freebsd-skq/sys/dev/mpt/mpt_freebsd.c

1531 lines
40 KiB
C
Raw Normal View History

/*
* FreeBSD/CAM specific routines for LSI '909 FC adapters.
* FreeBSD Version.
*
* Copyright (c) 2000, 2001 by Greg Ansley
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice immediately at the beginning of the file, without modification,
* this list of conditions, and the following disclaimer.
* 2. The name of the author may not be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR
* ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
/*
* Additional Copyright (c) 2002 by Matthew Jacob under same license.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <dev/mpt/mpt_freebsd.h>
static void mpt_poll(struct cam_sim *);
static timeout_t mpttimeout;
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
static timeout_t mpttimeout2;
static void mpt_action(struct cam_sim *, union ccb *);
static int mpt_setwidth(mpt_softc_t *, int, int);
static int mpt_setsync(mpt_softc_t *, int, int, int);
void
mpt_cam_attach(mpt_softc_t *mpt)
{
struct cam_devq *devq;
struct cam_sim *sim;
int maxq;
mpt->bus = 0;
maxq = (mpt->mpt_global_credits < MPT_MAX_REQUESTS(mpt))?
mpt->mpt_global_credits : MPT_MAX_REQUESTS(mpt);
/*
* Create the device queue for our SIM(s).
*/
devq = cam_simq_alloc(maxq);
if (devq == NULL) {
return;
}
/*
* Construct our SIM entry.
*/
sim = cam_sim_alloc(mpt_action, mpt_poll, "mpt", mpt,
mpt->unit, 1, maxq, devq);
if (sim == NULL) {
cam_simq_free(devq);
return;
}
/*
* Register exactly the bus.
*/
if (xpt_bus_register(sim, 0) != CAM_SUCCESS) {
cam_sim_free(sim, TRUE);
return;
}
if (xpt_create_path(&mpt->path, NULL, cam_sim_path(sim),
CAM_TARGET_WILDCARD, CAM_LUN_WILDCARD) != CAM_REQ_CMP) {
xpt_bus_deregister(cam_sim_path(sim));
cam_sim_free(sim, TRUE);
return;
}
mpt->sim = sim;
}
void
mpt_cam_detach(mpt_softc_t *mpt)
{
if (mpt->sim != NULL) {
xpt_free_path(mpt->path);
xpt_bus_deregister(cam_sim_path(mpt->sim));
cam_sim_free(mpt->sim, TRUE);
mpt->sim = NULL;
}
}
/* This routine is used after a system crash to dump core onto the
* swap device.
*/
static void
mpt_poll(struct cam_sim *sim)
{
mpt_softc_t *mpt = (mpt_softc_t *) cam_sim_softc(sim);
MPT_LOCK(mpt);
mpt_intr(mpt);
MPT_UNLOCK(mpt);
}
/*
* This routine is called if the 9x9 does not return completion status
* for a command after a CAM specified time.
*/
static void
mpttimeout(void *arg)
{
request_t *req;
union ccb *ccb = arg;
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
u_int32_t oseq;
mpt_softc_t *mpt;
mpt = ccb->ccb_h.ccb_mpt_ptr;
MPT_LOCK(mpt);
req = ccb->ccb_h.ccb_req_ptr;
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
oseq = req->sequence;
mpt->timeouts++;
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
if (mpt_intr(mpt)) {
if (req->sequence != oseq) {
mpt_prt(mpt, "bullet missed in timeout");
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
MPT_UNLOCK(mpt);
return;
}
mpt_prt(mpt, "bullet U-turned in timeout: got us");
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
}
mpt_prt(mpt,
"time out on request index = 0x%02x sequence = 0x%08x",
req->index, req->sequence);
mpt_check_doorbell(mpt);
mpt_prt(mpt, "Status %08x; Mask %08x; Doorbell %08x",
mpt_read(mpt, MPT_OFFSET_INTR_STATUS),
mpt_read(mpt, MPT_OFFSET_INTR_MASK),
mpt_read(mpt, MPT_OFFSET_DOORBELL) );
printf("request state %s\n", mpt_req_state(req->debug));
if (ccb != req->ccb) {
printf("time out: ccb %p != req->ccb %p\n",
ccb,req->ccb);
}
mpt_print_scsi_io_request((MSG_SCSI_IO_REQUEST *)req->req_vbuf);
req->debug = REQ_TIMEOUT;
req->ccb = NULL;
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
req->link.sle_next = (void *) mpt;
(void) timeout(mpttimeout2, (caddr_t)req, hz / 10);
ccb->ccb_h.status = CAM_CMD_TIMEOUT;
ccb->ccb_h.status |= CAM_RELEASE_SIMQ;
mpt->outofbeer = 0;
MPTLOCK_2_CAMLOCK(mpt);
xpt_done(ccb);
CAMLOCK_2_MPTLOCK(mpt);
MPT_UNLOCK(mpt);
}
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
static void
mpttimeout2(void *arg)
{
request_t *req = arg;
if (req->debug == REQ_TIMEOUT) {
mpt_softc_t *mpt = (mpt_softc_t *) req->link.sle_next;
MPT_LOCK(mpt);
mpt_free_request(mpt, req);
MPT_UNLOCK(mpt);
}
}
/*
* Callback routine from "bus_dmamap_load" or in simple case called directly.
*
* Takes a list of physical segments and builds the SGL for SCSI IO command
* and forwards the commard to the IOC after one last check that CAM has not
* aborted the transaction.
*/
static void
mpt_execute_req(void *arg, bus_dma_segment_t *dm_segs, int nseg, int error)
{
request_t *req;
union ccb *ccb;
mpt_softc_t *mpt;
MSG_SCSI_IO_REQUEST *mpt_req;
SGE_SIMPLE32 *se;
req = (request_t *)arg;
ccb = req->ccb;
mpt = ccb->ccb_h.ccb_mpt_ptr;
req = ccb->ccb_h.ccb_req_ptr;
mpt_req = req->req_vbuf;
if (error == 0 && nseg > MPT_SGL_MAX) {
error = EFBIG;
}
if (error != 0) {
if (error != EFBIG)
mpt_prt(mpt, "bus_dmamap_load returned %d", error);
if (ccb->ccb_h.status == CAM_REQ_INPROG) {
xpt_freeze_devq(ccb->ccb_h.path, 1);
ccb->ccb_h.status = CAM_DEV_QFRZN;
if (error == EFBIG)
ccb->ccb_h.status |= CAM_REQ_TOO_BIG;
else
ccb->ccb_h.status |= CAM_REQ_CMP_ERR;
}
ccb->ccb_h.status &= ~CAM_SIM_QUEUED;
xpt_done(ccb);
CAMLOCK_2_MPTLOCK(mpt);
mpt_free_request(mpt, req);
MPTLOCK_2_CAMLOCK(mpt);
return;
}
if (nseg > MPT_NSGL_FIRST(mpt)) {
int i, nleft = nseg;
u_int32_t flags;
bus_dmasync_op_t op;
SGE_CHAIN32 *ce;
mpt_req->DataLength = ccb->csio.dxfer_len;
flags = MPI_SGE_FLAGS_SIMPLE_ELEMENT;
if ((ccb->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_OUT)
flags |= MPI_SGE_FLAGS_HOST_TO_IOC;
se = (SGE_SIMPLE32 *) &mpt_req->SGL;
for (i = 0; i < MPT_NSGL_FIRST(mpt) - 1; i++, se++, dm_segs++) {
u_int32_t tf;
bzero(se, sizeof (*se));
se->Address = dm_segs->ds_addr;
MPI_pSGE_SET_LENGTH(se, dm_segs->ds_len);
tf = flags;
if (i == MPT_NSGL_FIRST(mpt) - 2) {
tf |= MPI_SGE_FLAGS_LAST_ELEMENT;
}
MPI_pSGE_SET_FLAGS(se, tf);
nleft -= 1;
}
/*
* Tell the IOC where to find the first chain element
*/
mpt_req->ChainOffset = ((char *)se - (char *)mpt_req) >> 2;
/*
* Until we're finished with all segments...
