
isp->isp_name, XS_TGT(xs)));
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_SETERR(xs, HBA_SELTIMEOUT);
return;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
PRINTF("%s: command incomplete for target %d lun %d, state "
"0x%x\n", isp->isp_name, XS_TGT(xs), XS_LUN(xs),
sp->req_state_flags);
break;
case RQCS_DMA_ERROR:
PRINTF("%s: DMA error for command on target %d, lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
break;
A little cleanup and print more error messages.
1997-09-29 05:41:30 +04:00
case RQCS_TRANSPORT_ERROR:
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF("%s: transport error\n", isp->isp_name);
A little cleanup and print more error messages.
1997-09-29 05:41:30 +04:00
isp_prtstst(sp);
break;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
case RQCS_RESET_OCCURRED:
IDPRINTF(2, ("%s: bus reset destroyed command for target %d "
"lun %d\n", isp->isp_name, XS_TGT(xs), XS_LUN(xs)));
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
/*
* XXX: Get port number for bus
*/
isp->isp_sendmarker = 3;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
XS_SETERR(xs, HBA_BUSRESET);
return;
case RQCS_ABORTED:
PRINTF("%s: command aborted for target %d lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
/*
* XXX: Get port number for bus
*/
isp->isp_sendmarker = 3;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
XS_SETERR(xs, HBA_ABORTED);
return;
case RQCS_TIMEOUT:
IDPRINTF(2, ("%s: command timed out for target %d lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs)));
XS_SETERR(xs, HBA_CMDTIMEOUT);
return;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
case RQCS_DATA_OVERRUN:
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (IS_FC(isp)) {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_RESID(xs) = sp->req_resid;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
break;
}
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_SETERR(xs, HBA_DATAOVR);
return;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
case RQCS_COMMAND_OVERRUN:
PRINTF("%s: command overrun for command on target %d, lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_STATUS_OVERRUN:
PRINTF("%s: status overrun for command on target %d, lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_BAD_MESSAGE:
PRINTF("%s: message not COMMAND COMPLETE after status on "
"target %d, lun %d\n", isp->isp_name, XS_TGT(xs),
XS_LUN(xs));
break;
case RQCS_NO_MESSAGE_OUT:
PRINTF("%s: No MESSAGE OUT phase after selection on "
"target %d, lun %d\n", isp->isp_name, XS_TGT(xs),
XS_LUN(xs));
break;
case RQCS_EXT_ID_FAILED:
PRINTF("%s: EXTENDED IDENTIFY failed on target %d, lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_IDE_MSG_FAILED:
PRINTF("%s: target %d lun %d rejected INITIATOR DETECTED "
"ERROR message\n", isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_ABORT_MSG_FAILED:
PRINTF("%s: target %d lun %d rejected ABORT message\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_REJECT_MSG_FAILED:
PRINTF("%s: target %d lun %d rejected MESSAGE REJECT message\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_NOP_MSG_FAILED:
PRINTF("%s: target %d lun %d rejected NOP message\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_PARITY_ERROR_MSG_FAILED:
PRINTF("%s: target %d lun %d rejected MESSAGE PARITY ERROR "
"message\n", isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_DEVICE_RESET_MSG_FAILED:
PRINTF("%s: target %d lun %d rejected BUS DEVICE RESET "
"message\n", isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_ID_MSG_FAILED:
PRINTF("%s: target %d lun %d rejected IDENTIFY "
"message\n", isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_UNEXP_BUS_FREE:
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
PRINTF("%s: target %d lun %d had an unexpected bus free\n",
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
case RQCS_DATA_UNDERRUN:
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (IS_FC(isp)) {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_RESID(xs) = sp->req_resid;
/* an UNDERRUN is not a botch ??? */
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_SETERR(xs, HBA_NOERROR);
return;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
case RQCS_XACT_ERR1:
PRINTF("%s: HBA attempted queued transaction with disconnect "
"not set for target %d lun %d\n", isp->isp_name, XS_TGT(xs),
XS_LUN(xs));
break;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
case RQCS_XACT_ERR2:
PRINTF("%s: HBA attempted queued transaction to target "
"routine %d on target %d\n", isp->isp_name, XS_LUN(xs),
XS_TGT(xs));
break;
case RQCS_XACT_ERR3:
PRINTF("%s: HBA attempted queued transaction for target %d lun "
"%d when queueing disabled\n", isp->isp_name, XS_TGT(xs),
XS_LUN(xs));
break;
case RQCS_BAD_ENTRY:
PRINTF("%s: invalid IOCB entry type detected\n", isp->isp_name);
break;
case RQCS_QUEUE_FULL:
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
IDPRINTF(3, ("%s: internal queues full for target %d lun %d "
+ Some basic deadchip detection. + Enable FIFO bursts, but also detect bogus 1040A with busted FIFO. + Use new MEMZERO crossplatform define. + Handle RQCS_QUEUE_FULL status case and let upper layer parse SCSI ststus byte if nonzero (should be 0x28- Queue Full status) + Fold ISP_NVRAM_FIFO_THRESHOLD_128 into isp_fifo_threshold tag.
1999-01-10 08:07:53 +03:00
"status 0x%x\n", isp->isp_name, XS_TGT(xs), XS_LUN(xs),
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
XS_STS(xs)));
+ Some basic deadchip detection. + Enable FIFO bursts, but also detect bogus 1040A with busted FIFO. + Use new MEMZERO crossplatform define. + Handle RQCS_QUEUE_FULL status case and let upper layer parse SCSI ststus byte if nonzero (should be 0x28- Queue Full status) + Fold ISP_NVRAM_FIFO_THRESHOLD_128 into isp_fifo_threshold tag.
1999-01-10 08:07:53 +03:00
/*
* If QFULL or some other status byte is set, then this
* isn't an error, per se.
*/
if (XS_STS(xs) != 0) {
XS_SETERR(xs, HBA_NOERROR);
return;
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
break;
case RQCS_PHASE_SKIPPED:
PRINTF("%s: SCSI phase skipped (e.g., COMMAND COMPLETE w/o "
"STATUS phase) for target %d lun %d\n", isp->isp_name,
XS_TGT(xs), XS_LUN(xs));
break;
case RQCS_ARQS_FAILED:
PRINTF("%s: Auto Request Sense failed for target %d lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
XS_SETERR(xs, HBA_ARQFAIL);
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
return;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
case RQCS_WIDE_FAILED:
PRINTF("%s: Wide Negotiation failed for target %d lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
if (IS_SCSI(isp)) {
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
sdparam *sdp = isp->isp_param;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
sdp += XS_CHANNEL(xs);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
sdp->isp_devparam[XS_TGT(xs)].dev_flags &= ~DPARM_WIDE;
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
sdp->isp_devparam[XS_TGT(xs)].dev_update = 1;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp->isp_update = XS_CHANNEL(xs)+1;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
}
XS_SETERR(xs, HBA_NOERROR);
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
return;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
case RQCS_SYNCXFER_FAILED:
PRINTF("%s: SDTR Message failed for target %d lun %d\n",
isp->isp_name, XS_TGT(xs), XS_LUN(xs));
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
if (IS_SCSI(isp)) {
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
sdparam *sdp = isp->isp_param;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
sdp += XS_CHANNEL(xs);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
sdp->isp_devparam[XS_TGT(xs)].dev_flags &= ~DPARM_SYNC;
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
sdp->isp_devparam[XS_TGT(xs)].dev_update = 1;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp->isp_update = XS_CHANNEL(xs)+1;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
}
break;
case RQCS_LVD_BUSERR:
PRINTF("%s: Bad LVD Bus condition while talking to target %d "
"lun %d\n", isp->isp_name, XS_TGT(xs), XS_LUN(xs));
break;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
case RQCS_PORT_UNAVAILABLE:
/*
* No such port on the loop. Moral equivalent of SELTIMEO
*/
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
IDPRINTF(3, ("%s: Port Unavailable for target %d\n",
isp->isp_name, XS_TGT(xs)));
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_SETERR(xs, HBA_SELTIMEOUT);
return;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
case RQCS_PORT_LOGGED_OUT:
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
/*
* It was there (maybe)- treat as a selection timeout.
*/
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
IDPRINTF(2, ("%s: port logout for target %d\n",
isp->isp_name, XS_TGT(xs)));
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
XS_SETERR(xs, HBA_SELTIMEOUT);
return;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
case RQCS_PORT_CHANGED:
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF("%s: port changed for target %d\n",
isp->isp_name, XS_TGT(xs));
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
break;
case RQCS_PORT_BUSY:
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF("%s: port busy for target %d\n",
isp->isp_name, XS_TGT(xs));
XS_SETERR(xs, HBA_TGTBSY);
return;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
default:
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF("%s: comp status %x\n", isp->isp_name,
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
sp->req_completion_status);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
break;
}
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_SETERR(xs, HBA_BOTCH);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
static void
isp_fastpost_complete(isp, fph)
struct ispsoftc *isp;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
u_int32_t fph;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
{
ISP_SCSI_XFER_T *xs;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (fph < 1) {
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
return;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
}
xs = isp_find_xs(isp, fph);
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
if (xs == NULL) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
PRINTF("%s: command for fast posting handle 0x%x not found\n",
isp->isp_name, fph);
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
return;
}
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
isp_destroy_handle(isp, fph);
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
/*
* Since we don't have a result queue entry item,
* we must believe that SCSI status is zero and
* that all data transferred.
