struct CommandList *c, u32 ioaccel_handle, u8 *cdb, int cdb_len,
u8 *scsi3addr, struct hpsa_scsi_dev_t *phys_disk);
static void hpsa_command_resubmit_worker(struct work_struct *work);
+static u32 lockup_detected(struct ctlr_info *h);
+static int detect_controller_lockup(struct ctlr_info *h);
static inline struct ctlr_info *sdev_to_hba(struct scsi_device *sdev)
{
/* List of controllers which cannot be hard reset on kexec with reset_devices */
static u32 unresettable_controller[] = {
0x324a103C, /* Smart Array P712m */
- 0x324b103C, /* SmartArray P711m */
+ 0x324b103C, /* Smart Array P711m */
0x3223103C, /* Smart Array P800 */
0x3234103C, /* Smart Array P400 */
0x3235103C, /* Smart Array P400i */
0x409D0E11, /* Smart Array 6400 EM */
};
-static int ctlr_is_hard_resettable(u32 board_id)
+static u32 needs_abort_tags_swizzled[] = {
+ 0x323D103C, /* Smart Array P700m */
+ 0x324a103C, /* Smart Array P712m */
+ 0x324b103C, /* SmartArray P711m */
+};
+
+static int board_id_in_array(u32 a[], int nelems, u32 board_id)
{
int i;
- for (i = 0; i < ARRAY_SIZE(unresettable_controller); i++)
- if (unresettable_controller[i] == board_id)
- return 0;
- return 1;
+ for (i = 0; i < nelems; i++)
+ if (a[i] == board_id)
+ return 1;
+ return 0;
}
-static int ctlr_is_soft_resettable(u32 board_id)
+static int ctlr_is_hard_resettable(u32 board_id)
{
- int i;
+ return !board_id_in_array(unresettable_controller,
+ ARRAY_SIZE(unresettable_controller), board_id);
+}
- for (i = 0; i < ARRAY_SIZE(soft_unresettable_controller); i++)
- if (soft_unresettable_controller[i] == board_id)
- return 0;
- return 1;
+static int ctlr_is_soft_resettable(u32 board_id)
+{
+ return !board_id_in_array(soft_unresettable_controller,
+ ARRAY_SIZE(soft_unresettable_controller), board_id);
}
static int ctlr_is_resettable(u32 board_id)
ctlr_is_soft_resettable(board_id);
}
+static int ctlr_needs_abort_tags_swizzled(u32 board_id)
+{
+ return board_id_in_array(needs_abort_tags_swizzled,
+ ARRAY_SIZE(needs_abort_tags_swizzled), board_id);
+}
+
static ssize_t host_show_resettable(struct device *dev,
struct device_attribute *attr, char *buf)
{
* a separate special register for submitting commands.
*/
-/* set_performant_mode: Modify the tag for cciss performant
+/*
+ * set_performant_mode: Modify the tag for cciss performant
* set bit 0 for pull model, bits 3-1 for block fetch
* register number
*/
-static void set_performant_mode(struct ctlr_info *h, struct CommandList *c)
+#define DEFAULT_REPLY_QUEUE (-1)
+static void set_performant_mode(struct ctlr_info *h, struct CommandList *c,
+ int reply_queue)
{
if (likely(h->transMethod & CFGTBL_Trans_Performant)) {
c->busaddr |= 1 | (h->blockFetchTable[c->Header.SGList] << 1);
- if (likely(h->msix_vector > 0))
+ if (unlikely(!h->msix_vector))
+ return;
+ if (likely(reply_queue == DEFAULT_REPLY_QUEUE))
c->Header.ReplyQueue =
raw_smp_processor_id() % h->nreply_queues;
+ else
+ c->Header.ReplyQueue = reply_queue % h->nreply_queues;
}
}
static void set_ioaccel1_performant_mode(struct ctlr_info *h,
- struct CommandList *c)
+ struct CommandList *c,
+ int reply_queue)
{
struct io_accel1_cmd *cp = &h->ioaccel_cmd_pool[c->cmdindex];
- /* Tell the controller to post the reply to the queue for this
+ /*
+ * Tell the controller to post the reply to the queue for this
* processor. This seems to give the best I/O throughput.
*/
- cp->ReplyQueue = smp_processor_id() % h->nreply_queues;
- /* Set the bits in the address sent down to include:
+ if (likely(reply_queue == DEFAULT_REPLY_QUEUE))
+ cp->ReplyQueue = smp_processor_id() % h->nreply_queues;
+ else
+ cp->ReplyQueue = reply_queue % h->nreply_queues;
+ /*
+ * Set the bits in the address sent down to include:
* - performant mode bit (bit 0)
* - pull count (bits 1-3)
* - command type (bits 4-6)
}
static void set_ioaccel2_performant_mode(struct ctlr_info *h,
- struct CommandList *c)
+ struct CommandList *c,
+ int reply_queue)
{
struct io_accel2_cmd *cp = &h->ioaccel2_cmd_pool[c->cmdindex];
- /* Tell the controller to post the reply to the queue for this
+ /*
+ * Tell the controller to post the reply to the queue for this
* processor. This seems to give the best I/O throughput.
