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

1150 lines
28 KiB
C
Raw Normal View History

netbsd.org -> NetBSD.org This was the last commit of this kind to src/sys, which is now totally "NetBSD.org clean". Thanks for the patiance, and sorry for all the commits.
2003-12-04 16:57:30 +03:00
/* $NetBSD: isp_netbsd.c,v 1.64 2003/12/04 13:57:30 keihan Exp $ */
Add a maintenance note. Change all splbio's to use the ISP_LOCK/ISP_UNLOCK or ISP_ILOCK/ISP_IUNLOCK macros.
2000-08-14 11:08:12 +04:00
/*
* This driver, which is contained in NetBSD in the files:
*
* sys/dev/ic/isp.c
Fix pathnames in comment.
2000-12-23 04:37:57 +03:00
* sys/dev/ic/isp_inline.h
* sys/dev/ic/isp_netbsd.c
* sys/dev/ic/isp_netbsd.h
* sys/dev/ic/isp_target.c
* sys/dev/ic/isp_target.h
* sys/dev/ic/isp_tpublic.h
* sys/dev/ic/ispmbox.h
* sys/dev/ic/ispreg.h
* sys/dev/ic/ispvar.h
Add a maintenance note. Change all splbio's to use the ISP_LOCK/ISP_UNLOCK or ISP_ILOCK/ISP_IUNLOCK macros.
2000-08-14 11:08:12 +04:00
* sys/microcode/isp/asm_sbus.h
* sys/microcode/isp/asm_1040.h
* sys/microcode/isp/asm_1080.h
* sys/microcode/isp/asm_12160.h
* sys/microcode/isp/asm_2100.h
* sys/microcode/isp/asm_2200.h
* sys/pci/isp_pci.c
* sys/sbus/isp_sbus.c
*
netbsd.org -> NetBSD.org This was the last commit of this kind to src/sys, which is now totally "NetBSD.org clean". Thanks for the patiance, and sorry for all the commits.
2003-12-04 16:57:30 +03:00
* Is being actively maintained by Matthew Jacob (mjacob@NetBSD.org).
Add a maintenance note. Change all splbio's to use the ISP_LOCK/ISP_UNLOCK or ISP_ILOCK/ISP_IUNLOCK macros.
2000-08-14 11:08:12 +04:00
* This driver also is shared source with FreeBSD, OpenBSD, Linux, Solaris,
* Linux versions. This tends to be an interesting maintenance problem.
*
* Please coordinate with Matthew Jacob on changes you wish to make here.
*/
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
/*
* Platform (NetBSD) dependent common attachment code for Qlogic adapters.
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
* Matthew Jacob <mjacob@nas.nasa.gov>
*/
/*
* Copyright (C) 1997, 1998, 1999 National Aeronautics & Space Administration
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
* 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.
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
* 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
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +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
* 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.
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
*/
add/cleanup RCSID
2001-11-13 16:14:31 +03:00
#include <sys/cdefs.h>
netbsd.org -> NetBSD.org This was the last commit of this kind to src/sys, which is now totally "NetBSD.org clean". Thanks for the patiance, and sorry for all the commits.
2003-12-04 16:57:30 +03:00
__KERNEL_RCSID(0, "$NetBSD: isp_netbsd.c,v 1.64 2003/12/04 13:57:30 keihan Exp $");
add/cleanup RCSID
2001-11-13 16:14:31 +03:00
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
#include <dev/ic/isp_netbsd.h>
Add in an ioctl entry point so scsictl mediated bus resets will work. Redo how we start commands- do a 'slow' start function which then looks to see when we're done the configuration process at which point it *then* enables sync/wide mode. Set the max openings amount to the true max openings- not a synthetic. Add a timeout driven command requeue function so that Loop Down events well freeze things until a later point in time where they might be restarted.
1999-10-14 06:31:11 +04:00
#include <sys/scsiio.h>
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
Gah. Bad port from FreeBSD code (thanks Robert Elz) for setting up default timeout for watchdogging commands.
2000-07-07 07:14:53 +04:00
/*
* Set a timeout for the watchdogging of a command.
*
* The dimensional analysis is
*
* milliseconds * (seconds/millisecond) * (ticks/second) = ticks
*
* =
*
* (milliseconds / 1000) * hz = ticks
*
*
* For timeouts less than 1 second, we'll get zero. Because of this, and
* because we want to establish *our* timeout to be longer than what the
* firmware might do, we just add 3 seconds at the back end.
*/
#define _XT(xs) ((((xs)->timeout/1000) * hz) + (3 * hz))
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
Defer turning off the no_mbox_ints flag until after the system is ready for interrupts. Handle FW crashes in outer layer.
2001-07-06 20:17:17 +04:00
static void isp_config_interrupts(struct device *);
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
static void ispminphys_1020(struct buf *);
Handles are now 16 bits. ANSIfy.
2001-03-14 08:44:21 +03:00
static void ispminphys(struct buf *);
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
static INLINE void ispcmd(struct ispsoftc *, XS_T *);
static void isprequest(struct scsipi_channel *, scsipi_adapter_req_t, void *);
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
static int
ispioctl(struct scsipi_channel *, u_long, caddr_t, int, struct proc *);
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
static void isp_polled_cmd(struct ispsoftc *, XS_T *);
Handles are now 16 bits. ANSIfy.
2001-03-14 08:44:21 +03:00
static void isp_dog(void *);
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
static void isp_create_fc_worker(void *);
static void isp_fc_worker(void *);
Do SCSI Bus resets in this layer (now). Don't do it for Fibre Channel yet (we get LOOP DOWN events, and we'll hang on that at this time). Add other isp_async cases- ISPASYNC_LOOP_DOWN and ISPASYNC_LOOP_UP. DOWN will cause internal queuing until UP, whereupon a timeout will fire up any pending xfers. It doesn't really keep commands from getting destroyed by loop down events, but at least minimizes the damage. This was much easier to implement with CAM.
1999-02-09 03:42:22 +03:00
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
/*
* Complete attachment of hardware, include subdevices.
*/
void
Handles are now 16 bits. ANSIfy.
2001-03-14 08:44:21 +03:00
isp_attach(struct ispsoftc *isp)
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
{
isp->isp_state = ISP_RUNSTATE;
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
isp->isp_osinfo._adapter.adapt_dev = &isp->isp_osinfo._dev;
isp->isp_osinfo._adapter.adapt_nchannels = IS_DUALBUS(isp) ? 2 : 1;
isp->isp_osinfo._adapter.adapt_openings = isp->isp_maxcmds;
/*
* It's not stated whether max_periph is limited by SPI
* tag uage, but let's assume that it is.
*/
isp->isp_osinfo._adapter.adapt_max_periph = min(isp->isp_maxcmds, 255);
isp->isp_osinfo._adapter.adapt_ioctl = ispioctl;
isp->isp_osinfo._adapter.adapt_request = isprequest;
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
if (isp->isp_type <= ISP_HA_SCSI_1020A) {
isp->isp_osinfo._adapter.adapt_minphys = ispminphys_1020;
} else {
isp->isp_osinfo._adapter.adapt_minphys = ispminphys;
}
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
isp->isp_osinfo._chan.chan_adapter = &isp->isp_osinfo._adapter;
isp->isp_osinfo._chan.chan_bustype = &scsi_bustype;
isp->isp_osinfo._chan.chan_channel = 0;
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
/*
* Until the midlayer is fixed to use REPORT LUNS, limit to 8 luns.
*/
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
isp->isp_osinfo._chan.chan_nluns = min(isp->isp_maxluns, 8);
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_FC(isp)) {
Remove NOSETTLE flag from channel in FibreChannel case. It makes sense that it should be there, but in practice, on some systems an ugly race condition rears its head: SCSI commands are issued before the FC thread ever runs, and fail from then on forever. Yuck.
2003-09-27 07:43:08 +04:00
#if 0 /* XXX channel "settle" time seems to sidestep some nasty race */
Fibre Channel doesn't need bus reset settle time.
2003-03-21 21:05:16 +03:00
isp->isp_osinfo._chan.chan_flags = SCSIPI_CHAN_NOSETTLE;
Remove NOSETTLE flag from channel in FibreChannel case. It makes sense that it should be there, but in practice, on some systems an ugly race condition rears its head: SCSI commands are issued before the FC thread ever runs, and fail from then on forever. Yuck.
2003-09-27 07:43:08 +04:00
#endif
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
isp->isp_osinfo._chan.chan_ntargets = MAX_FC_TARG;
isp->isp_osinfo._chan.chan_id = MAX_FC_TARG;
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
isp->isp_osinfo.threadwork = 1;
/*
* Note that isp_create_fc_worker won't get called
* until much much later (after proc0 is created).
*/
kthread_create(isp_create_fc_worker, isp);
Implement F/W crashdumps for 2200 && 2300 cards (kernel compile option). Distinguish between 2312 and 2300 cards (they *are* different). Enable RIO (Reduced Interrupt Operation) for the LVD cards (hey- I've seen batched completions of the 30 commands at a time with this,....)... If we get a Port Logout on local loop topologies, we have to force the f/w to log back in. The easiest way (for us) to do this is to force a LIP. This also will wake up the disk that probably just had a f/w crash. Implement mailbox 'continuations'- this allows interrupts to re-drive a mailbox command if it's one that just essentially repeats the previous mailbox command (e.g., f/w download). This saves a boatload of sleep/wakeup twitches. If we're not a 2300 and we're about to return with a 'bogus interrupt'- check the semaphore register to be non-zero at all and outgoing mailbox 0- this seems to be where some of the lost ISP1080 commands came from.