*/
XS_RESID(xs) = 0;
XS_STS(xs) = 0;
if (XS_XFRLEN(xs)) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
ISP_DMAFREE(isp, xs, fph);
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
}
XS_CMD_DONE(xs);
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (isp->isp_nactive)
isp->isp_nactive--;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
#define HINIB(x) ((x) >> 0x4)
#define LONIB(x) ((x) & 0xf)
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
#define MAKNIB(a, b) (((a) << 4) | (b))
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
static u_int8_t mbpcnt[] = {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
MAKNIB(1, 1), /* 0x00: MBOX_NO_OP */
MAKNIB(5, 5), /* 0x01: MBOX_LOAD_RAM */
MAKNIB(2, 0), /* 0x02: MBOX_EXEC_FIRMWARE */
MAKNIB(5, 5), /* 0x03: MBOX_DUMP_RAM */
MAKNIB(3, 3), /* 0x04: MBOX_WRITE_RAM_WORD */
MAKNIB(2, 3), /* 0x05: MBOX_READ_RAM_WORD */
MAKNIB(6, 6), /* 0x06: MBOX_MAILBOX_REG_TEST */
MAKNIB(2, 3), /* 0x07: MBOX_VERIFY_CHECKSUM */
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
MAKNIB(1, 4), /* 0x08: MBOX_ABOUT_FIRMWARE */
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
MAKNIB(0, 0), /* 0x09: */
MAKNIB(0, 0), /* 0x0a: */
MAKNIB(0, 0), /* 0x0b: */
MAKNIB(0, 0), /* 0x0c: */
MAKNIB(0, 0), /* 0x0d: */
MAKNIB(1, 2), /* 0x0e: MBOX_CHECK_FIRMWARE */
MAKNIB(0, 0), /* 0x0f: */
MAKNIB(5, 5), /* 0x10: MBOX_INIT_REQ_QUEUE */
MAKNIB(6, 6), /* 0x11: MBOX_INIT_RES_QUEUE */
MAKNIB(4, 4), /* 0x12: MBOX_EXECUTE_IOCB */
MAKNIB(2, 2), /* 0x13: MBOX_WAKE_UP */
MAKNIB(1, 6), /* 0x14: MBOX_STOP_FIRMWARE */
MAKNIB(4, 4), /* 0x15: MBOX_ABORT */
MAKNIB(2, 2), /* 0x16: MBOX_ABORT_DEVICE */
MAKNIB(3, 3), /* 0x17: MBOX_ABORT_TARGET */
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
MAKNIB(3, 1), /* 0x18: MBOX_BUS_RESET */
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
MAKNIB(2, 3), /* 0x19: MBOX_STOP_QUEUE */
MAKNIB(2, 3), /* 0x1a: MBOX_START_QUEUE */
MAKNIB(2, 3), /* 0x1b: MBOX_SINGLE_STEP_QUEUE */
MAKNIB(2, 3), /* 0x1c: MBOX_ABORT_QUEUE */
MAKNIB(2, 4), /* 0x1d: MBOX_GET_DEV_QUEUE_STATUS */
MAKNIB(0, 0), /* 0x1e: */
MAKNIB(1, 3), /* 0x1f: MBOX_GET_FIRMWARE_STATUS */
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
MAKNIB(1, 4), /* 0x20: MBOX_GET_INIT_SCSI_ID, MBOX_GET_LOOP_ID */
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
MAKNIB(1, 3), /* 0x21: MBOX_GET_SELECT_TIMEOUT */
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
MAKNIB(1, 3), /* 0x22: MBOX_GET_RETRY_COUNT */
MAKNIB(1, 2), /* 0x23: MBOX_GET_TAG_AGE_LIMIT */
MAKNIB(1, 2), /* 0x24: MBOX_GET_CLOCK_RATE */
MAKNIB(1, 2), /* 0x25: MBOX_GET_ACT_NEG_STATE */
MAKNIB(1, 2), /* 0x26: MBOX_GET_ASYNC_DATA_SETUP_TIME */
MAKNIB(1, 3), /* 0x27: MBOX_GET_PCI_PARAMS */
MAKNIB(2, 4), /* 0x28: MBOX_GET_TARGET_PARAMS */
MAKNIB(2, 4), /* 0x29: MBOX_GET_DEV_QUEUE_PARAMS */
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
MAKNIB(1, 2), /* 0x2a: MBOX_GET_RESET_DELAY_PARAMS */
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
MAKNIB(0, 0), /* 0x2b: */
MAKNIB(0, 0), /* 0x2c: */
MAKNIB(0, 0), /* 0x2d: */
MAKNIB(0, 0), /* 0x2e: */
MAKNIB(0, 0), /* 0x2f: */
MAKNIB(2, 2), /* 0x30: MBOX_SET_INIT_SCSI_ID */
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
MAKNIB(2, 3), /* 0x31: MBOX_SET_SELECT_TIMEOUT */
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
MAKNIB(3, 3), /* 0x32: MBOX_SET_RETRY_COUNT */
MAKNIB(2, 2), /* 0x33: MBOX_SET_TAG_AGE_LIMIT */
MAKNIB(2, 2), /* 0x34: MBOX_SET_CLOCK_RATE */
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
MAKNIB(2, 2), /* 0x35: MBOX_SET_ACT_NEG_STATE */
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
MAKNIB(2, 2), /* 0x36: MBOX_SET_ASYNC_DATA_SETUP_TIME */
MAKNIB(3, 3), /* 0x37: MBOX_SET_PCI_CONTROL_PARAMS */
MAKNIB(4, 4), /* 0x38: MBOX_SET_TARGET_PARAMS */
MAKNIB(4, 4), /* 0x39: MBOX_SET_DEV_QUEUE_PARAMS */
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
MAKNIB(1, 2), /* 0x3a: MBOX_SET_RESET_DELAY_PARAMS */
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
MAKNIB(0, 0), /* 0x3b: */
MAKNIB(0, 0), /* 0x3c: */
MAKNIB(0, 0), /* 0x3d: */
MAKNIB(0, 0), /* 0x3e: */
MAKNIB(0, 0), /* 0x3f: */
MAKNIB(1, 2), /* 0x40: MBOX_RETURN_BIOS_BLOCK_ADDR */
MAKNIB(6, 1), /* 0x41: MBOX_WRITE_FOUR_RAM_WORDS */
MAKNIB(2, 3), /* 0x42: MBOX_EXEC_BIOS_IOCB */
MAKNIB(0, 0), /* 0x43: */
MAKNIB(0, 0), /* 0x44: */
MAKNIB(0, 0), /* 0x45: */
MAKNIB(0, 0), /* 0x46: */
MAKNIB(0, 0), /* 0x47: */
MAKNIB(0, 0), /* 0x48: */
MAKNIB(0, 0), /* 0x49: */
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
MAKNIB(2, 1), /* 0x4a: MBOX_SET_FIRMWARE_FEATURES */
MAKNIB(1, 2), /* 0x4b: MBOX_GET_FIRMWARE_FEATURES */
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
MAKNIB(0, 0), /* 0x4c: */
MAKNIB(0, 0), /* 0x4d: */
MAKNIB(0, 0), /* 0x4e: */
MAKNIB(0, 0), /* 0x4f: */
MAKNIB(0, 0), /* 0x50: */
MAKNIB(0, 0), /* 0x51: */
MAKNIB(0, 0), /* 0x52: */
MAKNIB(0, 0), /* 0x53: */
MAKNIB(8, 0), /* 0x54: MBOX_EXEC_COMMAND_IOCB_A64 */
MAKNIB(0, 0), /* 0x55: */
MAKNIB(0, 0), /* 0x56: */
MAKNIB(0, 0), /* 0x57: */
MAKNIB(0, 0), /* 0x58: */
MAKNIB(0, 0), /* 0x59: */
MAKNIB(0, 0), /* 0x5a: */
MAKNIB(0, 0), /* 0x5b: */
MAKNIB(0, 0), /* 0x5c: */
MAKNIB(0, 0), /* 0x5d: */
MAKNIB(0, 0), /* 0x5e: */
MAKNIB(0, 0), /* 0x5f: */
MAKNIB(8, 6), /* 0x60: MBOX_INIT_FIRMWARE */
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
MAKNIB(0, 0), /* 0x61: */
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
MAKNIB(2, 1), /* 0x62: MBOX_INIT_LIP */
MAKNIB(8, 1), /* 0x63: MBOX_GET_FC_AL_POSITION_MAP */
MAKNIB(8, 1), /* 0x64: MBOX_GET_PORT_DB */
MAKNIB(3, 1), /* 0x65: MBOX_CLEAR_ACA */
MAKNIB(3, 1), /* 0x66: MBOX_TARGET_RESET */
MAKNIB(3, 1), /* 0x67: MBOX_CLEAR_TASK_SET */
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
MAKNIB(3, 1), /* 0x68: MBOX_ABORT_TASK_SET */
MAKNIB(1, 2), /* 0x69: MBOX_GET_FW_STATE */
MAKNIB(2, 8), /* 0x6a: MBOX_GET_PORT_NAME */
MAKNIB(8, 1), /* 0x6b: MBOX_GET_LINK_STATUS */
MAKNIB(4, 4), /* 0x6c: MBOX_INIT_LIP_RESET */
MAKNIB(0, 0), /* 0x6d: */
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
MAKNIB(8, 1), /* 0x6e: MBOX_SEND_SNS */
MAKNIB(4, 3), /* 0x6f: MBOX_FABRIC_LOGIN */
MAKNIB(2, 1), /* 0x70: MBOX_SEND_CHANGE_REQUEST */
MAKNIB(2, 1), /* 0x71: MBOX_FABRIC_LOGOUT */
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
MAKNIB(4, 1) /* 0x72: MBOX_INIT_LIP_LOGIN */
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
};
#define NMBCOM (sizeof (mbpcnt) / sizeof (mbpcnt[0]))
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
static void
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
isp_mboxcmd(isp, mbp)
struct ispsoftc *isp;
mbreg_t *mbp;
{
int outparam, inparam;
Locking is now in the outer framework for isp_init/isp_reset- this allows us to call isp_reset/isp_init internally with impunity. Rename isp_phoenix to isp_restart and make it global. Clean it up a tad. If we get an ASYNC_SYSTEM_ERROR code in isp_intr, call isp_restart and return- the f/w is toasted at this point (usually), so we have to bring things back to a known state. In isp_mboxcmd, when we don't see the HOST_INT bit go clear, try and find out if the isp is trying to tell us something and try again. This may avoid a potential deadlock where the previous mailbox command hasn't been cleared by the ISP. In isp_init don't try and get device parameters if we already have them- this typically doesn't work if we're in the middle of an isp_restart.
1998-07-19 01:10:16 +04:00
int loops, dld = 0;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
u_int8_t opcode;
if (mbp->param[0] == ISP2100_SET_PCI_PARAM) {
opcode = mbp->param[0] = MBOX_SET_PCI_PARAMETERS;
inparam = 4;
outparam = 4;
goto command_known;
} else if (mbp->param[0] > NMBCOM) {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF("%s: bad command %x\n", isp->isp_name, mbp->param[0]);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
return;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
opcode = mbp->param[0];
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
inparam = HINIB(mbpcnt[mbp->param[0]]);
outparam = LONIB(mbpcnt[mbp->param[0]]);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (inparam == 0 && outparam == 0) {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF("%s: no parameters for %x\n", isp->isp_name,
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
mbp->param[0]);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
return;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
/*
* Check for variants
*/
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
#ifdef ISP2100_SCCLUN
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (IS_FC(isp)) {
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
switch (mbp->param[0]) {
case MBOX_ABORT:
inparam = 7;
break;
case MBOX_ABORT_DEVICE:
case MBOX_START_QUEUE:
case MBOX_STOP_QUEUE:
case MBOX_SINGLE_STEP_QUEUE:
case MBOX_ABORT_QUEUE:
case MBOX_GET_DEV_QUEUE_STATUS:
inparam = 3;
break;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
case MBOX_BUS_RESET:
inparam = 2;
break;
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
default:
break;
}
}
#endif
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
command_known:
Locking is now in the outer framework for isp_init/isp_reset- this allows us to call isp_reset/isp_init internally with impunity. Rename isp_phoenix to isp_restart and make it global. Clean it up a tad. If we get an ASYNC_SYSTEM_ERROR code in isp_intr, call isp_restart and return- the f/w is toasted at this point (usually), so we have to bring things back to a known state. In isp_mboxcmd, when we don't see the HOST_INT bit go clear, try and find out if the isp is trying to tell us something and try again. This may avoid a potential deadlock where the previous mailbox command hasn't been cleared by the ISP. In isp_init don't try and get device parameters if we already have them- this typically doesn't work if we're in the middle of an isp_restart.
