#include <linux/kthread.h>
#define CCISS_DRIVER_VERSION(maj,min,submin) ((maj<<16)|(min<<8)|(submin))
-#define DRIVER_NAME "HP CISS Driver (v 3.6.20)"
-#define DRIVER_VERSION CCISS_DRIVER_VERSION(3, 6, 20)
+#define DRIVER_NAME "HP CISS Driver (v 3.6.26)"
+#define DRIVER_VERSION CCISS_DRIVER_VERSION(3, 6, 26)
/* Embedded module documentation macros - see modules.h */
MODULE_AUTHOR("Hewlett-Packard Company");
MODULE_DESCRIPTION("Driver for HP Smart Array Controllers");
-MODULE_SUPPORTED_DEVICE("HP SA5i SA5i+ SA532 SA5300 SA5312 SA641 SA642 SA6400"
- " SA6i P600 P800 P400 P400i E200 E200i E500 P700m"
- " Smart Array G2 Series SAS/SATA Controllers");
-MODULE_VERSION("3.6.20");
+MODULE_SUPPORTED_DEVICE("HP Smart Array Controllers");
+MODULE_VERSION("3.6.26");
MODULE_LICENSE("GPL");
static int cciss_allow_hpsa;
{PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x3249},
{PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x324A},
{PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x324B},
+ {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x3250},
+ {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x3251},
+ {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x3252},
+ {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x3253},
+ {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x3254},
{0,}
};
{0x3249103C, "Smart Array P812", &SA5_access},
{0x324A103C, "Smart Array P712m", &SA5_access},
{0x324B103C, "Smart Array P711m", &SA5_access},
+ {0x3250103C, "Smart Array", &SA5_access},
+ {0x3251103C, "Smart Array", &SA5_access},
+ {0x3252103C, "Smart Array", &SA5_access},
+ {0x3253103C, "Smart Array", &SA5_access},
+ {0x3254103C, "Smart Array", &SA5_access},
};
/* How long to wait (in milliseconds) for board to go into simple mode */
static LIST_HEAD(scan_q);
static void do_cciss_request(struct request_queue *q);
-static irqreturn_t do_cciss_intr(int irq, void *dev_id);
+static irqreturn_t do_cciss_intx(int irq, void *dev_id);
+static irqreturn_t do_cciss_msix_intr(int irq, void *dev_id);
static int cciss_open(struct block_device *bdev, fmode_t mode);
+static int cciss_unlocked_open(struct block_device *bdev, fmode_t mode);
static int cciss_release(struct gendisk *disk, fmode_t mode);
+static int do_ioctl(struct block_device *bdev, fmode_t mode,
+ unsigned int cmd, unsigned long arg);
static int cciss_ioctl(struct block_device *bdev, fmode_t mode,
unsigned int cmd, unsigned long arg);
static int cciss_getgeo(struct block_device *bdev, struct hd_geometry *geo);
sector_t total_size,
unsigned int block_size, InquiryData_struct *inq_buff,
drive_info_struct *drv);
-static void __devinit cciss_interrupt_mode(ctlr_info_t *, struct pci_dev *,
- __u32);
+static void __devinit cciss_interrupt_mode(ctlr_info_t *);
static void start_io(ctlr_info_t *h);
static int sendcmd_withirq(__u8 cmd, int ctlr, void *buff, size_t size,
__u8 page_code, unsigned char scsi3addr[],
int attempt_retry);
static int process_sendcmd_error(ctlr_info_t *h, CommandList_struct *c);
-static void fail_all_cmds(unsigned long ctlr);
static int add_to_scan_list(struct ctlr_info *h);
static int scan_thread(void *data);
static int check_for_unit_attention(ctlr_info_t *h, CommandList_struct *c);
static void cciss_device_release(struct device *dev);
static void cciss_free_gendisk(ctlr_info_t *h, int drv_index);
static void cciss_free_drive_info(ctlr_info_t *h, int drv_index);
+static inline u32 next_command(ctlr_info_t *h);
+static int __devinit cciss_find_cfg_addrs(struct pci_dev *pdev,
+ void __iomem *vaddr, u32 *cfg_base_addr, u64 *cfg_base_addr_index,
+ u64 *cfg_offset);
+static int __devinit cciss_pci_find_memory_BAR(struct pci_dev *pdev,
+ unsigned long *memory_bar);
+
+
+/* performant mode helper functions */
+static void calc_bucket_map(int *bucket, int num_buckets, int nsgs,
+ int *bucket_map);
+static void cciss_put_controller_into_performant_mode(ctlr_info_t *h);
#ifdef CONFIG_PROC_FS
static void cciss_procinit(int i);
static const struct block_device_operations cciss_fops = {
.owner = THIS_MODULE,
- .open = cciss_open,
+ .open = cciss_unlocked_open,
.release = cciss_release,
- .locked_ioctl = cciss_ioctl,
+ .ioctl = do_ioctl,
.getgeo = cciss_getgeo,
#ifdef CONFIG_COMPAT
.compat_ioctl = cciss_compat_ioctl,
.revalidate_disk = cciss_revalidate,
};
+/* 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(ctlr_info_t *h, CommandList_struct *c)
+{
+ if (likely(h->transMethod == CFGTBL_Trans_Performant))
+ c->busaddr |= 1 | (h->blockFetchTable[c->Header.SGList] << 1);
+}
+
/*
* Enqueuing and dequeuing functions for cmdlists.