*/
while (nleft) {
int ntodo;
/*
* Construct the chain element that point to the
* next segment.
*/
ce = (SGE_CHAIN32 *) se++;
if (nleft > MPT_NSGL(mpt)) {
ntodo = MPT_NSGL(mpt) - 1;
ce->NextChainOffset = (MPT_RQSL(mpt) -
sizeof (SGE_SIMPLE32)) >> 2;
2004-08-25 17:54:19 +00:00
ce->Length = MPT_NSGL(mpt) *
sizeof (SGE_SIMPLE32);
} else {
ntodo = nleft;
ce->NextChainOffset = 0;
ce->Length = ntodo * sizeof (SGE_SIMPLE32);
}
ce->Address = req->req_pbuf +
((char *)se - (char *)mpt_req);
ce->Flags = MPI_SGE_FLAGS_CHAIN_ELEMENT;
for (i = 0; i < ntodo; i++, se++, dm_segs++) {
u_int32_t tf;
bzero(se, sizeof (*se));
se->Address = dm_segs->ds_addr;
MPI_pSGE_SET_LENGTH(se, dm_segs->ds_len);
tf = flags;
if (i == ntodo - 1) {
tf |= MPI_SGE_FLAGS_LAST_ELEMENT;
if (ce->NextChainOffset == 0) {
tf |=
MPI_SGE_FLAGS_END_OF_LIST |
MPI_SGE_FLAGS_END_OF_BUFFER;
}
}
MPI_pSGE_SET_FLAGS(se, tf);
nleft -= 1;
}
}
if ((ccb->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_IN)
op = BUS_DMASYNC_PREREAD;
else
op = BUS_DMASYNC_PREWRITE;
if (!(ccb->ccb_h.flags & (CAM_SG_LIST_PHYS|CAM_DATA_PHYS))) {
bus_dmamap_sync(mpt->buffer_dmat, req->dmap, op);
}
} else if (nseg > 0) {
int i;
u_int32_t flags;
bus_dmasync_op_t op;
mpt_req->DataLength = ccb->csio.dxfer_len;
flags = MPI_SGE_FLAGS_SIMPLE_ELEMENT;
if ((ccb->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_OUT)
flags |= MPI_SGE_FLAGS_HOST_TO_IOC;
/* Copy the segments into our SG list */
se = (SGE_SIMPLE32 *) &mpt_req->SGL;
for (i = 0; i < nseg; i++, se++, dm_segs++) {
u_int32_t tf;
bzero(se, sizeof (*se));
se->Address = dm_segs->ds_addr;
MPI_pSGE_SET_LENGTH(se, dm_segs->ds_len);
tf = flags;
if (i == nseg - 1) {
tf |=
MPI_SGE_FLAGS_LAST_ELEMENT |
MPI_SGE_FLAGS_END_OF_BUFFER |
MPI_SGE_FLAGS_END_OF_LIST;
}
MPI_pSGE_SET_FLAGS(se, tf);
}
if ((ccb->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_IN)
op = BUS_DMASYNC_PREREAD;
else
op = BUS_DMASYNC_PREWRITE;
if (!(ccb->ccb_h.flags & (CAM_SG_LIST_PHYS|CAM_DATA_PHYS))) {
bus_dmamap_sync(mpt->buffer_dmat, req->dmap, op);
}
} else {
se = (SGE_SIMPLE32 *) &mpt_req->SGL;
/*
* No data to transfer so we just make a single simple SGL
* with zero length.
*/
MPI_pSGE_SET_FLAGS(se,
(MPI_SGE_FLAGS_LAST_ELEMENT | MPI_SGE_FLAGS_END_OF_BUFFER |
MPI_SGE_FLAGS_SIMPLE_ELEMENT | MPI_SGE_FLAGS_END_OF_LIST));
}
/*
* Last time we need to check if this CCB needs to be aborted.
*/
if (ccb->ccb_h.status != CAM_REQ_INPROG) {
if (nseg && (ccb->ccb_h.flags & CAM_SG_LIST_PHYS) == 0)
bus_dmamap_unload(mpt->buffer_dmat, req->dmap);
CAMLOCK_2_MPTLOCK(mpt);
mpt_free_request(mpt, req);
MPTLOCK_2_CAMLOCK(mpt);
xpt_done(ccb);
return;
}
ccb->ccb_h.status |= CAM_SIM_QUEUED;
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
MPTLOCK_2_CAMLOCK(mpt);
if (ccb->ccb_h.timeout != CAM_TIME_INFINITY) {
ccb->ccb_h.timeout_ch =
timeout(mpttimeout, (caddr_t)ccb,
(ccb->ccb_h.timeout * hz) / 1000);
} else {
callout_handle_init(&ccb->ccb_h.timeout_ch);
}
if (mpt->verbose > 1)
mpt_print_scsi_io_request(mpt_req);
mpt_send_cmd(mpt, req);
MPTLOCK_2_CAMLOCK(mpt);
}
static void
mpt_start(union ccb *ccb)
{
request_t *req;
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
struct mpt_softc *mpt;
MSG_SCSI_IO_REQUEST *mpt_req;
struct ccb_scsiio *csio = &ccb->csio;
struct ccb_hdr *ccbh = &ccb->ccb_h;
/* Get the pointer for the physical addapter */
mpt = ccb->ccb_h.ccb_mpt_ptr;
CAMLOCK_2_MPTLOCK(mpt);
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
/* Get a request structure off the free list */
if ((req = mpt_get_request(mpt)) == NULL) {
if (mpt->outofbeer == 0) {
mpt->outofbeer = 1;
xpt_freeze_simq(mpt->sim, 1);
if (mpt->verbose > 1) {
mpt_prt(mpt, "FREEZEQ");
}
}
MPTLOCK_2_CAMLOCK(mpt);
ccb->ccb_h.status = CAM_REQUEUE_REQ;
xpt_done(ccb);
return;
}
MPTLOCK_2_CAMLOCK(mpt);
/* Link the ccb and the request structure so we can find */
/* the other knowing either the request or the ccb */
req->ccb = ccb;
ccb->ccb_h.ccb_req_ptr = req;
/* Now we build the command for the IOC */
mpt_req = req->req_vbuf;
bzero(mpt_req, sizeof *mpt_req);
mpt_req->Function = MPI_FUNCTION_SCSI_IO_REQUEST;
mpt_req->Bus = mpt->bus;
mpt_req->SenseBufferLength =
(csio->sense_len < MPT_SENSE_SIZE) ?