1998-07-19 01:10:16 +04:00
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
/*
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
* Set semaphore on mailbox registers to win any races to acquire them.
*/
ISP_WRITE(isp, BIU_SEMA, 1);
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
/*
* Qlogic Errata for the ISP2100 says that there is a necessary
* debounce between between writing the semaphore register
* and reading a mailbox register. I believe we're okay here.
*/
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
/*
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
* Make sure we can send some words.
* Check to see if there's an async mbox event pending.
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
*/
loops = MBOX_DELAY_COUNT;
while ((ISP_READ(isp, HCCR) & HCCR_HOST_INT) != 0) {
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
if (ISP_READ(isp, BIU_SEMA) & 1) {
int fph;
u_int16_t mbox = ISP_READ(isp, OUTMAILBOX0);
/*
* We have a pending MBOX async event.
*/
if (mbox & 0x8000) {
fph = isp_parse_async(isp, (int) mbox);
Some stylistic changes, but also use the new INT_PENDING macro. Start tracking LIP sequence numbers.
1999-10-17 05:38:27 +04:00
IDPRINTF(5, ("%s: line %d, fph %d\n",
isp->isp_name, __LINE__, fph));
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
ISP_WRITE(isp, BIU_SEMA, 0);
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
if (fph < 0) {
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
return;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
} else if (fph > 0) {
isp_fastpost_complete(isp, fph);
}
SYS_DELAY(100);
goto command_known;
}
/*
* We have a pending MBOX completion? Might be
* from a previous command. We can't (sometimes)
* just clear HOST INTERRUPT, so we'll just silently
* eat this here.
*/
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
if (mbox & 0x4000) {
Some stylistic changes, but also use the new INT_PENDING macro. Start tracking LIP sequence numbers.
1999-10-17 05:38:27 +04:00
IDPRINTF(5, ("%s: line %d, mbox 0x%x\n",
isp->isp_name, __LINE__, mbox));
Locking is now in the outer framework for isp_init/isp_reset- this allows us to call isp_reset/isp_init internally with impunity. Rename isp_phoenix to isp_restart and make it global. Clean it up a tad. If we get an ASYNC_SYSTEM_ERROR code in isp_intr, call isp_restart and return- the f/w is toasted at this point (usually), so we have to bring things back to a known state. In isp_mboxcmd, when we don't see the HOST_INT bit go clear, try and find out if the isp is trying to tell us something and try again. This may avoid a potential deadlock where the previous mailbox command hasn't been cleared by the ISP. In isp_init don't try and get device parameters if we already have them- this typically doesn't work if we're in the middle of an isp_restart.
1998-07-19 01:10:16 +04:00
ISP_WRITE(isp, BIU_SEMA, 0);
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
SYS_DELAY(100);
goto command_known;
Locking is now in the outer framework for isp_init/isp_reset- this allows us to call isp_reset/isp_init internally with impunity. Rename isp_phoenix to isp_restart and make it global. Clean it up a tad. If we get an ASYNC_SYSTEM_ERROR code in isp_intr, call isp_restart and return- the f/w is toasted at this point (usually), so we have to bring things back to a known state. In isp_mboxcmd, when we don't see the HOST_INT bit go clear, try and find out if the isp is trying to tell us something and try again. This may avoid a potential deadlock where the previous mailbox command hasn't been cleared by the ISP. In isp_init don't try and get device parameters if we already have them- this typically doesn't work if we're in the middle of an isp_restart.
1998-07-19 01:10:16 +04:00
}
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
}
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
SYS_DELAY(100);
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
if (--loops < 0) {
if (dld++ > 10) {
PRINTF("%s: isp_mboxcmd could not get command "
"started\n", isp->isp_name);
return;
}
ISP_WRITE(isp, BIU_SEMA, 0);
Locking is now in the outer framework for isp_init/isp_reset- this allows us to call isp_reset/isp_init internally with impunity. Rename isp_phoenix to isp_restart and make it global. Clean it up a tad. If we get an ASYNC_SYSTEM_ERROR code in isp_intr, call isp_restart and return- the f/w is toasted at this point (usually), so we have to bring things back to a known state. In isp_mboxcmd, when we don't see the HOST_INT bit go clear, try and find out if the isp is trying to tell us something and try again. This may avoid a potential deadlock where the previous mailbox command hasn't been cleared by the ISP. In isp_init don't try and get device parameters if we already have them- this typically doesn't work if we're in the middle of an isp_restart.
1998-07-19 01:10:16 +04:00
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
goto command_known;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
}
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
/*
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
* Write input parameters.
*
* Special case some of the setups for the dual port SCSI cards.
* XXX Eventually will be fixed by converting register write/read
* XXX counts to bitmasks.
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
*/
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
if (IS_12X0(isp)) {
switch (opcode) {
case MBOX_GET_RETRY_COUNT:
case MBOX_SET_RETRY_COUNT:
ISP_WRITE(isp, INMAILBOX7, mbp->param[7]);
mbp->param[7] = 0;
ISP_WRITE(isp, INMAILBOX6, mbp->param[6]);
mbp->param[6] = 0;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
break;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
case MBOX_SET_ASYNC_DATA_SETUP_TIME:
case MBOX_SET_ACT_NEG_STATE:
case MBOX_SET_TAG_AGE_LIMIT:
case MBOX_SET_SELECT_TIMEOUT:
ISP_WRITE(isp, INMAILBOX2, mbp->param[2]);
Some stylistic changes, but also use the new INT_PENDING macro. Start tracking LIP sequence numbers.
1999-10-17 05:38:27 +04:00
mbp->param[2] = 0;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
break;
}
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
switch (inparam) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
case 8: ISP_WRITE(isp, INMAILBOX7, mbp->param[7]); mbp->param[7] = 0;
case 7: ISP_WRITE(isp, INMAILBOX6, mbp->param[6]); mbp->param[6] = 0;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
case 6:
/*
* The Qlogic 2100 cannot have registers 4 and 5 written to
* after initialization or BAD THINGS HAPPEN (tm).
*/
if (IS_SCSI(isp) || mbp->param[0] == MBOX_INIT_FIRMWARE)
ISP_WRITE(isp, INMAILBOX5, mbp->param[5]);
mbp->param[5] = 0;
case 5:
if (IS_SCSI(isp) || mbp->param[0] == MBOX_INIT_FIRMWARE)
ISP_WRITE(isp, INMAILBOX4, mbp->param[4]);
mbp->param[4] = 0;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
case 4: ISP_WRITE(isp, INMAILBOX3, mbp->param[3]); mbp->param[3] = 0;
case 3: ISP_WRITE(isp, INMAILBOX2, mbp->param[2]); mbp->param[2] = 0;
case 2: ISP_WRITE(isp, INMAILBOX1, mbp->param[1]); mbp->param[1] = 0;
case 1: ISP_WRITE(isp, INMAILBOX0, mbp->param[0]); mbp->param[0] = 0;
}
/*
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
* Clear RISC int condition.
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
*/
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
/*
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
* Clear semaphore on mailbox registers so that the Qlogic
* may update outgoing registers.
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
*/
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
ISP_WRITE(isp, BIU_SEMA, 0);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
/*
* Set Host Interrupt condition so that RISC will pick up mailbox regs.
*/
ISP_WRITE(isp, HCCR, HCCR_CMD_SET_HOST_INT);
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
/*
* Wait until HOST INT has gone away (meaning that the Qlogic
* has picked up the mailbox command. Wait a long time.
*/
loops = MBOX_DELAY_COUNT * 5;
while ((ISP_READ(isp, HCCR) & HCCR_CMD_CLEAR_RISC_INT) != 0) {
SYS_DELAY(100);
if (--loops < 0) {
PRINTF("%s: isp_mboxcmd timeout #2\n", isp->isp_name);
return;
}
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
/*
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
* While the Semaphore registers isn't set, wait for the Qlogic
* to process the mailbox command. Again- wait a long time.
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
*/
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
loops = MBOX_DELAY_COUNT * 5;
while ((ISP_READ(isp, BIU_SEMA) & 1) == 0) {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
SYS_DELAY(100);
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
/*
* Wierd- I've seen the case where the semaphore register
* isn't getting set- sort of a violation of the protocol..
*/
if (ISP_READ(isp, OUTMAILBOX0) & 0x4000)
break;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
if (--loops < 0) {
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
PRINTF("%s: isp_mboxcmd timeout #3\n", isp->isp_name);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
return;
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
}
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
/*
* Make sure that the MBOX_BUSY has gone away
*/
loops = MBOX_DELAY_COUNT;
for (;;) {
u_int16_t mbox = ISP_READ(isp, OUTMAILBOX0);
if (mbox == MBOX_BUSY) {
if (--loops < 0) {
PRINTF("%s: isp_mboxcmd timeout #4\n",
isp->isp_name);
return;
}
SYS_DELAY(100);
continue;
}
/*
* We have a pending MBOX async event.
*/
if (mbox & 0x8000) {
int fph = isp_parse_async(isp, (int) mbox);
ISP_WRITE(isp, BIU_SEMA, 0);
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
if (fph < 0) {
return;
} else if (fph > 0) {
isp_fastpost_complete(isp, fph);
}
SYS_DELAY(100);
continue;
}
break;
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
/*
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
* Pick up output parameters. Special case some of the readbacks
Some stylistic changes, but also use the new INT_PENDING macro. Start tracking LIP sequence numbers.
1999-10-17 05:38:27 +04:00
* for the dual port SCSI cards.
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
*/
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
if (IS_12X0(isp)) {
switch (opcode) {
case MBOX_GET_RETRY_COUNT:
case MBOX_SET_RETRY_COUNT:
mbp->param[7] = ISP_READ(isp, OUTMAILBOX7);
mbp->param[6] = ISP_READ(isp, OUTMAILBOX6);
break;
case MBOX_GET_TAG_AGE_LIMIT:
case MBOX_SET_TAG_AGE_LIMIT:
case MBOX_GET_ACT_NEG_STATE:
case MBOX_SET_ACT_NEG_STATE:
case MBOX_SET_ASYNC_DATA_SETUP_TIME:
case MBOX_GET_ASYNC_DATA_SETUP_TIME:
case MBOX_GET_RESET_DELAY_PARAMS:
case MBOX_SET_RESET_DELAY_PARAMS:
mbp->param[2] = ISP_READ(isp, OUTMAILBOX2);
break;
}
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
switch (outparam) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
case 8: mbp->param[7] = ISP_READ(isp, OUTMAILBOX7);
case 7: mbp->param[6] = ISP_READ(isp, OUTMAILBOX6);
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
case 6: mbp->param[5] = ISP_READ(isp, OUTMAILBOX5);
case 5: mbp->param[4] = ISP_READ(isp, OUTMAILBOX4);
case 4: mbp->param[3] = ISP_READ(isp, OUTMAILBOX3);
case 3: mbp->param[2] = ISP_READ(isp, OUTMAILBOX2);
case 2: mbp->param[1] = ISP_READ(isp, OUTMAILBOX1);
case 1: mbp->param[0] = ISP_READ(isp, OUTMAILBOX0);
}
/*
* Clear RISC int.
