linux/drivers/message/fusion/mptfc.c

1021 lines
27 KiB
C
Raw Normal View History

/*
* linux/drivers/message/fusion/mptfc.c
* For use with LSI Logic PCI chip/adapter(s)
* running LSI Logic Fusion MPT (Message Passing Technology) firmware.
*
* Copyright (c) 1999-2005 LSI Logic Corporation
* (mailto:mpt_linux_developer@lsil.com)
*
*/
/*=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=*/
/*
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; version 2 of the License.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
NO WARRANTY
THE PROGRAM IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OR
CONDITIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED INCLUDING, WITHOUT
LIMITATION, ANY WARRANTIES OR CONDITIONS OF TITLE, NON-INFRINGEMENT,
MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Each Recipient is
solely responsible for determining the appropriateness of using and
distributing the Program and assumes all risks associated with its
exercise of rights under this Agreement, including but not limited to
the risks and costs of program errors, damage to or loss of data,
programs or equipment, and unavailability or interruption of operations.
DISCLAIMER OF LIABILITY
NEITHER RECIPIENT NOR ANY CONTRIBUTORS SHALL HAVE ANY LIABILITY FOR ANY
DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING WITHOUT LIMITATION LOST PROFITS), 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 OR DISTRIBUTION OF THE PROGRAM OR THE EXERCISE OF ANY RIGHTS GRANTED
HEREUNDER, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGES
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
/*=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=*/
#include "linux_compat.h" /* linux-2.6 tweaks */
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/errno.h>
#include <linux/kdev_t.h>
#include <linux/blkdev.h>
#include <linux/delay.h> /* for mdelay */
#include <linux/interrupt.h> /* needed for in_interrupt() proto */
#include <linux/reboot.h> /* notifier code */
#include <linux/sched.h>
#include <linux/workqueue.h>
#include <linux/sort.h>
#include <scsi/scsi.h>
#include <scsi/scsi_cmnd.h>
#include <scsi/scsi_device.h>
#include <scsi/scsi_host.h>
#include <scsi/scsi_tcq.h>
#include <scsi/scsi_transport_fc.h>
#include "mptbase.h"
#include "mptscsih.h"
/*=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=*/
#define my_NAME "Fusion MPT FC Host driver"
#define my_VERSION MPT_LINUX_VERSION_COMMON
#define MYNAM "mptfc"
MODULE_AUTHOR(MODULEAUTHOR);
MODULE_DESCRIPTION(my_NAME);
MODULE_LICENSE("GPL");
/* Command line args */
static int mpt_pq_filter = 0;
module_param(mpt_pq_filter, int, 0);
MODULE_PARM_DESC(mpt_pq_filter, " Enable peripheral qualifier filter: enable=1 (default=0)");
#define MPTFC_DEV_LOSS_TMO (60)
static int mptfc_dev_loss_tmo = MPTFC_DEV_LOSS_TMO; /* reasonable default */
module_param(mptfc_dev_loss_tmo, int, 0);
MODULE_PARM_DESC(mptfc_dev_loss_tmo, " Initial time the driver programs the "
" transport to wait for an rport to "
" return following a device loss event."
" Default=60.");
static int mptfcDoneCtx = -1;
static int mptfcTaskCtx = -1;
static int mptfcInternalCtx = -1; /* Used only for internal commands */
static int mptfc_target_alloc(struct scsi_target *starget);
static int mptfc_slave_alloc(struct scsi_device *sdev);
static int mptfc_qcmd(struct scsi_cmnd *SCpnt,
void (*done)(struct scsi_cmnd *));
static void mptfc_target_destroy(struct scsi_target *starget);
static void mptfc_set_rport_loss_tmo(struct fc_rport *rport, uint32_t timeout);
static void __devexit mptfc_remove(struct pci_dev *pdev);
static struct scsi_host_template mptfc_driver_template = {
.module = THIS_MODULE,
.proc_name = "mptfc",
.proc_info = mptscsih_proc_info,
.name = "MPT FC Host",
.info = mptscsih_info,
.queuecommand = mptfc_qcmd,
.target_alloc = mptfc_target_alloc,
.slave_alloc = mptfc_slave_alloc,
.slave_configure = mptscsih_slave_configure,
.target_destroy = mptfc_target_destroy,
.slave_destroy = mptscsih_slave_destroy,
.change_queue_depth = mptscsih_change_queue_depth,
.eh_abort_handler = mptscsih_abort,
.eh_device_reset_handler = mptscsih_dev_reset,
.eh_bus_reset_handler = mptscsih_bus_reset,
.eh_host_reset_handler = mptscsih_host_reset,
.bios_param = mptscsih_bios_param,
.can_queue = MPT_FC_CAN_QUEUE,
.this_id = -1,
.sg_tablesize = MPT_SCSI_SG_DEPTH,
.max_sectors = 8192,
.cmd_per_lun = 7,
.use_clustering = ENABLE_CLUSTERING,
};
/****************************************************************************
* Supported hardware
*/
static struct pci_device_id mptfc_pci_table[] = {
{ PCI_VENDOR_ID_LSI_LOGIC, PCI_DEVICE_ID_LSI_FC909,
PCI_ANY_ID, PCI_ANY_ID },
{ PCI_VENDOR_ID_LSI_LOGIC, PCI_DEVICE_ID_LSI_FC919,
PCI_ANY_ID, PCI_ANY_ID },
{ PCI_VENDOR_ID_LSI_LOGIC, PCI_DEVICE_ID_LSI_FC929,
PCI_ANY_ID, PCI_ANY_ID },
{ PCI_VENDOR_ID_LSI_LOGIC, PCI_DEVICE_ID_LSI_FC919X,
PCI_ANY_ID, PCI_ANY_ID },
{ PCI_VENDOR_ID_LSI_LOGIC, PCI_DEVICE_ID_LSI_FC929X,
PCI_ANY_ID, PCI_ANY_ID },
{ PCI_VENDOR_ID_LSI_LOGIC, PCI_DEVICE_ID_LSI_FC939X,
PCI_ANY_ID, PCI_ANY_ID },
{ PCI_VENDOR_ID_LSI_LOGIC, PCI_DEVICE_ID_LSI_FC949X,
PCI_ANY_ID, PCI_ANY_ID },
{ PCI_VENDOR_ID_LSI_LOGIC, PCI_DEVICE_ID_LSI_FC949ES,
PCI_ANY_ID, PCI_ANY_ID },
{0} /* Terminating entry */
};
MODULE_DEVICE_TABLE(pci, mptfc_pci_table);
static struct scsi_transport_template *mptfc_transport_template = NULL;
static struct fc_function_template mptfc_transport_functions = {
.dd_fcrport_size = 8,
.show_host_node_name = 1,
.show_host_port_name = 1,
.show_host_supported_classes = 1,
.show_host_port_id = 1,
.show_rport_supported_classes = 1,
.show_starget_node_name = 1,
.show_starget_port_name = 1,
.show_starget_port_id = 1,
.set_rport_dev_loss_tmo = mptfc_set_rport_loss_tmo,
.show_rport_dev_loss_tmo = 1,
};
/* FIXME! values controlling firmware RESCAN event
* need to be set low to allow dev_loss_tmo to
* work as expected. Currently, firmware doesn't
* notify driver of RESCAN event until some number
* of seconds elapse. This value can be set via
* lsiutil.