csio->sense_len : MPT_SENSE_SIZE;
/* We use the message context to find the request structure when we */
/* Get the command competion interrupt from the FC IOC. */
mpt_req->MsgContext = req->index;
/* Which physical device to do the I/O on */
mpt_req->TargetID = ccb->ccb_h.target_id;
mpt_req->LUN[1] = ccb->ccb_h.target_lun;
/* Set the direction of the transfer */
if ((ccb->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_IN)
mpt_req->Control = MPI_SCSIIO_CONTROL_READ;
else if ((ccb->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_OUT)
mpt_req->Control = MPI_SCSIIO_CONTROL_WRITE;
else
mpt_req->Control = MPI_SCSIIO_CONTROL_NODATATRANSFER;
if ((ccb->ccb_h.flags & CAM_TAG_ACTION_VALID) != 0) {
switch(ccb->csio.tag_action) {
case MSG_HEAD_OF_Q_TAG:
mpt_req->Control |= MPI_SCSIIO_CONTROL_HEADOFQ;
break;
case MSG_ACA_TASK:
mpt_req->Control |= MPI_SCSIIO_CONTROL_ACAQ;
break;
case MSG_ORDERED_Q_TAG:
mpt_req->Control |= MPI_SCSIIO_CONTROL_ORDEREDQ;
break;
case MSG_SIMPLE_Q_TAG:
default:
mpt_req->Control |= MPI_SCSIIO_CONTROL_SIMPLEQ;
break;
}
} else {
if (mpt->is_fc)
mpt_req->Control |= MPI_SCSIIO_CONTROL_SIMPLEQ;
else
mpt_req->Control |= MPI_SCSIIO_CONTROL_UNTAGGED;
}
if (mpt->is_fc == 0) {
if (ccb->ccb_h.flags & CAM_DIS_DISCONNECT) {
mpt_req->Control |= MPI_SCSIIO_CONTROL_NO_DISCONNECT;
}
}
/* Copy the scsi command block into place */
if ((ccb->ccb_h.flags & CAM_CDB_POINTER) != 0)
bcopy(csio->cdb_io.cdb_ptr, mpt_req->CDB, csio->cdb_len);
else
bcopy(csio->cdb_io.cdb_bytes, mpt_req->CDB, csio->cdb_len);
mpt_req->CDBLength = csio->cdb_len;
mpt_req->DataLength = csio->dxfer_len;
mpt_req->SenseBufferLowAddr = req->sense_pbuf;
/*
* If we have any data to send with this command,
* map it into bus space.
*/
if ((ccbh->flags & CAM_DIR_MASK) != CAM_DIR_NONE) {
if ((ccbh->flags & CAM_SCATTER_VALID) == 0) {
/*
* We've been given a pointer to a single buffer.
*/
if ((ccbh->flags & CAM_DATA_PHYS) == 0) {
/*
* Virtual address that needs to translated into
* one or more physical pages.
*/
int error;
error = bus_dmamap_load(mpt->buffer_dmat,
req->dmap, csio->data_ptr, csio->dxfer_len,
mpt_execute_req, req, 0);
if (error == EINPROGRESS) {
/*
* So as to maintain ordering,
* freeze the controller queue
* until our mapping is
* returned.
*/
xpt_freeze_simq(mpt->sim, 1);
ccbh->status |= CAM_RELEASE_SIMQ;
}
} else {
/*
* We have been given a pointer to single
* physical buffer.
*/
struct bus_dma_segment seg;
seg.ds_addr =
(bus_addr_t)(vm_offset_t)csio->data_ptr;
seg.ds_len = csio->dxfer_len;
mpt_execute_req(req, &seg, 1, 0);
}
} else {
/*
* We have been given a list of addresses.
* These case could be easily done but they are not
* currently generated by the CAM subsystem so there
* is no point in wasting the time right now.
*/
struct bus_dma_segment *segs;
if ((ccbh->flags & CAM_SG_LIST_PHYS) == 0) {
mpt_execute_req(req, NULL, 0, EFAULT);
} else {
/* Just use the segments provided */
segs = (struct bus_dma_segment *)csio->data_ptr;
mpt_execute_req(req, segs, csio->sglist_cnt,
(csio->sglist_cnt < MPT_SGL_MAX)?
0 : EFBIG);
}
}
} else {
mpt_execute_req(req, NULL, 0, 0);
}
}
static int
mpt_bus_reset(union ccb *ccb)
{
int error;
request_t *req;
mpt_softc_t *mpt;
MSG_SCSI_TASK_MGMT *reset_req;
/* Get the pointer for the physical adapter */
mpt = ccb->ccb_h.ccb_mpt_ptr;
/* Get a request structure off the free list */
if ((req = mpt_get_request(mpt)) == NULL) {
return (CAM_REQUEUE_REQ);
}
/* Link the ccb and the request structure so we can find */
/* the other knowing either the request or the ccb */
req->ccb = ccb;
ccb->ccb_h.ccb_req_ptr = req;
reset_req = req->req_vbuf;
bzero(reset_req, sizeof *reset_req);
reset_req->Function = MPI_FUNCTION_SCSI_TASK_MGMT;
reset_req->MsgContext = req->index;
reset_req->TaskType = MPI_SCSITASKMGMT_TASKTYPE_RESET_BUS;
if (mpt->is_fc) {
/*
* Should really be TARGET_RESET_OPTION
*/
reset_req->MsgFlags =
MPI_SCSITASKMGMT_MSGFLAGS_LIP_RESET_OPTION;
}
/* Which physical device Reset */
reset_req->TargetID = ccb->ccb_h.target_id;
reset_req->LUN[1] = ccb->ccb_h.target_lun;
ccb->ccb_h.status |= CAM_SIM_QUEUED;
error = mpt_send_handshake_cmd(mpt,
sizeof (MSG_SCSI_TASK_MGMT), reset_req);
if (error) {
mpt_prt(mpt,
"mpt_bus_reset: mpt_send_handshake return %d", error);
return (CAM_REQ_CMP_ERR);
} else {
return (CAM_REQ_CMP);
}
}
/*
* Process an asynchronous event from the IOC.