*/
ISP_WRITE(isp, HCCR, HCCR_CMD_CLEAR_RISC_INT);
/*
* Release semaphore on mailbox registers
*/
ISP_WRITE(isp, BIU_SEMA, 0);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
/*
* Just to be chatty here...
*/
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
switch (mbp->param[0]) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
case MBOX_COMMAND_COMPLETE:
break;
case MBOX_INVALID_COMMAND:
Some stylistic changes, but also use the new INT_PENDING macro. Start tracking LIP sequence numbers.
1999-10-17 05:38:27 +04:00
IDPRINTF(2, ("%s: mbox cmd %x failed with INVALID_COMMAND\n",
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
isp->isp_name, opcode));
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
break;
case MBOX_HOST_INTERFACE_ERROR:
Some stylistic changes, but also use the new INT_PENDING macro. Start tracking LIP sequence numbers.
1999-10-17 05:38:27 +04:00
PRINTF("%s: mbox cmd %x failed with HOST_INTERFACE_ERROR\n",
isp->isp_name, opcode);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
break;
case MBOX_TEST_FAILED:
Some stylistic changes, but also use the new INT_PENDING macro. Start tracking LIP sequence numbers.
1999-10-17 05:38:27 +04:00
PRINTF("%s: mbox cmd %x failed with TEST_FAILED\n",
isp->isp_name, opcode);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
break;
case MBOX_COMMAND_ERROR:
Some stylistic changes, but also use the new INT_PENDING macro. Start tracking LIP sequence numbers.
1999-10-17 05:38:27 +04:00
if (opcode != MBOX_ABOUT_FIRMWARE)
PRINTF("%s: mbox cmd %x failed with COMMAND_ERROR\n",
isp->isp_name, opcode);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
break;
case MBOX_COMMAND_PARAM_ERROR:
Some stylistic changes, but also use the new INT_PENDING macro. Start tracking LIP sequence numbers.
1999-10-17 05:38:27 +04:00
switch (opcode) {
case MBOX_GET_PORT_DB:
case MBOX_GET_PORT_NAME:
case MBOX_GET_DEV_QUEUE_PARAMS:
break;
default:
PRINTF("%s: mbox cmd %x failed with "
"COMMAND_PARAM_ERROR\n", isp->isp_name, opcode);
}
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
break;
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
/*
* Be silent about these...
*/
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
case ASYNC_PDB_CHANGED:
((fcparam *) isp->isp_param)->isp_loopstate = LOOP_PDB_RCVD;
break;
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
case ASYNC_LIP_OCCURRED:
Some stylistic changes, but also use the new INT_PENDING macro. Start tracking LIP sequence numbers.
1999-10-17 05:38:27 +04:00
((fcparam *) isp->isp_param)->isp_lipseq = mbp->param[1];
/* FALLTHROUGH */
case ASYNC_LOOP_UP:
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
((fcparam *) isp->isp_param)->isp_fwstate = FW_CONFIG_WAIT;
((fcparam *) isp->isp_param)->isp_loopstate = LOOP_LIP_RCVD;
break;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
case ASYNC_LOOP_DOWN:
case ASYNC_LOOP_RESET:
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
((fcparam *) isp->isp_param)->isp_fwstate = FW_CONFIG_WAIT;
((fcparam *) isp->isp_param)->isp_loopstate = LOOP_NIL;
/* FALLTHROUGH */
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
case ASYNC_CHANGE_NOTIFY:
break;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
default:
/*
* The expected return of EXEC_FIRMWARE is zero.
*/
if ((opcode == MBOX_EXEC_FIRMWARE && mbp->param[0] != 0) ||
(opcode != MBOX_EXEC_FIRMWARE)) {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF("%s: mbox cmd %x failed with error %x\n",
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp->isp_name, opcode, mbp->param[0]);
}
break;
}
ISP 10x0 driver from Matthew Jacob of NASA Ames Research Center. (March 12, 1997 version).
1997-03-12 23:44:50 +03:00
}
A) Attempt to get clock rate from ISP chip prior to bashing it over the head in reset. If none there, try and get from the bus/platform specific code. If a nonzero value for either, set the clock rate. This is why the PCI card versions weren't working- they need to be set at 60MHZ, rather than the default 40MHZ (which worked fine for the internal ISP chips on the Alpha 8X00). B) If a isp_poll returns failure (command never completed) to the caller and no error is set in the xs struct, set XS_SELTIMEOUT. And then call... C) Added isp_lostcmd function to try and ask the ISP chip about it's current state as well as the state of commands for a particular target/lun. This is going along to try and figure out why the very first command to the ISP always seems to get swallowed up.
1997-04-05 06:53:22 +04:00
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
void
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
isp_lostcmd(isp, xs)
struct ispsoftc *isp;
ISP_SCSI_XFER_T *xs;
A) Attempt to get clock rate from ISP chip prior to bashing it over the head in reset. If none there, try and get from the bus/platform specific code. If a nonzero value for either, set the clock rate. This is why the PCI card versions weren't working- they need to be set at 60MHZ, rather than the default 40MHZ (which worked fine for the internal ISP chips on the Alpha 8X00). B) If a isp_poll returns failure (command never completed) to the caller and no error is set in the xs struct, set XS_SELTIMEOUT. And then call... C) Added isp_lostcmd function to try and ask the ISP chip about it's current state as well as the state of commands for a particular target/lun. This is going along to try and figure out why the very first command to the ISP always seems to get swallowed up.
1997-04-05 06:53:22 +04:00
{
mbreg_t mbs;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
mbs.param[0] = MBOX_GET_FIRMWARE_STATUS;
isp_mboxcmd(isp, &mbs);
A) Attempt to get clock rate from ISP chip prior to bashing it over the head in reset. If none there, try and get from the bus/platform specific code. If a nonzero value for either, set the clock rate. This is why the PCI card versions weren't working- they need to be set at 60MHZ, rather than the default 40MHZ (which worked fine for the internal ISP chips on the Alpha 8X00). B) If a isp_poll returns failure (command never completed) to the caller and no error is set in the xs struct, set XS_SELTIMEOUT. And then call... C) Added isp_lostcmd function to try and ask the ISP chip about it's current state as well as the state of commands for a particular target/lun. This is going along to try and figure out why the very first command to the ISP always seems to get swallowed up.
1997-04-05 06:53:22 +04:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_dumpregs(isp, "couldn't GET FIRMWARE STATUS");
A) Attempt to get clock rate from ISP chip prior to bashing it over the head in reset. If none there, try and get from the bus/platform specific code. If a nonzero value for either, set the clock rate. This is why the PCI card versions weren't working- they need to be set at 60MHZ, rather than the default 40MHZ (which worked fine for the internal ISP chips on the Alpha 8X00). B) If a isp_poll returns failure (command never completed) to the caller and no error is set in the xs struct, set XS_SELTIMEOUT. And then call... C) Added isp_lostcmd function to try and ask the ISP chip about it's current state as well as the state of commands for a particular target/lun. This is going along to try and figure out why the very first command to the ISP always seems to get swallowed up.
1997-04-05 06:53:22 +04:00
return;
}
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
if (mbs.param[1]) {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF("%s: %d commands on completion queue\n",
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
isp->isp_name, mbs.param[1]);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
if (XS_NULL(xs))
A) Attempt to get clock rate from ISP chip prior to bashing it over the head in reset. If none there, try and get from the bus/platform specific code. If a nonzero value for either, set the clock rate. This is why the PCI card versions weren't working- they need to be set at 60MHZ, rather than the default 40MHZ (which worked fine for the internal ISP chips on the Alpha 8X00). B) If a isp_poll returns failure (command never completed) to the caller and no error is set in the xs struct, set XS_SELTIMEOUT. And then call... C) Added isp_lostcmd function to try and ask the ISP chip about it's current state as well as the state of commands for a particular target/lun. This is going along to try and figure out why the very first command to the ISP always seems to get swallowed up.
1997-04-05 06:53:22 +04:00
return;
mbs.param[0] = MBOX_GET_DEV_QUEUE_STATUS;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
mbs.param[1] = (XS_TGT(xs) << 8) | XS_LUN(xs); /* XXX: WHICH BUS? */
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_mboxcmd(isp, &mbs);
A) Attempt to get clock rate from ISP chip prior to bashing it over the head in reset. If none there, try and get from the bus/platform specific code. If a nonzero value for either, set the clock rate. This is why the PCI card versions weren't working- they need to be set at 60MHZ, rather than the default 40MHZ (which worked fine for the internal ISP chips on the Alpha 8X00). B) If a isp_poll returns failure (command never completed) to the caller and no error is set in the xs struct, set XS_SELTIMEOUT. And then call... C) Added isp_lostcmd function to try and ask the ISP chip about it's current state as well as the state of commands for a particular target/lun. This is going along to try and figure out why the very first command to the ISP always seems to get swallowed up.
1997-04-05 06:53:22 +04:00
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_dumpregs(isp, "couldn't GET DEVICE QUEUE STATUS");
return;
}
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF("%s: lost command for target %d lun %d, %d active of %d, "
"Queue State: %x\n", isp->isp_name, XS_TGT(xs),
XS_LUN(xs), mbs.param[2], mbs.param[3], mbs.param[1]);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_dumpregs(isp, "lost command");
/*
* XXX: Need to try and do something to recover.