2002-02-22 01:32:40 +03:00
#ifdef ISP_FW_CRASH_DUMP
if (IS_2200(isp)) {
FCPARAM(isp)->isp_dump_data =
malloc(QLA2200_RISC_IMAGE_DUMP_SIZE, M_DEVBUF,
M_NOWAIT);
} else if (IS_23XX(isp)) {
FCPARAM(isp)->isp_dump_data =
malloc(QLA2300_RISC_IMAGE_DUMP_SIZE, M_DEVBUF,
M_NOWAIT);
}
if (FCPARAM(isp)->isp_dump_data)
FCPARAM(isp)->isp_dump_data[0] = 0;
#endif
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
} else {
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
int bus = 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
sdparam *sdp = isp->isp_param;
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
isp->isp_osinfo._chan.chan_ntargets = MAX_TARGETS;
isp->isp_osinfo._chan.chan_id = sdp->isp_initiator_id;
Add in an ioctl entry point so scsictl mediated bus resets will work. Redo how we start commands- do a 'slow' start function which then looks to see when we're done the configuration process at which point it *then* enables sync/wide mode. Set the max openings amount to the true max openings- not a synthetic. Add a timeout driven command requeue function so that Loop Down events well freeze things until a later point in time where they might be restarted.
1999-10-14 06:31:11 +04:00
isp->isp_osinfo.discovered[0] = 1 << sdp->isp_initiator_id;
Dual LVD (1280) support.
1999-12-16 08:35:42 +03:00
if (IS_DUALBUS(isp)) {
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
isp->isp_osinfo._chan_b = isp->isp_osinfo._chan;
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++;
some comments added
1999-10-17 05:23:21 +04:00
isp->isp_osinfo.discovered[1] =
1 << sdp->isp_initiator_id;
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
isp->isp_osinfo._chan_b.chan_id = sdp->isp_initiator_id;
isp->isp_osinfo._chan_b.chan_channel = 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
}
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
ISP_LOCK(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
(void) isp_control(isp, ISPCTL_RESET_BUS, &bus);
Dual LVD (1280) support.
1999-12-16 08:35:42 +03:00
if (IS_DUALBUS(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
bus++;
(void) isp_control(isp, ISPCTL_RESET_BUS, &bus);
}
Remove ispcmd_slow routine- SCBUSACCEL routine will set/clear tag/wide/sync. Make some changes about where some things sit in the softc.
2000-08-09 02:58:30 +04:00
ISP_UNLOCK(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
}
Do SCSI Bus resets in this layer (now). Don't do it for Fibre Channel yet (we get LOOP DOWN events, and we'll hang on that at this time). Add other isp_async cases- ISPASYNC_LOOP_DOWN and ISPASYNC_LOOP_UP. DOWN will cause internal queuing until UP, whereupon a timeout will fire up any pending xfers. It doesn't really keep commands from getting destroyed by loop down events, but at least minimizes the damage. This was much easier to implement with CAM.
1999-02-09 03:42:22 +03:00
Defer turning off the no_mbox_ints flag until after the system is ready for interrupts. Handle FW crashes in outer layer.
2001-07-06 20:17:17 +04:00
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
/*
Defer turning off the no_mbox_ints flag until after the system is ready for interrupts. Handle FW crashes in outer layer.
2001-07-06 20:17:17 +04:00
* Defer enabling mailbox interrupts until later.
*/
config_interrupts((struct device *) isp, isp_config_interrupts);
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
/*
* And attach children (if any).
*/
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
config_found((void *)isp, &isp->isp_chanA, scsiprint);
Dual LVD (1280) support.
1999-12-16 08:35:42 +03:00
if (IS_DUALBUS(isp)) {
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
config_found((void *)isp, &isp->isp_chanB, scsiprint);
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
}
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
}
Defer turning off the no_mbox_ints flag until after the system is ready for interrupts. Handle FW crashes in outer layer.
2001-07-06 20:17:17 +04:00
static void
isp_config_interrupts(struct device *self)
{
struct ispsoftc *isp = (struct ispsoftc *) self;
/*
* After this point, we'll be doing the new configuration
Do not enable interrupt driven mailbox commands for non-FC scsi.
2002-09-02 02:30:09 +04:00
* schema which allows interrupts, so we can do tsleep/wakeup
* for mailbox stuff at that point, if that's allowed.
Defer turning off the no_mbox_ints flag until after the system is ready for interrupts. Handle FW crashes in outer layer.
2001-07-06 20:17:17 +04:00
*/
Do not enable interrupt driven mailbox commands for non-FC scsi.
2002-09-02 02:30:09 +04:00
if (IS_FC(isp)) {
isp->isp_osinfo.no_mbox_ints = 0;
}
Defer turning off the no_mbox_ints flag until after the system is ready for interrupts. Handle FW crashes in outer layer.
2001-07-06 20:17:17 +04:00
}
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
/*
* minphys our xfers
*/
static void
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
ispminphys_1020(struct buf *bp)
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
{
if (bp->b_bcount >= (1 << 24)) {
bp->b_bcount = (1 << 24);
}
minphys(bp);
}
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
static void
ispminphys(struct buf *bp)
{
if (bp->b_bcount >= (1 << 30)) {
bp->b_bcount = (1 << 30);
}
minphys(bp);
}
Add in an ioctl entry point so scsictl mediated bus resets will work. Redo how we start commands- do a 'slow' start function which then looks to see when we're done the configuration process at which point it *then* enables sync/wide mode. Set the max openings amount to the true max openings- not a synthetic. Add a timeout driven command requeue function so that Loop Down events well freeze things until a later point in time where they might be restarted.
1999-10-14 06:31:11 +04:00
static int
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
ispioctl(struct scsipi_channel *chan, u_long cmd, caddr_t addr, int flag,
struct proc *p)
Add in an ioctl entry point so scsictl mediated bus resets will work. Redo how we start commands- do a 'slow' start function which then looks to see when we're done the configuration process at which point it *then* enables sync/wide mode. Set the max openings amount to the true max openings- not a synthetic. Add a timeout driven command requeue function so that Loop Down events well freeze things until a later point in time where they might be restarted.
1999-10-14 06:31:11 +04:00
{
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
struct ispsoftc *isp = (void *)chan->chan_adapter->adapt_dev;
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
int retval = ENOTTY;
Add in an ioctl entry point so scsictl mediated bus resets will work. Redo how we start commands- do a 'slow' start function which then looks to see when we're done the configuration process at which point it *then* enables sync/wide mode. Set the max openings amount to the true max openings- not a synthetic. Add a timeout driven command requeue function so that Loop Down events well freeze things until a later point in time where they might be restarted.
1999-10-14 06:31:11 +04:00
switch (cmd) {
Implement F/W crashdumps for 2200 && 2300 cards (kernel compile option). Distinguish between 2312 and 2300 cards (they *are* different). Enable RIO (Reduced Interrupt Operation) for the LVD cards (hey- I've seen batched completions of the 30 commands at a time with this,....)... If we get a Port Logout on local loop topologies, we have to force the f/w to log back in. The easiest way (for us) to do this is to force a LIP. This also will wake up the disk that probably just had a f/w crash. Implement mailbox 'continuations'- this allows interrupts to re-drive a mailbox command if it's one that just essentially repeats the previous mailbox command (e.g., f/w download). This saves a boatload of sleep/wakeup twitches. If we're not a 2300 and we're about to return with a 'bogus interrupt'- check the semaphore register to be non-zero at all and outgoing mailbox 0- this seems to be where some of the lost ISP1080 commands came from.
2002-02-22 01:32:40 +03:00
#ifdef ISP_FW_CRASH_DUMP
case ISP_GET_FW_CRASH_DUMP:
{
u_int16_t *ptr = FCPARAM(isp)->isp_dump_data;
size_t sz;
retval = 0;
if (IS_2200(isp))
sz = QLA2200_RISC_IMAGE_DUMP_SIZE;
Add in an ioctl entry point so scsictl mediated bus resets will work. Redo how we start commands- do a 'slow' start function which then looks to see when we're done the configuration process at which point it *then* enables sync/wide mode. Set the max openings amount to the true max openings- not a synthetic. Add a timeout driven command requeue function so that Loop Down events well freeze things until a later point in time where they might be restarted.
1999-10-14 06:31:11 +04:00
else
Implement F/W crashdumps for 2200 && 2300 cards (kernel compile option). Distinguish between 2312 and 2300 cards (they *are* different). Enable RIO (Reduced Interrupt Operation) for the LVD cards (hey- I've seen batched completions of the 30 commands at a time with this,....)... If we get a Port Logout on local loop topologies, we have to force the f/w to log back in. The easiest way (for us) to do this is to force a LIP. This also will wake up the disk that probably just had a f/w crash. Implement mailbox 'continuations'- this allows interrupts to re-drive a mailbox command if it's one that just essentially repeats the previous mailbox command (e.g., f/w download). This saves a boatload of sleep/wakeup twitches. If we're not a 2300 and we're about to return with a 'bogus interrupt'- check the semaphore register to be non-zero at all and outgoing mailbox 0- this seems to be where some of the lost ISP1080 commands came from.