*/
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
static void mpt_ctlop(mpt_softc_t *, void *, u_int32_t);
static void mpt_event_notify_reply(mpt_softc_t *mpt, MSG_EVENT_NOTIFY_REPLY *);
static void
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
mpt_ctlop(mpt_softc_t *mpt, void *vmsg, u_int32_t reply)
{
MSG_DEFAULT_REPLY *dmsg = vmsg;
if (dmsg->Function == MPI_FUNCTION_EVENT_NOTIFICATION) {
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
mpt_event_notify_reply(mpt, vmsg);
mpt_free_reply(mpt, (reply << 1));
} else if (dmsg->Function == MPI_FUNCTION_EVENT_ACK) {
mpt_free_reply(mpt, (reply << 1));
} else if (dmsg->Function == MPI_FUNCTION_PORT_ENABLE) {
MSG_PORT_ENABLE_REPLY *msg = vmsg;
int index = msg->MsgContext & ~0x80000000;
if (mpt->verbose > 1) {
mpt_prt(mpt, "enable port reply idx %d", index);
}
if (index >= 0 && index < MPT_MAX_REQUESTS(mpt)) {
request_t *req = &mpt->request_pool[index];
req->debug = REQ_DONE;
}
mpt_free_reply(mpt, (reply << 1));
} else if (dmsg->Function == MPI_FUNCTION_CONFIG) {
MSG_CONFIG_REPLY *msg = vmsg;
int index = msg->MsgContext & ~0x80000000;
if (index >= 0 && index < MPT_MAX_REQUESTS(mpt)) {
request_t *req = &mpt->request_pool[index];
req->debug = REQ_DONE;
req->sequence = reply;
} else {
mpt_free_reply(mpt, (reply << 1));
}
} else {
mpt_prt(mpt, "unknown mpt_ctlop: %x", dmsg->Function);
}
}
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
static void
mpt_event_notify_reply(mpt_softc_t *mpt, MSG_EVENT_NOTIFY_REPLY *msg)
{
switch(msg->Event) {
case MPI_EVENT_LOG_DATA:
/* Some error occured that LSI wants logged */
printf("\tEvtLogData: IOCLogInfo: 0x%08x\n", msg->IOCLogInfo);
printf("\tEvtLogData: Event Data:");
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
{
int i;
for (i = 0; i < msg->EventDataLength; i++) {
printf(" %08x", msg->Data[i]);
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
}
}
printf("\n");
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
break;
case MPI_EVENT_UNIT_ATTENTION:
mpt_prt(mpt, "Bus: 0x%02x TargetID: 0x%02x",
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
(msg->Data[0] >> 8) & 0xff, msg->Data[0] & 0xff);
break;
case MPI_EVENT_IOC_BUS_RESET:
/* We generated a bus reset */
mpt_prt(mpt, "IOC Bus Reset Port: %d",
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
(msg->Data[0] >> 8) & 0xff);
break;
case MPI_EVENT_EXT_BUS_RESET:
/* Someone else generated a bus reset */
mpt_prt(mpt, "Ext Bus Reset");
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
/*
* These replies don't return EventData like the MPI
* spec says they do
*/
/* xpt_async(AC_BUS_RESET, path, NULL); */
break;
case MPI_EVENT_RESCAN:
/*
* In general this means a device has been added
* to the loop.
*/
mpt_prt(mpt, "Rescan Port: %d", (msg->Data[0] >> 8) & 0xff);
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
/* xpt_async(AC_FOUND_DEVICE, path, NULL); */
break;
case MPI_EVENT_LINK_STATUS_CHANGE:
mpt_prt(mpt, "Port %d: LinkState: %s",
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
(msg->Data[1] >> 8) & 0xff,
((msg->Data[0] & 0xff) == 0)? "Failed" : "Active");
break;
case MPI_EVENT_LOOP_STATE_CHANGE:
switch ((msg->Data[0] >> 16) & 0xff) {
case 0x01:
mpt_prt(mpt,
"Port 0x%x: FC LinkEvent: LIP(%02x,%02x) (Loop Initialization)\n",
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
(msg->Data[1] >> 8) & 0xff,
(msg->Data[0] >> 8) & 0xff,
(msg->Data[0] ) & 0xff);
switch ((msg->Data[0] >> 8) & 0xff) {
case 0xF7:
if ((msg->Data[0] & 0xff) == 0xF7) {
printf("Device needs AL_PA\n");
} else {
printf("Device %02x doesn't like FC performance\n",
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
msg->Data[0] & 0xFF);
}
break;
case 0xF8:
if ((msg->Data[0] & 0xff) == 0xF7) {
printf("Device had loop failure at its receiver prior to acquiring AL_PA\n");
} else {
printf("Device %02x detected loop failure at its receiver\n",
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
msg->Data[0] & 0xFF);
}
break;
default:
printf("Device %02x requests that device %02x reset itself\n",
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
msg->Data[0] & 0xFF,
(msg->Data[0] >> 8) & 0xFF);
break;
}
break;
case 0x02:
mpt_prt(mpt, "Port 0x%x: FC LinkEvent: LPE(%02x,%02x) (Loop Port Enable)",
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
(msg->Data[1] >> 8) & 0xff, /* Port */
(msg->Data[0] >> 8) & 0xff, /* Character 3 */
(msg->Data[0] ) & 0xff /* Character 4 */
);
break;
case 0x03:
mpt_prt(mpt, "Port 0x%x: FC LinkEvent: LPB(%02x,%02x) (Loop Port Bypass)",
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
(msg->Data[1] >> 8) & 0xff, /* Port */
(msg->Data[0] >> 8) & 0xff, /* Character 3 */
(msg->Data[0] ) & 0xff /* Character 4 */
);
break;
default:
mpt_prt(mpt, "Port 0x%x: FC LinkEvent: Unknown FC event (%02x %02x %02x)",
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
(msg->Data[1] >> 8) & 0xff, /* Port */
(msg->Data[0] >> 16) & 0xff, /* Event */
(msg->Data[0] >> 8) & 0xff, /* Character 3 */
(msg->Data[0] ) & 0xff /* Character 4 */
);
}
break;
case MPI_EVENT_LOGOUT:
mpt_prt(mpt, "FC Logout Port: %d N_PortID: %02x",
(msg->Data[1] >> 8) & 0xff, msg->Data[0]);
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
break;
case MPI_EVENT_EVENT_CHANGE:
/* This is just an acknowledgement of our
mpt_send_event_request */
break;
default:
mpt_prt(mpt, "Unknown event 0x%x\n", msg->Event);
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
}
if (msg->AckRequired) {
MSG_EVENT_ACK *ackp;
request_t *req;
if ((req = mpt_get_request(mpt)) == NULL) {
panic("unable to get request to acknowledge notify");
}
ackp = (MSG_EVENT_ACK *) req->req_vbuf;
bzero(ackp, sizeof *ackp);
ackp->Function = MPI_FUNCTION_EVENT_ACK;
ackp->Event = msg->Event;
ackp->EventContext = msg->EventContext;
ackp->MsgContext = req->index | 0x80000000;
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
mpt_check_doorbell(mpt);
mpt_send_cmd(mpt, req);
}
}
void
mpt_done(mpt_softc_t *mpt, u_int32_t reply)
{
int index;
request_t *req;
union ccb *ccb;
MSG_REQUEST_HEADER *mpt_req;
MSG_SCSI_IO_REPLY *mpt_reply;
index = -1; /* Shutup the complier */
if ((reply & MPT_CONTEXT_REPLY) == 0) {
/* context reply */
mpt_reply = NULL;
index = reply & MPT_CONTEXT_MASK;
} else {
unsigned *pReply;
bus_dmamap_sync(mpt->reply_dmat, mpt->reply_dmap,
BUS_DMASYNC_POSTREAD);
/* address reply (Error) */
mpt_reply = MPT_REPLY_PTOV(mpt, reply);
if (mpt->verbose > 1) {
pReply = (unsigned *) mpt_reply;
mpt_prt(mpt, "Address Reply (index %u)",
mpt_reply->MsgContext & 0xffff);
printf("%08x %08x %08x %08x\n",
pReply[0], pReply[1], pReply[2], pReply[3]);
printf("%08x %08x %08x %08x\n",
pReply[4], pReply[5], pReply[6], pReply[7]);
printf("%08x %08x %08x %08x\n\n",
pReply[8], pReply[9], pReply[10], pReply[11]);
}
index = mpt_reply->MsgContext;
}
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
/*
* Address reply with MessageContext high bit set
* This is most likely a notify message so we try
* to process it then free it
*/
if ((index & 0x80000000) != 0) {
if (mpt_reply != NULL) {
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
mpt_ctlop(mpt, mpt_reply, reply);
} else {
mpt_prt(mpt, "mpt_done: index 0x%x, NULL reply", index);
}
return;
}
/* Did we end up with a valid index into the table? */
if (index < 0 || index >= MPT_MAX_REQUESTS(mpt)) {
mpt_prt(mpt, "mpt_done: invalid index (%x) in reply", index);
return;
}
req = &mpt->request_pool[index];
/* Make sure memory hasn't been trashed */
if (req->index != index) {
printf("mpt_done: corrupted request struct");
return;
}
/* Short cut for task management replys; nothing more for us to do */
mpt_req = req->req_vbuf;
if (mpt_req->Function == MPI_FUNCTION_SCSI_TASK_MGMT) {
if (mpt->verbose > 1) {
mpt_prt(mpt, "mpt_done: TASK MGMT");
}
goto done;
}
if (mpt_req->Function == MPI_FUNCTION_PORT_ENABLE) {
goto done;
}
/*
* At this point it better be a SCSI IO command, but don't
* crash if it isn't
*/
if (mpt_req->Function != MPI_FUNCTION_SCSI_IO_REQUEST) {
goto done;
}
/* Recover the CAM control block from the request structure */
ccb = req->ccb;
/* Can't have had a SCSI command with out a CAM control block */
if (ccb == NULL || (ccb->ccb_h.status & CAM_SIM_QUEUED) == 0) {
mpt_prt(mpt,
"mpt_done: corrupted ccb, index = 0x%02x seq = 0x%08x",
req->index, req->sequence);
printf(" request state %s\nmpt_request:\n",
mpt_req_state(req->debug));
mpt_print_scsi_io_request((MSG_SCSI_IO_REQUEST *)req->req_vbuf);
if (mpt_reply != NULL) {
printf("\nmpt_done: reply:\n");
mpt_print_reply(MPT_REPLY_PTOV(mpt, reply));
} else {
printf("\nmpt_done: context reply: 0x%08x\n", reply);
}
goto done;
}
untimeout(mpttimeout, ccb, ccb->ccb_h.timeout_ch);
if ((ccb->ccb_h.flags & CAM_DIR_MASK) != CAM_DIR_NONE) {
bus_dmasync_op_t op;
if ((ccb->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_IN) {
op = BUS_DMASYNC_POSTREAD;
} else {
op = BUS_DMASYNC_POSTWRITE;
}
bus_dmamap_sync(mpt->buffer_dmat, req->dmap, op);
bus_dmamap_unload(mpt->buffer_dmat, req->dmap);
}
ccb->csio.resid = 0;
if (mpt_reply == NULL) {
/* Context reply; report that the command was successfull */
ccb->ccb_h.status = CAM_REQ_CMP;
ccb->csio.scsi_status = SCSI_STATUS_OK;
ccb->ccb_h.status &= ~CAM_SIM_QUEUED;
if (mpt->outofbeer) {
ccb->ccb_h.status |= CAM_RELEASE_SIMQ;
mpt->outofbeer = 0;
if (mpt->verbose > 1) {
mpt_prt(mpt, "THAWQ");
}
}
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
MPTLOCK_2_CAMLOCK(mpt);
xpt_done(ccb);
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
CAMLOCK_2_MPTLOCK(mpt);
goto done;
}
ccb->csio.scsi_status = mpt_reply->SCSIStatus;
switch(mpt_reply->IOCStatus) {
case MPI_IOCSTATUS_SCSI_DATA_OVERRUN:
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
ccb->ccb_h.status = CAM_DATA_RUN_ERR;
break;
case MPI_IOCSTATUS_SCSI_DATA_UNDERRUN:
/*
* Yikes, Tagged queue full comes through this path!
*
* So we'll change it to a status error and anything
* that returns status should probably be a status
* error as well.
*/
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
ccb->csio.resid =
ccb->csio.dxfer_len - mpt_reply->TransferCount;
if (mpt_reply->SCSIState & MPI_SCSI_STATE_NO_SCSI_STATUS) {
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
ccb->ccb_h.status = CAM_DATA_RUN_ERR;
break;
}
/* Fall through */
case MPI_IOCSTATUS_SUCCESS:
case MPI_IOCSTATUS_SCSI_RECOVERED_ERROR:
switch (ccb->csio.scsi_status) {
case SCSI_STATUS_OK:
ccb->ccb_h.status = CAM_REQ_CMP;
break;
default:
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
ccb->ccb_h.status = CAM_SCSI_STATUS_ERROR;
break;
}
break;
case MPI_IOCSTATUS_BUSY:
case MPI_IOCSTATUS_INSUFFICIENT_RESOURCES:
ccb->ccb_h.status = CAM_BUSY;
break;
case MPI_IOCSTATUS_SCSI_INVALID_BUS:
case MPI_IOCSTATUS_SCSI_INVALID_TARGETID:
case MPI_IOCSTATUS_SCSI_DEVICE_NOT_THERE:
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
ccb->ccb_h.status = CAM_DEV_NOT_THERE;
break;
case MPI_IOCSTATUS_SCSI_RESIDUAL_MISMATCH:
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
ccb->ccb_h.status = CAM_DATA_RUN_ERR;
break;
case MPI_IOCSTATUS_SCSI_PROTOCOL_ERROR:
case MPI_IOCSTATUS_SCSI_IO_DATA_ERROR:
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
ccb->ccb_h.status = CAM_UNCOR_PARITY;
break;
case MPI_IOCSTATUS_SCSI_TASK_TERMINATED:
ccb->ccb_h.status = CAM_REQ_CMP;
break;
case MPI_IOCSTATUS_SCSI_TASK_MGMT_FAILED:
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
ccb->ccb_h.status = CAM_UA_TERMIO;
break;
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
case MPI_IOCSTATUS_SCSI_IOC_TERMINATED:
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
ccb->ccb_h.status = CAM_REQ_TERMIO;
break;
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
case MPI_IOCSTATUS_SCSI_EXT_TERMINATED:
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
ccb->ccb_h.status = CAM_SCSI_BUS_RESET;
break;
default:
ccb->ccb_h.status = CAM_UNREC_HBA_ERROR;
break;
}
if ((mpt_reply->SCSIState & MPI_SCSI_STATE_AUTOSENSE_VALID) != 0) {
if (ccb->ccb_h.flags & (CAM_SENSE_PHYS | CAM_SENSE_PTR)) {
ccb->ccb_h.status |= CAM_AUTOSENSE_FAIL;
} else {
ccb->ccb_h.status |= CAM_AUTOSNS_VALID;
ccb->csio.sense_resid = mpt_reply->SenseCount;
bcopy(req->sense_vbuf, &ccb->csio.sense_data,
ccb->csio.sense_len);
}
} else if (mpt_reply->SCSIState & MPI_SCSI_STATE_AUTOSENSE_FAILED) {
ccb->ccb_h.status &= ~CAM_STATUS_MASK;
ccb->ccb_h.status |= CAM_AUTOSENSE_FAIL;
}
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
if ((ccb->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) {
if ((ccb->ccb_h.status & CAM_DEV_QFRZN) == 0) {
ccb->ccb_h.status |= CAM_DEV_QFRZN;
xpt_freeze_devq(ccb->ccb_h.path, 1);
}
}
ccb->ccb_h.status &= ~CAM_SIM_QUEUED;
if (mpt->outofbeer) {
ccb->ccb_h.status |= CAM_RELEASE_SIMQ;
mpt->outofbeer = 0;
if (mpt->verbose > 1) {
mpt_prt(mpt, "THAWQ");
}
}
MPTLOCK_2_CAMLOCK(mpt);
xpt_done(ccb);
CAMLOCK_2_MPTLOCK(mpt);
done:
/* If IOC done with this request free it up */
if (mpt_reply == NULL || (mpt_reply->MsgFlags & 0x80) == 0)
mpt_free_request(mpt, req);
/* If address reply; give the buffer back to the IOC */
if (mpt_reply != NULL)
mpt_free_reply(mpt, (reply << 1));
}
static void
mpt_action(struct cam_sim *sim, union ccb *ccb)
{
int tgt, error;
mpt_softc_t *mpt;
struct ccb_trans_settings *cts;
CAM_DEBUG(ccb->ccb_h.path, CAM_DEBUG_TRACE, ("mpt_action\n"));
mpt = (mpt_softc_t *)cam_sim_softc(sim);
ccb->ccb_h.ccb_mpt_ptr = mpt;
switch (ccb->ccb_h.func_code) {
case XPT_RESET_BUS:
if (mpt->verbose > 1)
mpt_prt(mpt, "XPT_RESET_BUS");
CAMLOCK_2_MPTLOCK(mpt);
error = mpt_bus_reset(ccb);
switch (error) {
case CAM_REQ_INPROG:
MPTLOCK_2_CAMLOCK(mpt);
break;
case CAM_REQUEUE_REQ:
if (mpt->outofbeer == 0) {
mpt->outofbeer = 1;
xpt_freeze_simq(sim, 1);
if (mpt->verbose > 1) {
mpt_prt(mpt, "FREEZEQ");
}
}
ccb->ccb_h.status = CAM_REQUEUE_REQ;
MPTLOCK_2_CAMLOCK(mpt);
xpt_done(ccb);
break;
case CAM_REQ_CMP:
ccb->ccb_h.status &= ~CAM_SIM_QUEUED;
ccb->ccb_h.status |= CAM_REQ_CMP;
if (mpt->outofbeer) {
ccb->ccb_h.status |= CAM_RELEASE_SIMQ;
mpt->outofbeer = 0;
if (mpt->verbose > 1) {
mpt_prt(mpt, "THAWQ");
}
}
MPTLOCK_2_CAMLOCK(mpt);
xpt_done(ccb);
break;
default:
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
MPTLOCK_2_CAMLOCK(mpt);
xpt_done(ccb);
}
break;
case XPT_SCSI_IO: /* Execute the requested I/O operation */
/*
* Do a couple of preliminary checks...
*/
if ((ccb->ccb_h.flags & CAM_CDB_POINTER) != 0) {
if ((ccb->ccb_h.flags & CAM_CDB_PHYS) != 0) {
ccb->ccb_h.status = CAM_REQ_INVALID;
xpt_done(ccb);
break;
}
}
/* Max supported CDB length is 16 bytes */
if (ccb->csio.cdb_len >
sizeof (((PTR_MSG_SCSI_IO_REQUEST)0)->CDB)) {
ccb->ccb_h.status = CAM_REQ_INVALID;
xpt_done(ccb);
return;
}
ccb->csio.scsi_status = SCSI_STATUS_OK;
mpt_start(ccb);
break;
case XPT_ABORT:
/*
* XXX: Need to implement
*/
ccb->ccb_h.status = CAM_UA_ABORT;
xpt_done(ccb);
break;
#ifdef CAM_NEW_TRAN_CODE
#define IS_CURRENT_SETTINGS(c) (c->type == CTS_TYPE_CURRENT_SETTINGS)
#else
#define IS_CURRENT_SETTINGS(c) (c->flags & CCB_TRANS_CURRENT_SETTINGS)
#endif
#define DP_DISC_ENABLE 0x1
#define DP_DISC_DISABL 0x2
#define DP_DISC (DP_DISC_ENABLE|DP_DISC_DISABL)
#define DP_TQING_ENABLE 0x4
#define DP_TQING_DISABL 0x8
#define DP_TQING (DP_TQING_ENABLE|DP_TQING_DISABL)
#define DP_WIDE 0x10
#define DP_NARROW 0x20
#define DP_WIDTH (DP_WIDE|DP_NARROW)
#define DP_SYNC 0x40
case XPT_SET_TRAN_SETTINGS: /* Nexus Settings */
cts = &ccb->cts;
if (!IS_CURRENT_SETTINGS(cts)) {
ccb->ccb_h.status = CAM_REQ_INVALID;
xpt_done(ccb);
break;
}
tgt = cts->ccb_h.target_id;
if (mpt->is_fc == 0) {
u_int8_t dval = 0;
u_int period = 0, offset = 0;
#ifndef CAM_NEW_TRAN_CODE
if (cts->valid & CCB_TRANS_DISC_VALID) {
dval |= DP_DISC_ENABLE;
}
if (cts->valid & CCB_TRANS_TQ_VALID) {
dval |= DP_TQING_ENABLE;
}
if (cts->valid & CCB_TRANS_BUS_WIDTH_VALID) {
if (cts->bus_width)
dval |= DP_WIDE;
else
dval |= DP_NARROW;
}
/*
* Any SYNC RATE of nonzero and SYNC_OFFSET
* of nonzero will cause us to go to the
* selected (from NVRAM) maximum value for
* this device. At a later point, we'll
* allow finer control.