*/
}
static void
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
isp_dumpregs(isp, msg)
struct ispsoftc *isp;
const char *msg;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
{
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF("%s: %s\n", isp->isp_name, msg);
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (IS_SCSI(isp))
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF(" biu_conf1=%x", ISP_READ(isp, BIU_CONF1));
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
else
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF(" biu_csr=%x", ISP_READ(isp, BIU2100_CSR));
PRINTF(" biu_icr=%x biu_isr=%x biu_sema=%x ", ISP_READ(isp, BIU_ICR),
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
ISP_READ(isp, BIU_ISR), ISP_READ(isp, BIU_SEMA));
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF("risc_hccr=%x\n", ISP_READ(isp, HCCR));
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (IS_SCSI(isp)) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
ISP_WRITE(isp, HCCR, HCCR_CMD_PAUSE);
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF(" cdma_conf=%x cdma_sts=%x cdma_fifostat=%x\n",
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
ISP_READ(isp, CDMA_CONF), ISP_READ(isp, CDMA_STATUS),
ISP_READ(isp, CDMA_FIFO_STS));
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF(" ddma_conf=%x ddma_sts=%x ddma_fifostat=%x\n",
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
ISP_READ(isp, DDMA_CONF), ISP_READ(isp, DDMA_STATUS),
ISP_READ(isp, DDMA_FIFO_STS));
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
PRINTF(" sxp_int=%x sxp_gross=%x sxp(scsi_ctrl)=%x\n",
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
ISP_READ(isp, SXP_INTERRUPT),
ISP_READ(isp, SXP_GROSS_ERR),
ISP_READ(isp, SXP_PINS_CONTROL));
ISP_WRITE(isp, HCCR, HCCR_CMD_RELEASE);
}
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
PRINTF(" mbox regs: %x %x %x %x %x\n",
ISP_READ(isp, OUTMAILBOX0), ISP_READ(isp, OUTMAILBOX1),
ISP_READ(isp, OUTMAILBOX2), ISP_READ(isp, OUTMAILBOX3),
ISP_READ(isp, OUTMAILBOX4));
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
ISP_DUMPREGS(isp);
}
static void
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
isp_fw_state(isp)
struct ispsoftc *isp;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
{
mbreg_t mbs;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (IS_FC(isp)) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
int once = 0;
fcparam *fcp = isp->isp_param;
again:
mbs.param[0] = MBOX_GET_FW_STATE;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
IDPRINTF(0, ("%s: isp_fw_state 0x%x\n", isp->isp_name,
mbs.param[0]));
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
switch (mbs.param[0]) {
case ASYNC_PDB_CHANGED:
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
if (once++ < 10) {
goto again;
}
fcp->isp_fwstate = FW_CONFIG_WAIT;
fcp->isp_loopstate = LOOP_PDB_RCVD;
goto again;
case ASYNC_LIP_OCCURRED:
Some stylistic changes, but also use the new INT_PENDING macro. Start tracking LIP sequence numbers.
1999-10-17 05:38:27 +04:00
fcp->isp_lipseq = mbs.param[1];
/* FALLTHROUGH */
case ASYNC_LOOP_UP:
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
fcp->isp_fwstate = FW_CONFIG_WAIT;
fcp->isp_loopstate = LOOP_LIP_RCVD;
if (once++ < 10) {
goto again;
}
break;
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
case ASYNC_LOOP_RESET:
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
case ASYNC_LOOP_DOWN:
fcp->isp_fwstate = FW_CONFIG_WAIT;
fcp->isp_loopstate = LOOP_NIL;
/* FALLTHROUGH */
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
case ASYNC_CHANGE_NOTIFY:
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
if (once++ < 10) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
goto again;
}
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
break;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
PRINTF("%s: GET FIRMWARE STATE failed (0x%x)\n",
isp->isp_name, mbs.param[0]);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
return;
}
fcp->isp_fwstate = mbs.param[1];
}
}
static void
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
isp_update(isp)
struct ispsoftc *isp;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
{
int bus;
for (bus = 0; isp->isp_update != 0; bus++) {
if (isp->isp_update & (1 << bus)) {
isp_update_bus(isp, bus);
isp->isp_update ^= (1 << bus);
}
}
}
static void
isp_update_bus(isp, bus)
struct ispsoftc *isp;
int bus;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
{
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
int tgt;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
mbreg_t mbs;
sdparam *sdp;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
if (IS_FC(isp)) {
A) Attempt to get clock rate from ISP chip prior to bashing it over the head in reset. If none there, try and get from the bus/platform specific code. If a nonzero value for either, set the clock rate. This is why the PCI card versions weren't working- they need to be set at 60MHZ, rather than the default 40MHZ (which worked fine for the internal ISP chips on the Alpha 8X00). B) If a isp_poll returns failure (command never completed) to the caller and no error is set in the xs struct, set XS_SELTIMEOUT. And then call... C) Added isp_lostcmd function to try and ask the ISP chip about it's current state as well as the state of commands for a particular target/lun. This is going along to try and figure out why the very first command to the ISP always seems to get swallowed up.
1997-04-05 06:53:22 +04:00
return;
}
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
sdp = isp->isp_param;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
sdp += bus;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
for (tgt = 0; tgt < MAX_TARGETS; tgt++) {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
u_int16_t flags, period, offset;
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
int get;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
if (sdp->isp_devparam[tgt].dev_enable == 0) {
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
IDPRINTF(1, ("%s: skipping target %d bus %d update\n",
isp->isp_name, tgt, bus));
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
continue;
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
/*
* If the goal is to update the status of the device,
* take what's in dev_flags and try and set the device
* toward that. Otherwise, if we're just refreshing the
* current device state, get the current parameters.
*/
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
if (sdp->isp_devparam[tgt].dev_update) {
mbs.param[0] = MBOX_SET_TARGET_PARAMS;
mbs.param[2] = sdp->isp_devparam[tgt].dev_flags;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
/*
* Insist that PARITY must be enabled if SYNC
* is enabled.
*/
if (mbs.param[2] & DPARM_SYNC) {
mbs.param[2] |= DPARM_PARITY;
}
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
mbs.param[3] =
(sdp->isp_devparam[tgt].sync_offset << 8) |
(sdp->isp_devparam[tgt].sync_period);
sdp->isp_devparam[tgt].dev_update = 0;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
/*
* A command completion later that has
* RQSTF_NEGOTIATION set will cause
* the dev_refresh/announce cycle.
*
* Note: It is really important to update our current
* flags with at least the state of TAG capabilities-
* otherwise we might try and send a tagged command
* when we have it all turned off. So change it here
* to say that current already matches goal.
*/
sdp->isp_devparam[tgt].cur_dflags &= ~DPARM_TQING;
sdp->isp_devparam[tgt].cur_dflags |=
(sdp->isp_devparam[tgt].dev_flags & DPARM_TQING);
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
sdp->isp_devparam[tgt].dev_refresh = 1;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
IDPRINTF(3, ("%s: bus %d set tgt %d flags 0x%x off 0x%x"
" period 0x%x\n", isp->isp_name, bus, tgt,
mbs.param[2], mbs.param[3] >> 8,
mbs.param[3] & 0xff));
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
get = 0;
} else if (sdp->isp_devparam[tgt].dev_refresh) {
mbs.param[0] = MBOX_GET_TARGET_PARAMS;
sdp->isp_devparam[tgt].dev_refresh = 0;
get = 1;
} else {
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
continue;
}
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
mbs.param[1] = (bus << 15) | (tgt << 8) ;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
PRINTF("%s: failed to %cet SCSI parameters for "
"target %d\n", isp->isp_name, (get)? 'g' : 's',
tgt);
continue;
}
if (get == 0) {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp->isp_sendmarker |= (1 << bus);
Implement and use Fast Posting for both parallel && fibre. Redo a bit of the startup code. Implement a call to outer framework function so that asynchronous events can be handled (e.g., speed negotiation, target mode). Roll internal release tags.
1999-01-30 10:31:50 +03:00
continue;
}
flags = mbs.param[2];
period = mbs.param[3] & 0xff;
offset = mbs.param[3] >> 8;
sdp->isp_devparam[tgt].cur_dflags = flags;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
sdp->isp_devparam[tgt].cur_period = period;
sdp->isp_devparam[tgt].cur_offset = offset;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
get = (bus << 16) | tgt;
(void) isp_async(isp, ISPASYNC_NEW_TGT_PARAMS, &get);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
}
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
static void
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp_setdfltparm(isp, channel)
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
struct ispsoftc *isp;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
int channel;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
{
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
int tgt;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
mbreg_t mbs;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
sdparam *sdp, *sdp_chan0, *sdp_chan1;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
if (IS_FC(isp)) {
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
fcparam *fcp = (fcparam *) isp->isp_param;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
fcp += channel;
if (fcp->isp_gotdparms) {
return;
}
fcp->isp_gotdparms = 1;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
fcp->isp_maxfrmlen = ICB_DFLT_FRMLEN;
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
fcp->isp_maxalloc = ICB_DFLT_ALLOC;
fcp->isp_execthrottle = ICB_DFLT_THROTTLE;
fcp->isp_retry_delay = ICB_DFLT_RDELAY;
fcp->isp_retry_count = ICB_DFLT_RCOUNT;
/* Platform specific.... */
fcp->isp_loopid = DEFAULT_LOOPID(isp);
fcp->isp_nodewwn = DEFAULT_WWN(isp);
fcp->isp_portwwn = DEFAULT_WWN(isp);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
/*
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
* Now try and read NVRAM
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
*/
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
if ((isp->isp_confopts & ISP_CFG_NONVRAM) == 0) {
if (isp_read_nvram(isp)) {
PRINTF("%s: using default WWN 0x%08x%08x\n",
isp->isp_name,
(u_int32_t)(fcp->isp_portwwn >> 32),
(u_int32_t)(fcp->isp_portwwn & 0xffffffff));
}
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
return;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
sdp_chan0 = (sdparam *) isp->isp_param;
sdp_chan1 = sdp_chan0 + 1;
sdp = sdp_chan0 + channel;
/*
* Been there, done that, got the T-shirt...
*/
if (sdp->isp_gotdparms) {
return;
}
sdp->isp_gotdparms = 1;
/*
* If we've not been told to avoid reading NVRAM, try and read it.
* If we're successful reading it, we can return since NVRAM will
* tell us the right thing to do. Otherwise, establish some reasonable
* defaults.
*/
if ((isp->isp_confopts & ISP_CFG_NONVRAM) == 0) {
if (isp_read_nvram(isp) == 0) {
return;
}
}
/*
* Now try and see whether we have specific values for them.
*/
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
mbs.param[0] = MBOX_GET_ACT_NEG_STATE;
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
IDPRINTF(2, ("could not GET ACT NEG STATE\n"));
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
sdp_chan0->isp_req_ack_active_neg = 1;
sdp_chan0->isp_data_line_active_neg = 1;
if (IS_12X0(isp)) {
sdp_chan1->isp_req_ack_active_neg = 1;
sdp_chan1->isp_data_line_active_neg = 1;
}
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
} else {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
sdp_chan0->isp_req_ack_active_neg = (mbs.param[1] >> 4) & 0x1;
sdp_chan0->isp_data_line_active_neg = (mbs.param[1] >> 5) & 0x1;
if (IS_12X0(isp)) {
sdp_chan1->isp_req_ack_active_neg =
(mbs.param[2] >> 4) & 0x1;
sdp_chan1->isp_data_line_active_neg =
(mbs.param[2] >> 5) & 0x1;
}
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
/*
* The trick here is to establish a default for the default (honk!)