*/
hlist_del_init(&c->list);
}
+static void enqueue_cmd_and_start_io(ctlr_info_t *h,
+ CommandList_struct *c)
+{
+ unsigned long flags;
+ set_performant_mode(h, c);
+ spin_lock_irqsave(&h->lock, flags);
+ addQ(&h->reqQ, c);
+ h->Qdepth++;
+ start_io(h);
+ spin_unlock_irqrestore(&h->lock, flags);
+}
+
static void cciss_free_sg_chain_blocks(SGDescriptor_struct **cmd_sg_list,
int nr_cmds)
{
h->product_name,
(unsigned long)h->board_id,
h->firm_ver[0], h->firm_ver[1], h->firm_ver[2],
- h->firm_ver[3], (unsigned int)h->intr[SIMPLE_MODE_INT],
+ h->firm_ver[3], (unsigned int)h->intr[PERF_MODE_INT],
h->num_luns,
h->Qdepth, h->commands_outstanding,
h->maxQsinceinit, h->max_outstanding, h->maxSG);
return 0;
}
+static int cciss_unlocked_open(struct block_device *bdev, fmode_t mode)
+{
+ int ret;
+
+ lock_kernel();
+ ret = cciss_open(bdev, mode);
+ unlock_kernel();
+
+ return ret;
+}
+
/*
* Close. Sync first.
*/
static int cciss_release(struct gendisk *disk, fmode_t mode)
{
- ctlr_info_t *host = get_host(disk);
- drive_info_struct *drv = get_drv(disk);
+ ctlr_info_t *host;
+ drive_info_struct *drv;
+
+ lock_kernel();
+ host = get_host(disk);
+ drv = get_drv(disk);
#ifdef CCISS_DEBUG
printk(KERN_DEBUG "cciss_release %s\n", disk->disk_name);
drv->usage_count--;
host->usage_count--;
+ unlock_kernel();
return 0;
}
-#ifdef CONFIG_COMPAT
-
static int do_ioctl(struct block_device *bdev, fmode_t mode,
unsigned cmd, unsigned long arg)
{
return ret;
}
+#ifdef CONFIG_COMPAT
+
static int cciss_ioctl32_passthru(struct block_device *bdev, fmode_t mode,
unsigned cmd, unsigned long arg);
static int cciss_ioctl32_big_passthru(struct block_device *bdev, fmode_t mode,
CommandList_struct *c;
char *buff = NULL;
u64bit temp64;
- unsigned long flags;
DECLARE_COMPLETION_ONSTACK(wait);
if (!arg)
}
c->waiting = &wait;
- /* Put the request on the tail of the request queue */
- spin_lock_irqsave(CCISS_LOCK(ctlr), flags);
- addQ(&host->reqQ, c);
- host->Qdepth++;
- start_io(host);
- spin_unlock_irqrestore(CCISS_LOCK(ctlr), flags);
-
+ enqueue_cmd_and_start_io(host, c);
wait_for_completion(&wait);
/* unlock the buffers from DMA */
unsigned char **buff = NULL;
int *buff_size = NULL;
u64bit temp64;
- unsigned long flags;
BYTE sg_used = 0;
int status = 0;
int i;
}
}
c->waiting = &wait;
- /* Put the request on the tail of the request queue */
- spin_lock_irqsave(CCISS_LOCK(ctlr), flags);
- addQ(&host->reqQ, c);
- host->Qdepth++;
- start_io(host);
- spin_unlock_irqrestore(CCISS_LOCK(ctlr), flags);
+ enqueue_cmd_and_start_io(host, c);
wait_for_completion(&wait);
/* unlock the buffers from DMA */
for (i = 0; i < sg_used; i++) {
CommandList_struct *cmd = rq->completion_data;
ctlr_info_t *h = hba[cmd->ctlr];