*/
- cp->reply_queue = smp_processor_id() % h->nreply_queues;
- /* Set the bits in the address sent down to include:
+ if (likely(reply_queue == DEFAULT_REPLY_QUEUE))
+ cp->reply_queue = smp_processor_id() % h->nreply_queues;
+ else
+ cp->reply_queue = reply_queue % h->nreply_queues;
+ /*
+ * Set the bits in the address sent down to include:
* - performant mode bit not used in ioaccel mode 2
* - pull count (bits 0-3)
* - command type isn't needed for ioaccel2
h->heartbeat_sample_interval = HEARTBEAT_SAMPLE_INTERVAL;
}
-static void enqueue_cmd_and_start_io(struct ctlr_info *h,
- struct CommandList *c)
+static void __enqueue_cmd_and_start_io(struct ctlr_info *h,
+ struct CommandList *c, int reply_queue)
{
dial_down_lockup_detection_during_fw_flash(h, c);
atomic_inc(&h->commands_outstanding);
switch (c->cmd_type) {
case CMD_IOACCEL1:
- set_ioaccel1_performant_mode(h, c);
+ set_ioaccel1_performant_mode(h, c, reply_queue);
writel(c->busaddr, h->vaddr + SA5_REQUEST_PORT_OFFSET);
break;
case CMD_IOACCEL2:
- set_ioaccel2_performant_mode(h, c);
+ set_ioaccel2_performant_mode(h, c, reply_queue);
writel(c->busaddr, h->vaddr + IOACCEL2_INBOUND_POSTQ_32);
break;
default:
- set_performant_mode(h, c);
+ set_performant_mode(h, c, reply_queue);
h->access.submit_command(h, c);
}
}
+static void enqueue_cmd_and_start_io(struct ctlr_info *h,
+ struct CommandList *c)
+{
+ __enqueue_cmd_and_start_io(h, c, DEFAULT_REPLY_QUEUE);
+}
+
static inline int is_hba_lunid(unsigned char scsi3addr[])
{
return memcmp(scsi3addr, RAID_CTLR_LUNID, 8) == 0;
return !found;
}
+static inline void hpsa_show_dev_msg(const char *level, struct ctlr_info *h,
+ struct hpsa_scsi_dev_t *dev, char *description)
+{
+ dev_printk(level, &h->pdev->dev,
+ "scsi %d:%d:%d:%d: %s %s %.8s %.16s RAID-%s SSDSmartPathCap%c En%c Exp=%d\n",
+ h->scsi_host->host_no, dev->bus, dev->target, dev->lun,
+ description,
+ scsi_device_type(dev->devtype),
+ dev->vendor,
+ dev->model,
+ dev->raid_level > RAID_UNKNOWN ?
+ "RAID-?" : raid_label[dev->raid_level],
+ dev->offload_config ? '+' : '-',
+ dev->offload_enabled ? '+' : '-',
+ dev->expose_state);
+}
+
/* Add an entry into h->dev[] array. */
static int hpsa_scsi_add_entry(struct ctlr_info *h, int hostno,
struct hpsa_scsi_dev_t *device,
device->offload_enabled = 0;
added[*nadded] = device;
(*nadded)++;
-
- /* initially, (before registering with scsi layer) we don't
- * know our hostno and we don't want to print anything first
- * time anyway (the scsi layer's inquiries will show that info)
- */
- /* if (hostno != -1) */
- dev_info(&h->pdev->dev, "%s device c%db%dt%dl%d added.\n",
- scsi_device_type(device->devtype), hostno,
- device->bus, device->target, device->lun);
+ hpsa_show_dev_msg(KERN_INFO, h, device,
+ device->expose_state & HPSA_SCSI_ADD ? "added" : "masked");
return 0;
}
if (!new_entry->offload_enabled)
h->dev[entry]->offload_enabled = 0;
+ hpsa_show_dev_msg(KERN_INFO, h, h->dev[entry], "updated");
}
/* Replace an entry from h->dev[] array. */
h->dev[entry] = new_entry;
added[*nadded] = new_entry;
(*nadded)++;
- dev_info(&h->pdev->dev, "%s device c%db%dt%dl%d changed.\n",
- scsi_device_type(new_entry->devtype), hostno, new_entry->bus,
- new_entry->target, new_entry->lun);
+ hpsa_show_dev_msg(KERN_INFO, h, new_entry, "replaced");
}
/* Remove an entry from h->dev[] array. */
for (i = entry; i < h->ndevices-1; i++)
h->dev[i] = h->dev[i+1];
h->ndevices--;
- dev_info(&h->pdev->dev, "%s device c%db%dt%dl%d removed.\n",
- scsi_device_type(sd->devtype), hostno, sd->bus, sd->target,
- sd->lun);
+ hpsa_show_dev_msg(KERN_INFO, h, sd, "removed");
}
#define SCSI3ADDR_EQ(a, b) ( \
*/
if (sd[i]->volume_offline) {
hpsa_show_volume_status(h, sd[i]);
- dev_info(&h->pdev->dev, "c%db%dt%dl%d: temporarily offline\n",
- h->scsi_host->host_no,
- sd[i]->bus, sd[i]->target, sd[i]->lun);
+ hpsa_show_dev_msg(KERN_INFO, h, sd[i], "offline");
continue;
}
* future cmds to this device will get selection
* timeout as if the device was gone.