2002-02-22 01:32:40 +03:00
sz = QLA2300_RISC_IMAGE_DUMP_SIZE;
ISP_LOCK(isp);
if (ptr && *ptr) {
void *uaddr = *((void **) addr);
if (copyout(ptr, uaddr, sz)) {
retval = EFAULT;
} else {
*ptr = 0;
}
} else {
retval = ENXIO;
}
ISP_UNLOCK(isp);
break;
}
case ISP_FORCE_CRASH_DUMP:
ISP_LOCK(isp);
if (isp->isp_osinfo.blocked == 0) {
isp->isp_osinfo.blocked = 1;
scsipi_channel_freeze(&isp->isp_chanA, 1);
}
isp_fw_dump(isp);
isp_reinit(isp);
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
ISP_UNLOCK(isp);
Implement F/W crashdumps for 2200 && 2300 cards (kernel compile option). Distinguish between 2312 and 2300 cards (they *are* different). Enable RIO (Reduced Interrupt Operation) for the LVD cards (hey- I've seen batched completions of the 30 commands at a time with this,....)... If we get a Port Logout on local loop topologies, we have to force the f/w to log back in. The easiest way (for us) to do this is to force a LIP. This also will wake up the disk that probably just had a f/w crash. Implement mailbox 'continuations'- this allows interrupts to re-drive a mailbox command if it's one that just essentially repeats the previous mailbox command (e.g., f/w download). This saves a boatload of sleep/wakeup twitches. If we're not a 2300 and we're about to return with a 'bogus interrupt'- check the semaphore register to be non-zero at all and outgoing mailbox 0- this seems to be where some of the lost ISP1080 commands came from.
2002-02-22 01:32:40 +03:00
retval = 0;
break;
#endif
Add initial implementation of ISP_SDBLEV, ISP_RESETHBA, ISP_FC_RESCAN, ISP_FC_LIP and ISP_FC_GETDINFO ioctls.
2001-04-11 01:52:00 +04:00
case ISP_SDBLEV:
{
int olddblev = isp->isp_dblev;
isp->isp_dblev = *(int *)addr;
*(int *)addr = olddblev;
retval = 0;
break;
}
case ISP_RESETHBA:
ISP_LOCK(isp);
isp_reinit(isp);
ISP_UNLOCK(isp);
retval = 0;
break;
Implement F/W crashdumps for 2200 && 2300 cards (kernel compile option). Distinguish between 2312 and 2300 cards (they *are* different). Enable RIO (Reduced Interrupt Operation) for the LVD cards (hey- I've seen batched completions of the 30 commands at a time with this,....)... If we get a Port Logout on local loop topologies, we have to force the f/w to log back in. The easiest way (for us) to do this is to force a LIP. This also will wake up the disk that probably just had a f/w crash. Implement mailbox 'continuations'- this allows interrupts to re-drive a mailbox command if it's one that just essentially repeats the previous mailbox command (e.g., f/w download). This saves a boatload of sleep/wakeup twitches. If we're not a 2300 and we're about to return with a 'bogus interrupt'- check the semaphore register to be non-zero at all and outgoing mailbox 0- this seems to be where some of the lost ISP1080 commands came from.
2002-02-22 01:32:40 +03:00
case ISP_RESCAN:
Add initial implementation of ISP_SDBLEV, ISP_RESETHBA, ISP_FC_RESCAN, ISP_FC_LIP and ISP_FC_GETDINFO ioctls.
2001-04-11 01:52:00 +04:00
if (IS_FC(isp)) {
ISP_LOCK(isp);
if (isp_fc_runstate(isp, 5 * 1000000)) {
retval = EIO;
} else {
retval = 0;
}
ISP_UNLOCK(isp);
}
break;
case ISP_FC_LIP:
if (IS_FC(isp)) {
ISP_LOCK(isp);
if (isp_control(isp, ISPCTL_SEND_LIP, 0)) {
retval = EIO;
} else {
retval = 0;
}
ISP_UNLOCK(isp);
}
break;
case ISP_FC_GETDINFO:
{
struct isp_fc_device *ifc = (struct isp_fc_device *) addr;
struct lportdb *lp;
if (ifc->loopid < 0 || ifc->loopid >= MAX_FC_TARG) {
retval = EINVAL;
break;
}
ISP_LOCK(isp);
lp = &FCPARAM(isp)->portdb[ifc->loopid];
if (lp->valid) {
ifc->loopid = lp->loopid;
ifc->portid = lp->portid;
ifc->node_wwn = lp->node_wwn;
ifc->port_wwn = lp->port_wwn;
retval = 0;
} else {
retval = ENODEV;
}
ISP_UNLOCK(isp);
break;
}
Add support for ISP_FC_GETHINFO which can tell you connection topology, current speed, loopid, etc.
2002-06-15 04:20:18 +04:00
case ISP_GET_STATS:
{
isp_stats_t *sp = (isp_stats_t *) addr;
MEMZERO(sp, sizeof (*sp));
sp->isp_stat_version = ISP_STATS_VERSION;
sp->isp_type = isp->isp_type;
sp->isp_revision = isp->isp_revision;
ISP_LOCK(isp);
sp->isp_stats[ISP_INTCNT] = isp->isp_intcnt;
sp->isp_stats[ISP_INTBOGUS] = isp->isp_intbogus;
sp->isp_stats[ISP_INTMBOXC] = isp->isp_intmboxc;
sp->isp_stats[ISP_INGOASYNC] = isp->isp_intoasync;
sp->isp_stats[ISP_RSLTCCMPLT] = isp->isp_rsltccmplt;
sp->isp_stats[ISP_FPHCCMCPLT] = isp->isp_fphccmplt;
sp->isp_stats[ISP_RSCCHIWAT] = isp->isp_rscchiwater;
sp->isp_stats[ISP_FPCCHIWAT] = isp->isp_fpcchiwater;
ISP_UNLOCK(isp);
retval = 0;
break;
}
case ISP_CLR_STATS:
ISP_LOCK(isp);
isp->isp_intcnt = 0;
isp->isp_intbogus = 0;
isp->isp_intmboxc = 0;
isp->isp_intoasync = 0;
isp->isp_rsltccmplt = 0;
isp->isp_fphccmplt = 0;
isp->isp_rscchiwater = 0;
isp->isp_fpcchiwater = 0;
ISP_UNLOCK(isp);
retval = 0;
break;
case ISP_FC_GETHINFO:
{
struct isp_hba_device *hba = (struct isp_hba_device *) addr;
MEMZERO(hba, sizeof (*hba));
ISP_LOCK(isp);
hba->fc_speed = FCPARAM(isp)->isp_gbspeed;
hba->fc_scsi_supported = 1;
hba->fc_topology = FCPARAM(isp)->isp_topo + 1;
hba->fc_loopid = FCPARAM(isp)->isp_loopid;
Note both active and nvram WWNN/WWPNs Fix the bug where the thread worker constantly looped. It was a race as to whether or not the thread fired up or a polled/probed command happened first, or maybe it's due to other scsipi changes, but we never set the value that caused the fc thread worker to actually go to sleep until an initial loop up. This was rather ugly as it consumed nearly all CPU time available from thence on. Bad.
2003-08-07 05:10:53 +04:00
hba->nvram_node_wwn = FCPARAM(isp)->isp_nodewwn;
hba->nvram_port_wwn = FCPARAM(isp)->isp_portwwn;
hba->active_node_wwn = ISP_NODEWWN(isp);
hba->active_port_wwn = ISP_PORTWWN(isp);
Add support for ISP_FC_GETHINFO which can tell you connection topology, current speed, loopid, etc.
2002-06-15 04:20:18 +04:00
ISP_UNLOCK(isp);
break;
}
Implement F/W crashdumps for 2200 && 2300 cards (kernel compile option). Distinguish between 2312 and 2300 cards (they *are* different). Enable RIO (Reduced Interrupt Operation) for the LVD cards (hey- I've seen batched completions of the 30 commands at a time with this,....)... If we get a Port Logout on local loop topologies, we have to force the f/w to log back in. The easiest way (for us) to do this is to force a LIP. This also will wake up the disk that probably just had a f/w crash. Implement mailbox 'continuations'- this allows interrupts to re-drive a mailbox command if it's one that just essentially repeats the previous mailbox command (e.g., f/w download). This saves a boatload of sleep/wakeup twitches. If we're not a 2300 and we're about to return with a 'bogus interrupt'- check the semaphore register to be non-zero at all and outgoing mailbox 0- this seems to be where some of the lost ISP1080 commands came from.
2002-02-22 01:32:40 +03:00
case SCBUSIORESET:
ISP_LOCK(isp);
if (isp_control(isp, ISPCTL_RESET_BUS, &chan->chan_channel))
retval = EIO;
else
retval = 0;
ISP_UNLOCK(isp);
break;
Add in an ioctl entry point so scsictl mediated bus resets will work. Redo how we start commands- do a 'slow' start function which then looks to see when we're done the configuration process at which point it *then* enables sync/wide mode. Set the max openings amount to the true max openings- not a synthetic. Add a timeout driven command requeue function so that Loop Down events well freeze things until a later point in time where they might be restarted.
1999-10-14 06:31:11 +04:00
default:
break;
}
return (retval);
}
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
static INLINE void
ispcmd(struct ispsoftc *isp, XS_T *xs)
Add in an ioctl entry point so scsictl mediated bus resets will work. Redo how we start commands- do a 'slow' start function which then looks to see when we're done the configuration process at which point it *then* enables sync/wide mode. Set the max openings amount to the true max openings- not a synthetic. Add a timeout driven command requeue function so that Loop Down events well freeze things until a later point in time where they might be restarted.
1999-10-14 06:31:11 +04:00
{
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
ISP_LOCK(isp);
if (isp->isp_state < ISP_RUNSTATE) {
DISABLE_INTS(isp);
isp_init(isp);
if (isp->isp_state != ISP_INITSTATE) {
ENABLE_INTS(isp);
ISP_UNLOCK(isp);
Various changes to allow for overriding just loopid and/or iid. Fix a random 23XX ISR reading bug. Be chattier about HBA_BOTCH errors. More f/w cashdump layering.