SGDescriptor_struct *curr_sg = cmd->SG;
- unsigned long flags;
u64bit temp64;
+ unsigned long flags;
int i, ddir;
int sg_index = 0;
#endif /* CCISS_DEBUG */
/* set the residual count for pc requests */
- if (blk_pc_request(rq))
+ if (rq->cmd_type == REQ_TYPE_BLOCK_PC)
rq->resid_len = cmd->err_info->ResidualCnt;
blk_end_request_all(rq, (rq->errors == 0) ? 0 : -EIO);
{
DECLARE_COMPLETION_ONSTACK(wait);
u64bit buff_dma_handle;
- unsigned long flags;
int return_status = IO_OK;
resend_cmd2:
c->waiting = &wait;
- /* Put the request on the tail of the queue and send it */
- spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags);
- addQ(&h->reqQ, c);
- h->Qdepth++;
- start_io(h);
- spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags);
+ enqueue_cmd_and_start_io(h, c);
wait_for_completion(&wait);
driver_byte = DRIVER_OK;
msg_byte = cmd->err_info->CommandStatus; /* correct? seems too device specific */
- if (blk_pc_request(cmd->rq))
+ if (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC)
host_byte = DID_PASSTHROUGH;
else
host_byte = DID_OK;
host_byte, driver_byte);
if (cmd->err_info->ScsiStatus != SAM_STAT_CHECK_CONDITION) {
- if (!blk_pc_request(cmd->rq))
+ if (cmd->rq->cmd_type != REQ_TYPE_BLOCK_PC)
printk(KERN_WARNING "cciss: cmd %p "
"has SCSI Status 0x%x\n",
cmd, cmd->err_info->ScsiStatus);
/* check the sense key */
sense_key = 0xf & cmd->err_info->SenseInfo[2];
/* no status or recovered error */
- if (((sense_key == 0x0) || (sense_key == 0x1)) && !blk_pc_request(cmd->rq))
+ if (((sense_key == 0x0) || (sense_key == 0x1)) &&
+ (cmd->rq->cmd_type != REQ_TYPE_BLOCK_PC))
error_value = 0;
if (check_for_unit_attention(h, cmd)) {
- *retry_cmd = !blk_pc_request(cmd->rq);
+ *retry_cmd = !(cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC);
return 0;
}
- if (!blk_pc_request(cmd->rq)) { /* Not SG_IO or similar? */
+ /* Not SG_IO or similar? */
+ if (cmd->rq->cmd_type != REQ_TYPE_BLOCK_PC) {
if (error_value != 0)
printk(KERN_WARNING "cciss: cmd %p has CHECK CONDITION"
" sense key = 0x%x\n", cmd, sense_key);
rq->errors = evaluate_target_status(h, cmd, &retry_cmd);
break;
case CMD_DATA_UNDERRUN:
- if (blk_fs_request(cmd->rq)) {
+ if (cmd->rq->cmd_type == REQ_TYPE_FS) {
printk(KERN_WARNING "cciss: cmd %p has"
" completed with data underrun "
"reported\n", cmd);
}
break;
case CMD_DATA_OVERRUN:
- if (blk_fs_request(cmd->rq))
+ if (cmd->rq->cmd_type == REQ_TYPE_FS)
printk(KERN_WARNING "cciss: cmd %p has"
" completed with data overrun "
"reported\n", cmd);
"reported invalid\n", cmd);
rq->errors = make_status_bytes(SAM_STAT_GOOD,
cmd->err_info->CommandStatus, DRIVER_OK,
- blk_pc_request(cmd->rq) ? DID_PASSTHROUGH : DID_ERROR);
+ (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ?
