NetBSD/sys/dev/ic/isp_netbsd.c

1145 lines
28 KiB
C
Raw Normal View History

/* $NetBSD: isp_netbsd.c,v 1.59 2003/03/21 18:05:16 mjacob Exp $ */
/*
* This driver, which is contained in NetBSD in the files:
*
* sys/dev/ic/isp.c
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
* 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
*
* Is being actively maintained by Matthew Jacob (mjacob@netbsd.org).
* 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.
*/
/*
* Platform (NetBSD) dependent common attachment code for Qlogic adapters.
* Matthew Jacob <mjacob@nas.nasa.gov>
*/
/*
* Copyright (C) 1997, 1998, 1999 National Aeronautics & Space Administration
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. The name of the author may not be used to endorse or promote products
* derived from this software without specific prior written permission
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``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.
*/
2001-11-13 16:14:31 +03:00
#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: isp_netbsd.c,v 1.59 2003/03/21 18:05:16 mjacob Exp $");
2001-11-13 16:14:31 +03:00
#include <dev/ic/isp_netbsd.h>
#include <sys/scsiio.h>
/*
* 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))
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 *);
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 *);
static int
ispioctl(struct scsipi_channel *, u_long, caddr_t, int, struct proc *);
static void isp_polled_cmd(struct ispsoftc *, XS_T *);
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 *);
/*
* Complete attachment of hardware, include subdevices.
*/
void
2001-03-14 08:44:21 +03:00
isp_attach(struct ispsoftc *isp)
{
isp->isp_state = ISP_RUNSTATE;
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;
}
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;
/*
* Until the midlayer is fixed to use REPORT LUNS, limit to 8 luns.
*/
isp->isp_osinfo._chan.chan_nluns = min(isp->isp_maxluns, 8);
if (IS_FC(isp)) {
isp->isp_osinfo._chan.chan_flags = SCSIPI_CHAN_NOSETTLE;
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);
#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
} 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;
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
isp->isp_osinfo._chan.chan_ntargets = MAX_TARGETS;
isp->isp_osinfo._chan.chan_id = sdp->isp_initiator_id;
isp->isp_osinfo.discovered[0] = 1 << sdp->isp_initiator_id;
1999-12-16 08:35:42 +03:00
if (IS_DUALBUS(isp)) {
isp->isp_osinfo._chan_b = isp->isp_osinfo._chan;
sdp++;
1999-10-17 05:23:21 +04:00
isp->isp_osinfo.discovered[1] =
1 << sdp->isp_initiator_id;
isp->isp_osinfo._chan_b.chan_id = sdp->isp_initiator_id;
isp->isp_osinfo._chan_b.chan_channel = 1;
}
ISP_LOCK(isp);
(void) isp_control(isp, ISPCTL_RESET_BUS, &bus);
1999-12-16 08:35:42 +03:00
if (IS_DUALBUS(isp)) {
bus++;
(void) isp_control(isp, ISPCTL_RESET_BUS, &bus);
}
ISP_UNLOCK(isp);
}
/*
* Defer enabling mailbox interrupts until later.
*/
config_interrupts((struct device *) isp, isp_config_interrupts);
/*
* 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);
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);
}
}
static void
isp_config_interrupts(struct device *self)
{
struct ispsoftc *isp = (struct ispsoftc *) self;
/*
* After this point, we'll be doing the new configuration
* schema which allows interrupts, so we can do tsleep/wakeup
* for mailbox stuff at that point, if that's allowed.