2002-08-13 01:33:39 +04:00
isp_prt(isp, ISP_LOGERR, "isp not at init state");
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
XS_SETERR(xs, HBA_BOTCH);
scsipi_done(xs);
return;
}
isp->isp_state = ISP_RUNSTATE;
ENABLE_INTS(isp);
}
/*
* Handle the case of a FC card where the FC thread hasn't
* fired up yet and we have loop state to clean up. If we
* can't clear things up and we've never seen loop up, bounce
* the command.
*/
if (IS_FC(isp) && isp->isp_osinfo.threadwork &&
isp->isp_osinfo.thread == 0) {
volatile u_int8_t ombi = isp->isp_osinfo.no_mbox_ints;
int delay_time;
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
if (xs->xs_control & XS_CTL_POLL) {
isp->isp_osinfo.no_mbox_ints = 1;
}
if (isp->isp_osinfo.loop_checked == 0) {
delay_time = 10 * 1000000;
isp->isp_osinfo.loop_checked = 1;
} else {
delay_time = 250000;
}
if (isp_fc_runstate(isp, delay_time) != 0) {
if (xs->xs_control & XS_CTL_POLL) {
isp->isp_osinfo.no_mbox_ints = ombi;
}
if (FCPARAM(isp)->loop_seen_once == 0) {
XS_SETERR(xs, HBA_SELTIMEOUT);
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
scsipi_done(xs);
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
ISP_UNLOCK(isp);
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
return;
}
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
/*
* Otherwise, fall thru to be queued up for later.
*/
} else {
int wasblocked =
(isp->isp_osinfo.blocked || isp->isp_osinfo.paused);
isp->isp_osinfo.threadwork = 0;
isp->isp_osinfo.blocked =
isp->isp_osinfo.paused = 0;
If I've told myself once, I've told myself 1000 times- *NEVER* commit w/o test compiling first. Argh. I nuked one extra line that I shouldn't have.
2001-07-07 05:44:21 +04:00
if (wasblocked) {
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
scsipi_channel_thaw(&isp->isp_chanA, 1);
}
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
}
Update for SCSIPI changes. Note that capabilities update is disabled until Matt Jacob has a change to update the driver for the latest firmware, etc. since update device parameters seems to fail once the system is really up and running, and eventually causes the controller to wedge. This may be due to a firmware bug. Per discussion with Matt.
1999-10-01 03:06:18 +04:00
if (xs->xs_control & XS_CTL_POLL) {
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
isp->isp_osinfo.no_mbox_ints = ombi;
Do SCSI Bus resets in this layer (now). Don't do it for Fibre Channel yet (we get LOOP DOWN events, and we'll hang on that at this time). Add other isp_async cases- ISPASYNC_LOOP_DOWN and ISPASYNC_LOOP_UP. DOWN will cause internal queuing until UP, whereupon a timeout will fire up any pending xfers. It doesn't really keep commands from getting destroyed by loop down events, but at least minimizes the damage. This was much easier to implement with CAM.
1999-02-09 03:42:22 +03:00
}
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
}
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
if (isp->isp_osinfo.paused) {
isp_prt(isp, ISP_LOGWARN, "I/O while paused");
xs->error = XS_RESOURCE_SHORTAGE;
scsipi_done(xs);
ISP_UNLOCK(isp);
return;
}
if (isp->isp_osinfo.blocked) {
isp_prt(isp, ISP_LOGWARN, "I/O while blocked");
xs->error = XS_REQUEUE;
scsipi_done(xs);
ISP_UNLOCK(isp);
return;
}
if (xs->xs_control & XS_CTL_POLL) {
volatile u_int8_t ombi = isp->isp_osinfo.no_mbox_ints;
isp->isp_osinfo.no_mbox_ints = 1;
isp_polled_cmd(isp, xs);
isp->isp_osinfo.no_mbox_ints = ombi;
ISP_UNLOCK(isp);
return;
}
switch (isp_start(xs)) {
case CMD_QUEUED:
if (xs->timeout) {
callout_reset(&xs->xs_callout, _XT(xs), isp_dog, xs);
}
break;
case CMD_EAGAIN:
isp->isp_osinfo.paused = 1;
xs->error = XS_RESOURCE_SHORTAGE;
scsipi_channel_freeze(&isp->isp_chanA, 1);
if (IS_DUALBUS(isp)) {
scsipi_channel_freeze(&isp->isp_chanB, 1);
}
scsipi_done(xs);
break;
case CMD_RQLATER:
/*
* We can only get RQLATER from FC devices (1 channel only)
*
* Also, if we've never seen loop up, bounce the command
* (somebody has booted with no FC cable connected)
*/
if (FCPARAM(isp)->loop_seen_once == 0) {
XS_SETERR(xs, HBA_SELTIMEOUT);
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
scsipi_done(xs);
break;
}
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
if (isp->isp_osinfo.blocked == 0) {
isp->isp_osinfo.blocked = 1;
scsipi_channel_freeze(&isp->isp_chanA, 1);
}
xs->error = XS_REQUEUE;
scsipi_done(xs);
break;
case CMD_COMPLETE:
scsipi_done(xs);
break;
}
ISP_UNLOCK(isp);
}
static void
isprequest(struct scsipi_channel *chan, scsipi_adapter_req_t req, void *arg)
{
struct ispsoftc *isp = (void *)chan->chan_adapter->adapt_dev;
switch (req) {
case ADAPTER_REQ_RUN_XFER:
ispcmd(isp, (XS_T *) arg);
break;
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
case ADAPTER_REQ_GROW_RESOURCES:
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
/* Not supported. */
break;
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
case ADAPTER_REQ_SET_XFER_MODE:
if (IS_SCSI(isp)) {
struct scsipi_xfer_mode *xm = arg;
int dflags = 0;
sdparam *sdp = SDPARAM(isp);
sdp += chan->chan_channel;
if (xm->xm_mode & PERIPH_CAP_TQING)
dflags |= DPARM_TQING;
if (xm->xm_mode & PERIPH_CAP_WIDE16)
dflags |= DPARM_WIDE;
if (xm->xm_mode & PERIPH_CAP_SYNC)
dflags |= DPARM_SYNC;
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
ISP_LOCK(isp);
Add support for 2 Gigabit cards (2300/2312). This necessitated a change in how interrupts are down- the 23XX has not only a different place to check for an interrupt, but unlike all other QLogic cards, you have to read the status as a 32 bit word- not 16 bit words. Rather than have device specific functions as called from the core module (in isp_intr), it makes more sense to have the platform/bus modules do the gruntwork of splitting out the isr, semaphore register and the first outgoing mailbox register (if needed) *prior* to calling isp_intr (if calling isp_intr is necessary at all).
2001-09-01 11:12:23 +04:00
sdp->isp_devparam[xm->xm_target].goal_flags |= dflags;
dflags = sdp->isp_devparam[xm->xm_target].goal_flags;
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
sdp->isp_devparam[xm->xm_target].dev_update = 1;
isp->isp_update |= (1 << chan->chan_channel);
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
ISP_UNLOCK(isp);
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
isp_prt(isp, ISP_LOGDEBUG1,
"ispioctl: device flags 0x%x for %d.%d.X",
dflags, chan->chan_channel, xm->xm_target);
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
break;
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
}
default:
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
break;
}
}
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
static void
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
isp_polled_cmd(struct ispsoftc *isp, XS_T *xs)
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
{
int result;
int infinite = 0, mswait;
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
result = isp_start(xs);
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
Add in an ioctl entry point so scsictl mediated bus resets will work. Redo how we start commands- do a 'slow' start function which then looks to see when we're done the configuration process at which point it *then* enables sync/wide mode. Set the max openings amount to the true max openings- not a synthetic. Add a timeout driven command requeue function so that Loop Down events well freeze things until a later point in time where they might be restarted.
1999-10-14 06:31:11 +04:00
switch (result) {
case CMD_QUEUED:
break;
case CMD_RQLATER:
if (XS_NOERR(xs)) {
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
xs->error = XS_REQUEUE;
Add in an ioctl entry point so scsictl mediated bus resets will work. Redo how we start commands- do a 'slow' start function which then looks to see when we're done the configuration process at which point it *then* enables sync/wide mode. Set the max openings amount to the true max openings- not a synthetic. Add a timeout driven command requeue function so that Loop Down events well freeze things until a later point in time where they might be restarted.
1999-10-14 06:31:11 +04:00
}
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
case CMD_EAGAIN:
if (XS_NOERR(xs)) {
xs->error = XS_RESOURCE_SHORTAGE;
}
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
/* FALLTHROUGH */
Add in an ioctl entry point so scsictl mediated bus resets will work. Redo how we start commands- do a 'slow' start function which then looks to see when we're done the configuration process at which point it *then* enables sync/wide mode. Set the max openings amount to the true max openings- not a synthetic. Add a timeout driven command requeue function so that Loop Down events well freeze things until a later point in time where they might be restarted.
1999-10-14 06:31:11 +04:00
case CMD_COMPLETE:
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
scsipi_done(xs);
return;
Add in an ioctl entry point so scsictl mediated bus resets will work. Redo how we start commands- do a 'slow' start function which then looks to see when we're done the configuration process at which point it *then* enables sync/wide mode. Set the max openings amount to the true max openings- not a synthetic. Add a timeout driven command requeue function so that Loop Down events well freeze things until a later point in time where they might be restarted.
1999-10-14 06:31:11 +04:00
}
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
/*
* If we can't use interrupts, poll on completion.