*/
static void
mptfc_set_rport_loss_tmo(struct fc_rport *rport, uint32_t timeout)
{
if (timeout > 0)
rport->dev_loss_tmo = timeout;
else
rport->dev_loss_tmo = mptfc_dev_loss_tmo;
}
static int
mptfc_FcDevPage0_cmp_func(const void *a, const void *b)
{
FCDevicePage0_t **aa = (FCDevicePage0_t **)a;
FCDevicePage0_t **bb = (FCDevicePage0_t **)b;
if ((*aa)->CurrentBus == (*bb)->CurrentBus) {
if ((*aa)->CurrentTargetID == (*bb)->CurrentTargetID)
return 0;
if ((*aa)->CurrentTargetID < (*bb)->CurrentTargetID)
return -1;
return 1;
}
if ((*aa)->CurrentBus < (*bb)->CurrentBus)
return -1;
return 1;
}
static int
mptfc_GetFcDevPage0(MPT_ADAPTER *ioc, int ioc_port,
void(*func)(MPT_ADAPTER *ioc,int channel, FCDevicePage0_t *arg))
{
ConfigPageHeader_t hdr;
CONFIGPARMS cfg;
FCDevicePage0_t *ppage0_alloc, *fc;
dma_addr_t page0_dma;
int data_sz;
int ii;
FCDevicePage0_t *p0_array=NULL, *p_p0;
FCDevicePage0_t **pp0_array=NULL, **p_pp0;
int rc = -ENOMEM;
U32 port_id = 0xffffff;
int num_targ = 0;
int max_bus = ioc->facts.MaxBuses;
int max_targ = ioc->facts.MaxDevices;
if (max_bus == 0 || max_targ == 0)
goto out;
data_sz = sizeof(FCDevicePage0_t) * max_bus * max_targ;
p_p0 = p0_array = kzalloc(data_sz, GFP_KERNEL);
if (!p0_array)
goto out;
data_sz = sizeof(FCDevicePage0_t *) * max_bus * max_targ;
p_pp0 = pp0_array = kzalloc(data_sz, GFP_KERNEL);
if (!pp0_array)
goto out;
do {
/* Get FC Device Page 0 header */
hdr.PageVersion = 0;
hdr.PageLength = 0;
hdr.PageNumber = 0;
hdr.PageType = MPI_CONFIG_PAGETYPE_FC_DEVICE;
cfg.cfghdr.hdr = &hdr;
cfg.physAddr = -1;
cfg.action = MPI_CONFIG_ACTION_PAGE_HEADER;
cfg.dir = 0;
cfg.pageAddr = port_id;
cfg.timeout = 0;
if ((rc = mpt_config(ioc, &cfg)) != 0)
break;
if (hdr.PageLength <= 0)
break;
data_sz = hdr.PageLength * 4;
ppage0_alloc = pci_alloc_consistent(ioc->pcidev, data_sz,
&page0_dma);
rc = -ENOMEM;
if (!ppage0_alloc)
break;
cfg.physAddr = page0_dma;
cfg.action = MPI_CONFIG_ACTION_PAGE_READ_CURRENT;
if ((rc = mpt_config(ioc, &cfg)) == 0) {
ppage0_alloc->PortIdentifier =
le32_to_cpu(ppage0_alloc->PortIdentifier);
ppage0_alloc->WWNN.Low =
le32_to_cpu(ppage0_alloc->WWNN.Low);
ppage0_alloc->WWNN.High =
le32_to_cpu(ppage0_alloc->WWNN.High);
ppage0_alloc->WWPN.Low =
le32_to_cpu(ppage0_alloc->WWPN.Low);
ppage0_alloc->WWPN.High =
le32_to_cpu(ppage0_alloc->WWPN.High);
ppage0_alloc->BBCredit =
le16_to_cpu(ppage0_alloc->BBCredit);
ppage0_alloc->MaxRxFrameSize =
le16_to_cpu(ppage0_alloc->MaxRxFrameSize);
port_id = ppage0_alloc->PortIdentifier;
num_targ++;
*p_p0 = *ppage0_alloc; /* save data */
*p_pp0++ = p_p0++; /* save addr */
}
pci_free_consistent(ioc->pcidev, data_sz,
(u8 *) ppage0_alloc, page0_dma);
if (rc != 0)
break;
} while (port_id <= 0xff0000);
if (num_targ) {
/* sort array */
if (num_targ > 1)
sort (pp0_array, num_targ, sizeof(FCDevicePage0_t *),
mptfc_FcDevPage0_cmp_func, NULL);
/* call caller's func for each targ */
for (ii = 0; ii < num_targ; ii++) {
fc = *(pp0_array+ii);
func(ioc, ioc_port, fc);
}
}
out:
if (pp0_array)
kfree(pp0_array);
if (p0_array)
kfree(p0_array);
return rc;
}
static int
mptfc_generate_rport_ids(FCDevicePage0_t *pg0, struct fc_rport_identifiers *rid)
{
/* not currently usable */
if (pg0->Flags & (MPI_FC_DEVICE_PAGE0_FLAGS_PLOGI_INVALID |
MPI_FC_DEVICE_PAGE0_FLAGS_PRLI_INVALID))
return -1;
if (!(pg0->Flags & MPI_FC_DEVICE_PAGE0_FLAGS_TARGETID_BUS_VALID))
return -1;
if (!(pg0->Protocol & MPI_FC_DEVICE_PAGE0_PROT_FCP_TARGET))
return -1;
/*
* board data structure already normalized to platform endianness
* shifted to avoid unaligned access on 64 bit architecture
*/
rid->node_name = ((u64)pg0->WWNN.High) << 32 | (u64)pg0->WWNN.Low;
rid->port_name = ((u64)pg0->WWPN.High) << 32 | (u64)pg0->WWPN.Low;
rid->port_id = pg0->PortIdentifier;
rid->roles = FC_RPORT_ROLE_UNKNOWN;
rid->roles |= FC_RPORT_ROLE_FCP_TARGET;
if (pg0->Protocol & MPI_FC_DEVICE_PAGE0_PROT_FCP_INITIATOR)
rid->roles |= FC_RPORT_ROLE_FCP_INITIATOR;
return 0;
}
static void
mptfc_register_dev(MPT_ADAPTER *ioc, int channel, FCDevicePage0_t *pg0)
{
struct fc_rport_identifiers rport_ids;
struct fc_rport *rport;
struct mptfc_rport_info *ri;
int new_ri = 1;
[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations mptbase.h bump version number to 3.03.09 remove unneeded flags define workq and remove old fc specific locks mptbase.c initialize new lock and don't initialize two removed locks mptscsih.c when firmware reports target is no longer there, return DID_REQUEUE for fc hosts so that i/o doesn't get killed until the transport has an opportunity to manage the loss via its dev loss timer when the "eh_abort" routine is called, check to see if the driver has the command or not before looking to see if a reset is pending. James Smart and I talked about this and believe that the API for this routine is: if driver doesn't have command, return SUCCESS. This change helps prevent a target from being taken offline. SUCCESS is returned because it's likely that the command completed after error recovery timed it out but before it could be aborted. provide a routine to queue work to newly created workq, and use it. remove "ioc" from mptscsih_abort() it was only used one time. the other references were via hd->ioc, so I just moved it.... net change in references to ioc via hd->ioc is zero move hd->resetPending test and hd->timeouts increment to after the test for whether the command to be aborted remains known to the driver Make certain that the workq exists before queuing work to it. mptfc.c no longer need to lock rport data structures as I was able to single thread the code! I fixed up the debug code to eliminate compilation messages due to type mismatch in the printk. Got rid of some no longer needed rport flags. Initialize and destroy the workq used for the rescan work. simplify the logic regarding the increment of fc_rescan_work_count. use post increment and test for zero vs. pre increment and test for one; eliminate work_count variable: queue_work can be called with the work_lock held as it doesn't sleep Signed-off-by: Michael Reed <mdr@sgi.com> Signed-off-by: Eric Moore <Eric.Moore@lsil.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-04-22 06:14:35 +08:00
u64 pn, nn;
VirtTarget *vtarget;
if (mptfc_generate_rport_ids(pg0, &rport_ids) < 0)
return;
/* scan list looking for a match */
list_for_each_entry(ri, &ioc->fc_rports, list) {
pn = (u64)ri->pg0.WWPN.High << 32 | (u64)ri->pg0.WWPN.Low;
if (pn == rport_ids.port_name) { /* match */
list_move_tail(&ri->list, &ioc->fc_rports);
new_ri = 0;
break;
}
}
if (new_ri) { /* allocate one */
ri = kzalloc(sizeof(struct mptfc_rport_info), GFP_KERNEL);
if (!ri)
return;
list_add_tail(&ri->list, &ioc->fc_rports);
}
ri->pg0 = *pg0; /* add/update pg0 data */
ri->flags &= ~MPT_RPORT_INFO_FLAGS_MISSING;
/* MPT_RPORT_INFO_FLAGS_REGISTERED - rport not previously deleted */
if (!(ri->flags & MPT_RPORT_INFO_FLAGS_REGISTERED)) {
ri->flags |= MPT_RPORT_INFO_FLAGS_REGISTERED;
rport = fc_remote_port_add(ioc->sh, channel, &rport_ids);
if (rport) {
ri->rport = rport;
if (new_ri) /* may have been reset by user */
rport->dev_loss_tmo = mptfc_dev_loss_tmo;
/*
* if already mapped, remap here. If not mapped,
* target_alloc will allocate vtarget and map,
* slave_alloc will fill in vdev from vtarget.