*/
if (IS_FC(isp)) {
isp->isp_osinfo.no_mbox_ints = 0;
}
}
/*
* 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)
{
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);
}
static int
ispioctl(struct scsipi_channel *chan, u_long cmd, caddr_t addr, int flag,
struct proc *p)
{
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;
switch (cmd) {
#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;
else
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);
retval = 0;
break;
#endif
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;
case ISP_RESCAN:
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;
}
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;
hba->active_node_wwn = FCPARAM(isp)->isp_nodewwn;
hba->active_port_wwn = FCPARAM(isp)->isp_portwwn;
ISP_UNLOCK(isp);
break;
}
case SCBUSIORESET:
ISP_LOCK(isp);
if (isp_control(isp, ISPCTL_RESET_BUS, &chan->chan_channel))
retval = EIO;
else
retval = 0;
ISP_UNLOCK(isp);
break;
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)
{
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);
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;
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);
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);
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 (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);
}
}
if (xs->xs_control & XS_CTL_POLL) {
isp->isp_osinfo.no_mbox_ints = ombi;
}
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
}
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);
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;
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;
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);
sdp->isp_devparam[xm->xm_target].goal_flags |= dflags;
dflags = sdp->isp_devparam[xm->xm_target].goal_flags;
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);
isp_prt(isp, ISP_LOGDEBUG1,
"ispioctl: device flags 0x%x for %d.%d.X",
dflags, chan->chan_channel, xm->xm_target);
break;
}
default:
break;
}
}
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)
{
int result;
int infinite = 0, mswait;
result = isp_start(xs);
switch (result) {
case CMD_QUEUED:
break;
case CMD_RQLATER:
if (XS_NOERR(xs)) {
xs->error = XS_REQUEUE;
}
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;
}
/* FALLTHROUGH */
case CMD_COMPLETE:
scsipi_done(xs);
return;
}
/*
* If we can't use interrupts, poll on completion.
*/
if ((mswait = XS_TIME(xs)) == 0)
infinite = 1;
while (mswait || infinite) {
u_int16_t isr, sema, mbox;
if (ISP_READ_ISR(isp, &isr, &sema, &mbox)) {
isp_intr(isp, isr, sema, mbox);
if (XS_CMD_DONE_P(xs)) {
break;
}
}
USEC_DELAY(1000);
mswait -= 1;
}
/*
* 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)) {
isp_reinit(isp);
}
if (XS_NOERR(xs)) {
isp_prt(isp, ISP_LOGERR, "polled command timed out");
XS_SETERR(xs, HBA_BOTCH);
}
}
scsipi_done(xs);
}
void
2001-03-14 08:44:21 +03:00
isp_done(XS_T *xs)
{
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)) {
isp_prt(isp, ISP_LOGDEBUG1,
"finished command on borrowed time");
}
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);
}
}
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));
}
scsipi_done(xs);
}
}
static void
2001-03-14 08:44:21 +03:00
isp_dog(void *arg)
{
XS_T *xs = arg;
struct ispsoftc *isp = XS_ISP(xs);
2001-03-14 08:44:21 +03:00
u_int16_t handle;
ISP_ILOCK(isp);
/*
* 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.
*/
handle = isp_find_handle(isp, xs);
if (handle) {
u_int16_t isr, mbox, sema;
if (XS_CMD_DONE_P(xs)) {
isp_prt(isp, ISP_LOGDEBUG1,
"watchdog found done cmd (handle 0x%x)", handle);
ISP_IUNLOCK(isp);
return;
}
if (XS_CMD_WDOG_P(xs)) {
isp_prt(isp, ISP_LOGDEBUG1,
"recursive watchdog (handle 0x%x)", handle);
ISP_IUNLOCK(isp);
return;
}
XS_CMD_S_WDOG(xs);
if (ISP_READ_ISR(isp, &isr, &sema, &mbox)) {
isp_intr(isp, isr, sema, mbox);
}
if (XS_CMD_DONE_P(xs)) {
isp_prt(isp, ISP_LOGDEBUG1,
"watchdog cleanup for handle 0x%x", handle);
XS_CMD_C_WDOG(xs);
isp_done(xs);
} else if (XS_CMD_GRACE_P(xs)) {
isp_prt(isp, ISP_LOGDEBUG1,
"watchdog timeout for handle 0x%x", handle);
/*
* 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);
/*
* After this point, the comamnd is really dead.
*/
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;
isp_prt(isp, ISP_LOGDEBUG2,
"possible command timeout on handle %x", handle);
XS_CMD_C_WDOG(xs);
callout_reset(&xs->xs_callout, hz, isp_dog, xs);
2002-11-25 04:44:21 +03:00
if (isp_getrqentry(isp, &nxti, &optr, (void *) &qe)) {
ISP_UNLOCK(isp);
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);
}
} else {
isp_prt(isp, ISP_LOGDEBUG0, "watchdog with no command");
}
ISP_IUNLOCK(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
/*
* 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;
if (isp_fc_runstate(isp, 10 * 1000000) == 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;
}
if (isp->isp_osinfo.loop_checked &&
FCPARAM(isp)->loop_seen_once == 0) {
splx(s);
goto 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
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;
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;
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);
}
/*
* 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
2001-03-14 08:44:21 +03:00
isp_uninit(struct ispsoftc *isp)
{
isp_lock(isp);
/*
* Leave with interrupts disabled.