*/
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
if ((mswait = XS_TIME(xs)) == 0)
infinite = 1;
while (mswait || infinite) {
Add support for 2 Gigabit cards (2300/2312). This necessitated a change in how interrupts are down- the 23XX has not only a different place to check for an interrupt, but unlike all other QLogic cards, you have to read the status as a 32 bit word- not 16 bit words. Rather than have device specific functions as called from the core module (in isp_intr), it makes more sense to have the platform/bus modules do the gruntwork of splitting out the isr, semaphore register and the first outgoing mailbox register (if needed) *prior* to calling isp_intr (if calling isp_intr is necessary at all).
2001-09-01 11:12:23 +04:00
u_int16_t isr, sema, mbox;
if (ISP_READ_ISR(isp, &isr, &sema, &mbox)) {
isp_intr(isp, isr, sema, mbox);
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
if (XS_CMD_DONE_P(xs)) {
break;
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
}
}
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
USEC_DELAY(1000);
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
mswait -= 1;
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
}
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
/*
* If no other error occurred but we didn't finish,
* something bad happened.
*/
if (XS_CMD_DONE_P(xs) == 0) {
if (isp_control(isp, ISPCTL_ABORT_CMD, xs)) {
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
isp_reinit(isp);
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
}
if (XS_NOERR(xs)) {
Various changes to allow for overriding just loopid and/or iid. Fix a random 23XX ISR reading bug. Be chattier about HBA_BOTCH errors. More f/w cashdump layering.
2002-08-13 01:33:39 +04:00
isp_prt(isp, ISP_LOGERR, "polled command timed out");
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
XS_SETERR(xs, HBA_BOTCH);
}
}
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
scsipi_done(xs);
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
}
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
void
Handles are now 16 bits. ANSIfy.
2001-03-14 08:44:21 +03:00
isp_done(XS_T *xs)
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
{
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
XS_CMD_S_DONE(xs);
if (XS_CMD_WDOG_P(xs) == 0) {
struct ispsoftc *isp = XS_ISP(xs);
callout_stop(&xs->xs_callout);
if (XS_CMD_GRACE_P(xs)) {
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
isp_prt(isp, ISP_LOGDEBUG1,
"finished command on borrowed time");
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
}
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
XS_CMD_S_CLEAR(xs);
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
/*
* Fixup- if we get a QFULL, we need
* to set XS_BUSY as the error.
*/
if (xs->status == SCSI_QUEUE_FULL) {
xs->error = XS_BUSY;
}
if (isp->isp_osinfo.paused) {
isp->isp_osinfo.paused = 0;
scsipi_channel_timed_thaw(&isp->isp_chanA);
if (IS_DUALBUS(isp)) {
scsipi_channel_timed_thaw(&isp->isp_chanB);
}
}
Various changes to allow for overriding just loopid and/or iid. Fix a random 23XX ISR reading bug. Be chattier about HBA_BOTCH errors. More f/w cashdump layering.
2002-08-13 01:33:39 +04:00
if (xs->error == XS_DRIVER_STUFFUP) {
isp_prt(isp, ISP_LOGERR, "BOTCHED cmd for %d.%d.%d cmd 0x%x datalen %ld",
XS_CHANNEL(xs), XS_TGT(xs), XS_LUN(xs), XS_CDBP(xs)[0], (long) XS_XFRLEN(xs));
}
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
scsipi_done(xs);
NetBSD OS specific routines and definitions.
1998-07-15 23:44:04 +04:00
}
}
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
static void
Handles are now 16 bits. ANSIfy.
2001-03-14 08:44:21 +03:00
isp_dog(void *arg)
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
{
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
XS_T *xs = arg;
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
struct ispsoftc *isp = XS_ISP(xs);
Handles are now 16 bits. ANSIfy.
2001-03-14 08:44:21 +03:00
u_int16_t handle;
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
Add a maintenance note. Change all splbio's to use the ISP_LOCK/ISP_UNLOCK or ISP_ILOCK/ISP_IUNLOCK macros.
2000-08-14 11:08:12 +04:00
ISP_ILOCK(isp);
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
/*
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
* We've decided this command is dead. Make sure we're not trying
* to kill a command that's already dead by getting it's handle and
* and seeing whether it's still alive.
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
*/
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
handle = isp_find_handle(isp, xs);
if (handle) {
Add support for 2 Gigabit cards (2300/2312). This necessitated a change in how interrupts are down- the 23XX has not only a different place to check for an interrupt, but unlike all other QLogic cards, you have to read the status as a 32 bit word- not 16 bit words. Rather than have device specific functions as called from the core module (in isp_intr), it makes more sense to have the platform/bus modules do the gruntwork of splitting out the isr, semaphore register and the first outgoing mailbox register (if needed) *prior* to calling isp_intr (if calling isp_intr is necessary at all).
2001-09-01 11:12:23 +04:00
u_int16_t isr, mbox, sema;
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
if (XS_CMD_DONE_P(xs)) {
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
isp_prt(isp, ISP_LOGDEBUG1,
"watchdog found done cmd (handle 0x%x)", handle);
Add a maintenance note. Change all splbio's to use the ISP_LOCK/ISP_UNLOCK or ISP_ILOCK/ISP_IUNLOCK macros.
2000-08-14 11:08:12 +04:00
ISP_IUNLOCK(isp);
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
return;
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
}
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
if (XS_CMD_WDOG_P(xs)) {
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
isp_prt(isp, ISP_LOGDEBUG1,
"recursive watchdog (handle 0x%x)", handle);
Add a maintenance note. Change all splbio's to use the ISP_LOCK/ISP_UNLOCK or ISP_ILOCK/ISP_IUNLOCK macros.
2000-08-14 11:08:12 +04:00
ISP_IUNLOCK(isp);
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
return;
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
}
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
XS_CMD_S_WDOG(xs);
Add support for 2 Gigabit cards (2300/2312). This necessitated a change in how interrupts are down- the 23XX has not only a different place to check for an interrupt, but unlike all other QLogic cards, you have to read the status as a 32 bit word- not 16 bit words. Rather than have device specific functions as called from the core module (in isp_intr), it makes more sense to have the platform/bus modules do the gruntwork of splitting out the isr, semaphore register and the first outgoing mailbox register (if needed) *prior* to calling isp_intr (if calling isp_intr is necessary at all).
2001-09-01 11:12:23 +04:00
if (ISP_READ_ISR(isp, &isr, &sema, &mbox)) {
isp_intr(isp, isr, sema, mbox);
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
Add support for 2 Gigabit cards (2300/2312). This necessitated a change in how interrupts are down- the 23XX has not only a different place to check for an interrupt, but unlike all other QLogic cards, you have to read the status as a 32 bit word- not 16 bit words. Rather than have device specific functions as called from the core module (in isp_intr), it makes more sense to have the platform/bus modules do the gruntwork of splitting out the isr, semaphore register and the first outgoing mailbox register (if needed) *prior* to calling isp_intr (if calling isp_intr is necessary at all).
2001-09-01 11:12:23 +04:00
}
if (XS_CMD_DONE_P(xs)) {
isp_prt(isp, ISP_LOGDEBUG1,
"watchdog cleanup for handle 0x%x", handle);
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
XS_CMD_C_WDOG(xs);
isp_done(xs);
} else if (XS_CMD_GRACE_P(xs)) {
Add support for 2 Gigabit cards (2300/2312). This necessitated a change in how interrupts are down- the 23XX has not only a different place to check for an interrupt, but unlike all other QLogic cards, you have to read the status as a 32 bit word- not 16 bit words. Rather than have device specific functions as called from the core module (in isp_intr), it makes more sense to have the platform/bus modules do the gruntwork of splitting out the isr, semaphore register and the first outgoing mailbox register (if needed) *prior* to calling isp_intr (if calling isp_intr is necessary at all).
2001-09-01 11:12:23 +04:00
isp_prt(isp, ISP_LOGDEBUG1,
"watchdog timeout for handle 0x%x", handle);
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
/*
* Make sure the command is *really* dead before we
* release the handle (and DMA resources) for reuse.
*/
(void) isp_control(isp, ISPCTL_ABORT_CMD, arg);
/*
Fix some typos. From Tom Cosgrove via jmc@openbsd.
2003-11-02 14:07:44 +03:00
* After this point, the command is really dead.
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
*/
if (XS_XFRLEN(xs)) {
ISP_DMAFREE(isp, xs, handle);
}
isp_destroy_handle(isp, handle);
XS_SETERR(xs, XS_TIMEOUT);
XS_CMD_S_CLEAR(xs);
isp_done(xs);
} else {
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. For NetBSD, it does a ddi_dmamap_sync as appropriate. This gets us out of the explicit ddi_dmamap_sync on the whole response queue that we did for SBus cards at each interrupt. Set things up so that platforms that cannot have an SBus don't get a lot of the SBus code checks (dead coded out). Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. Nuke some additional __P macros.
2001-12-14 03:13:44 +03:00
u_int16_t nxti, optr;
ispreq_t local, *mp = &local, *qe;
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
isp_prt(isp, ISP_LOGDEBUG2,
Add support for 2 Gigabit cards (2300/2312). This necessitated a change in how interrupts are down- the 23XX has not only a different place to check for an interrupt, but unlike all other QLogic cards, you have to read the status as a 32 bit word- not 16 bit words. Rather than have device specific functions as called from the core module (in isp_intr), it makes more sense to have the platform/bus modules do the gruntwork of splitting out the isr, semaphore register and the first outgoing mailbox register (if needed) *prior* to calling isp_intr (if calling isp_intr is necessary at all).
2001-09-01 11:12:23 +04:00
"possible command timeout on handle %x", handle);
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
XS_CMD_C_WDOG(xs);
callout_reset(&xs->xs_callout, hz, isp_dog, xs);
Avoid strict-alias warnings.