*/
if (ri->starget) {
vtarget = ri->starget->hostdata;
if (vtarget) {
vtarget->target_id = pg0->CurrentTargetID;
vtarget->bus_id = pg0->CurrentBus;
}
}
[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations mptbase.h bump version number to 3.03.09 remove unneeded flags define workq and remove old fc specific locks mptbase.c initialize new lock and don't initialize two removed locks mptscsih.c when firmware reports target is no longer there, return DID_REQUEUE for fc hosts so that i/o doesn't get killed until the transport has an opportunity to manage the loss via its dev loss timer when the "eh_abort" routine is called, check to see if the driver has the command or not before looking to see if a reset is pending. James Smart and I talked about this and believe that the API for this routine is: if driver doesn't have command, return SUCCESS. This change helps prevent a target from being taken offline. SUCCESS is returned because it's likely that the command completed after error recovery timed it out but before it could be aborted. provide a routine to queue work to newly created workq, and use it. remove "ioc" from mptscsih_abort() it was only used one time. the other references were via hd->ioc, so I just moved it.... net change in references to ioc via hd->ioc is zero move hd->resetPending test and hd->timeouts increment to after the test for whether the command to be aborted remains known to the driver Make certain that the workq exists before queuing work to it. mptfc.c no longer need to lock rport data structures as I was able to single thread the code! I fixed up the debug code to eliminate compilation messages due to type mismatch in the printk. Got rid of some no longer needed rport flags. Initialize and destroy the workq used for the rescan work. simplify the logic regarding the increment of fc_rescan_work_count. use post increment and test for zero vs. pre increment and test for one; eliminate work_count variable: queue_work can be called with the work_lock held as it doesn't sleep Signed-off-by: Michael Reed <mdr@sgi.com> Signed-off-by: Eric Moore <Eric.Moore@lsil.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-04-22 06:14:35 +08:00
/* once dd_data is filled in, commands will issue to hardware */
*((struct mptfc_rport_info **)rport->dd_data) = ri;
pn = (u64)ri->pg0.WWPN.High << 32 | (u64)ri->pg0.WWPN.Low;
nn = (u64)ri->pg0.WWNN.High << 32 | (u64)ri->pg0.WWNN.Low;
dfcprintk ((MYIOC_s_INFO_FMT
"mptfc_reg_dev.%d: %x, %llx / %llx, tid %d, "
"rport tid %d, tmo %d\n",
ioc->name,
ioc->sh->host_no,
pg0->PortIdentifier,
[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations mptbase.h bump version number to 3.03.09 remove unneeded flags define workq and remove old fc specific locks mptbase.c initialize new lock and don't initialize two removed locks mptscsih.c when firmware reports target is no longer there, return DID_REQUEUE for fc hosts so that i/o doesn't get killed until the transport has an opportunity to manage the loss via its dev loss timer when the "eh_abort" routine is called, check to see if the driver has the command or not before looking to see if a reset is pending. James Smart and I talked about this and believe that the API for this routine is: if driver doesn't have command, return SUCCESS. This change helps prevent a target from being taken offline. SUCCESS is returned because it's likely that the command completed after error recovery timed it out but before it could be aborted. provide a routine to queue work to newly created workq, and use it. remove "ioc" from mptscsih_abort() it was only used one time. the other references were via hd->ioc, so I just moved it.... net change in references to ioc via hd->ioc is zero move hd->resetPending test and hd->timeouts increment to after the test for whether the command to be aborted remains known to the driver Make certain that the workq exists before queuing work to it. mptfc.c no longer need to lock rport data structures as I was able to single thread the code! I fixed up the debug code to eliminate compilation messages due to type mismatch in the printk. Got rid of some no longer needed rport flags. Initialize and destroy the workq used for the rescan work. simplify the logic regarding the increment of fc_rescan_work_count. use post increment and test for zero vs. pre increment and test for one; eliminate work_count variable: queue_work can be called with the work_lock held as it doesn't sleep Signed-off-by: Michael Reed <mdr@sgi.com> Signed-off-by: Eric Moore <Eric.Moore@lsil.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-04-22 06:14:35 +08:00
(unsigned long long)nn,
(unsigned long long)pn,
pg0->CurrentTargetID,
ri->rport->scsi_target_id,
ri->rport->dev_loss_tmo));
} else {
list_del(&ri->list);
kfree(ri);
ri = NULL;
}
}
}
/*
* OS entry point to allow for host driver to free allocated memory
* Called if no device present or device being unloaded
*/
static void
mptfc_target_destroy(struct scsi_target *starget)
{
struct fc_rport *rport;
struct mptfc_rport_info *ri;
rport = starget_to_rport(starget);
if (rport) {
ri = *((struct mptfc_rport_info **)rport->dd_data);
if (ri) /* better be! */
ri->starget = NULL;
}
if (starget->hostdata)
kfree(starget->hostdata);
starget->hostdata = NULL;
}
/*
* OS entry point to allow host driver to alloc memory
* for each scsi target. Called once per device the bus scan.
* Return non-zero if allocation fails.
*/
static int
mptfc_target_alloc(struct scsi_target *starget)
{
VirtTarget *vtarget;
struct fc_rport *rport;
struct mptfc_rport_info *ri;
int rc;
vtarget = kzalloc(sizeof(VirtTarget), GFP_KERNEL);
if (!vtarget)
return -ENOMEM;
starget->hostdata = vtarget;
rc = -ENODEV;
rport = starget_to_rport(starget);
if (rport) {
ri = *((struct mptfc_rport_info **)rport->dd_data);
if (ri) { /* better be! */
vtarget->target_id = ri->pg0.CurrentTargetID;
vtarget->bus_id = ri->pg0.CurrentBus;
ri->starget = starget;
rc = 0;
}
}
if (rc != 0) {
kfree(vtarget);
starget->hostdata = NULL;
}
return rc;
}
/*
* OS entry point to allow host driver to alloc memory
* for each scsi device. Called once per device the bus scan.
* Return non-zero if allocation fails.
* Init memory once per LUN.