*/
if ((cts->valid & CCB_TRANS_SYNC_RATE_VALID) &&
(cts->valid & CCB_TRANS_SYNC_OFFSET_VALID)) {
dval |= DP_SYNC;
period = cts->sync_period;
offset = cts->sync_offset;
}
#else
struct ccb_trans_settings_scsi *scsi =
&cts->proto_specific.scsi;
struct ccb_trans_settings_spi *spi =
&cts->xport_specific.spi;
if ((spi->valid & CTS_SPI_VALID_DISC) != 0) {
if ((spi->flags & CTS_SPI_FLAGS_DISC_ENB) != 0)
dval |= DP_DISC_ENABLE;
else
dval |= DP_DISC_DISABL;
}
if ((scsi->valid & CTS_SCSI_VALID_TQ) != 0) {
if ((scsi->flags & CTS_SCSI_FLAGS_TAG_ENB) != 0)
dval |= DP_TQING_ENABLE;
else
dval |= DP_TQING_DISABL;
}
if ((spi->valid & CTS_SPI_VALID_BUS_WIDTH) != 0) {
if (spi->bus_width == MSG_EXT_WDTR_BUS_16_BIT)
dval |= DP_WIDE;
else
dval |= DP_NARROW;
}
if ((spi->valid & CTS_SPI_VALID_SYNC_OFFSET) &&
(spi->valid & CTS_SPI_VALID_SYNC_RATE) &&
(spi->sync_period && spi->sync_offset)) {
dval |= DP_SYNC;
period = spi->sync_period;
offset = spi->sync_offset;
}
#endif
CAMLOCK_2_MPTLOCK(mpt);
if (dval & DP_DISC_ENABLE) {
mpt->mpt_disc_enable |= (1 << tgt);
} else if (dval & DP_DISC_DISABL) {
mpt->mpt_disc_enable &= ~(1 << tgt);
}
if (dval & DP_TQING_ENABLE) {
mpt->mpt_tag_enable |= (1 << tgt);
} else if (dval & DP_TQING_DISABL) {
mpt->mpt_tag_enable &= ~(1 << tgt);
}
if (dval & DP_WIDTH) {
if (mpt_setwidth(mpt, tgt, dval & DP_WIDE)) {
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
MPTLOCK_2_CAMLOCK(mpt);
xpt_done(ccb);
break;
}
}
if (dval & DP_SYNC) {
if (mpt_setsync(mpt, tgt, period, offset)) {
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
MPTLOCK_2_CAMLOCK(mpt);
xpt_done(ccb);
break;
}
}
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
MPTLOCK_2_CAMLOCK(mpt);
if (mpt->verbose > 1) {
mpt_prt(mpt,
"SET tgt %d flags %x period %x off %x",
tgt, dval, period, offset);
}
}
ccb->ccb_h.status = CAM_REQ_CMP;
xpt_done(ccb);
break;
case XPT_GET_TRAN_SETTINGS:
cts = &ccb->cts;
tgt = cts->ccb_h.target_id;
if (mpt->is_fc) {
#ifndef CAM_NEW_TRAN_CODE
/*
* a lot of normal SCSI things don't make sense.
*/
cts->flags = CCB_TRANS_TAG_ENB | CCB_TRANS_DISC_ENB;
cts->valid = CCB_TRANS_DISC_VALID | CCB_TRANS_TQ_VALID;
/*
* How do you measure the width of a high
* speed serial bus? Well, in bytes.
*
* Offset and period make no sense, though, so we set
* (above) a 'base' transfer speed to be gigabit.
*/
cts->bus_width = MSG_EXT_WDTR_BUS_8_BIT;
#else
struct ccb_trans_settings_fc *fc =
&cts->xport_specific.fc;
cts->protocol = PROTO_SCSI;
cts->protocol_version = SCSI_REV_2;
cts->transport = XPORT_FC;
cts->transport_version = 0;
fc->valid = CTS_FC_VALID_SPEED;
fc->bitrate = 100000; /* XXX: Need for 2Gb/s */
/* XXX: need a port database for each target */
#endif
} else {
#ifdef CAM_NEW_TRAN_CODE
struct ccb_trans_settings_scsi *scsi =
&cts->proto_specific.scsi;
struct ccb_trans_settings_spi *spi =
&cts->xport_specific.spi;
#endif
u_int8_t dval, pval, oval;
/*
* We aren't going off of Port PAGE2 params for
* tagged queuing or disconnect capabilities
* for current settings. For goal settings,
* we assert all capabilities- we've had some
* problems with reading NVRAM data.
*/
if (IS_CURRENT_SETTINGS(cts)) {
CONFIG_PAGE_SCSI_DEVICE_0 tmp;
dval = 0;
tmp = mpt->mpt_dev_page0[tgt];
CAMLOCK_2_MPTLOCK(mpt);
if (mpt_read_cfg_page(mpt, tgt, &tmp.Header)) {
mpt_prt(mpt,
"cannot get target %d DP0", tgt);
} else {
if (mpt->verbose > 1) {
mpt_prt(mpt,
"SPI Tgt %d Page 0: NParms %x Information %x",
tgt,
tmp.NegotiatedParameters,
tmp.Information);
}
}
MPTLOCK_2_CAMLOCK(mpt);
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
if (tmp.NegotiatedParameters &
MPI_SCSIDEVPAGE0_NP_WIDE)
dval |= DP_WIDE;
if (mpt->mpt_disc_enable & (1 << tgt)) {
dval |= DP_DISC_ENABLE;
}
if (mpt->mpt_tag_enable & (1 << tgt)) {
dval |= DP_TQING_ENABLE;
}
Wads more cleanup... In mpttimeout, call mpt_intr just on the offchance that we missed an interrupt. We can check to see whether or not the command that is timing out got completed. When we *do* decide to timeout a command, set the command state to REQ_TIMEOUT and then invoke another timeout (hz/10)- mpttimeout2. This allows us to catch a couple cases we've seen where the command we timed out on in fact is ready to be completed by the firmware. In any case, it's only after mpttimeout2 is called that we actually take down the private state and free the request itself. CAM has been notified in mpttimeout anyway. This whole area should be redone, but that will take 105% of my available game time for this month. Fix a couple of missing (and not useful, at presnet) CAMLOCK_2_MPTLOCK and MPTLOCK_2_CAMLOCK locations. Split mpt_notify into mpt_ctlop, which handles all reply completions that have 0x800000000 or'd into the ContextID. This function can, in fact, call mpt_event_notify_reply, which handles the traditional async event notifications. While we're at it, put in the extremely important (but currently untested) code that send back an Ack to an Event Notification (if the Event Notification is marked with AckRequired). Note that an Ack also generates another ctlop completion, tra la. Fix up mpt_done substantially to try and get how we plug into CAM correctly done. Remove bogus CAM_RELEASE_SIMQ settings. Do some cleanups in mpt_action that are related to speed negotiation for Ultra4 cards. This is an area that is still quite fragile and worrisome as config data being read back often doesn't make sense or jibe with the documentation. At any rate, after these changes were done, I was finally able to get Lars Eggert's dual 320M disk system to stay up under load all weekend- hopefully we're in good enough for now shape. MFC after: 1 week
2002-09-23 05:25:59 +00:00
oval = (tmp.NegotiatedParameters >> 16) & 0xff;
pval = (tmp.NegotiatedParameters >> 8) & 0xff;
} else {
/*
* XXX: Fix wrt NVRAM someday. Attempts
* XXX: to read port page2 device data
* XXX: just returns zero in these areas.