2002-11-25 04:44:21 +03:00
if (isp_getrqentry(isp, &nxti, &optr, (void *) &qe)) {
Finally fix this driver to be sensible about the ENDIAN dance. It's not quite simply a question of the Qlogic being little endian and having to have stuff swapped on big endian machines- it also has to do with the fact that the SBus and PCI DMA layouts are wierd with respect to this. At any rate, now finally fixed- works on Mac G4, tested it on a SS10 for sparc, checked on alpha to see if I've broken anything, and as soon as I get another spare afternoon I'll finally install a sparc64 version which should just work (as it'll be like the Mac).
2000-12-09 11:06:31 +03:00
ISP_UNLOCK(isp);
Redo watchdogs to handle cases of false death of commands- time each command now.. Get rid of SCCLUn stuff. Use an isp_done routine to handle cases of watchdog and isp_done racing to completion.
2000-07-06 02:23:05 +04:00
return;
}
XS_CMD_S_GRACE(xs);
MEMZERO((void *) mp, sizeof (*mp));
mp->req_header.rqs_entry_count = 1;
mp->req_header.rqs_entry_type = RQSTYPE_MARKER;
mp->req_modifier = SYNC_ALL;
mp->req_target = XS_CHANNEL(xs) << 7;
Major restructuring for swizzling to the request queue and unswizzling from the response queue. Instead of the ad hoc ISP_SWIZZLE_REQUEST, we now have a complete set of inline functions in isp_inline.h. Each platform is responsible for providing just one of a set of ISP_IOX_{GET,PUT}{8,16,32} macros. The reason this needs to be done is that we need to have a single set of functions that will work correctly on multiple architectures for both little and big endian machines. It also needs to work correctly in the case that we have the request or response queues in memory that has to be treated specially (e.g., have ddi_dma_sync called on it for Solaris after we update it or before we read from it). One thing that falls out of this is that we no longer build requests in the request queue itself. Instead, we build the request locally (e.g., on the stack) and then as part of the swizzling operation, copy it to the request queue entry we've allocated. I thought long and hard about whether this was too expensive a change to make as it in a lot of cases requires an extra copy. On balance, the flexbility is worth it. With any luck, the entry that we build locally stays in a processor writeback cache (after all, it's only 64 bytes) so that the cost of actually flushing it to the memory area that is the shared queue with the PCI device is not all that expensive. We may examine this again and try to get clever in the future to try and avoid copies. Another change that falls out of this is that MEMORYBARRIER should be taken a lot more seriously. The macro ISP_ADD_REQUEST does a MEMORYBARRIER on the entry being added. But there had been many other places this had been missing. It's now very important that it be done. For NetBSD, it does a ddi_dmamap_sync as appropriate. This gets us out of the explicit ddi_dmamap_sync on the whole response queue that we did for SBus cards at each interrupt. Set things up so that platforms that cannot have an SBus don't get a lot of the SBus code checks (dead coded out). Additional changes: Fix a longstanding buglet of sorts. When we get an entry via isp_getrqentry, the iptr value that gets returned is the value we intend to eventually plug into the ISP registers as the entry *one past* the last one we've written- *not* the current entry we're updating. All along we've been calling sync functions on the wrong index value. Argh. The 'fix' here is to rename all 'iptr' variables as 'nxti' to remember that this is the 'next' pointer- not the current pointer. Devote a single bit to mboxbsy- and set aside bits for output mbox registers that we need to pick up- we can have at least one command which does not have any defined output registers (MBOX_EXECUTE_FIRMWARE). Explicitly decode GetAllNext SNS Response back *as* a GetAllNext response. Otherwise, we won't unswizzle it correctly. Nuke some additional __P macros.
2001-12-14 03:13:44 +03:00
isp_put_request(isp, mp, qe);
ISP_ADD_REQUEST(isp, nxti);
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
}
Add a maintenance note. Change all splbio's to use the ISP_LOCK/ISP_UNLOCK or ISP_ILOCK/ISP_IUNLOCK macros.
2000-08-14 11:08:12 +04:00
} else {
isp_prt(isp, ISP_LOGDEBUG0, "watchdog with no command");
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
}
Add a maintenance note. Change all splbio's to use the ISP_LOCK/ISP_UNLOCK or ISP_ILOCK/ISP_IUNLOCK macros.
2000-08-14 11:08:12 +04:00
ISP_IUNLOCK(isp);
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
}
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
/*
* Fibre Channel state cleanup thread
*/
static void
isp_create_fc_worker(void *arg)
{
struct ispsoftc *isp = arg;
if (kthread_create1(isp_fc_worker, isp, &isp->isp_osinfo.thread,
"%s:fc_thrd", isp->isp_name)) {
isp_prt(isp, ISP_LOGERR, "unable to create FC worker thread");
panic("isp_create_fc_worker");
}
}
static void
isp_fc_worker(void *arg)
{
void scsipi_run_queue(struct scsipi_channel *);
struct ispsoftc *isp = arg;
for (;;) {
int s;
/*
* Note we do *not* use the ISP_LOCK/ISP_UNLOCK macros here.
*/
s = splbio();
while (isp->isp_osinfo.threadwork) {
isp->isp_osinfo.threadwork = 0;
In isp_fc_worker() reduce the timeout value used to poll for the link status from 10 seconds to 250 milli seconds for the second and subsequent attempts to determine the link status. This prevents unnecessary long stalls after the device attach routines have completed. Tested on a dual port QLogic 3212. Approved by mjacob@feral.com.
2003-10-08 23:51:01 +04:00
if (isp_fc_runstate(isp, 250000) == 0) {
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
break;
}
Add support for 2 Gigabit cards (2300/2312). This necessitated a change in how interrupts are down- the 23XX has not only a different place to check for an interrupt, but unlike all other QLogic cards, you have to read the status as a 32 bit word- not 16 bit words. Rather than have device specific functions as called from the core module (in isp_intr), it makes more sense to have the platform/bus modules do the gruntwork of splitting out the isr, semaphore register and the first outgoing mailbox register (if needed) *prior* to calling isp_intr (if calling isp_intr is necessary at all).
2001-09-01 11:12:23 +04:00
if (isp->isp_osinfo.loop_checked &&
FCPARAM(isp)->loop_seen_once == 0) {
splx(s);
goto skip;
}
Note both active and nvram WWNN/WWPNs Fix the bug where the thread worker constantly looped. It was a race as to whether or not the thread fired up or a polled/probed command happened first, or maybe it's due to other scsipi changes, but we never set the value that caused the fc thread worker to actually go to sleep until an initial loop up. This was rather ugly as it consumed nearly all CPU time available from thence on. Bad.
2003-08-07 05:10:53 +04:00
isp->isp_osinfo.loop_checked = 1;
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
isp->isp_osinfo.threadwork = 1;
splx(s);
delay(500 * 1000);
s = splbio();
}
if (FCPARAM(isp)->isp_fwstate != FW_READY ||
FCPARAM(isp)->isp_loopstate != LOOP_READY) {
isp_prt(isp, ISP_LOGINFO, "isp_fc_runstate in vain");
isp->isp_osinfo.threadwork = 1;
splx(s);
continue;
}
if (isp->isp_osinfo.blocked) {
isp->isp_osinfo.blocked = 0;
Add support for 2 Gigabit cards (2300/2312). This necessitated a change in how interrupts are down- the 23XX has not only a different place to check for an interrupt, but unlike all other QLogic cards, you have to read the status as a 32 bit word- not 16 bit words. Rather than have device specific functions as called from the core module (in isp_intr), it makes more sense to have the platform/bus modules do the gruntwork of splitting out the isr, semaphore register and the first outgoing mailbox register (if needed) *prior* to calling isp_intr (if calling isp_intr is necessary at all).
2001-09-01 11:12:23 +04:00
isp_prt(isp, ISP_LOGDEBUG0,
"restarting queues (freeze count %d)",
isp->isp_chanA.chan_qfreeze);
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
scsipi_channel_thaw(&isp->isp_chanA, 1);
}
if (isp->isp_osinfo.thread == NULL)
break;
Add support for 2 Gigabit cards (2300/2312). This necessitated a change in how interrupts are down- the 23XX has not only a different place to check for an interrupt, but unlike all other QLogic cards, you have to read the status as a 32 bit word- not 16 bit words. Rather than have device specific functions as called from the core module (in isp_intr), it makes more sense to have the platform/bus modules do the gruntwork of splitting out the isr, semaphore register and the first outgoing mailbox register (if needed) *prior* to calling isp_intr (if calling isp_intr is necessary at all).
2001-09-01 11:12:23 +04:00
skip:
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
(void) tsleep(&isp->isp_osinfo.thread, PRIBIO, "fcclnup", 0);
splx(s);
}
/* In case parent is waiting for us to exit. */
wakeup(&isp->isp_osinfo.thread);
kthread_exit(0);
}
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
/*
* Free any associated resources prior to decommissioning and
* set the card to a known state (so it doesn't wake up and kick
* us when we aren't expecting it to).
*
* Locks are held before coming here.
*/
void
Handles are now 16 bits. ANSIfy.
2001-03-14 08:44:21 +03:00
isp_uninit(struct ispsoftc *isp)
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
{
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
isp_lock(isp);
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +03:00
/*
* Leave with interrupts disabled.
*/
DISABLE_INTS(isp);
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
isp_unlock(isp);
clean up headers; move uninit/watch to outer layers
1998-12-28 22:10:43 +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
int
Handles are now 16 bits. ANSIfy.