*/
static int
mptfc_slave_alloc(struct scsi_device *sdev)
{
MPT_SCSI_HOST *hd;
VirtTarget *vtarget;
VirtDevice *vdev;
struct scsi_target *starget;
struct fc_rport *rport;
[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations mptbase.h bump version number to 3.03.09 remove unneeded flags define workq and remove old fc specific locks mptbase.c initialize new lock and don't initialize two removed locks mptscsih.c when firmware reports target is no longer there, return DID_REQUEUE for fc hosts so that i/o doesn't get killed until the transport has an opportunity to manage the loss via its dev loss timer when the "eh_abort" routine is called, check to see if the driver has the command or not before looking to see if a reset is pending. James Smart and I talked about this and believe that the API for this routine is: if driver doesn't have command, return SUCCESS. This change helps prevent a target from being taken offline. SUCCESS is returned because it's likely that the command completed after error recovery timed it out but before it could be aborted. provide a routine to queue work to newly created workq, and use it. remove "ioc" from mptscsih_abort() it was only used one time. the other references were via hd->ioc, so I just moved it.... net change in references to ioc via hd->ioc is zero move hd->resetPending test and hd->timeouts increment to after the test for whether the command to be aborted remains known to the driver Make certain that the workq exists before queuing work to it. mptfc.c no longer need to lock rport data structures as I was able to single thread the code! I fixed up the debug code to eliminate compilation messages due to type mismatch in the printk. Got rid of some no longer needed rport flags. Initialize and destroy the workq used for the rescan work. simplify the logic regarding the increment of fc_rescan_work_count. use post increment and test for zero vs. pre increment and test for one; eliminate work_count variable: queue_work can be called with the work_lock held as it doesn't sleep Signed-off-by: Michael Reed <mdr@sgi.com> Signed-off-by: Eric Moore <Eric.Moore@lsil.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-04-22 06:14:35 +08:00
starget = scsi_target(sdev);
rport = starget_to_rport(starget);
if (!rport || fc_remote_port_chkready(rport))
return -ENXIO;
hd = (MPT_SCSI_HOST *)sdev->host->hostdata;
vdev = kzalloc(sizeof(VirtDevice), GFP_KERNEL);
if (!vdev) {
printk(MYIOC_s_ERR_FMT "slave_alloc kmalloc(%zd) FAILED!\n",
hd->ioc->name, sizeof(VirtDevice));
return -ENOMEM;
}
sdev->hostdata = vdev;
vtarget = starget->hostdata;
if (vtarget->num_luns == 0) {
vtarget->ioc_id = hd->ioc->id;
vtarget->tflags = MPT_TARGET_FLAGS_Q_YES |
MPT_TARGET_FLAGS_VALID_INQUIRY;
hd->Targets[sdev->id] = vtarget;
}
vdev->vtarget = vtarget;
vdev->lun = sdev->lun;
vtarget->num_luns++;
[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations mptbase.h bump version number to 3.03.09 remove unneeded flags define workq and remove old fc specific locks mptbase.c initialize new lock and don't initialize two removed locks mptscsih.c when firmware reports target is no longer there, return DID_REQUEUE for fc hosts so that i/o doesn't get killed until the transport has an opportunity to manage the loss via its dev loss timer when the "eh_abort" routine is called, check to see if the driver has the command or not before looking to see if a reset is pending. James Smart and I talked about this and believe that the API for this routine is: if driver doesn't have command, return SUCCESS. This change helps prevent a target from being taken offline. SUCCESS is returned because it's likely that the command completed after error recovery timed it out but before it could be aborted. provide a routine to queue work to newly created workq, and use it. remove "ioc" from mptscsih_abort() it was only used one time. the other references were via hd->ioc, so I just moved it.... net change in references to ioc via hd->ioc is zero move hd->resetPending test and hd->timeouts increment to after the test for whether the command to be aborted remains known to the driver Make certain that the workq exists before queuing work to it. mptfc.c no longer need to lock rport data structures as I was able to single thread the code! I fixed up the debug code to eliminate compilation messages due to type mismatch in the printk. Got rid of some no longer needed rport flags. Initialize and destroy the workq used for the rescan work. simplify the logic regarding the increment of fc_rescan_work_count. use post increment and test for zero vs. pre increment and test for one; eliminate work_count variable: queue_work can be called with the work_lock held as it doesn't sleep Signed-off-by: Michael Reed <mdr@sgi.com> Signed-off-by: Eric Moore <Eric.Moore@lsil.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-04-22 06:14:35 +08:00
#ifdef DMPT_DEBUG_FC
[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations mptbase.h bump version number to 3.03.09 remove unneeded flags define workq and remove old fc specific locks mptbase.c initialize new lock and don't initialize two removed locks mptscsih.c when firmware reports target is no longer there, return DID_REQUEUE for fc hosts so that i/o doesn't get killed until the transport has an opportunity to manage the loss via its dev loss timer when the "eh_abort" routine is called, check to see if the driver has the command or not before looking to see if a reset is pending. James Smart and I talked about this and believe that the API for this routine is: if driver doesn't have command, return SUCCESS. This change helps prevent a target from being taken offline. SUCCESS is returned because it's likely that the command completed after error recovery timed it out but before it could be aborted. provide a routine to queue work to newly created workq, and use it. remove "ioc" from mptscsih_abort() it was only used one time. the other references were via hd->ioc, so I just moved it.... net change in references to ioc via hd->ioc is zero move hd->resetPending test and hd->timeouts increment to after the test for whether the command to be aborted remains known to the driver Make certain that the workq exists before queuing work to it. mptfc.c no longer need to lock rport data structures as I was able to single thread the code! I fixed up the debug code to eliminate compilation messages due to type mismatch in the printk. Got rid of some no longer needed rport flags. Initialize and destroy the workq used for the rescan work. simplify the logic regarding the increment of fc_rescan_work_count. use post increment and test for zero vs. pre increment and test for one; eliminate work_count variable: queue_work can be called with the work_lock held as it doesn't sleep Signed-off-by: Michael Reed <mdr@sgi.com> Signed-off-by: Eric Moore <Eric.Moore@lsil.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-04-22 06:14:35 +08:00
{
u64 nn, pn;
struct mptfc_rport_info *ri;
ri = *((struct mptfc_rport_info **)rport->dd_data);
[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations mptbase.h bump version number to 3.03.09 remove unneeded flags define workq and remove old fc specific locks mptbase.c initialize new lock and don't initialize two removed locks mptscsih.c when firmware reports target is no longer there, return DID_REQUEUE for fc hosts so that i/o doesn't get killed until the transport has an opportunity to manage the loss via its dev loss timer when the "eh_abort" routine is called, check to see if the driver has the command or not before looking to see if a reset is pending. James Smart and I talked about this and believe that the API for this routine is: if driver doesn't have command, return SUCCESS. This change helps prevent a target from being taken offline. SUCCESS is returned because it's likely that the command completed after error recovery timed it out but before it could be aborted. provide a routine to queue work to newly created workq, and use it. remove "ioc" from mptscsih_abort() it was only used one time. the other references were via hd->ioc, so I just moved it.... net change in references to ioc via hd->ioc is zero move hd->resetPending test and hd->timeouts increment to after the test for whether the command to be aborted remains known to the driver Make certain that the workq exists before queuing work to it. mptfc.c no longer need to lock rport data structures as I was able to single thread the code! I fixed up the debug code to eliminate compilation messages due to type mismatch in the printk. Got rid of some no longer needed rport flags. Initialize and destroy the workq used for the rescan work. simplify the logic regarding the increment of fc_rescan_work_count. use post increment and test for zero vs. pre increment and test for one; eliminate work_count variable: queue_work can be called with the work_lock held as it doesn't sleep Signed-off-by: Michael Reed <mdr@sgi.com> Signed-off-by: Eric Moore <Eric.Moore@lsil.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-04-22 06:14:35 +08:00