+ DID_PASSTHROUGH : DID_ERROR);
break;
case CMD_PROTOCOL_ERR:
printk(KERN_WARNING "cciss: cmd %p has "
"protocol error \n", cmd);
rq->errors = make_status_bytes(SAM_STAT_GOOD,
cmd->err_info->CommandStatus, DRIVER_OK,
- blk_pc_request(cmd->rq) ? DID_PASSTHROUGH : DID_ERROR);
+ (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ?
+ DID_PASSTHROUGH : DID_ERROR);
break;
case CMD_HARDWARE_ERR:
printk(KERN_WARNING "cciss: cmd %p had "
" hardware error\n", cmd);
rq->errors = make_status_bytes(SAM_STAT_GOOD,
cmd->err_info->CommandStatus, DRIVER_OK,
- blk_pc_request(cmd->rq) ? DID_PASSTHROUGH : DID_ERROR);
+ (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ?
+ DID_PASSTHROUGH : DID_ERROR);
break;
case CMD_CONNECTION_LOST:
printk(KERN_WARNING "cciss: cmd %p had "
"connection lost\n", cmd);
rq->errors = make_status_bytes(SAM_STAT_GOOD,
cmd->err_info->CommandStatus, DRIVER_OK,
- blk_pc_request(cmd->rq) ? DID_PASSTHROUGH : DID_ERROR);
+ (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ?
+ DID_PASSTHROUGH : DID_ERROR);
break;
case CMD_ABORTED:
printk(KERN_WARNING "cciss: cmd %p was "
"aborted\n", cmd);
rq->errors = make_status_bytes(SAM_STAT_GOOD,
cmd->err_info->CommandStatus, DRIVER_OK,
- blk_pc_request(cmd->rq) ? DID_PASSTHROUGH : DID_ABORT);
+ (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ?
+ DID_PASSTHROUGH : DID_ABORT);
break;
case CMD_ABORT_FAILED:
printk(KERN_WARNING "cciss: cmd %p reports "
"abort failed\n", cmd);
rq->errors = make_status_bytes(SAM_STAT_GOOD,
cmd->err_info->CommandStatus, DRIVER_OK,
- blk_pc_request(cmd->rq) ? DID_PASSTHROUGH : DID_ERROR);
+ (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ?
+ DID_PASSTHROUGH : DID_ERROR);
break;
case CMD_UNSOLICITED_ABORT:
printk(KERN_WARNING "cciss%d: unsolicited "
"many times\n", h->ctlr, cmd);
rq->errors = make_status_bytes(SAM_STAT_GOOD,
cmd->err_info->CommandStatus, DRIVER_OK,
- blk_pc_request(cmd->rq) ? DID_PASSTHROUGH : DID_ABORT);
+ (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ?
+ DID_PASSTHROUGH : DID_ABORT);
break;
case CMD_TIMEOUT:
printk(KERN_WARNING "cciss: cmd %p timedout\n", cmd);
rq->errors = make_status_bytes(SAM_STAT_GOOD,
cmd->err_info->CommandStatus, DRIVER_OK,
- blk_pc_request(cmd->rq) ? DID_PASSTHROUGH : DID_ERROR);
+ (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ?
+ DID_PASSTHROUGH : DID_ERROR);
break;
default:
printk(KERN_WARNING "cciss: cmd %p returned "
cmd->err_info->CommandStatus);
rq->errors = make_status_bytes(SAM_STAT_GOOD,
cmd->err_info->CommandStatus, DRIVER_OK,
- blk_pc_request(cmd->rq) ? DID_PASSTHROUGH : DID_ERROR);
+ (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ?
+ DID_PASSTHROUGH : DID_ERROR);
}
after_error_processing:
blk_complete_request(cmd->rq);
}
+static inline u32 cciss_tag_contains_index(u32 tag)
+{
+#define DIRECT_LOOKUP_BIT 0x10
+ return tag & DIRECT_LOOKUP_BIT;
+}
+
+static inline u32 cciss_tag_to_index(u32 tag)
+{
+#define DIRECT_LOOKUP_SHIFT 5
+ return tag >> DIRECT_LOOKUP_SHIFT;
+}
+
+static inline u32 cciss_tag_discard_error_bits(u32 tag)
+{
+#define CCISS_ERROR_BITS 0x03
+ return tag & ~CCISS_ERROR_BITS;
+}
+
+static inline void cciss_mark_tag_indexed(u32 *tag)
+{
+ *tag |= DIRECT_LOOKUP_BIT;
+}
+
+static inline void cciss_set_tag_index(u32 *tag, u32 index)
+{
+ *tag |= (index << DIRECT_LOOKUP_SHIFT);
+}
+
/*
* Get a request and submit it to the controller.