*/
- dev_warn(&h->pdev->dev,
- "didn't find c%db%dt%dl%d for removal.\n",
- hostno, removed[i]->bus,
- removed[i]->target, removed[i]->lun);
+ hpsa_show_dev_msg(KERN_WARNING, h, removed[i],
+ "didn't find device for removal.");
}
}
kfree(removed[i]);
if (scsi_add_device(sh, added[i]->bus,
added[i]->target, added[i]->lun) == 0)
continue;
- dev_warn(&h->pdev->dev, "scsi_add_device c%db%dt%dl%d failed, "
- "device not added.\n", hostno, added[i]->bus,
- added[i]->target, added[i]->lun);
+ hpsa_show_dev_msg(KERN_WARNING, h, added[i],
+ "addition failed, device not added.");
/* now we have to remove it from h->dev,
* since it didn't get added to scsi mid layer
*/
if (cp->cmd_type == CMD_IOACCEL2 || cp->cmd_type == CMD_IOACCEL1)
atomic_dec(&cp->phys_disk->ioaccel_cmds_out);
+ /*
+ * We check for lockup status here as it may be set for
+ * CMD_SCSI, CMD_IOACCEL1 and CMD_IOACCEL2 commands by
+ * fail_all_oustanding_cmds()
+ */
+ if (unlikely(ei->CommandStatus == CMD_CTLR_LOCKUP)) {
+ /* DID_NO_CONNECT will prevent a retry */
+ cmd->result = DID_NO_CONNECT << 16;
+ cmd_free(h, cp);
+ cmd->scsi_done(cmd);
+ return;
+ }
+
if (cp->cmd_type == CMD_IOACCEL2)
return process_ioaccel2_completion(h, cp, cmd, dev);
return 0;
}
-static inline void hpsa_scsi_do_simple_cmd_core(struct ctlr_info *h,
- struct CommandList *c)
+#define NO_TIMEOUT ((unsigned long) -1)
+#define DEFAULT_TIMEOUT 30000 /* milliseconds */
+static int hpsa_scsi_do_simple_cmd_core(struct ctlr_info *h,
+ struct CommandList *c, int reply_queue, unsigned long timeout_msecs)
{
DECLARE_COMPLETION_ONSTACK(wait);
c->waiting = &wait;
- enqueue_cmd_and_start_io(h, c);
- wait_for_completion(&wait);
+ __enqueue_cmd_and_start_io(h, c, reply_queue);
+ if (timeout_msecs == NO_TIMEOUT) {
+ /* TODO: get rid of this no-timeout thing */
+ wait_for_completion_io(&wait);
+ return IO_OK;
+ }
+ if (!wait_for_completion_io_timeout(&wait,
+ msecs_to_jiffies(timeout_msecs))) {
+ dev_warn(&h->pdev->dev, "Command timed out.\n");
+ return -ETIMEDOUT;
+ }
+ return IO_OK;
+}
+
+static int hpsa_scsi_do_simple_cmd(struct ctlr_info *h, struct CommandList *c,
+ int reply_queue, unsigned long timeout_msecs)
+{
+ if (unlikely(lockup_detected(h))) {
+ c->err_info->CommandStatus = CMD_CTLR_LOCKUP;
+ return IO_OK;
+ }
+ return hpsa_scsi_do_simple_cmd_core(h, c, reply_queue, timeout_msecs);
}
static u32 lockup_detected(struct ctlr_info *h)
return rc;
}
-static void hpsa_scsi_do_simple_cmd_core_if_no_lockup(struct ctlr_info *h,
- struct CommandList *c)
-{
- /* If controller lockup detected, fake a hardware error. */
- if (unlikely(lockup_detected(h)))
- c->err_info->CommandStatus = CMD_HARDWARE_ERR;
- else
- hpsa_scsi_do_simple_cmd_core(h, c);
-}
-
#define MAX_DRIVER_CMD_RETRIES 25
-static void hpsa_scsi_do_simple_cmd_with_retry(struct ctlr_info *h,
- struct CommandList *c, int data_direction)
+static int hpsa_scsi_do_simple_cmd_with_retry(struct ctlr_info *h,
+ struct CommandList *c, int data_direction, unsigned long timeout_msecs)
{
int backoff_time = 10, retry_count = 0;
+ int rc;
do {
memset(c->err_info, 0, sizeof(*c->err_info));
- hpsa_scsi_do_simple_cmd_core(h, c);
+ rc = hpsa_scsi_do_simple_cmd(h, c, DEFAULT_REPLY_QUEUE,
+ timeout_msecs);
+ if (rc)
+ break;
retry_count++;
if (retry_count > 3) {
msleep(backoff_time);
check_for_busy(h, c)) &&
retry_count <= MAX_DRIVER_CMD_RETRIES);
hpsa_pci_unmap(h->pdev, c, 1, data_direction);
+ if (retry_count > MAX_DRIVER_CMD_RETRIES)
+ rc = -EIO;
+ return rc;
}
static void hpsa_print_cmd(struct ctlr_info *h, char *txt,
case CMD_UNABORTABLE:
hpsa_print_cmd(h, "unabortable", cp);
break;
+ case CMD_CTLR_LOCKUP:
+ hpsa_print_cmd(h, "controller lockup detected", cp);
+ break;
default:
hpsa_print_cmd(h, "unknown status", cp);
dev_warn(d, "Unknown command status %x\n",
rc = -1;
goto out;
}
- hpsa_scsi_do_simple_cmd_with_retry(h, c, PCI_DMA_FROMDEVICE);
+ rc = hpsa_scsi_do_simple_cmd_with_retry(h, c,
+ PCI_DMA_FROMDEVICE, NO_TIMEOUT);
+ if (rc)
+ goto out;
ei = c->err_info;
if (ei->CommandStatus != 0 && ei->CommandStatus != CMD_DATA_UNDERRUN) {
hpsa_scsi_interpret_error(h, c);
rc = -1;
goto out;
}
- hpsa_scsi_do_simple_cmd_with_retry(h, c, PCI_DMA_FROMDEVICE);
+ rc = hpsa_scsi_do_simple_cmd_with_retry(h, c,
+ PCI_DMA_FROMDEVICE, NO_TIMEOUT);
+ if (rc)
+ goto out;
ei = c->err_info;
if (ei->CommandStatus != 0 && ei->CommandStatus != CMD_DATA_UNDERRUN) {