* state (dev_flags). Then try and get the current status from
* the card to fill in the current state. We don't, in fact, set
* the default to the SAFE default state- that's not the goal state.
*/
for (tgt = 0; tgt < MAX_TARGETS; tgt++) {
sdp->isp_devparam[tgt].cur_offset = 0;
sdp->isp_devparam[tgt].cur_period = 0;
sdp->isp_devparam[tgt].dev_flags = DPARM_DEFAULT;
sdp->isp_devparam[tgt].cur_dflags = 0;
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
if (isp->isp_type < ISP_HA_SCSI_1040 ||
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
(isp->isp_clock && isp->isp_clock < 60)) {
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
sdp->isp_devparam[tgt].sync_offset =
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
ISP_10M_SYNCPARMS >> 8;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
sdp->isp_devparam[tgt].sync_period =
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
ISP_10M_SYNCPARMS & 0xff;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
} else if (IS_1080(isp)) {
sdp->isp_devparam[tgt].sync_offset =
ISP_40M_SYNCPARMS >> 8;
sdp->isp_devparam[tgt].sync_period =
ISP_40M_SYNCPARMS & 0xff;
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
} else {
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
sdp->isp_devparam[tgt].sync_offset =
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
ISP_20M_SYNCPARMS >> 8;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
sdp->isp_devparam[tgt].sync_period =
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
ISP_20M_SYNCPARMS & 0xff;
}
/*
* Don't get current target parameters if we've been
* told not to use NVRAM- it's really the same thing.
*/
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
if (isp->isp_confopts & ISP_CFG_NONVRAM) {
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
continue;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
mbs.param[0] = MBOX_GET_TARGET_PARAMS;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
mbs.param[1] = tgt << 8;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_mboxcmd(isp, &mbs);
if (mbs.param[0] != MBOX_COMMAND_COMPLETE) {
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
continue;
}
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
sdp->isp_devparam[tgt].cur_dflags = mbs.param[2];
sdp->isp_devparam[tgt].dev_flags = mbs.param[2];
sdp->isp_devparam[tgt].cur_period = mbs.param[3] & 0xff;
sdp->isp_devparam[tgt].cur_offset = mbs.param[3] >> 8;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
/*
* The maximum period we can really see
* here is 100 (decimal), or 400 ns.
* For some unknown reason we sometimes
* get back wildass numbers from the
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
* boot device's parameters (alpha only).
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
*/
if ((mbs.param[3] & 0xff) <= 0x64) {
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
sdp->isp_devparam[tgt].sync_period =
mbs.param[3] & 0xff;
sdp->isp_devparam[tgt].sync_offset =
mbs.param[3] >> 8;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
}
/*
* It is not safe to run Ultra Mode with a clock < 60.
*/
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
if (((isp->isp_clock && isp->isp_clock < 60) ||
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
(isp->isp_type < ISP_HA_SCSI_1020A)) &&
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
(sdp->isp_devparam[tgt].sync_period <=
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
(ISP_20M_SYNCPARMS & 0xff))) {
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
sdp->isp_devparam[tgt].sync_offset =
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
ISP_10M_SYNCPARMS >> 8;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
sdp->isp_devparam[tgt].sync_period =
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
ISP_10M_SYNCPARMS & 0xff;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
}
/*
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
* Establish default some more default parameters.
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
*/
sdp->isp_cmd_dma_burst_enable = 1;
sdp->isp_data_dma_burst_enabl = 1;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
sdp->isp_fifo_threshold = 0;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
sdp->isp_initiator_id = 7;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
/* XXXX This is probably based upon clock XXXX */
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
if (isp->isp_type >= ISP_HA_SCSI_1040) {
sdp->isp_async_data_setup = 9;
} else {
sdp->isp_async_data_setup = 6;
}
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
sdp->isp_selection_timeout = 250;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
sdp->isp_max_queue_depth = MAXISPREQUEST;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
sdp->isp_tag_aging = 8;
sdp->isp_bus_reset_delay = 3;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
sdp->isp_retry_count = 2;
sdp->isp_retry_delay = 2;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
for (tgt = 0; tgt < MAX_TARGETS; tgt++) {
sdp->isp_devparam[tgt].exc_throttle = 16;
sdp->isp_devparam[tgt].dev_enable = 1;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
}
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
/*
Locking is now in the outer framework for isp_init/isp_reset- this allows us to call isp_reset/isp_init internally with impunity. Rename isp_phoenix to isp_restart and make it global. Clean it up a tad. If we get an ASYNC_SYSTEM_ERROR code in isp_intr, call isp_restart and return- the f/w is toasted at this point (usually), so we have to bring things back to a known state. In isp_mboxcmd, when we don't see the HOST_INT bit go clear, try and find out if the isp is trying to tell us something and try again. This may avoid a potential deadlock where the previous mailbox command hasn't been cleared by the ISP. In isp_init don't try and get device parameters if we already have them- this typically doesn't work if we're in the middle of an isp_restart.
1998-07-19 01:10:16 +04:00
* Re-initialize the ISP and complete all orphaned commands
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
* with a 'botched' notice. The reset/init routines should
* not disturb an already active list of commands.
Locking is now in the outer framework for isp_init/isp_reset- this allows us to call isp_reset/isp_init internally with impunity. Rename isp_phoenix to isp_restart and make it global. Clean it up a tad. If we get an ASYNC_SYSTEM_ERROR code in isp_intr, call isp_restart and return- the f/w is toasted at this point (usually), so we have to bring things back to a known state. In isp_mboxcmd, when we don't see the HOST_INT bit go clear, try and find out if the isp is trying to tell us something and try again. This may avoid a potential deadlock where the previous mailbox command hasn't been cleared by the ISP. In isp_init don't try and get device parameters if we already have them- this typically doesn't work if we're in the middle of an isp_restart.
1998-07-19 01:10:16 +04:00
*
* Locks held prior to coming here.
*/
void
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
isp_restart(isp)
struct ispsoftc *isp;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
{
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
ISP_SCSI_XFER_T *xs;
u_int32_t handle;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
#if 0
Make firmware revision a triple. Clean up some FC init stuff for board versions with no BIOS. Separate mailbox interrupts from IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not after you clear it (potential race condition). Clear out older broken BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup if already in LVD mode. Note usage of presumptive 1040C revision. For all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state as unknown as well as marking the need to do a getpdb on targets- after a LIP for certain the f/w has to do PRLI/PLOGI for all targets again and marking f/w state as unknown gives us a fighting chance to (start to) hold up for that to complete.
1999-04-04 06:28:53 +04:00
isp->isp_gotdparms = 0;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
#endif
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
isp_reset(isp);
Locking is now in the outer framework for isp_init/isp_reset- this allows us to call isp_reset/isp_init internally with impunity. Rename isp_phoenix to isp_restart and make it global. Clean it up a tad. If we get an ASYNC_SYSTEM_ERROR code in isp_intr, call isp_restart and return- the f/w is toasted at this point (usually), so we have to bring things back to a known state. In isp_mboxcmd, when we don't see the HOST_INT bit go clear, try and find out if the isp is trying to tell us something and try again. This may avoid a potential deadlock where the previous mailbox command hasn't been cleared by the ISP. In isp_init don't try and get device parameters if we already have them- this typically doesn't work if we're in the middle of an isp_restart.
1998-07-19 01:10:16 +04:00
if (isp->isp_state == ISP_RESETSTATE) {
isp_init(isp);
if (isp->isp_state == ISP_INITSTATE) {
isp->isp_state = ISP_RUNSTATE;
}
}
if (isp->isp_state != ISP_RUNSTATE) {
PRINTF("%s: isp_restart cannot restart ISP\n", isp->isp_name);
}
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
isp->isp_nactive = 0;
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
for (handle = 1; handle <= (int) isp->isp_maxcmds; handle++) {
xs = isp_find_xs(isp, handle);
if (xs == NULL) {
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
continue;
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
}
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
isp_destroy_handle(isp, handle);
if (XS_XFRLEN(xs)) {
ISP_DMAFREE(isp, xs, handle);
XS_RESID(xs) = XS_XFRLEN(xs);
} else {
XS_RESID(xs) = 0;
}
Several changes, including some stub fibre target mode stuff. Establish a slightly different fibre startup (print ALPA now too). Change the way that return values from dma setup is done. Make debug messages out of some queue overflow situations. Turn PORT LOGGED OUT into Selection Timeout equivlaent. On isp_restart actions don't blow off the commands with HBA_BOTCH (XS_DRIVER_STUFFUP) - set them with HBA_BUSRESET (which is defined as XS_DRIVER_STUFFUP until someone decides whether the suggested change to the midlayer NetBSD is worthy of inclusion).
1998-09-18 03:05:46 +04:00
XS_SETERR(xs, HBA_BUSRESET);
Some major thrashing to handle building for other than NetBSD. There is one change of note- build a list of completing commands in ispintr and then say you're done- this avoids some re-entrancy issues that had surfaced.
1998-07-15 23:50:16 +04:00
XS_CMD_DONE(xs);
*Major* differences to support the Qlogic 2100 (Fibre Channel PCI), as well as being able to do microdifferentiation of the ISP SCSI chip; a couple of bug fixes, and the beginnings of some watchdog support.
1997-08-16 04:22:11 +04:00
}
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
/*
* NVRAM Routines
*/
static int
isp_read_nvram(isp)
struct ispsoftc *isp;
{
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
static char *tru = "true";
static char *not = "false";
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
int i, amt;
u_int8_t csum, minversion;
union {
u_int8_t _x[ISP2100_NVRAM_SIZE];
u_int16_t _s[ISP2100_NVRAM_SIZE>>1];
} _n;
#define nvram_data _n._x
#define nvram_words _n._s
Update the driver with some infrastructure for the 1080. Fix an embarrassing clock botch bug. Additional infrastructure for PDB change stuff.
1999-03-17 09:15:47 +03:00
if (IS_FC(isp)) {
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
amt = ISP2100_NVRAM_SIZE;
minversion = 1;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
} else if (IS_1080(isp) || IS_12X0(isp)) {
amt = ISP1080_NVRAM_SIZE;
minversion = 0;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
} else {
amt = ISP_NVRAM_SIZE;
minversion = 2;
}
/*
* Just read the first two words first to see if we have a valid
* NVRAM to continue reading the rest with.