pn = (u64)ri->pg0.WWPN.High << 32 | (u64)ri->pg0.WWPN.Low;
nn = (u64)ri->pg0.WWNN.High << 32 | (u64)ri->pg0.WWNN.Low;
dfcprintk ((MYIOC_s_INFO_FMT
"mptfc_slv_alloc.%d: num_luns %d, sdev.id %d, "
"CurrentTargetID %d, %x %llx %llx\n",
hd->ioc->name,
sdev->host->host_no,
vtarget->num_luns,
sdev->id, ri->pg0.CurrentTargetID,
[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations mptbase.h bump version number to 3.03.09 remove unneeded flags define workq and remove old fc specific locks mptbase.c initialize new lock and don't initialize two removed locks mptscsih.c when firmware reports target is no longer there, return DID_REQUEUE for fc hosts so that i/o doesn't get killed until the transport has an opportunity to manage the loss via its dev loss timer when the "eh_abort" routine is called, check to see if the driver has the command or not before looking to see if a reset is pending. James Smart and I talked about this and believe that the API for this routine is: if driver doesn't have command, return SUCCESS. This change helps prevent a target from being taken offline. SUCCESS is returned because it's likely that the command completed after error recovery timed it out but before it could be aborted. provide a routine to queue work to newly created workq, and use it. remove "ioc" from mptscsih_abort() it was only used one time. the other references were via hd->ioc, so I just moved it.... net change in references to ioc via hd->ioc is zero move hd->resetPending test and hd->timeouts increment to after the test for whether the command to be aborted remains known to the driver Make certain that the workq exists before queuing work to it. mptfc.c no longer need to lock rport data structures as I was able to single thread the code! I fixed up the debug code to eliminate compilation messages due to type mismatch in the printk. Got rid of some no longer needed rport flags. Initialize and destroy the workq used for the rescan work. simplify the logic regarding the increment of fc_rescan_work_count. use post increment and test for zero vs. pre increment and test for one; eliminate work_count variable: queue_work can be called with the work_lock held as it doesn't sleep Signed-off-by: Michael Reed <mdr@sgi.com> Signed-off-by: Eric Moore <Eric.Moore@lsil.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-04-22 06:14:35 +08:00
ri->pg0.PortIdentifier,
(unsigned long long)pn,
(unsigned long long)nn));
}
#endif
return 0;
}
static int
mptfc_qcmd(struct scsi_cmnd *SCpnt, void (*done)(struct scsi_cmnd *))
{
struct mptfc_rport_info *ri;
struct fc_rport *rport = starget_to_rport(scsi_target(SCpnt->device));
int err;
err = fc_remote_port_chkready(rport);
if (unlikely(err)) {
SCpnt->result = err;
done(SCpnt);
return 0;
}
[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations mptbase.h bump version number to 3.03.09 remove unneeded flags define workq and remove old fc specific locks mptbase.c initialize new lock and don't initialize two removed locks mptscsih.c when firmware reports target is no longer there, return DID_REQUEUE for fc hosts so that i/o doesn't get killed until the transport has an opportunity to manage the loss via its dev loss timer when the "eh_abort" routine is called, check to see if the driver has the command or not before looking to see if a reset is pending. James Smart and I talked about this and believe that the API for this routine is: if driver doesn't have command, return SUCCESS. This change helps prevent a target from being taken offline. SUCCESS is returned because it's likely that the command completed after error recovery timed it out but before it could be aborted. provide a routine to queue work to newly created workq, and use it. remove "ioc" from mptscsih_abort() it was only used one time. the other references were via hd->ioc, so I just moved it.... net change in references to ioc via hd->ioc is zero move hd->resetPending test and hd->timeouts increment to after the test for whether the command to be aborted remains known to the driver Make certain that the workq exists before queuing work to it. mptfc.c no longer need to lock rport data structures as I was able to single thread the code! I fixed up the debug code to eliminate compilation messages due to type mismatch in the printk. Got rid of some no longer needed rport flags. Initialize and destroy the workq used for the rescan work. simplify the logic regarding the increment of fc_rescan_work_count. use post increment and test for zero vs. pre increment and test for one; eliminate work_count variable: queue_work can be called with the work_lock held as it doesn't sleep Signed-off-by: Michael Reed <mdr@sgi.com> Signed-off-by: Eric Moore <Eric.Moore@lsil.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-04-22 06:14:35 +08:00
/* dd_data is null until finished adding target */
ri = *((struct mptfc_rport_info **)rport->dd_data);
[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations mptbase.h bump version number to 3.03.09 remove unneeded flags define workq and remove old fc specific locks mptbase.c initialize new lock and don't initialize two removed locks mptscsih.c when firmware reports target is no longer there, return DID_REQUEUE for fc hosts so that i/o doesn't get killed until the transport has an opportunity to manage the loss via its dev loss timer when the "eh_abort" routine is called, check to see if the driver has the command or not before looking to see if a reset is pending. James Smart and I talked about this and believe that the API for this routine is: if driver doesn't have command, return SUCCESS. This change helps prevent a target from being taken offline. SUCCESS is returned because it's likely that the command completed after error recovery timed it out but before it could be aborted. provide a routine to queue work to newly created workq, and use it. remove "ioc" from mptscsih_abort() it was only used one time. the other references were via hd->ioc, so I just moved it.... net change in references to ioc via hd->ioc is zero move hd->resetPending test and hd->timeouts increment to after the test for whether the command to be aborted remains known to the driver Make certain that the workq exists before queuing work to it. mptfc.c no longer need to lock rport data structures as I was able to single thread the code! I fixed up the debug code to eliminate compilation messages due to type mismatch in the printk. Got rid of some no longer needed rport flags. Initialize and destroy the workq used for the rescan work. simplify the logic regarding the increment of fc_rescan_work_count. use post increment and test for zero vs. pre increment and test for one; eliminate work_count variable: queue_work can be called with the work_lock held as it doesn't sleep Signed-off-by: Michael Reed <mdr@sgi.com> Signed-off-by: Eric Moore <Eric.Moore@lsil.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-04-22 06:14:35 +08:00
if (unlikely(!ri)) {
dfcprintk ((MYIOC_s_INFO_FMT
"mptfc_qcmd.%d: %d:%d, dd_data is null.\n",
((MPT_SCSI_HOST *) SCpnt->device->host->hostdata)->ioc->name,
((MPT_SCSI_HOST *) SCpnt->device->host->hostdata)->ioc->sh->host_no,
SCpnt->device->id,SCpnt->device->lun));
SCpnt->result = DID_IMM_RETRY << 16;
done(SCpnt);
return 0;
}
[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations mptbase.h bump version number to 3.03.09 remove unneeded flags define workq and remove old fc specific locks mptbase.c initialize new lock and don't initialize two removed locks mptscsih.c when firmware reports target is no longer there, return DID_REQUEUE for fc hosts so that i/o doesn't get killed until the transport has an opportunity to manage the loss via its dev loss timer when the "eh_abort" routine is called, check to see if the driver has the command or not before looking to see if a reset is pending. James Smart and I talked about this and believe that the API for this routine is: if driver doesn't have command, return SUCCESS. This change helps prevent a target from being taken offline. SUCCESS is returned because it's likely that the command completed after error recovery timed it out but before it could be aborted. provide a routine to queue work to newly created workq, and use it. remove "ioc" from mptscsih_abort() it was only used one time. the other references were via hd->ioc, so I just moved it.... net change in references to ioc via hd->ioc is zero move hd->resetPending test and hd->timeouts increment to after the test for whether the command to be aborted remains known to the driver Make certain that the workq exists before queuing work to it. mptfc.c no longer need to lock rport data structures as I was able to single thread the code! I fixed up the debug code to eliminate compilation messages due to type mismatch in the printk. Got rid of some no longer needed rport flags. Initialize and destroy the workq used for the rescan work. simplify the logic regarding the increment of fc_rescan_work_count. use post increment and test for zero vs. pre increment and test for one; eliminate work_count variable: queue_work can be called with the work_lock held as it doesn't sleep Signed-off-by: Michael Reed <mdr@sgi.com> Signed-off-by: Eric Moore <Eric.Moore@lsil.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-04-22 06:14:35 +08:00