hpsa_scsi_interpret_error(h, c);
}
static int hpsa_send_reset(struct ctlr_info *h, unsigned char *scsi3addr,
- u8 reset_type)
+ u8 reset_type, int reply_queue)
{
int rc = IO_OK;
struct CommandList *c;
(void) fill_cmd(c, HPSA_DEVICE_RESET_MSG, h, NULL, 0, 0,
scsi3addr, TYPE_MSG);
c->Request.CDB[1] = reset_type; /* fill_cmd defaults to LUN reset */
- hpsa_scsi_do_simple_cmd_core(h, c);
+ rc = hpsa_scsi_do_simple_cmd(h, c, reply_queue, NO_TIMEOUT);
+ if (rc) {
+ dev_warn(&h->pdev->dev, "Failed to send reset command\n");
+ goto out;
+ }
/* no unmap needed here because no data xfer. */
ei = c->err_info;
hpsa_scsi_interpret_error(h, c);
rc = -1;
}
+out:
cmd_free(h, c);
return rc;
}
sizeof(this_device->raid_map), 0,
scsi3addr, TYPE_CMD)) {
dev_warn(&h->pdev->dev, "Out of memory in hpsa_get_raid_map()\n");
- cmd_free(h, c);
- return -ENOMEM;
+ rc = -ENOMEM;
+ goto out;
}
- hpsa_scsi_do_simple_cmd_with_retry(h, c, PCI_DMA_FROMDEVICE);
+ rc = hpsa_scsi_do_simple_cmd_with_retry(h, c,
+ PCI_DMA_FROMDEVICE, NO_TIMEOUT);
+ if (rc)
+ goto out;
ei = c->err_info;
if (ei->CommandStatus != 0 && ei->CommandStatus != CMD_DATA_UNDERRUN) {
hpsa_scsi_interpret_error(h, c);
- cmd_free(h, c);
- return -1;
+ rc = -1;
+ goto out;
}
cmd_free(h, c);
}
hpsa_debug_map_buff(h, rc, &this_device->raid_map);
return rc;
+out:
+ cmd_free(h, c);
+ return rc;
}
static int hpsa_bmic_id_physical_device(struct ctlr_info *h,
c->Request.CDB[2] = bmic_device_index & 0xff;
c->Request.CDB[9] = (bmic_device_index >> 8) & 0xff;
- hpsa_scsi_do_simple_cmd_with_retry(h, c, PCI_DMA_FROMDEVICE);
+ hpsa_scsi_do_simple_cmd_with_retry(h, c, PCI_DMA_FROMDEVICE,
+ NO_TIMEOUT);
ei = c->err_info;
if (ei->CommandStatus != 0 && ei->CommandStatus != CMD_DATA_UNDERRUN) {
hpsa_scsi_interpret_error(h, c);
}
if (extended_response)
c->Request.CDB[1] = extended_response;
- hpsa_scsi_do_simple_cmd_with_retry(h, c, PCI_DMA_FROMDEVICE);
+ rc = hpsa_scsi_do_simple_cmd_with_retry(h, c,
+ PCI_DMA_FROMDEVICE, NO_TIMEOUT);
+ if (rc)
+ goto out;
ei = c->err_info;
if (ei->CommandStatus != 0 &&
ei->CommandStatus != CMD_DATA_UNDERRUN) {
{
struct CommandList *c;
unsigned char *sense, sense_key, asc, ascq;
- int ldstat = 0;
+ int rc, ldstat = 0;
u16 cmd_status;
u8 scsi_status;
#define ASC_LUN_NOT_READY 0x04
if (!c)
return 0;
(void) fill_cmd(c, TEST_UNIT_READY, h, NULL, 0, 0, scsi3addr, TYPE_CMD);
- hpsa_scsi_do_simple_cmd_core(h, c);
+ rc = hpsa_scsi_do_simple_cmd(h, c, DEFAULT_REPLY_QUEUE, NO_TIMEOUT);
+ if (rc) {
+ cmd_free(h, c);
+ return 0;
+ }
sense = c->err_info->SenseInfo;
sense_key = sense[2];
asc = sense[12];
return 0;
}
+/*
+ * Find out if a logical device supports aborts by simply trying one.
+ * Smart Array may claim not to support aborts on logical drives, but
+ * if a MSA2000 * is connected, the drives on that will be presented
+ * by the Smart Array as logical drives, and aborts may be sent to
+ * those devices successfully. So the simplest way to find out is
+ * to simply try an abort and see how the device responds.
+ */
+static int hpsa_device_supports_aborts(struct ctlr_info *h,
+ unsigned char *scsi3addr)
+{
+ struct CommandList *c;
+ struct ErrorInfo *ei;
+ int rc = 0;
+
+ u64 tag = (u64) -1; /* bogus tag */
+
+ /* Assume that physical devices support aborts */
+ if (!is_logical_dev_addr_mode(scsi3addr))
+ return 1;
+
+ c = cmd_alloc(h);
+ if (!c)
+ return -ENOMEM;
+ (void) fill_cmd(c, HPSA_ABORT_MSG, h, &tag, 0, 0, scsi3addr, TYPE_MSG);
+ (void) hpsa_scsi_do_simple_cmd(h, c, DEFAULT_REPLY_QUEUE, NO_TIMEOUT);
+ /* no unmap needed here because no data xfer. */
+ ei = c->err_info;
+ switch (ei->CommandStatus) {
+ case CMD_INVALID:
+ rc = 0;
+ break;
+ case CMD_UNABORTABLE:
+ case CMD_ABORT_FAILED:
+ rc = 1;
+ break;
+ default:
+ rc = 0;
+ break;
+ }
+ cmd_free(h, c);
+ return rc;
+}
+
static int hpsa_update_device_info(struct ctlr_info *h,
unsigned char scsi3addr[], struct hpsa_scsi_dev_t *this_device,
unsigned char *is_OBDR_device)
strncmp(obdr_sig, OBDR_TAPE_SIG,
OBDR_SIG_LEN) == 0);
}
-
kfree(inq_buff);
return 0;
return 1;
}
+static void hpsa_update_device_supports_aborts(struct ctlr_info *h,
+ struct hpsa_scsi_dev_t *dev, u8 *scsi3addr)
+{
+ unsigned long flags;
+ int rc, entry;
+ /*
+ * See if this device supports aborts. If we already know
+ * the device, we already know if it supports aborts, otherwise
+ * we have to find out if it supports aborts by trying one.