*/
dval = DP_WIDE|DP_DISC|DP_TQING;
oval = (mpt->mpt_port_page0.Capabilities >> 16);
pval = (mpt->mpt_port_page0.Capabilities >> 8);
}
#ifndef CAM_NEW_TRAN_CODE
cts->flags &= ~(CCB_TRANS_DISC_ENB|CCB_TRANS_TAG_ENB);
if (dval & DP_DISC_ENABLE) {
cts->flags |= CCB_TRANS_DISC_ENB;
}
if (dval & DP_TQING_ENABLE) {
cts->flags |= CCB_TRANS_TAG_ENB;
}
if (dval & DP_WIDE) {
cts->bus_width = MSG_EXT_WDTR_BUS_16_BIT;
} else {
cts->bus_width = MSG_EXT_WDTR_BUS_8_BIT;
}
cts->valid = CCB_TRANS_BUS_WIDTH_VALID |
CCB_TRANS_DISC_VALID | CCB_TRANS_TQ_VALID;
if (oval) {
cts->sync_period = pval;
cts->sync_offset = oval;
cts->valid |=
CCB_TRANS_SYNC_RATE_VALID |
CCB_TRANS_SYNC_OFFSET_VALID;
}
#else
cts->protocol = PROTO_SCSI;
cts->protocol_version = SCSI_REV_2;
cts->transport = XPORT_SPI;
cts->transport_version = 2;
scsi->flags &= ~CTS_SCSI_FLAGS_TAG_ENB;
spi->flags &= ~CTS_SPI_FLAGS_DISC_ENB;
if (dval & DP_DISC_ENABLE) {
spi->flags |= CTS_SPI_FLAGS_DISC_ENB;
}
if (dval & DP_TQING_ENABLE) {
scsi->flags |= CTS_SCSI_FLAGS_TAG_ENB;
}
if (oval && pval) {
spi->sync_offset = oval;
spi->sync_period = pval;
spi->valid |= CTS_SPI_VALID_SYNC_OFFSET;
spi->valid |= CTS_SPI_VALID_SYNC_RATE;
}
spi->valid |= CTS_SPI_VALID_BUS_WIDTH;
if (dval & DP_WIDE) {
spi->bus_width = MSG_EXT_WDTR_BUS_16_BIT;
} else {
spi->bus_width = MSG_EXT_WDTR_BUS_8_BIT;
}
if (cts->ccb_h.target_lun != CAM_LUN_WILDCARD) {
scsi->valid = CTS_SCSI_VALID_TQ;
spi->valid |= CTS_SPI_VALID_DISC;
} else {
scsi->valid = 0;
}
#endif
if (mpt->verbose > 1) {
mpt_prt(mpt,
"GET %s tgt %d flags %x period %x off %x",
IS_CURRENT_SETTINGS(cts)? "ACTIVE" :
"NVRAM", tgt, dval, pval, oval);
}
}
ccb->ccb_h.status = CAM_REQ_CMP;
xpt_done(ccb);
break;
case XPT_CALC_GEOMETRY:
{
struct ccb_calc_geometry *ccg;
ccg = &ccb->ccg;
if (ccg->block_size == 0) {
ccb->ccb_h.status = CAM_REQ_INVALID;
xpt_done(ccb);
break;
}
cam_calc_geometry(ccg, /*extended*/1);
xpt_done(ccb);
break;
}
case XPT_PATH_INQ: /* Path routing inquiry */
{
struct ccb_pathinq *cpi = &ccb->cpi;
cpi->version_num = 1;
cpi->target_sprt = 0;
cpi->hba_eng_cnt = 0;
cpi->max_lun = 7;
cpi->bus_id = cam_sim_bus(sim);
if (mpt->is_fc) {
cpi->max_target = 255;
cpi->hba_misc = PIM_NOBUSRESET;
cpi->initiator_id = cpi->max_target + 1;
cpi->base_transfer_speed = 100000;
cpi->hba_inquiry = PI_TAG_ABLE;
} else {
cpi->initiator_id = mpt->mpt_ini_id;
cpi->base_transfer_speed = 3300;
cpi->hba_inquiry = PI_SDTR_ABLE|PI_TAG_ABLE|PI_WIDE_16;
cpi->hba_misc = 0;
cpi->max_target = 15;
}
strncpy(cpi->sim_vid, "FreeBSD", SIM_IDLEN);
strncpy(cpi->hba_vid, "LSI", HBA_IDLEN);
strncpy(cpi->dev_name, cam_sim_name(sim), DEV_IDLEN);
cpi->unit_number = cam_sim_unit(sim);
cpi->ccb_h.status = CAM_REQ_CMP;
xpt_done(ccb);
break;
}
default:
ccb->ccb_h.status = CAM_REQ_INVALID;
xpt_done(ccb);
break;
}
}
static int
mpt_setwidth(mpt_softc_t *mpt, int tgt, int onoff)
{
CONFIG_PAGE_SCSI_DEVICE_1 tmp;
tmp = mpt->mpt_dev_page1[tgt];
if (onoff) {
tmp.RequestedParameters |= MPI_SCSIDEVPAGE1_RP_WIDE;
} else {
tmp.RequestedParameters &= ~MPI_SCSIDEVPAGE1_RP_WIDE;
}
if (mpt_write_cfg_page(mpt, tgt, &tmp.Header)) {
return (-1);
}
if (mpt_read_cfg_page(mpt, tgt, &tmp.Header)) {
return (-1);
}
mpt->mpt_dev_page1[tgt] = tmp;
if (mpt->verbose > 1) {
mpt_prt(mpt,
"SPI Target %d Page 1: RequestedParameters %x Config %x",
tgt, mpt->mpt_dev_page1[tgt].RequestedParameters,
mpt->mpt_dev_page1[tgt].Configuration);
}
return (0);
}
static int
mpt_setsync(mpt_softc_t *mpt, int tgt, int period, int offset)
{
CONFIG_PAGE_SCSI_DEVICE_1 tmp;
tmp = mpt->mpt_dev_page1[tgt];
tmp.RequestedParameters &=
~MPI_SCSIDEVPAGE1_RP_MIN_SYNC_PERIOD_MASK;
tmp.RequestedParameters &=
~MPI_SCSIDEVPAGE1_RP_MAX_SYNC_OFFSET_MASK;
tmp.RequestedParameters &=
~MPI_SCSIDEVPAGE1_RP_DT;
tmp.RequestedParameters &=
~MPI_SCSIDEVPAGE1_RP_QAS;
tmp.RequestedParameters &=
~MPI_SCSIDEVPAGE1_RP_IU;
/*
* XXX: For now, we're ignoring specific settings
*/
if (period && offset) {
int factor, offset, np;
factor = (mpt->mpt_port_page0.Capabilities >> 8) & 0xff;
offset = (mpt->mpt_port_page0.Capabilities >> 16) & 0xff;
np = 0;
if (factor < 0x9) {
np |= MPI_SCSIDEVPAGE1_RP_QAS;
np |= MPI_SCSIDEVPAGE1_RP_IU;
}
if (factor < 0xa) {
np |= MPI_SCSIDEVPAGE1_RP_DT;
}
np |= (factor << 8) | (offset << 16);
tmp.RequestedParameters |= np;
}
if (mpt_write_cfg_page(mpt, tgt, &tmp.Header)) {
return (-1);
}
if (mpt_read_cfg_page(mpt, tgt, &tmp.Header)) {
return (-1);
}
mpt->mpt_dev_page1[tgt] = tmp;
if (mpt->verbose > 1) {
mpt_prt(mpt,
"SPI Target %d Page 1: RParams %x Config %x",
tgt, mpt->mpt_dev_page1[tgt].RequestedParameters,
mpt->mpt_dev_page1[tgt].Configuration);
}
return (0);
}