*/
/* got command from pool, so use the command block index instead */
/* for direct lookups. */
/* The first 2 bits are reserved for controller error reporting. */
- c->Header.Tag.lower = (c->cmdindex << 3);
- c->Header.Tag.lower |= 0x04; /* flag for direct lookup. */
+ cciss_set_tag_index(&c->Header.Tag.lower, c->cmdindex);
+ cciss_mark_tag_indexed(&c->Header.Tag.lower);
memcpy(&c->Header.LUN, drv->LunID, sizeof(drv->LunID));
c->Request.CDBLen = 10; /* 12 byte commands not in FW yet; */
c->Request.Type.Type = TYPE_CMD; /* It is a command. */
blk_rq_sectors(creq), seg, chained);
#endif /* CCISS_DEBUG */
- c->Header.SGList = c->Header.SGTotal = seg + chained;
- if (seg > h->max_cmd_sgentries)
+ c->Header.SGTotal = seg + chained;
+ if (seg <= h->max_cmd_sgentries)
+ c->Header.SGList = c->Header.SGTotal;
+ else
c->Header.SGList = h->max_cmd_sgentries;
+ set_performant_mode(h, c);
- if (likely(blk_fs_request(creq))) {
+ if (likely(creq->cmd_type == REQ_TYPE_FS)) {
if(h->cciss_read == CCISS_READ_10) {
c->Request.CDB[1] = 0;
c->Request.CDB[2] = (start_blk >> 24) & 0xff; /* MSB */
c->Request.CDB[13]= blk_rq_sectors(creq) & 0xff;
c->Request.CDB[14] = c->Request.CDB[15] = 0;
}
- } else if (blk_pc_request(creq)) {
+ } else if (creq->cmd_type == REQ_TYPE_BLOCK_PC) {
c->Request.CDBLen = creq->cmd_len;
memcpy(c->Request.CDB, creq->cmd, BLK_MAX_CDB);
} else {
static inline long interrupt_not_for_us(ctlr_info_t *h)
{
- return (((h->access.intr_pending(h) == 0) ||
- (h->interrupts_enabled == 0)));
+ return !(h->msi_vector || h->msix_vector) &&
+ ((h->access.intr_pending(h) == 0) ||
+ (h->interrupts_enabled == 0));
}
-static irqreturn_t do_cciss_intr(int irq, void *dev_id)
+static inline int bad_tag(ctlr_info_t *h, u32 tag_index,
+ u32 raw_tag)
{
- ctlr_info_t *h = dev_id;
+ if (unlikely(tag_index >= h->nr_cmds)) {
+ dev_warn(&h->pdev->dev, "bad tag 0x%08x ignored.\n", raw_tag);
+ return 1;
+ }
+ return 0;
+}
+
+static inline void finish_cmd(ctlr_info_t *h, CommandList_struct *c,
+ u32 raw_tag)
+{
+ removeQ(c);
+ if (likely(c->cmd_type == CMD_RWREQ))
+ complete_command(h, c, 0);
+ else if (c->cmd_type == CMD_IOCTL_PEND)
+ complete(c->waiting);
+#ifdef CONFIG_CISS_SCSI_TAPE
+ else if (c->cmd_type == CMD_SCSI)
+ complete_scsi_command(c, 0, raw_tag);
+#endif
+}
+
+static inline u32 next_command(ctlr_info_t *h)
+{
+ u32 a;
+
+ if (unlikely(h->transMethod != CFGTBL_Trans_Performant))
+ return h->access.command_completed(h);
+
+ if ((*(h->reply_pool_head) & 1) == (h->reply_pool_wraparound)) {
+ a = *(h->reply_pool_head); /* Next cmd in ring buffer */
+ (h->reply_pool_head)++;
+ h->commands_outstanding--;
+ } else {
+ a = FIFO_EMPTY;
+ }
+ /* Check for wraparound */
+ if (h->reply_pool_head == (h->reply_pool + h->max_commands)) {
+ h->reply_pool_head = h->reply_pool;
+ h->reply_pool_wraparound ^= 1;
+ }
+ return a;
+}
+
+/* process completion of an indexed ("direct lookup") command */
+static inline u32 process_indexed_cmd(ctlr_info_t *h, u32 raw_tag)
+{
+ u32 tag_index;
CommandList_struct *c;
+
+ tag_index = cciss_tag_to_index(raw_tag);