+ */
+ spin_lock_irqsave(&h->devlock, flags);
+ rc = hpsa_scsi_find_entry(dev, h->dev, h->ndevices, &entry);
+ if ((rc == DEVICE_SAME || rc == DEVICE_UPDATED) &&
+ entry >= 0 && entry < h->ndevices) {
+ dev->supports_aborts = h->dev[entry]->supports_aborts;
+ spin_unlock_irqrestore(&h->devlock, flags);
+ } else {
+ spin_unlock_irqrestore(&h->devlock, flags);
+ dev->supports_aborts =
+ hpsa_device_supports_aborts(h, scsi3addr);
+ if (dev->supports_aborts < 0)
+ dev->supports_aborts = 0;
+ }
+}
+
static unsigned char *ext_target_model[] = {
"MSA2012",
"MSA2024",
(*n_ext_target_devs)++;
hpsa_set_bus_target_lun(this_device,
tmpdevice->bus, tmpdevice->target, 0);
+ hpsa_update_device_supports_aborts(h, this_device, scsi3addr);
set_bit(tmpdevice->target, lunzerobits);
return 1;
}
&is_OBDR))
continue; /* skip it if we can't talk to it. */
figure_bus_target_lun(h, lunaddrbytes, tmpdevice);
+ hpsa_update_device_supports_aborts(h, tmpdevice, lunaddrbytes);
this_device = currentsd[ncurrent];
/*
dev->phys_disk[map_index]);
}
-/* Submit commands down the "normal" RAID stack path */
+/*
+ * Submit commands down the "normal" RAID stack path
+ * All callers to hpsa_ciss_submit must check lockup_detected
+ * beforehand, before (opt.) and after calling cmd_alloc
+ */
static int hpsa_ciss_submit(struct ctlr_info *h,
struct CommandList *c, struct scsi_cmnd *cmd,
unsigned char scsi3addr[])
memcpy(scsi3addr, dev->scsi3addr, sizeof(scsi3addr));
if (unlikely(lockup_detected(h))) {
- cmd->result = DID_ERROR << 16;
+ cmd->result = DID_NO_CONNECT << 16;
cmd->scsi_done(cmd);
return 0;
}
return SCSI_MLQUEUE_HOST_BUSY;
}
if (unlikely(lockup_detected(h))) {
- cmd->result = DID_ERROR << 16;
+ cmd->result = DID_NO_CONNECT << 16;
cmd_free(h, c);
cmd->scsi_done(cmd);
return 0;
/* Send the Test Unit Ready, fill_cmd can't fail, no mapping */
(void) fill_cmd(c, TEST_UNIT_READY, h,
NULL, 0, 0, lunaddr, TYPE_CMD);
- hpsa_scsi_do_simple_cmd_core(h, c);
+ rc = hpsa_scsi_do_simple_cmd(h, c, DEFAULT_REPLY_QUEUE,
+ NO_TIMEOUT);
+ if (rc)
+ goto do_it_again;
/* no unmap needed here because no data xfer. */
if (c->err_info->CommandStatus == CMD_SUCCESS)
(c->err_info->SenseInfo[2] == NO_SENSE ||
c->err_info->SenseInfo[2] == UNIT_ATTENTION))
break;
-
+do_it_again:
dev_warn(&h->pdev->dev, "waiting %d secs "
"for device to become ready.\n", waittime);
rc = 1; /* device not ready. */
"device lookup failed.\n");
return FAILED;
}
- dev_warn(&h->pdev->dev, "resetting device %d:%d:%d:%d\n",
- h->scsi_host->host_no, dev->bus, dev->target, dev->lun);
+
+ /* if controller locked up, we can guarantee command won't complete */
+ if (lockup_detected(h)) {
+ dev_warn(&h->pdev->dev,
+ "scsi %d:%d:%d:%d RESET FAILED, lockup detected\n",
+ h->scsi_host->host_no, dev->bus, dev->target,
+ dev->lun);
+ return FAILED;
+ }
+
+ /* this reset request might be the result of a lockup; check */
+ if (detect_controller_lockup(h)) {
+ dev_warn(&h->pdev->dev,
+ "scsi %d:%d:%d:%d RESET FAILED, new lockup detected\n",
+ h->scsi_host->host_no, dev->bus, dev->target,
+ dev->lun);
+ return FAILED;
+ }
+
+ hpsa_show_dev_msg(KERN_WARNING, h, dev, "resetting");
+
/* send a reset to the SCSI LUN which the command was sent to */
- rc = hpsa_send_reset(h, dev->scsi3addr, HPSA_RESET_TYPE_LUN);
+ rc = hpsa_send_reset(h, dev->scsi3addr, HPSA_RESET_TYPE_LUN,
+ DEFAULT_REPLY_QUEUE);
if (rc == 0 && wait_for_device_to_become_ready(h, dev->scsi3addr) == 0)
return SUCCESS;
- dev_warn(&h->pdev->dev, "resetting device failed.\n");
+ dev_warn(&h->pdev->dev,
+ "scsi %d:%d:%d:%d reset failed\n",
+ h->scsi_host->host_no, dev->bus, dev->target, dev->lun);
return FAILED;
}
}
static int hpsa_send_abort(struct ctlr_info *h, unsigned char *scsi3addr,