*/
DISABLE_INTS(isp);
isp_unlock(isp);
}
int
2001-03-14 08:44:21 +03:00
isp_async(struct ispsoftc *isp, ispasync_t cmd, void *arg)
{
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
switch (cmd) {
case ISPASYNC_NEW_TGT_PARAMS:
if (IS_SCSI(isp) && isp->isp_dblev) {
sdparam *sdp = isp->isp_param;
int flags;
struct scsipi_xfer_mode xm;
tgt = *((int *) arg);
bus = (tgt >> 16) & 0xffff;
tgt &= 0xffff;
1999-12-16 08:35:42 +03:00
sdp += bus;
flags = sdp->isp_devparam[tgt].actv_flags;
1999-12-16 08:35:42 +03:00
xm.xm_mode = 0;
xm.xm_period = sdp->isp_devparam[tgt].actv_period;
xm.xm_offset = sdp->isp_devparam[tgt].actv_offset;
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);
break;
}
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);
isp_prt(isp, ISP_LOGINFO, "SCSI bus %d reset detected", bus);
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;
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.
*/
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);
}
}
isp_prt(isp, ISP_LOGINFO, "Loop DOWN");
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...
*/
isp_prt(isp, ISP_LOGINFO, "Loop UP");
break;
case ISPASYNC_PROMENADE:
if (IS_FC(isp) && isp->isp_dblev) {
2002-06-08 20:51:38 +04:00
static const char fmt[] = "Target %d (Loop 0x%x) Port ID 0x%x "
"(role %s) %s\n Port WWN 0x%08x%08x\n Node WWN 0x%08x%08x";
2002-06-08 20:51:38 +04:00
const static char *const roles[4] = {
"None", "Target", "Initiator", "Target/Initiator"
};
fcparam *fcp = isp->isp_param;
int tgt = *((int *) arg);
struct lportdb *lp = &fcp->portdb[tgt];
isp_prt(isp, ISP_LOGINFO, fmt, tgt, lp->loopid, lp->portid,
roles[lp->roles & 0x3],
(lp->valid)? "Arrived" : "Departed",
(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));
break;
}
case ISPASYNC_CHANGE_NOTIFY:
if (arg == ISPASYNC_CHANGE_PDB) {
isp_prt(isp, ISP_LOGINFO, "Port Database Changed");
} else if (arg == ISPASYNC_CHANGE_SNS) {
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");
}
break;
case ISPASYNC_FABRIC_DEV:
{
int target, base, lim;
fcparam *fcp = isp->isp_param;
struct lportdb *lp = NULL;
struct lportdb *clp = (struct lportdb *) arg;
char *pt;
switch (clp->port_type) {
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:
pt = " ";
break;
}
isp_prt(isp, ISP_LOGINFO,
"%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;
}
/*
* Is this entry for us? If so, we bail.
*/
if (fcp->isp_portid == clp->portid) {
break;
}
/*
* 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;
else
base = 0;
if (fcp->isp_topo == TOPO_N_PORT)
lim = 1;
else
lim = MAX_FC_TARG;
/*
* Is it already in our list?
*/
for (target = base; target < lim; target++) {
if (target >= FL_PORT_ID && target <= FC_SNS_ID) {
continue;
}
lp = &fcp->portdb[target];
if (lp->port_wwn == clp->port_wwn &&
lp->node_wwn == clp->node_wwn) {
lp->fabric_dev = 1;
break;
}
}
if (target < lim) {
break;
}
for (target = base; target < lim; target++) {
if (target >= FL_PORT_ID && target <= FC_SNS_ID) {
continue;
}
lp = &fcp->portdb[target];
if (lp->port_wwn == 0) {
break;
}
}
if (target == lim) {
isp_prt(isp, ISP_LOGWARN,
"out of space for fabric devices");
break;
}
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;
lp->fabric_dev = 1;
break;
}
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,
"Internal Firmware Error on bus %d @ RISC Address 0x%x",
mbox6, mbox1);
#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);
}
isp_reinit(isp);
isp_async(isp, ISPASYNC_FW_RESTART, NULL);
#endif
break;
}
default:
break;
}
return (0);
}
#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");
}