*/
for (i = 0; i < 2; i++) {
isp_rdnvram_word(isp, i, &nvram_words[i]);
}
if (nvram_data[0] != 'I' || nvram_data[1] != 'S' ||
nvram_data[2] != 'P') {
if (isp->isp_bustype != ISP_BT_SBUS) {
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
PRINTF("%s: invalid NVRAM header (%x,%x,%x,%x)\n",
isp->isp_name, nvram_data[0], nvram_data[1],
nvram_data[2], nvram_data[3]);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
}
return (-1);
}
for (i = 2; i < amt>>1; i++) {
isp_rdnvram_word(isp, i, &nvram_words[i]);
}
for (csum = 0, i = 0; i < amt; i++) {
csum += nvram_data[i];
}
if (csum != 0) {
PRINTF("%s: invalid NVRAM checksum\n", isp->isp_name);
return (-1);
}
if (ISP_NVRAM_VERSION(nvram_data) < minversion) {
PRINTF("%s: version %d NVRAM not understood\n", isp->isp_name,
ISP_NVRAM_VERSION(nvram_data));
return (-1);
}
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
if (IS_1080(isp) || IS_12X0(isp)) {
int bus;
sdparam *sdp = (sdparam *) isp->isp_param;
for (bus = 0; bus < (IS_1080(isp)? 1 : 2); bus++, sdp++) {
sdp->isp_fifo_threshold =
ISP1080_NVRAM_FIFO_THRESHOLD(nvram_data);
sdp->isp_initiator_id =
ISP1080_NVRAM_INITIATOR_ID(nvram_data, bus);
sdp->isp_bus_reset_delay =
ISP1080_NVRAM_BUS_RESET_DELAY(nvram_data, bus);
sdp->isp_retry_count =
ISP1080_NVRAM_BUS_RETRY_COUNT(nvram_data, bus);
sdp->isp_retry_delay =
ISP1080_NVRAM_BUS_RETRY_DELAY(nvram_data, bus);
sdp->isp_async_data_setup =
ISP1080_NVRAM_ASYNC_DATA_SETUP_TIME(nvram_data,
bus);
sdp->isp_req_ack_active_neg =
ISP1080_NVRAM_REQ_ACK_ACTIVE_NEGATION(nvram_data,
bus);
sdp->isp_data_line_active_neg =
ISP1080_NVRAM_DATA_LINE_ACTIVE_NEGATION(nvram_data,
bus);
sdp->isp_data_dma_burst_enabl =
ISP1080_NVRAM_BURST_ENABLE(nvram_data);
sdp->isp_cmd_dma_burst_enable =
ISP1080_NVRAM_BURST_ENABLE(nvram_data);
sdp->isp_selection_timeout =
ISP1080_NVRAM_SELECTION_TIMEOUT(nvram_data, bus);
sdp->isp_max_queue_depth =
ISP1080_NVRAM_MAX_QUEUE_DEPTH(nvram_data, bus);
if (isp->isp_dblev >= 3) {
PRINTF("%s: ISP1080 bus %d NVRAM values:\n",
isp->isp_name, bus);
PRINTF(" Initiator ID = %d\n",
sdp->isp_initiator_id);
PRINTF(" Fifo Threshold = 0x%x\n",
sdp->isp_fifo_threshold);
PRINTF(" Bus Reset Delay = %d\n",
sdp->isp_bus_reset_delay);
PRINTF(" Retry Count = %d\n",
sdp->isp_retry_count);
PRINTF(" Retry Delay = %d\n",
sdp->isp_retry_delay);
PRINTF(" Tag Age Limit = %d\n",
sdp->isp_tag_aging);
PRINTF(" Selection Timeout = %d\n",
sdp->isp_selection_timeout);
PRINTF(" Max Queue Depth = %d\n",
sdp->isp_max_queue_depth);
PRINTF(" Async Data Setup = 0x%x\n",
sdp->isp_async_data_setup);
PRINTF(" REQ/ACK Active Negation = %s\n",
sdp->isp_req_ack_active_neg? tru : not);
PRINTF(" Data Line Active Negation = %s\n",
sdp->isp_data_line_active_neg? tru : not);
PRINTF(" Cmd DMA Burst Enable = %s\n",
sdp->isp_cmd_dma_burst_enable? tru : not);
}
for (i = 0; i < MAX_TARGETS; i++) {
sdp->isp_devparam[i].dev_enable =
ISP1080_NVRAM_TGT_DEVICE_ENABLE(nvram_data, i, bus);
sdp->isp_devparam[i].exc_throttle =
ISP1080_NVRAM_TGT_EXEC_THROTTLE(nvram_data, i, bus);
sdp->isp_devparam[i].sync_offset =
ISP1080_NVRAM_TGT_SYNC_OFFSET(nvram_data, i, bus);
sdp->isp_devparam[i].sync_period =
ISP1080_NVRAM_TGT_SYNC_PERIOD(nvram_data, i, bus);
sdp->isp_devparam[i].dev_flags = 0;
if (ISP1080_NVRAM_TGT_RENEG(nvram_data, i, bus))
sdp->isp_devparam[i].dev_flags |= DPARM_RENEG;
if (ISP1080_NVRAM_TGT_QFRZ(nvram_data, i, bus)) {
PRINTF("%s: not supporting QFRZ option "
"for target %d bus %d\n",
isp->isp_name, i, bus);
}
sdp->isp_devparam[i].dev_flags |= DPARM_ARQ;
if (ISP1080_NVRAM_TGT_ARQ(nvram_data, i, bus) == 0) {
PRINTF("%s: not disabling ARQ option "
"for target %d bus %d\n",
isp->isp_name, i, bus);
}
if (ISP1080_NVRAM_TGT_TQING(nvram_data, i, bus))
sdp->isp_devparam[i].dev_flags |= DPARM_TQING;
if (ISP1080_NVRAM_TGT_SYNC(nvram_data, i, bus))
sdp->isp_devparam[i].dev_flags |= DPARM_SYNC;
if (ISP1080_NVRAM_TGT_WIDE(nvram_data, i, bus))
sdp->isp_devparam[i].dev_flags |= DPARM_WIDE;
if (ISP1080_NVRAM_TGT_PARITY(nvram_data, i, bus))
sdp->isp_devparam[i].dev_flags |= DPARM_PARITY;
if (ISP1080_NVRAM_TGT_DISC(nvram_data, i, bus))
sdp->isp_devparam[i].dev_flags |= DPARM_DISC;
sdp->isp_devparam[i].cur_dflags = 0;
if (isp->isp_dblev >= 3) {
PRINTF(" Target %d: Ena %d Throttle "
"%d Offset %d Period %d Flags "
"0x%x\n", i,
sdp->isp_devparam[i].dev_enable,
sdp->isp_devparam[i].exc_throttle,
sdp->isp_devparam[i].sync_offset,
sdp->isp_devparam[i].sync_period,
sdp->isp_devparam[i].dev_flags);
}
}
}
} else if (IS_SCSI(isp)) {
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
sdparam *sdp = (sdparam *) isp->isp_param;
sdp->isp_fifo_threshold =
+ Some basic deadchip detection. + Enable FIFO bursts, but also detect bogus 1040A with busted FIFO. + Use new MEMZERO crossplatform define. + Handle RQCS_QUEUE_FULL status case and let upper layer parse SCSI ststus byte if nonzero (should be 0x28- Queue Full status) + Fold ISP_NVRAM_FIFO_THRESHOLD_128 into isp_fifo_threshold tag.
1999-01-10 08:07:53 +03:00
ISP_NVRAM_FIFO_THRESHOLD(nvram_data) |
(ISP_NVRAM_FIFO_THRESHOLD_128(nvram_data) << 2);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
sdp->isp_initiator_id =
ISP_NVRAM_INITIATOR_ID(nvram_data);
sdp->isp_bus_reset_delay =
ISP_NVRAM_BUS_RESET_DELAY(nvram_data);
sdp->isp_retry_count =
ISP_NVRAM_BUS_RETRY_COUNT(nvram_data);
sdp->isp_retry_delay =
ISP_NVRAM_BUS_RETRY_DELAY(nvram_data);
sdp->isp_async_data_setup =
ISP_NVRAM_ASYNC_DATA_SETUP_TIME(nvram_data);
if (isp->isp_type >= ISP_HA_SCSI_1040) {
if (sdp->isp_async_data_setup < 9) {
sdp->isp_async_data_setup = 9;
}
} else {
if (sdp->isp_async_data_setup != 6) {
sdp->isp_async_data_setup = 6;
}
}
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
sdp->isp_req_ack_active_neg =
ISP_NVRAM_REQ_ACK_ACTIVE_NEGATION(nvram_data);
sdp->isp_data_line_active_neg =
ISP_NVRAM_DATA_LINE_ACTIVE_NEGATION(nvram_data);
sdp->isp_data_dma_burst_enabl =
ISP_NVRAM_DATA_DMA_BURST_ENABLE(nvram_data);
sdp->isp_cmd_dma_burst_enable =
ISP_NVRAM_CMD_DMA_BURST_ENABLE(nvram_data);
sdp->isp_tag_aging =
ISP_NVRAM_TAG_AGE_LIMIT(nvram_data);
sdp->isp_selection_timeout =
ISP_NVRAM_SELECTION_TIMEOUT(nvram_data);
sdp->isp_max_queue_depth =
ISP_NVRAM_MAX_QUEUE_DEPTH(nvram_data);
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp->isp_fast_mttr = ISP_NVRAM_FAST_MTTR_ENABLE(nvram_data);
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
if (isp->isp_dblev > 2) {
PRINTF("%s: NVRAM values:\n", isp->isp_name);
PRINTF(" Fifo Threshold = 0x%x\n",
sdp->isp_fifo_threshold);
PRINTF(" Bus Reset Delay = %d\n",
sdp->isp_bus_reset_delay);
PRINTF(" Retry Count = %d\n",
sdp->isp_retry_count);
PRINTF(" Retry Delay = %d\n",
sdp->isp_retry_delay);
PRINTF(" Tag Age Limit = %d\n",
sdp->isp_tag_aging);
PRINTF(" Selection Timeout = %d\n",
sdp->isp_selection_timeout);
PRINTF(" Max Queue Depth = %d\n",
sdp->isp_max_queue_depth);
PRINTF(" Async Data Setup = 0x%x\n",
sdp->isp_async_data_setup);
PRINTF(" REQ/ACK Active Negation = %s\n",
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
sdp->isp_req_ack_active_neg? tru : not);
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
PRINTF(" Data Line Active Negation = %s\n",
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
sdp->isp_data_line_active_neg? tru : not);
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
PRINTF(" Data DMA Burst Enable = %s\n",
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
sdp->isp_data_dma_burst_enabl? tru : not);
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
PRINTF(" Cmd DMA Burst Enable = %s\n",
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
sdp->isp_cmd_dma_burst_enable? tru : not);
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
PRINTF(" Fast MTTR = %s\n",
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
isp->isp_fast_mttr? tru : not);
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
}
for (i = 0; i < MAX_TARGETS; i++) {
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
sdp->isp_devparam[i].dev_enable =
ISP_NVRAM_TGT_DEVICE_ENABLE(nvram_data, i);
sdp->isp_devparam[i].exc_throttle =
ISP_NVRAM_TGT_EXEC_THROTTLE(nvram_data, i);
sdp->isp_devparam[i].sync_offset =
ISP_NVRAM_TGT_SYNC_OFFSET(nvram_data, i);
sdp->isp_devparam[i].sync_period =
ISP_NVRAM_TGT_SYNC_PERIOD(nvram_data, i);
if (isp->isp_type < ISP_HA_SCSI_1040) {
/*
* If we're not ultra, we can't possibly
* be a shorter period than this.