- struct CommandList *abort, int swizzle)
+ struct CommandList *abort, int reply_queue)
{
int rc = IO_OK;
struct CommandList *c;
}
/* fill_cmd can't fail here, no buffer to map */
- (void) fill_cmd(c, HPSA_ABORT_MSG, h, abort,
+ (void) fill_cmd(c, HPSA_ABORT_MSG, h, &abort->Header.tag,
0, 0, scsi3addr, TYPE_MSG);
- if (swizzle)
+ if (h->needs_abort_tags_swizzled)
swizzle_abort_tag(&c->Request.CDB[4]);
- hpsa_scsi_do_simple_cmd_core(h, c);
+ (void) hpsa_scsi_do_simple_cmd(h, c, reply_queue, NO_TIMEOUT);
hpsa_get_tag(h, abort, &taglower, &tagupper);
- dev_dbg(&h->pdev->dev, "%s: Tag:0x%08x:%08x: do_simple_cmd_core completed.\n",
+ dev_dbg(&h->pdev->dev, "%s: Tag:0x%08x:%08x: do_simple_cmd(abort) completed.\n",
__func__, tagupper, taglower);
/* no unmap needed here because no data xfer. */
*/
static int hpsa_send_reset_as_abort_ioaccel2(struct ctlr_info *h,
- unsigned char *scsi3addr, struct CommandList *abort)
+ unsigned char *scsi3addr, struct CommandList *abort, int reply_queue)
{
int rc = IO_OK;
struct scsi_cmnd *scmd; /* scsi command within request being aborted */
if (h->raid_offload_debug > 0)
dev_info(&h->pdev->dev,
- "Reset as abort: Abort requested on C%d:B%d:T%d:L%d scsi3addr 0x%02x%02x%02x%02x%02x%02x%02x%02x\n",
+ "scsi %d:%d:%d:%d %s scsi3addr 0x%02x%02x%02x%02x%02x%02x%02x%02x\n",
h->scsi_host->host_no, dev->bus, dev->target, dev->lun,
+ "Reset as abort",
scsi3addr[0], scsi3addr[1], scsi3addr[2], scsi3addr[3],
scsi3addr[4], scsi3addr[5], scsi3addr[6], scsi3addr[7]);
"Reset as abort: Resetting physical device at scsi3addr 0x%02x%02x%02x%02x%02x%02x%02x%02x\n",
psa[0], psa[1], psa[2], psa[3],
psa[4], psa[5], psa[6], psa[7]);
- rc = hpsa_send_reset(h, psa, HPSA_RESET_TYPE_TARGET);
+ rc = hpsa_send_reset(h, psa, HPSA_RESET_TYPE_TARGET, reply_queue);
if (rc != 0) {
dev_warn(&h->pdev->dev,
"Reset as abort: Failed on physical device at scsi3addr 0x%02x%02x%02x%02x%02x%02x%02x%02x\n",
return rc; /* success */
}
-/* Some Smart Arrays need the abort tag swizzled, and some don't. It's hard to
- * tell which kind we're dealing with, so we send the abort both ways. There
- * shouldn't be any collisions between swizzled and unswizzled tags due to the
- * way we construct our tags but we check anyway in case the assumptions which
- * make this true someday become false.
- */
static int hpsa_send_abort_both_ways(struct ctlr_info *h,
- unsigned char *scsi3addr, struct CommandList *abort)
+ unsigned char *scsi3addr, struct CommandList *abort, int reply_queue)
{
/* ioccelerator mode 2 commands should be aborted via the
* accelerated path, since RAID path is unaware of these commands,
* Change abort to physical device reset.
*/
if (abort->cmd_type == CMD_IOACCEL2)
- return hpsa_send_reset_as_abort_ioaccel2(h, scsi3addr, abort);
+ return hpsa_send_reset_as_abort_ioaccel2(h, scsi3addr,
+ abort, reply_queue);
+ return hpsa_send_abort(h, scsi3addr, abort, reply_queue);
+}
- return hpsa_send_abort(h, scsi3addr, abort, 0) &&
- hpsa_send_abort(h, scsi3addr, abort, 1);
+/* Find out which reply queue a command was meant to return on */
+static int hpsa_extract_reply_queue(struct ctlr_info *h,
+ struct CommandList *c)
+{
+ if (c->cmd_type == CMD_IOACCEL2)
+ return h->ioaccel2_cmd_pool[c->cmdindex].reply_queue;
+ return c->Header.ReplyQueue;
+}
+
+/*
+ * Limit concurrency of abort commands to prevent
+ * over-subscription of commands
+ */
+static inline int wait_for_available_abort_cmd(struct ctlr_info *h)
+{
+#define ABORT_CMD_WAIT_MSECS 5000
+ return !wait_event_timeout(h->abort_cmd_wait_queue,
+ atomic_dec_if_positive(&h->abort_cmds_available) >= 0,
+ msecs_to_jiffies(ABORT_CMD_WAIT_MSECS));
}
/* Send an abort for the specified command.