2001-03-14 08:44:21 +03:00
isp_async(struct ispsoftc *isp, ispasync_t cmd, void *arg)
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
{
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, tgt;
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +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
switch (cmd) {
case ISPASYNC_NEW_TGT_PARAMS:
Add in an ioctl entry point so scsictl mediated bus resets will work. Redo how we start commands- do a 'slow' start function which then looks to see when we're done the configuration process at which point it *then* enables sync/wide mode. Set the max openings amount to the true max openings- not a synthetic. Add a timeout driven command requeue function so that Loop Down events well freeze things until a later point in time where they might be restarted.
1999-10-14 06:31:11 +04:00
if (IS_SCSI(isp) && isp->isp_dblev) {
sdparam *sdp = isp->isp_param;
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
int flags;
struct scsipi_xfer_mode xm;
Add in an ioctl entry point so scsictl mediated bus resets will work. Redo how we start commands- do a 'slow' start function which then looks to see when we're done the configuration process at which point it *then* enables sync/wide mode. Set the max openings amount to the true max openings- not a synthetic. Add a timeout driven command requeue function so that Loop Down events well freeze things until a later point in time where they might be restarted.
1999-10-14 06:31:11 +04:00
tgt = *((int *) arg);
bus = (tgt >> 16) & 0xffff;
tgt &= 0xffff;
Dual LVD (1280) support.
1999-12-16 08:35:42 +03:00
sdp += bus;
Add support for 2 Gigabit cards (2300/2312). This necessitated a change in how interrupts are down- the 23XX has not only a different place to check for an interrupt, but unlike all other QLogic cards, you have to read the status as a 32 bit word- not 16 bit words. Rather than have device specific functions as called from the core module (in isp_intr), it makes more sense to have the platform/bus modules do the gruntwork of splitting out the isr, semaphore register and the first outgoing mailbox register (if needed) *prior* to calling isp_intr (if calling isp_intr is necessary at all).
2001-09-01 11:12:23 +04:00
flags = sdp->isp_devparam[tgt].actv_flags;
Dual LVD (1280) support.
1999-12-16 08:35:42 +03:00
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
xm.xm_mode = 0;
Add support for 2 Gigabit cards (2300/2312). This necessitated a change in how interrupts are down- the 23XX has not only a different place to check for an interrupt, but unlike all other QLogic cards, you have to read the status as a 32 bit word- not 16 bit words. Rather than have device specific functions as called from the core module (in isp_intr), it makes more sense to have the platform/bus modules do the gruntwork of splitting out the isr, semaphore register and the first outgoing mailbox register (if needed) *prior* to calling isp_intr (if calling isp_intr is necessary at all).
2001-09-01 11:12:23 +04:00
xm.xm_period = sdp->isp_devparam[tgt].actv_period;
xm.xm_offset = sdp->isp_devparam[tgt].actv_offset;
Pull up the thorpej_scsipi branch to main branch. This is a completely rewritten scsipi_xfer execution engine, and the associated changes to HBA drivers. Overview of changes & features: - All xfers are queued in the mid-layer, rather than doing so in an ad-hoc fashion in individual adapter drivers. - Adapter/channel resource management in the mid-layer, avoids even trying to start running an xfer if the adapter/channel doesn't have the resources. - Better communication between the mid-layer and the adapters. - Asynchronous event notification mechanism from adapter to mid-layer and peripherals. - Better peripheral queue management: freeze/thaw, sorted requeueing during recovery, etc. - Clean separation of peripherals, adapters, and adapter channels (no more scsipi_link). - Kernel thread for each scsipi_channel makes error recovery much easier (no more dealing with interrupt context when recovering from an error). - Mid-layer support for tagged queueing: commands can have the tag type set explicitly, tag IDs are allocated in the mid-layer (thus eliminating the need to use buggy tag ID allocation schemes in many adapter drivers). - support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command will be requeued, or a REQUEST SENSE will be sent as appropriate. Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge
2001-04-25 21:53:04 +04:00
xm.xm_target = tgt;
if ((flags & DPARM_SYNC) && xm.xm_period && xm.xm_offset)
xm.xm_mode |= PERIPH_CAP_SYNC;
if (flags & DPARM_WIDE)
xm.xm_mode |= PERIPH_CAP_WIDE16;
if (flags & DPARM_TQING)
xm.xm_mode |= PERIPH_CAP_TQING;
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
scsipi_async_event(bus? &isp->isp_chanB : &isp->isp_chanA,
ASYNC_EVENT_XFER_MODE, &xm);
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
break;
Add in an ioctl entry point so scsictl mediated bus resets will work. Redo how we start commands- do a 'slow' start function which then looks to see when we're done the configuration process at which point it *then* enables sync/wide mode. Set the max openings amount to the true max openings- not a synthetic. Add a timeout driven command requeue function so that Loop Down events well freeze things until a later point in time where they might be restarted.
1999-10-14 06:31:11 +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
case ISPASYNC_BUS_RESET:
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
bus = *((int *) arg);
scsipi_async_event(bus? &isp->isp_chanB : &isp->isp_chanA,
ASYNC_EVENT_RESET, NULL);
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
isp_prt(isp, ISP_LOGINFO, "SCSI bus %d reset detected", 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
break;
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
case ISPASYNC_LIP:
/*
* Don't do queue freezes or blockage until we have the
* thread running that can unfreeze/unblock us.
*/
if (isp->isp_osinfo.blocked == 0) {
if (isp->isp_osinfo.thread) {
isp->isp_osinfo.blocked = 1;
scsipi_channel_freeze(&isp->isp_chanA, 1);
}
}
isp_prt(isp, ISP_LOGINFO, "LIP Received");
break;
case ISPASYNC_LOOP_RESET:
/*
* Don't do queue freezes or blockage until we have the
* thread running that can unfreeze/unblock us.
*/
if (isp->isp_osinfo.blocked == 0) {
if (isp->isp_osinfo.thread) {
isp->isp_osinfo.blocked = 1;
scsipi_channel_freeze(&isp->isp_chanA, 1);
}
}
isp_prt(isp, ISP_LOGINFO, "Loop Reset Received");
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 ISPASYNC_LOOP_DOWN:
/*
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
* Don't do queue freezes or blockage until we have the
* thread running that can unfreeze/unblock us.
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
*/
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
if (isp->isp_osinfo.blocked == 0) {
if (isp->isp_osinfo.thread) {
isp->isp_osinfo.blocked = 1;
scsipi_channel_freeze(&isp->isp_chanA, 1);
}
}
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
isp_prt(isp, ISP_LOGINFO, "Loop DOWN");
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 ISPASYNC_LOOP_UP:
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
/*
* Let the subsequent ISPASYNC_CHANGE_NOTIFY invoke
* the FC worker thread. When the FC worker thread
* is done, let *it* call scsipi_channel_thaw...
*/
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
isp_prt(isp, ISP_LOGINFO, "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
break;
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
case ISPASYNC_PROMENADE:
Add in an ioctl entry point so scsictl mediated bus resets will work. Redo how we start commands- do a 'slow' start function which then looks to see when we're done the configuration process at which point it *then* enables sync/wide mode. Set the max openings amount to the true max openings- not a synthetic. Add a timeout driven command requeue function so that Loop Down events well freeze things until a later point in time where they might be restarted.
1999-10-14 06:31:11 +04:00
if (IS_FC(isp) && isp->isp_dblev) {
make auto const variables static.
2002-06-08 20:51:38 +04:00
static const char fmt[] = "Target %d (Loop 0x%x) Port ID 0x%x "
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
"(role %s) %s\n Port WWN 0x%08x%08x\n Node WWN 0x%08x%08x";
make auto const variables static.
2002-06-08 20:51:38 +04:00
const static char *const roles[4] = {
It's 'role None', not 'role No'. Also, pay attention to the difference in the ISPASYNC_CHANGE_NOTIFY between loop and fabric changes.
2001-05-16 07:52:10 +04:00
"None", "Target", "Initiator", "Target/Initiator"
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
fcparam *fcp = isp->isp_param;
int tgt = *((int *) arg);
struct lportdb *lp = &fcp->portdb[tgt];
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
isp_prt(isp, ISP_LOGINFO, fmt, tgt, lp->loopid, lp->portid,
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
roles[lp->roles & 0x3],
(lp->valid)? "Arrived" : "Departed",
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
(u_int32_t) (lp->port_wwn >> 32),
(u_int32_t) (lp->port_wwn & 0xffffffffLL),
(u_int32_t) (lp->node_wwn >> 32),
(u_int32_t) (lp->node_wwn & 0xffffffffLL));
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 ISPASYNC_CHANGE_NOTIFY:
It's 'role None', not 'role No'. Also, pay attention to the difference in the ISPASYNC_CHANGE_NOTIFY between loop and fabric changes.
2001-05-16 07:52:10 +04:00
if (arg == ISPASYNC_CHANGE_PDB) {
isp_prt(isp, ISP_LOGINFO, "Port Database Changed");
} else if (arg == ISPASYNC_CHANGE_SNS) {
ISPASYNC_PDB_CHANGED->ISPASYNC_LOGGED_INOUT. And also, ISPASYNC_CHANGE_NOTIFY is for both local loop and fabric cases now.