+ if (bad_tag(h, tag_index, raw_tag))
+ return next_command(h);
+ c = h->cmd_pool + tag_index;
+ finish_cmd(h, c, raw_tag);
+ return next_command(h);
+}
+
+/* process completion of a non-indexed command */
+static inline u32 process_nonindexed_cmd(ctlr_info_t *h, u32 raw_tag)
+{
+ u32 tag;
+ CommandList_struct *c = NULL;
+ struct hlist_node *tmp;
+ __u32 busaddr_masked, tag_masked;
+
+ tag = cciss_tag_discard_error_bits(raw_tag);
+ hlist_for_each_entry(c, tmp, &h->cmpQ, list) {
+ busaddr_masked = cciss_tag_discard_error_bits(c->busaddr);
+ tag_masked = cciss_tag_discard_error_bits(tag);
+ if (busaddr_masked == tag_masked) {
+ finish_cmd(h, c, raw_tag);
+ return next_command(h);
+ }
+ }
+ bad_tag(h, h->nr_cmds + 1, raw_tag);
+ return next_command(h);
+}
+
+static irqreturn_t do_cciss_intx(int irq, void *dev_id)
+{
+ ctlr_info_t *h = dev_id;
unsigned long flags;
- __u32 a, a1, a2;
+ u32 raw_tag;
if (interrupt_not_for_us(h))
return IRQ_NONE;
*/
spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags);
while (interrupt_pending(h)) {
- while ((a = get_next_completion(h)) != FIFO_EMPTY) {
- a1 = a;
- if ((a & 0x04)) {
- a2 = (a >> 3);
- if (a2 >= h->nr_cmds) {
- printk(KERN_WARNING
- "cciss: controller cciss%d failed, stopping.\n",
- h->ctlr);
- spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags);
- fail_all_cmds(h->ctlr);
- return IRQ_HANDLED;
- }
+ raw_tag = get_next_completion(h);
+ while (raw_tag != FIFO_EMPTY) {
+ if (cciss_tag_contains_index(raw_tag))
+ raw_tag = process_indexed_cmd(h, raw_tag);
+ else
+ raw_tag = process_nonindexed_cmd(h, raw_tag);
+ }
+ }
- c = h->cmd_pool + a2;
- a = c->busaddr;
+ spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags);
+ return IRQ_HANDLED;
+}
- } else {
- struct hlist_node *tmp;
+/* Add a second interrupt handler for MSI/MSI-X mode. In this mode we never
+ * check the interrupt pending register because it is not set.
+ */
+static irqreturn_t do_cciss_msix_intr(int irq, void *dev_id)
+{
+ ctlr_info_t *h = dev_id;
+ unsigned long flags;
+ u32 raw_tag;
- a &= ~3;
- c = NULL;
- hlist_for_each_entry(c, tmp, &h->cmpQ, list) {
- if (c->busaddr == a)
- break;
- }
- }
- /*
- * If we've found the command, take it off the
- * completion Q and free it
- */
- if (c && c->busaddr == a) {
- removeQ(c);
- if (c->cmd_type == CMD_RWREQ) {
- complete_command(h, c, 0);
- } else if (c->cmd_type == CMD_IOCTL_PEND) {
- complete(c->waiting);
- }
-# ifdef CONFIG_CISS_SCSI_TAPE
- else if (c->cmd_type == CMD_SCSI)
- complete_scsi_command(c, 0, a1);
-# endif
- continue;
- }
- }
+ if (interrupt_not_for_us(h))
+ return IRQ_NONE;
+ /*
+ * If there are completed commands in the completion queue,
+ * we had better do something about it.
+ */
+ spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags);
+ raw_tag = get_next_completion(h);
+ while (raw_tag != FIFO_EMPTY) {
+ if (cciss_tag_contains_index(raw_tag))
+ raw_tag = process_indexed_cmd(h, raw_tag);
+ else
+ raw_tag = process_nonindexed_cmd(h, raw_tag);
}
spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags);
* the io functions.