err = mptscsih_qcmd(SCpnt,done);
#ifdef DMPT_DEBUG_FC
if (unlikely(err)) {
dfcprintk ((MYIOC_s_INFO_FMT
"mptfc_qcmd.%d: %d:%d, mptscsih_qcmd returns non-zero.\n",
((MPT_SCSI_HOST *) SCpnt->device->host->hostdata)->ioc->name,
((MPT_SCSI_HOST *) SCpnt->device->host->hostdata)->ioc->sh->host_no,
SCpnt->device->id,SCpnt->device->lun));
}
#endif
return err;
}
static void
mptfc_init_host_attr(MPT_ADAPTER *ioc,int portnum)
{
unsigned class = 0, cos = 0;
/* don't know what to do as only one scsi (fc) host was allocated */
if (portnum != 0)
return;
class = ioc->fc_port_page0[portnum].SupportedServiceClass;
if (class & MPI_FCPORTPAGE0_SUPPORT_CLASS_1)
cos |= FC_COS_CLASS1;
if (class & MPI_FCPORTPAGE0_SUPPORT_CLASS_2)
cos |= FC_COS_CLASS2;
if (class & MPI_FCPORTPAGE0_SUPPORT_CLASS_3)
cos |= FC_COS_CLASS3;
fc_host_node_name(ioc->sh) =
(u64)ioc->fc_port_page0[portnum].WWNN.High << 32
| (u64)ioc->fc_port_page0[portnum].WWNN.Low;
fc_host_port_name(ioc->sh) =
(u64)ioc->fc_port_page0[portnum].WWPN.High << 32
| (u64)ioc->fc_port_page0[portnum].WWPN.Low;
fc_host_port_id(ioc->sh) = ioc->fc_port_page0[portnum].PortIdentifier;
fc_host_supported_classes(ioc->sh) = cos;
fc_host_tgtid_bind_type(ioc->sh) = FC_TGTID_BIND_BY_WWPN;
}
static void
mptfc_rescan_devices(void *arg)
{
MPT_ADAPTER *ioc = (MPT_ADAPTER *)arg;
int ii;
int work_to_do;
[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations mptbase.h bump version number to 3.03.09 remove unneeded flags define workq and remove old fc specific locks mptbase.c initialize new lock and don't initialize two removed locks mptscsih.c when firmware reports target is no longer there, return DID_REQUEUE for fc hosts so that i/o doesn't get killed until the transport has an opportunity to manage the loss via its dev loss timer when the "eh_abort" routine is called, check to see if the driver has the command or not before looking to see if a reset is pending. James Smart and I talked about this and believe that the API for this routine is: if driver doesn't have command, return SUCCESS. This change helps prevent a target from being taken offline. SUCCESS is returned because it's likely that the command completed after error recovery timed it out but before it could be aborted. provide a routine to queue work to newly created workq, and use it. remove "ioc" from mptscsih_abort() it was only used one time. the other references were via hd->ioc, so I just moved it.... net change in references to ioc via hd->ioc is zero move hd->resetPending test and hd->timeouts increment to after the test for whether the command to be aborted remains known to the driver Make certain that the workq exists before queuing work to it. mptfc.c no longer need to lock rport data structures as I was able to single thread the code! I fixed up the debug code to eliminate compilation messages due to type mismatch in the printk. Got rid of some no longer needed rport flags. Initialize and destroy the workq used for the rescan work. simplify the logic regarding the increment of fc_rescan_work_count. use post increment and test for zero vs. pre increment and test for one; eliminate work_count variable: queue_work can be called with the work_lock held as it doesn't sleep Signed-off-by: Michael Reed <mdr@sgi.com> Signed-off-by: Eric Moore <Eric.Moore@lsil.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-04-22 06:14:35 +08:00
u64 pn;
unsigned long flags;
struct mptfc_rport_info *ri;
do {
/* start by tagging all ports as missing */
list_for_each_entry(ri, &ioc->fc_rports, list) {
if (ri->flags & MPT_RPORT_INFO_FLAGS_REGISTERED) {
ri->flags |= MPT_RPORT_INFO_FLAGS_MISSING;
}
}
/*
* now rescan devices known to adapter,
* will reregister existing rports
*/
for (ii=0; ii < ioc->facts.NumberOfPorts; ii++) {
(void) mptbase_GetFcPortPage0(ioc, ii);
mptfc_init_host_attr(ioc,ii); /* refresh */
mptfc_GetFcDevPage0(ioc,ii,mptfc_register_dev);
}
/* delete devices still missing */
list_for_each_entry(ri, &ioc->fc_rports, list) {
/* if newly missing, delete it */
[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations mptbase.h bump version number to 3.03.09 remove unneeded flags define workq and remove old fc specific locks mptbase.c initialize new lock and don't initialize two removed locks mptscsih.c when firmware reports target is no longer there, return DID_REQUEUE for fc hosts so that i/o doesn't get killed until the transport has an opportunity to manage the loss via its dev loss timer when the "eh_abort" routine is called, check to see if the driver has the command or not before looking to see if a reset is pending. James Smart and I talked about this and believe that the API for this routine is: if driver doesn't have command, return SUCCESS. This change helps prevent a target from being taken offline. SUCCESS is returned because it's likely that the command completed after error recovery timed it out but before it could be aborted. provide a routine to queue work to newly created workq, and use it. remove "ioc" from mptscsih_abort() it was only used one time. the other references were via hd->ioc, so I just moved it.... net change in references to ioc via hd->ioc is zero move hd->resetPending test and hd->timeouts increment to after the test for whether the command to be aborted remains known to the driver Make certain that the workq exists before queuing work to it. mptfc.c no longer need to lock rport data structures as I was able to single thread the code! I fixed up the debug code to eliminate compilation messages due to type mismatch in the printk. Got rid of some no longer needed rport flags. Initialize and destroy the workq used for the rescan work. simplify the logic regarding the increment of fc_rescan_work_count. use post increment and test for zero vs. pre increment and test for one; eliminate work_count variable: queue_work can be called with the work_lock held as it doesn't sleep Signed-off-by: Michael Reed <mdr@sgi.com> Signed-off-by: Eric Moore <Eric.Moore@lsil.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-04-22 06:14:35 +08:00
if (ri->flags & MPT_RPORT_INFO_FLAGS_MISSING) {
ri->flags &= ~(MPT_RPORT_INFO_FLAGS_REGISTERED|
MPT_RPORT_INFO_FLAGS_MISSING);
[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations mptbase.h bump version number to 3.03.09 remove unneeded flags define workq and remove old fc specific locks mptbase.c initialize new lock and don't initialize two removed locks mptscsih.c when firmware reports target is no longer there, return DID_REQUEUE for fc hosts so that i/o doesn't get killed until the transport has an opportunity to manage the loss via its dev loss timer when the "eh_abort" routine is called, check to see if the driver has the command or not before looking to see if a reset is pending. James Smart and I talked about this and believe that the API for this routine is: if driver doesn't have command, return SUCCESS. This change helps prevent a target from being taken offline. SUCCESS is returned because it's likely that the command completed after error recovery timed it out but before it could be aborted. provide a routine to queue work to newly created workq, and use it. remove "ioc" from mptscsih_abort() it was only used one time. the other references were via hd->ioc, so I just moved it.... net change in references to ioc via hd->ioc is zero move hd->resetPending test and hd->timeouts increment to after the test for whether the command to be aborted remains known to the driver Make certain that the workq exists before queuing work to it. mptfc.c no longer need to lock rport data structures as I was able to single thread the code! I fixed up the debug code to eliminate compilation messages due to type mismatch in the printk. Got rid of some no longer needed rport flags. Initialize and destroy the workq used for the rescan work. simplify the logic regarding the increment of fc_rescan_work_count. use post increment and test for zero vs. pre increment and test for one; eliminate work_count variable: queue_work can be called with the work_lock held as it doesn't sleep Signed-off-by: Michael Reed <mdr@sgi.com> Signed-off-by: Eric Moore <Eric.Moore@lsil.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-04-22 06:14:35 +08:00