*/
if (sdp->isp_devparam[i].sync_period < 0x19) {
sdp->isp_devparam[i].sync_period =
0x19;
}
if (sdp->isp_devparam[i].sync_offset > 0xc) {
sdp->isp_devparam[i].sync_offset =
0x0c;
}
} else {
if (sdp->isp_devparam[i].sync_offset > 0x8) {
sdp->isp_devparam[i].sync_offset = 0x8;
}
}
sdp->isp_devparam[i].dev_flags = 0;
if (ISP_NVRAM_TGT_RENEG(nvram_data, i))
sdp->isp_devparam[i].dev_flags |= DPARM_RENEG;
if (ISP_NVRAM_TGT_QFRZ(nvram_data, i)) {
PRINTF("%s: not supporting QFRZ option for "
"target %d\n", isp->isp_name, i);
}
sdp->isp_devparam[i].dev_flags |= DPARM_ARQ;
if (ISP_NVRAM_TGT_ARQ(nvram_data, i) == 0) {
PRINTF("%s: not disabling ARQ option for "
"target %d\n", isp->isp_name, i);
}
if (ISP_NVRAM_TGT_TQING(nvram_data, i))
sdp->isp_devparam[i].dev_flags |= DPARM_TQING;
if (ISP_NVRAM_TGT_SYNC(nvram_data, i))
sdp->isp_devparam[i].dev_flags |= DPARM_SYNC;
if (ISP_NVRAM_TGT_WIDE(nvram_data, i))
sdp->isp_devparam[i].dev_flags |= DPARM_WIDE;
if (ISP_NVRAM_TGT_PARITY(nvram_data, i))
sdp->isp_devparam[i].dev_flags |= DPARM_PARITY;
if (ISP_NVRAM_TGT_DISC(nvram_data, i))
sdp->isp_devparam[i].dev_flags |= DPARM_DISC;
add isp1080 support and some basic PDB change stuff
1999-03-27 01:39:44 +03:00
sdp->isp_devparam[i].cur_dflags = 0; /* we don't know */
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
if (isp->isp_dblev > 2) {
PRINTF(" Target %d: Enabled %d Throttle %d "
"Offset %d Period %d Flags 0x%x\n", i,
sdp->isp_devparam[i].dev_enable,
sdp->isp_devparam[i].exc_throttle,
sdp->isp_devparam[i].sync_offset,
sdp->isp_devparam[i].sync_period,
sdp->isp_devparam[i].dev_flags);
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
}
} else {
fcparam *fcp = (fcparam *) isp->isp_param;
union {
struct {
#if BYTE_ORDER == BIG_ENDIAN
u_int32_t hi32;
u_int32_t lo32;
#else
u_int32_t lo32;
u_int32_t hi32;
#endif
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
} wd;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
u_int64_t full64;
} wwnstore;
wwnstore.full64 = ISP2100_NVRAM_NODE_NAME(nvram_data);
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
/*
* Broken PTI cards with nothing in the top nibble. Pah.
*/
if ((wwnstore.wd.hi32 >> 28) == 0) {
wwnstore.wd.hi32 |= (2 << 28);
CFGPRINTF("%s: (corrected) Adapter WWN 0x%08x%08x\n",
isp->isp_name, wwnstore.wd.hi32, wwnstore.wd.lo32);
} else {
CFGPRINTF("%s: Adapter WWN 0x%08x%08x\n", isp->isp_name,
wwnstore.wd.hi32, wwnstore.wd.lo32);
}
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
fcp->isp_nodewwn = wwnstore.full64;
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
Change to approved NASA/Ames copyright. Add Fabric support. Fix SCCLUN support. Add code that tries to track LoopID shifting.
1999-07-06 00:31:35 +04:00
/*
* If the Node WWN has 2 in the top nibble, we can
* authoritatively construct a Port WWN by adding
* our unit number (plus one to make it nonzero) and
* putting it into bits 59..56. If the top nibble isn't
* 2, then we just set them identically.
*/
if ((fcp->isp_nodewwn >> 60) == 2) {
fcp->isp_portwwn = fcp->isp_nodewwn |
(((u_int64_t)(isp->isp_unit+1)) << 56);
} else {
fcp->isp_portwwn = fcp->isp_nodewwn;
}
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
wwnstore.full64 = ISP2100_NVRAM_BOOT_NODE_NAME(nvram_data);
if (wwnstore.full64 != 0) {
Update HBAs to incorporate the new max_lun property.
1998-12-05 22:43:33 +03:00
PRINTF("%s: BOOT DEVICE WWN 0x%08x%08x\n",
remove most target mode stuff. Change the arguments to the isp_fastpost_complete function to include a handle. Do some isr register debouncing. Use new inline functions for xflist handle storage. Remove isp_dumpxflist function. Do some fixups of NVRAM from some broken cards. Use Full Login after LIP option for FC cards if f/w < 1.17 - there's a f/w bug that causes the port database to not be actually refreshed for local loop devices! Do the appropriate endian swizzling for the ICB. Ditto for SNS structures (these are no-ops until UltraSparc PCI needs them).
1999-10-14 06:27:12 +04:00
isp->isp_name, wwnstore.wd.hi32, wwnstore.wd.lo32);
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
}
fcp->isp_maxalloc =
ISP2100_NVRAM_MAXIOCBALLOCATION(nvram_data);
fcp->isp_maxfrmlen =
ISP2100_NVRAM_MAXFRAMELENGTH(nvram_data);
fcp->isp_retry_delay =
ISP2100_NVRAM_RETRY_DELAY(nvram_data);
fcp->isp_retry_count =
ISP2100_NVRAM_RETRY_COUNT(nvram_data);
fcp->isp_loopid =
ISP2100_NVRAM_HARDLOOPID(nvram_data);
fcp->isp_execthrottle =
ISP2100_NVRAM_EXECUTION_THROTTLE(nvram_data);
Roll internal release tag. Print out if we're in a 64 bit PCI slot. Use fast memory timing NVRAM parameter. Clean up and fix establishment of default target parameters. Don't use NVRAM if are flagged as not to do so (I had a busted NVRAM setup which I couldn't edit that enabled SYNC mode but disabled disconnect/reconnect and wide!!). Fix delays after resets. BUS resets not done in isp_init anymore- relegated to OS specific outer layers. Fix a buglet where you can get in a loop for a NULL xs in the completion list in isp_intr. Add in some defines that can disable fast posting. Add in code for Loop Up/Loop Down events that call into the outer layers as to what to do.
1999-02-09 03:52:23 +03:00
fcp->isp_fwoptions = ISP2100_NVRAM_OPTIONS(nvram_data);
if (isp->isp_dblev > 2) {
PRINTF("%s: NVRAM values:\n", isp->isp_name);
PRINTF(" Max IOCB Allocation = %d\n",
fcp->isp_maxalloc);
PRINTF(" Max Frame Length = %d\n",
fcp->isp_maxfrmlen);
PRINTF(" Execution Throttle = %d\n",
fcp->isp_execthrottle);
PRINTF(" Retry Count = %d\n",
fcp->isp_retry_count);
PRINTF(" Retry Delay = %d\n",
fcp->isp_retry_delay);
PRINTF(" Hard Loop ID = %d\n",
fcp->isp_loopid);
PRINTF(" Options = 0x%x\n",
fcp->isp_fwoptions);
PRINTF(" HBA Options = 0x%x\n",
ISP2100_NVRAM_HBA_OPTIONS(nvram_data));
}
}
IDPRINTF(3, ("%s: NVRAM is valid\n", isp->isp_name));
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
return (0);
}
static void
isp_rdnvram_word(isp, wo, rp)
struct ispsoftc *isp;
int wo;
u_int16_t *rp;
{
int i, cbits;
u_int16_t bit, rqst;
ISP_WRITE(isp, BIU_NVRAM, BIU_NVRAM_SELECT);
SYS_DELAY(2);
ISP_WRITE(isp, BIU_NVRAM, BIU_NVRAM_SELECT|BIU_NVRAM_CLOCK);
SYS_DELAY(2);
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
if (IS_FC(isp)) {
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
wo &= ((ISP2100_NVRAM_SIZE >> 1) - 1);
rqst = (ISP_NVRAM_READ << 8) | wo;
Do a fairly large internal restructuring to accomodate dual-bus host adapters (e.g., the 1240). Include the new 1080/1240 NVRAM layout reading code. Some moderately significant mailbox changes were necessary also to accomodate a second channel.
1999-05-12 22:59:23 +04:00
cbits = 10;
} else if (IS_1080(isp) || IS_12X0(isp)) {
wo &= ((ISP1080_NVRAM_SIZE >> 1) - 1);
rqst = (ISP_NVRAM_READ << 8) | wo;
Major reworking of initialization code, addition of NVRAM reading, usage of device speed updating, etc... A lot of FC and other case statement handling. Whew. Too much to really detail here.
1998-09-08 11:28:50 +04:00
cbits = 10;
} else {
wo &= ((ISP_NVRAM_SIZE >> 1) - 1);
rqst = (ISP_NVRAM_READ << 6) | wo;
cbits = 8;
}
/*
* Clock the word select request out...
*/
for (i = cbits; i >= 0; i--) {
if ((rqst >> i) & 1) {
bit = BIU_NVRAM_SELECT | BIU_NVRAM_DATAOUT;
} else {
bit = BIU_NVRAM_SELECT;
}
ISP_WRITE(isp, BIU_NVRAM, bit);
SYS_DELAY(2);
ISP_WRITE(isp, BIU_NVRAM, bit | BIU_NVRAM_CLOCK);
SYS_DELAY(2);
ISP_WRITE(isp, BIU_NVRAM, bit);
SYS_DELAY(2);
}
/*
* Now read the result back in (bits come back in MSB format).
*/
*rp = 0;
for (i = 0; i < 16; i++) {
u_int16_t rv;
*rp <<= 1;
ISP_WRITE(isp, BIU_NVRAM, BIU_NVRAM_SELECT|BIU_NVRAM_CLOCK);
SYS_DELAY(2);
rv = ISP_READ(isp, BIU_NVRAM);
if (rv & BIU_NVRAM_DATAIN) {
*rp |= 1;
}
SYS_DELAY(2);
ISP_WRITE(isp, BIU_NVRAM, BIU_NVRAM_SELECT);
SYS_DELAY(2);
}
ISP_WRITE(isp, BIU_NVRAM, 0);
SYS_DELAY(2);
#if BYTE_ORDER == BIG_ENDIAN
*rp = ((*rp >> 8) | ((*rp & 0xff) << 8));
#endif
}