char msg[256]; /* For debug messaging. */
int ml = 0;
__le32 tagupper, taglower;
- int refcount;
+ int refcount, reply_queue;
- /* Find the controller of the command to be aborted */
- h = sdev_to_hba(sc->device);
- if (WARN(h == NULL,
- "ABORT REQUEST FAILED, Controller lookup failed.\n"))
+ if (sc == NULL)
return FAILED;
- if (lockup_detected(h))
+ if (sc->device == NULL)
return FAILED;
- /* Check that controller supports some kind of task abort */
- if (!(HPSATMF_PHYS_TASK_ABORT & h->TMFSupportFlags) &&
- !(HPSATMF_LOG_TASK_ABORT & h->TMFSupportFlags))
+ /* Find the controller of the command to be aborted */
+ h = sdev_to_hba(sc->device);
+ if (h == NULL)
return FAILED;
- memset(msg, 0, sizeof(msg));
- ml += sprintf(msg+ml, "ABORT REQUEST on C%d:B%d:T%d:L%llu ",
- h->scsi_host->host_no, sc->device->channel,
- sc->device->id, sc->device->lun);
-
/* Find the device of the command to be aborted */
dev = sc->device->hostdata;
if (!dev) {
return FAILED;
}
+ /* If controller locked up, we can guarantee command won't complete */
+ if (lockup_detected(h)) {
+ hpsa_show_dev_msg(KERN_WARNING, h, dev,
+ "ABORT FAILED, lockup detected");
+ return FAILED;
+ }
+
+ /* This is a good time to check if controller lockup has occurred */
+ if (detect_controller_lockup(h)) {
+ hpsa_show_dev_msg(KERN_WARNING, h, dev,
+ "ABORT FAILED, new lockup detected");
+ return FAILED;
+ }
+
+ /* Check that controller supports some kind of task abort */
+ if (!(HPSATMF_PHYS_TASK_ABORT & h->TMFSupportFlags) &&
+ !(HPSATMF_LOG_TASK_ABORT & h->TMFSupportFlags))
+ return FAILED;
+
+ memset(msg, 0, sizeof(msg));
+ ml += sprintf(msg+ml, "scsi %d:%d:%d:%llu %s",
+ h->scsi_host->host_no, sc->device->channel,
+ sc->device->id, sc->device->lun,
+ "Aborting command");
+
/* Get SCSI command to be aborted */
abort = (struct CommandList *) sc->host_scribble;
if (abort == NULL) {
cmd_free(h, abort);
return SUCCESS;
}
+
+ /* Don't bother trying the abort if we know it won't work. */
+ if (abort->cmd_type != CMD_IOACCEL2 &&
+ abort->cmd_type != CMD_IOACCEL1 && !dev->supports_aborts) {
+ cmd_free(h, abort);
+ return FAILED;
+ }
+
hpsa_get_tag(h, abort, &taglower, &tagupper);
+ reply_queue = hpsa_extract_reply_queue(h, abort);
ml += sprintf(msg+ml, "Tag:0x%08x:%08x ", tagupper, taglower);
as = abort->scsi_cmd;
if (as != NULL)
ml += sprintf(msg+ml, "Command:0x%x SN:0x%lx ",
as->cmnd[0], as->serial_number);
dev_dbg(&h->pdev->dev, "%s\n", msg);
- dev_warn(&h->pdev->dev, "Abort request on C%d:B%d:T%d:L%d\n",
- h->scsi_host->host_no, dev->bus, dev->target, dev->lun);
+ hpsa_show_dev_msg(KERN_WARNING, h, dev, "Aborting command");
/*
* Command is in flight, or possibly already completed
* by the firmware (but not to the scsi mid layer) but we can't
* distinguish which. Send the abort down.
*/
- rc = hpsa_send_abort_both_ways(h, dev->scsi3addr, abort);
+ if (wait_for_available_abort_cmd(h)) {
+ dev_warn(&h->pdev->dev,
+ "Timed out waiting for an abort command to become available.\n");
+ cmd_free(h, abort);
+ return FAILED;
+ }
+ rc = hpsa_send_abort_both_ways(h, dev->scsi3addr, abort, reply_queue);
+ atomic_inc(&h->abort_cmds_available);
+ wake_up_all(&h->abort_cmd_wait_queue);
if (rc != 0) {
- dev_dbg(&h->pdev->dev, "%s Request FAILED.\n", msg);
- dev_warn(&h->pdev->dev, "FAILED abort on device C%d:B%d:T%d:L%d\n",
- h->scsi_host->host_no,
- dev->bus, dev->target, dev->lun);
+ hpsa_show_dev_msg(KERN_WARNING, h, dev,
+ "FAILED to abort command");
cmd_free(h, abort);
return FAILED;
}
c->SG[0].Len = cpu_to_le32(iocommand.buf_size);
c->SG[0].Ext = cpu_to_le32(HPSA_SG_LAST); /* not chaining */
}
- hpsa_scsi_do_simple_cmd_core_if_no_lockup(h, c);
+ rc = hpsa_scsi_do_simple_cmd(h, c, DEFAULT_REPLY_QUEUE, NO_TIMEOUT);
if (iocommand.buf_size > 0)
hpsa_pci_unmap(h->pdev, c, 1, PCI_DMA_BIDIRECTIONAL);
check_ioctl_unit_attention(h, c);
+ if (rc) {
+ rc = -EIO;
+ goto out;
+ }
/* Copy the error information out */
memcpy(&iocommand.error_info, c->err_info,
}
c->SG[--i].Ext = cpu_to_le32(HPSA_SG_LAST);
}
- hpsa_scsi_do_simple_cmd_core_if_no_lockup(h, c);
+ status = hpsa_scsi_do_simple_cmd(h, c, DEFAULT_REPLY_QUEUE, NO_TIMEOUT);
if (sg_used)
hpsa_pci_unmap(h->pdev, c, sg_used, PCI_DMA_BIDIRECTIONAL);
check_ioctl_unit_attention(h, c);
+ if (status) {
+ status = -EIO;
+ goto cleanup0;
+ }
+
/* Copy the error information out */
memcpy(&ioc->error_info, c->err_info, sizeof(ioc->error_info));
if (copy_to_user(argp, ioc, sizeof(*ioc))) {
int cmd_type)
{
int pci_dir = XFER_NONE;
- struct CommandList *a; /* for commands to be aborted */
+ u64 tag; /* for commands to be aborted */
c->cmd_type = CMD_IOCTL_PEND;
c->Header.ReplyQueue = 0;
c->Request.CDB[7] = 0x00;
break;
case HPSA_ABORT_MSG:
- a = buff; /* point to command to be aborted */
+ memcpy(&tag, buff, sizeof(tag));
dev_dbg(&h->pdev->dev,
- "Abort Tag:0x%016llx request Tag:0x%016llx",
- a->Header.tag, c->Header.tag);
+ "Abort Tag:0x%016llx using rqst Tag:0x%016llx",
+ tag, c->Header.tag);
c->Request.CDBLen = 16;
c->Request.type_attr_dir =
TYPE_ATTR_DIR(cmd_type,
c->Request.CDB[2] = 0x00; /* reserved */
c->Request.CDB[3] = 0x00; /* reserved */
/* Tag to abort goes in CDB[4]-CDB[11] */
- memcpy(&c->Request.CDB[4], &a->Header.tag,
- sizeof(a->Header.tag));
+ memcpy(&c->Request.CDB[4], &tag, sizeof(tag));
c->Request.CDB[12] = 0x00; /* reserved */
c->Request.CDB[13] = 0x00; /* reserved */
c->Request.CDB[14] = 0x00; /* reserved */
* as we enter this code.)