fc_remote_port_delete(ri->rport); /* won't sleep */
ri->rport = NULL;
[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations mptbase.h bump version number to 3.03.09 remove unneeded flags define workq and remove old fc specific locks mptbase.c initialize new lock and don't initialize two removed locks mptscsih.c when firmware reports target is no longer there, return DID_REQUEUE for fc hosts so that i/o doesn't get killed until the transport has an opportunity to manage the loss via its dev loss timer when the "eh_abort" routine is called, check to see if the driver has the command or not before looking to see if a reset is pending. James Smart and I talked about this and believe that the API for this routine is: if driver doesn't have command, return SUCCESS. This change helps prevent a target from being taken offline. SUCCESS is returned because it's likely that the command completed after error recovery timed it out but before it could be aborted. provide a routine to queue work to newly created workq, and use it. remove "ioc" from mptscsih_abort() it was only used one time. the other references were via hd->ioc, so I just moved it.... net change in references to ioc via hd->ioc is zero move hd->resetPending test and hd->timeouts increment to after the test for whether the command to be aborted remains known to the driver Make certain that the workq exists before queuing work to it. mptfc.c no longer need to lock rport data structures as I was able to single thread the code! I fixed up the debug code to eliminate compilation messages due to type mismatch in the printk. Got rid of some no longer needed rport flags. Initialize and destroy the workq used for the rescan work. simplify the logic regarding the increment of fc_rescan_work_count. use post increment and test for zero vs. pre increment and test for one; eliminate work_count variable: queue_work can be called with the work_lock held as it doesn't sleep Signed-off-by: Michael Reed <mdr@sgi.com> Signed-off-by: Eric Moore <Eric.Moore@lsil.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-04-22 06:14:35 +08:00
pn = (u64)ri->pg0.WWPN.High << 32 |
(u64)ri->pg0.WWPN.Low;
dfcprintk ((MYIOC_s_INFO_FMT
"mptfc_rescan.%d: %llx deleted\n",
ioc->name,
ioc->sh->host_no,
[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations mptbase.h bump version number to 3.03.09 remove unneeded flags define workq and remove old fc specific locks mptbase.c initialize new lock and don't initialize two removed locks mptscsih.c when firmware reports target is no longer there, return DID_REQUEUE for fc hosts so that i/o doesn't get killed until the transport has an opportunity to manage the loss via its dev loss timer when the "eh_abort" routine is called, check to see if the driver has the command or not before looking to see if a reset is pending. James Smart and I talked about this and believe that the API for this routine is: if driver doesn't have command, return SUCCESS. This change helps prevent a target from being taken offline. SUCCESS is returned because it's likely that the command completed after error recovery timed it out but before it could be aborted. provide a routine to queue work to newly created workq, and use it. remove "ioc" from mptscsih_abort() it was only used one time. the other references were via hd->ioc, so I just moved it.... net change in references to ioc via hd->ioc is zero move hd->resetPending test and hd->timeouts increment to after the test for whether the command to be aborted remains known to the driver Make certain that the workq exists before queuing work to it. mptfc.c no longer need to lock rport data structures as I was able to single thread the code! I fixed up the debug code to eliminate compilation messages due to type mismatch in the printk. Got rid of some no longer needed rport flags. Initialize and destroy the workq used for the rescan work. simplify the logic regarding the increment of fc_rescan_work_count. use post increment and test for zero vs. pre increment and test for one; eliminate work_count variable: queue_work can be called with the work_lock held as it doesn't sleep Signed-off-by: Michael Reed <mdr@sgi.com> Signed-off-by: Eric Moore <Eric.Moore@lsil.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-04-22 06:14:35 +08:00
(unsigned long long)pn));
}
}
/*
* allow multiple passes as target state
* might have changed during scan
*/
spin_lock_irqsave(&ioc->fc_rescan_work_lock, flags);
if (ioc->fc_rescan_work_count > 2) /* only need one more */
ioc->fc_rescan_work_count = 2;
work_to_do = --ioc->fc_rescan_work_count;
spin_unlock_irqrestore(&ioc->fc_rescan_work_lock, flags);
} while (work_to_do);
}
static int
mptfc_probe(struct pci_dev *pdev, const struct pci_device_id *id)
{
struct Scsi_Host *sh;
MPT_SCSI_HOST *hd;
MPT_ADAPTER *ioc;
unsigned long flags;
int ii;
int numSGE = 0;
int scale;
int ioc_cap;
int error=0;
int r;
if ((r = mpt_attach(pdev,id)) != 0)
return r;
ioc = pci_get_drvdata(pdev);
ioc->DoneCtx = mptfcDoneCtx;
ioc->TaskCtx = mptfcTaskCtx;
ioc->InternalCtx = mptfcInternalCtx;
/* Added sanity check on readiness of the MPT adapter.
*/
if (ioc->last_state != MPI_IOC_STATE_OPERATIONAL) {
printk(MYIOC_s_WARN_FMT
"Skipping because it's not operational!\n",
ioc->name);
error = -ENODEV;
goto out_mptfc_probe;
}
if (!ioc->active) {
printk(MYIOC_s_WARN_FMT "Skipping because it's disabled!\n",
ioc->name);
error = -ENODEV;
goto out_mptfc_probe;
}
/* Sanity check - ensure at least 1 port is INITIATOR capable
*/
ioc_cap = 0;
for (ii=0; ii < ioc->facts.NumberOfPorts; ii++) {
if (ioc->pfacts[ii].ProtocolFlags &
MPI_PORTFACTS_PROTOCOL_INITIATOR)
ioc_cap ++;
}
if (!ioc_cap) {
printk(MYIOC_s_WARN_FMT
"Skipping ioc=%p because SCSI Initiator mode is NOT enabled!\n",
ioc->name, ioc);
return -ENODEV;
}
sh = scsi_host_alloc(&mptfc_driver_template, sizeof(MPT_SCSI_HOST));
if (!sh) {
printk(MYIOC_s_WARN_FMT
"Unable to register controller with SCSI subsystem\n",
ioc->name);
error = -1;
goto out_mptfc_probe;
}
INIT_WORK(&ioc->fc_rescan_work, mptfc_rescan_devices,(void *)ioc);
spin_lock_irqsave(&ioc->FreeQlock, flags);
/* Attach the SCSI Host to the IOC structure
*/
ioc->sh = sh;
sh->io_port = 0;
sh->n_io_port = 0;
sh->irq = 0;
/* set 16 byte cdb's */
sh->max_cmd_len = 16;
sh->max_id = MPT_MAX_FC_DEVICES<256 ? MPT_MAX_FC_DEVICES : 255;
sh->max_lun = MPT_LAST_LUN + 1;
sh->max_channel = 0;
sh->this_id = ioc->pfacts[0].PortSCSIID;
/* Required entry.
*/
sh->unique_id = ioc->id;
/* Verify that we won't exceed the maximum
* number of chain buffers
* We can optimize: ZZ = req_sz/sizeof(SGE)
* For 32bit SGE's:
* numSGE = 1 + (ZZ-1)*(maxChain -1) + ZZ
* + (req_sz - 64)/sizeof(SGE)
* A slightly different algorithm is required for
* 64bit SGEs.
*/
scale = ioc->req_sz/(sizeof(dma_addr_t) + sizeof(u32));
if (sizeof(dma_addr_t) == sizeof(u64)) {
numSGE = (scale - 1) *
(ioc->facts.MaxChainDepth-1) + scale +
(ioc->req_sz - 60) / (sizeof(dma_addr_t) +
sizeof(u32));
} else {
numSGE = 1 + (scale - 1) *
(ioc->facts.MaxChainDepth-1) + scale +
(ioc->req_sz - 64) / (sizeof(dma_addr_t) +
sizeof(u32));
}
if (numSGE < sh->sg_tablesize) {
/* Reset this value */
dprintk((MYIOC_s_INFO_FMT
"Resetting sg_tablesize to %d from %d\n",
ioc->name, numSGE, sh->sg_tablesize));
sh->sg_tablesize = numSGE;
}
spin_unlock_irqrestore(&ioc->FreeQlock, flags);
hd = (MPT_SCSI_HOST *) sh->hostdata;
hd->ioc = ioc;
/* SCSI needs scsi_cmnd lookup table!
* (with size equal to req_depth*PtrSz!)
*/
hd->ScsiLookup = kcalloc(ioc->req_depth, sizeof(void *), GFP_ATOMIC);
if (!hd->ScsiLookup) {
error = -ENOMEM;
goto out_mptfc_probe;
}
dprintk((MYIOC_s_INFO_FMT "ScsiLookup @ %p\n",
ioc->name, hd->ScsiLookup));
/* Allocate memory for the device structures.
* A non-Null pointer at an offset
* indicates a device exists.
* max_id = 1 + maximum id (hosts.h)
*/
hd->Targets = kcalloc(sh->max_id, sizeof(void *), GFP_ATOMIC);
if (!hd->Targets) {
error = -ENOMEM;
goto out_mptfc_probe;
}
dprintk((KERN_INFO " vdev @ %p\n", hd->Targets));
/* Clear the TM flags
*/
hd->tmPending = 0;
hd->tmState = TM_STATE_NONE;
hd->resetPending = 0;
hd->abortSCpnt = NULL;
/* Clear the pointer used to store
* single-threaded commands, i.e., those
* issued during a bus scan, dv and
* configuration pages.