2001-01-09 21:54:53 +03:00
isp_prt(isp, ISP_LOGINFO,
"Name Server Database Changed");
}
Create a kernel thread for Fibre Channel cards. This is the new way of doing business- modulo some startup spasms and peculiarities of the way kthreads are started (*after* configuration, weird) and some strangeness with the freeze/thaw code, what now happens is that any of Loop Down, LIP, Loop Reset or Port Datbase or Name Server Database Changed ASYNC events cause the queues to freeze for this channel. The arrival of a Loop UP is not relevant. What *is* relevant is that the Port Datbase or Name Server Changed async event indicate that it's okay to go and (re)evaluate the state of the FC link and (re)probe local loop and fabric membership. We have a kthread do this because it's *sooooo* much nicer to be able to sleep while doing the 130-250 mailbox commands it'll take to re-evaluate things. When the state is well known again, we can unfreeze the channel queues. Then, as commands start arriving, we simply can start them or bounce them with XS_SELTIMEOUT (if the device in question has gone away). Previously, we did lazy evaluation, which meant that if a change occurred, we would wait until the very *next* command to go rebuild stuff. The reason this is not sensible is: a) Even with sleeping, you can hang up your system because you might be making some poor stat(2) call pay the price of re-evaluating the whole fabric. b) If we ever really want to get to dynamic attachment/detachment, we should find out sooner, rather than later, where things get to. Split off ispminphys_1020 from ispminphys- a 1020 has a 24 bit limit- not anything newer. Re-enable LIPs and Loop Resets as async events- this allows the outer layer to set policy about them. Roll platform major && minor. Remove bogus waitq (no longer used). Remove callout entry in softc (no longer used). Define some shorthands for channels. Clean up a variety of cruft left over from the thorpej_scsipi changeover.
2001-05-26 01:45:55 +04:00
/*
* We can set blocked here because we know it's now okay
* to try and run isp_fc_runstate (in order to build loop
* state). But we don't try and freeze the midlayer's queue
* if we have no thread that we can wake to later unfreeze
* it.
*/
if (isp->isp_osinfo.blocked == 0) {
isp->isp_osinfo.blocked = 1;
if (isp->isp_osinfo.thread) {
scsipi_channel_freeze(&isp->isp_chanA, 1);
}
}
/*
* Note that we have work for the thread to do, and
* if the thread is here already, wake it up.
*/
isp->isp_osinfo.threadwork++;
if (isp->isp_osinfo.thread) {
wakeup(&isp->isp_osinfo.thread);
} else {
isp_prt(isp, ISP_LOGDEBUG1, "no FC thread yet");
}
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;
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 ISPASYNC_FABRIC_DEV:
{
Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list.
2002-04-05 03:38:43 +04:00
int target, base, lim;
fcparam *fcp = isp->isp_param;
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
struct lportdb *lp = NULL;
Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list.
2002-04-05 03:38:43 +04:00
struct lportdb *clp = (struct lportdb *) arg;
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
char *pt;
Add in commented-out hiwater measurement. Remove egregious older bug which had us refusing to log into fabric devices unless they were NL ports. Whuff.
2000-10-16 09:15:55 +04:00
Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list.
2002-04-05 03:38:43 +04:00
switch (clp->port_type) {
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
case 1:
pt = " N_Port";
break;
case 2:
pt = " NL_Port";
break;
case 3:
pt = "F/NL_Port";
break;
case 0x7f:
pt = " Nx_Port";
break;
case 0x81:
pt = " F_port";
break;
case 0x82:
pt = " FL_Port";
break;
case 0x84:
pt = " E_port";
break;
default:
Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list.
2002-04-05 03:38:43 +04:00
pt = " ";
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
break;
}
Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list.
2002-04-05 03:38:43 +04:00
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
isp_prt(isp, ISP_LOGINFO,
Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list.
2002-04-05 03:38:43 +04:00
"%s Fabric Device @ PortID 0x%x", pt, clp->portid);
/*
* If we don't have an initiator role we bail.
*
* We just use ISPASYNC_FABRIC_DEV for announcement purposes.
*/
if ((isp->isp_role & ISP_ROLE_INITIATOR) == 0) {
break;
}
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
/*
Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list.
2002-04-05 03:38:43 +04:00
* Is this entry for us? If so, we bail.
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
*/
Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list.
2002-04-05 03:38:43 +04:00
if (fcp->isp_portid == clp->portid) {
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
break;
}
Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list.
2002-04-05 03:38:43 +04:00
/*
* Else, the default policy is to find room for it in
* our local port database. Later, when we execute
* the call to isp_pdb_sync either this newly arrived
* or already logged in device will be (re)announced.
*/
if (fcp->isp_topo == TOPO_FL_PORT)
base = FC_SNS_ID+1;
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
else
Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list.
2002-04-05 03:38:43 +04:00
base = 0;
if (fcp->isp_topo == TOPO_N_PORT)
lim = 1;
else
lim = MAX_FC_TARG;
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
/*
* Is it already in our list?
*/
Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list.
2002-04-05 03:38:43 +04:00
for (target = base; target < lim; target++) {
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
if (target >= FL_PORT_ID && target <= FC_SNS_ID) {
continue;
}
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
lp = &fcp->portdb[target];
Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list.
2002-04-05 03:38:43 +04:00
if (lp->port_wwn == clp->port_wwn &&
lp->node_wwn == clp->node_wwn) {
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
lp->fabric_dev = 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
break;
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +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
}
Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list.
2002-04-05 03:38:43 +04:00
if (target < lim) {
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;
}
Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list.
2002-04-05 03:38:43 +04:00
for (target = base; target < lim; target++) {
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
if (target >= FL_PORT_ID && target <= FC_SNS_ID) {
continue;
}
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
lp = &fcp->portdb[target];
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
if (lp->port_wwn == 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
break;
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +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
}
Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list.
2002-04-05 03:38:43 +04:00
if (target == lim) {
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
isp_prt(isp, ISP_LOGWARN,
Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list.
2002-04-05 03:38:43 +04:00
"out of space for fabric devices");
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
break;
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
}
Redo fabric evaluation to not use GET ALL NEXT (GA_NXT). Switches seem to be trying to wriggle out of supporting this well. Instead, use GID_FT to get a list of Port IDs and then use GPN_ID/GNN_ID to find the port and node wwn. This should make working on fabrics a bit cleaner and more stable. This also caused some cleanup of SNS subcommand canonicalization so that we can actually check for FS_ACC and FS_RJT, and if we get an FS_RJT, print out the reason and explanation codes. We'll keep the old GA_NXT method around if people want to uncomment a controlling definition in ispvar.h. This also had us clean up ISPASYNC_FABRICDEV to use a local lportdb argument and to have the caller explicitly say that a device is at the end of the fabric list.
2002-04-05 03:38:43 +04:00
lp->port_type = clp->port_type;
lp->fc4_type = clp->fc4_type;
lp->node_wwn = clp->node_wwn;
lp->port_wwn = clp->port_wwn;
lp->portid = clp->portid;
Do some cleanup based upon adapter role- mainly not enabling interrupts if we're ISP_ROLE_NONE. Change ISPASYNC_LOGGED_INOUT to ISPASYNC_PROMENADE. Make sure we note if something is a fabric device.
2001-02-13 02:30:12 +03:00
lp->fabric_dev = 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
break;
}
Defer turning off the no_mbox_ints flag until after the system is ready for interrupts. Handle FW crashes in outer layer.
2001-07-06 20:17:17 +04:00
case ISPASYNC_FW_CRASH:
{
u_int16_t mbox1, mbox6;
mbox1 = ISP_READ(isp, OUTMAILBOX1);
if (IS_DUALBUS(isp)) {
mbox6 = ISP_READ(isp, OUTMAILBOX6);
} else {
mbox6 = 0;
}
isp_prt(isp, ISP_LOGERR,
Fix a misspelled error message. Temp work around problems where if we allow for non-polled mailbox commands we got nailed by hardclock calling is for a timed thaw. Basically, this means we only enable non-polled mailbox commands in the FC kthread when it calls isp_fc_runstate.
2001-09-06 03:08:23 +04:00
"Internal Firmware Error on bus %d @ RISC Address 0x%x",
Defer turning off the no_mbox_ints flag until after the system is ready for interrupts. Handle FW crashes in outer layer.
2001-07-06 20:17:17 +04:00
mbox6, mbox1);
Various changes to allow for overriding just loopid and/or iid. Fix a random 23XX ISR reading bug. Be chattier about HBA_BOTCH errors. More f/w cashdump layering.
2002-08-13 01:33:39 +04:00
#ifdef ISP_FW_CRASH_DUMP
if (IS_FC(isp)) {
if (isp->isp_osinfo.blocked == 0) {
isp->isp_osinfo.blocked = 1;
scsipi_channel_freeze(&isp->isp_chanA, 1);
}
isp_fw_dump(isp);
}
Defer turning off the no_mbox_ints flag until after the system is ready for interrupts. Handle FW crashes in outer layer.
2001-07-06 20:17:17 +04:00
isp_reinit(isp);
Various changes to allow for overriding just loopid and/or iid. Fix a random 23XX ISR reading bug. Be chattier about HBA_BOTCH errors. More f/w cashdump layering.
2002-08-13 01:33:39 +04:00
isp_async(isp, ISPASYNC_FW_RESTART, NULL);
#endif
Defer turning off the no_mbox_ints flag until after the system is ready for interrupts. Handle FW crashes in outer layer.
2001-07-06 20:17:17 +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
default:
break;
}
return (0);
}
Core version 2.0 (platform version 1.0) rewrite of ISP driver. Some interace cleanups, some new common functions. The major impact that will be noticeable right away is that if you boot with not Fibre connected to the FC cards, you no longer hang indefinitely.
2000-08-02 03:55:09 +04:00
#include <machine/stdarg.h>
void
isp_prt(struct ispsoftc *isp, int level, const char *fmt, ...)
{
va_list ap;
if (level != ISP_LOGALL && (level & isp->isp_dblev) == 0) {
return;
}
printf("%s: ", isp->isp_name);
va_start(ap, fmt);
vprintf(fmt, ap);
va_end(ap);
printf("\n");
}