*/
for (i = 0; i < MAX_MODE_CHANGE_WAIT; i++) {
+ if (h->remove_in_progress)
+ goto done;
spin_lock_irqsave(&h->lock, flags);
doorbell_value = readl(h->vaddr + SA5_DOORBELL);
spin_unlock_irqrestore(&h->lock, flags);
h->product_name = products[prod_index].product_name;
h->access = *(products[prod_index].access);
+ h->needs_abort_tags_swizzled =
+ ctlr_needs_abort_tags_swizzled(h->board_id);
+
pci_disable_link_state(h->pdev, PCIE_LINK_STATE_L0S |
PCIE_LINK_STATE_L1 | PCIE_LINK_STATE_CLKPM);
{
int i, refcount;
struct CommandList *c;
+ int failcount = 0;
flush_workqueue(h->resubmit_wq); /* ensure all cmds are fully built */
for (i = 0; i < h->nr_cmds; i++) {
c = h->cmd_pool + i;
refcount = atomic_inc_return(&c->refcount);
if (refcount > 1) {
- c->err_info->CommandStatus = CMD_HARDWARE_ERR;
+ c->err_info->CommandStatus = CMD_CTLR_LOCKUP;
finish_cmd(c);
+ failcount++;
}
cmd_free(h, c);
}
+ dev_warn(&h->pdev->dev,
+ "failed %d commands in fail_all\n", failcount);
}
static void set_lockup_detected_for_all_cpus(struct ctlr_info *h, u32 value)
if (!lockup_detected) {
/* no heartbeat, but controller gave us a zero. */
dev_warn(&h->pdev->dev,
- "lockup detected but scratchpad register is zero\n");
+ "lockup detected after %d but scratchpad register is zero\n",
+ h->heartbeat_sample_interval / HZ);
lockup_detected = 0xffffffff;
}
set_lockup_detected_for_all_cpus(h, lockup_detected);
spin_unlock_irqrestore(&h->lock, flags);
- dev_warn(&h->pdev->dev, "Controller lockup detected: 0x%08x\n",
- lockup_detected);
+ dev_warn(&h->pdev->dev, "Controller lockup detected: 0x%08x after %d\n",
+ lockup_detected, h->heartbeat_sample_interval / HZ);
pci_disable_device(h->pdev);
fail_all_outstanding_cmds(h);
}
-static void detect_controller_lockup(struct ctlr_info *h)
+static int detect_controller_lockup(struct ctlr_info *h)
{
u64 now;
u32 heartbeat;
/* If we've received an interrupt recently, we're ok. */
if (time_after64(h->last_intr_timestamp +
(h->heartbeat_sample_interval), now))
- return;
+ return false;
/*
* If we've already checked the heartbeat recently, we're ok.
*/
if (time_after64(h->last_heartbeat_timestamp +
(h->heartbeat_sample_interval), now))
- return;
+ return false;
/* If heartbeat has not changed since we last looked, we're not ok. */
spin_lock_irqsave(&h->lock, flags);
spin_unlock_irqrestore(&h->lock, flags);
if (h->last_heartbeat == heartbeat) {
controller_lockup_detected(h);
- return;
+ return true;
}
/* We're ok. */
h->last_heartbeat = heartbeat;
h->last_heartbeat_timestamp = now;
+ return false;
}
static void hpsa_ack_ctlr_events(struct ctlr_info *h)
spin_lock_init(&h->offline_device_lock);
spin_lock_init(&h->scan_lock);
atomic_set(&h->passthru_cmds_avail, HPSA_MAX_CONCURRENT_PASSTHRUS);
+ atomic_set(&h->abort_cmds_available, HPSA_CMDS_RESERVED_FOR_ABORTS);
h->rescan_ctlr_wq = hpsa_create_controller_wq(h, "rescan");
if (!h->rescan_ctlr_wq) {
if (hpsa_allocate_sg_chain_blocks(h))
goto clean4;
init_waitqueue_head(&h->scan_wait_queue);
+ init_waitqueue_head(&h->abort_cmd_wait_queue);
h->scan_finished = 1; /* no scan currently in progress */
pci_set_drvdata(pdev, h);
{
char *flush_buf;
struct CommandList *c;
+ int rc;
/* Don't bother trying to flush the cache if locked up */
+ /* FIXME not necessary if do_simple_cmd does the check */
if (unlikely(lockup_detected(h)))
return;
flush_buf = kzalloc(4, GFP_KERNEL);
RAID_CTLR_LUNID, TYPE_CMD)) {
goto out;
}
- hpsa_scsi_do_simple_cmd_with_retry(h, c, PCI_DMA_TODEVICE);
+ rc = hpsa_scsi_do_simple_cmd_with_retry(h, c,
+ PCI_DMA_TODEVICE, NO_TIMEOUT);
+ if (rc)
+ goto out;
if (c->err_info->CommandStatus != 0)
out:
dev_warn(&h->pdev->dev,