*/
hd->cmdPtr = NULL;
/* Initialize this SCSI Hosts' timers
* To use, set the timer expires field
* and add_timer
*/
init_timer(&hd->timer);
hd->timer.data = (unsigned long) hd;
hd->timer.function = mptscsih_timer_expired;
hd->mpt_pq_filter = mpt_pq_filter;
ddvprintk((MYIOC_s_INFO_FMT
"mpt_pq_filter %x\n",
ioc->name,
mpt_pq_filter));
init_waitqueue_head(&hd->scandv_waitq);
hd->scandv_wait_done = 0;
hd->last_queue_full = 0;
sh->transportt = mptfc_transport_template;
error = scsi_add_host (sh, &ioc->pcidev->dev);
if(error) {
dprintk((KERN_ERR MYNAM
"scsi_add_host failed\n"));
goto out_mptfc_probe;
}
[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations mptbase.h bump version number to 3.03.09 remove unneeded flags define workq and remove old fc specific locks mptbase.c initialize new lock and don't initialize two removed locks mptscsih.c when firmware reports target is no longer there, return DID_REQUEUE for fc hosts so that i/o doesn't get killed until the transport has an opportunity to manage the loss via its dev loss timer when the "eh_abort" routine is called, check to see if the driver has the command or not before looking to see if a reset is pending. James Smart and I talked about this and believe that the API for this routine is: if driver doesn't have command, return SUCCESS. This change helps prevent a target from being taken offline. SUCCESS is returned because it's likely that the command completed after error recovery timed it out but before it could be aborted. provide a routine to queue work to newly created workq, and use it. remove "ioc" from mptscsih_abort() it was only used one time. the other references were via hd->ioc, so I just moved it.... net change in references to ioc via hd->ioc is zero move hd->resetPending test and hd->timeouts increment to after the test for whether the command to be aborted remains known to the driver Make certain that the workq exists before queuing work to it. mptfc.c no longer need to lock rport data structures as I was able to single thread the code! I fixed up the debug code to eliminate compilation messages due to type mismatch in the printk. Got rid of some no longer needed rport flags. Initialize and destroy the workq used for the rescan work. simplify the logic regarding the increment of fc_rescan_work_count. use post increment and test for zero vs. pre increment and test for one; eliminate work_count variable: queue_work can be called with the work_lock held as it doesn't sleep Signed-off-by: Michael Reed <mdr@sgi.com> Signed-off-by: Eric Moore <Eric.Moore@lsil.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-04-22 06:14:35 +08:00
/* initialize workqueue */
snprintf(ioc->fc_rescan_work_q_name, KOBJ_NAME_LEN, "mptfc_wq_%d",
sh->host_no);
ioc->fc_rescan_work_q =
create_singlethread_workqueue(ioc->fc_rescan_work_q_name);
if (!ioc->fc_rescan_work_q)
goto out_mptfc_probe;
/*
* scan for rports -
* by doing it via the workqueue, some locking is eliminated
*/
ioc->fc_rescan_work_count = 1;
queue_work(ioc->fc_rescan_work_q, &ioc->fc_rescan_work);
flush_workqueue(ioc->fc_rescan_work_q);
return 0;
out_mptfc_probe:
mptscsih_remove(pdev);
return error;
}
static struct pci_driver mptfc_driver = {
.name = "mptfc",
.id_table = mptfc_pci_table,
.probe = mptfc_probe,
.remove = __devexit_p(mptfc_remove),
.shutdown = mptscsih_shutdown,
#ifdef CONFIG_PM
.suspend = mptscsih_suspend,
.resume = mptscsih_resume,
#endif
};
/*=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=*/
/**
* mptfc_init - Register MPT adapter(s) as SCSI host(s) with
* linux scsi mid-layer.
*
* Returns 0 for success, non-zero for failure.
*/
static int __init
mptfc_init(void)
{
int error;
show_mptmod_ver(my_NAME, my_VERSION);
/* sanity check module parameter */
if (mptfc_dev_loss_tmo == 0)
mptfc_dev_loss_tmo = MPTFC_DEV_LOSS_TMO;
mptfc_transport_template =
fc_attach_transport(&mptfc_transport_functions);
if (!mptfc_transport_template)
return -ENODEV;
mptfcDoneCtx = mpt_register(mptscsih_io_done, MPTFC_DRIVER);
mptfcTaskCtx = mpt_register(mptscsih_taskmgmt_complete, MPTFC_DRIVER);
mptfcInternalCtx = mpt_register(mptscsih_scandv_complete, MPTFC_DRIVER);
if (mpt_event_register(mptfcDoneCtx, mptscsih_event_process) == 0) {
devtverboseprintk((KERN_INFO MYNAM
": Registered for IOC event notifications\n"));
}
if (mpt_reset_register(mptfcDoneCtx, mptscsih_ioc_reset) == 0) {
dprintk((KERN_INFO MYNAM
": Registered for IOC reset notifications\n"));
}
error = pci_register_driver(&mptfc_driver);
if (error)
fc_release_transport(mptfc_transport_template);
return error;
}
/*=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=*/
/**
* mptfc_remove - Removed fc infrastructure for devices
* @pdev: Pointer to pci_dev structure
*
*/
static void __devexit
mptfc_remove(struct pci_dev *pdev)
{
[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations mptbase.h bump version number to 3.03.09 remove unneeded flags define workq and remove old fc specific locks mptbase.c initialize new lock and don't initialize two removed locks mptscsih.c when firmware reports target is no longer there, return DID_REQUEUE for fc hosts so that i/o doesn't get killed until the transport has an opportunity to manage the loss via its dev loss timer when the "eh_abort" routine is called, check to see if the driver has the command or not before looking to see if a reset is pending. James Smart and I talked about this and believe that the API for this routine is: if driver doesn't have command, return SUCCESS. This change helps prevent a target from being taken offline. SUCCESS is returned because it's likely that the command completed after error recovery timed it out but before it could be aborted. provide a routine to queue work to newly created workq, and use it. remove "ioc" from mptscsih_abort() it was only used one time. the other references were via hd->ioc, so I just moved it.... net change in references to ioc via hd->ioc is zero move hd->resetPending test and hd->timeouts increment to after the test for whether the command to be aborted remains known to the driver Make certain that the workq exists before queuing work to it. mptfc.c no longer need to lock rport data structures as I was able to single thread the code! I fixed up the debug code to eliminate compilation messages due to type mismatch in the printk. Got rid of some no longer needed rport flags. Initialize and destroy the workq used for the rescan work. simplify the logic regarding the increment of fc_rescan_work_count. use post increment and test for zero vs. pre increment and test for one; eliminate work_count variable: queue_work can be called with the work_lock held as it doesn't sleep Signed-off-by: Michael Reed <mdr@sgi.com> Signed-off-by: Eric Moore <Eric.Moore@lsil.com> Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-04-22 06:14:35 +08:00
MPT_ADAPTER *ioc = pci_get_drvdata(pdev);
struct mptfc_rport_info *p, *n;
struct workqueue_struct *work_q;
unsigned long flags;
/* destroy workqueue */
if ((work_q=ioc->fc_rescan_work_q)) {
spin_lock_irqsave(&ioc->fc_rescan_work_lock, flags);
ioc->fc_rescan_work_q = NULL;
spin_unlock_irqrestore(&ioc->fc_rescan_work_lock, flags);
destroy_workqueue(work_q);
}
fc_remove_host(ioc->sh);
list_for_each_entry_safe(p, n, &ioc->fc_rports, list) {
list_del(&p->list);
kfree(p);
}
mptscsih_remove(pdev);
}
/*=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=*/
/*=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=*/
/**
* mptfc_exit - Unregisters MPT adapter(s)
*
*/
static void __exit
mptfc_exit(void)
{
pci_unregister_driver(&mptfc_driver);
fc_release_transport(mptfc_transport_template);
mpt_reset_deregister(mptfcDoneCtx);
dprintk((KERN_INFO MYNAM
": Deregistered for IOC reset notifications\n"));
mpt_event_deregister(mptfcDoneCtx);
dprintk((KERN_INFO MYNAM
": Deregistered for IOC event notifications\n"));
mpt_deregister(mptfcInternalCtx);
mpt_deregister(mptfcTaskCtx);
mpt_deregister(mptfcDoneCtx);
}
module_init(mptfc_init);
module_exit(mptfc_exit);