* This is for debug only.
*/
-#ifdef CCISS_DEBUG
static void print_cfg_table(CfgTable_struct *tb)
{
+#ifdef CCISS_DEBUG
int i;
char temp_name[17];
temp_name[16] = '\0';
printk(" Server Name = %s\n", temp_name);
printk(" Heartbeat Counter = 0x%x\n\n\n", readl(&(tb->HeartBeat)));
-}
#endif /* CCISS_DEBUG */
+}
static int find_PCI_BAR_index(struct pci_dev *pdev, unsigned long pci_bar_addr)
{
return -1;
}
+/* Fill in bucket_map[], given nsgs (the max number of
+ * scatter gather elements supported) and bucket[],
+ * which is an array of 8 integers. The bucket[] array
+ * contains 8 different DMA transfer sizes (in 16
+ * byte increments) which the controller uses to fetch
+ * commands. This function fills in bucket_map[], which
+ * maps a given number of scatter gather elements to one of
+ * the 8 DMA transfer sizes. The point of it is to allow the
+ * controller to only do as much DMA as needed to fetch the
+ * command, with the DMA transfer size encoded in the lower
+ * bits of the command address.
+ */
+static void calc_bucket_map(int bucket[], int num_buckets,
+ int nsgs, int *bucket_map)
+{
+ int i, j, b, size;
+
+ /* even a command with 0 SGs requires 4 blocks */
+#define MINIMUM_TRANSFER_BLOCKS 4
+#define NUM_BUCKETS 8
+ /* Note, bucket_map must have nsgs+1 entries. */
+ for (i = 0; i <= nsgs; i++) {
+ /* Compute size of a command with i SG entries */
+ size = i + MINIMUM_TRANSFER_BLOCKS;
+ b = num_buckets; /* Assume the biggest bucket */
+ /* Find the bucket that is just big enough */
+ for (j = 0; j < 8; j++) {
+ if (bucket[j] >= size) {
+ b = j;
+ break;
+ }
+ }
+ /* for a command with i SG entries, use bucket b. */
+ bucket_map[i] = b;
+ }
+}
+
+static void __devinit cciss_wait_for_mode_change_ack(ctlr_info_t *h)
+{
+ int i;
+
+ /* under certain very rare conditions, this can take awhile.
+ * (e.g.: hot replace a failed 144GB drive in a RAID 5 set right
+ * as we enter this code.) */
+ for (i = 0; i < MAX_CONFIG_WAIT; i++) {
+ if (!(readl(h->vaddr + SA5_DOORBELL) & CFGTBL_ChangeReq))
+ break;
+ msleep(10);
+ }
+}
+
+static __devinit void cciss_enter_performant_mode(ctlr_info_t *h)
+{
+ /* This is a bit complicated. There are 8 registers on
+ * the controller which we write to to tell it 8 different
+ * sizes of commands which there may be. It's a way of
+ * reducing the DMA done to fetch each command. Encoded into
+ * each command's tag are 3 bits which communicate to the controller
+ * which of the eight sizes that command fits within. The size of
+ * each command depends on how many scatter gather entries there are.
+ * Each SG entry requires 16 bytes. The eight registers are programmed
+ * with the number of 16-byte blocks a command of that size requires.
+ * The smallest command possible requires 5 such 16 byte blocks.
+ * the largest command possible requires MAXSGENTRIES + 4 16-byte
+ * blocks. Note, this only extends to the SG entries contained
+ * within the command block, and does not extend to chained blocks
+ * of SG elements. bft[] contains the eight values we write to
+ * the registers. They are not evenly distributed, but have more
+ * sizes for small commands, and fewer sizes for larger commands.
+ */
+ __u32 trans_offset;
+ int bft[8] = { 5, 6, 8, 10, 12, 20, 28, MAXSGENTRIES + 4};
+ /*
+ * 5 = 1 s/g entry or 4k
+ * 6 = 2 s/g entry or 8k
+ * 8 = 4 s/g entry or 16k
+ * 10 = 6 s/g entry or 24k
+ */
+ unsigned long register_value;
+ BUILD_BUG_ON(28 > MAXSGENTRIES + 4);
+
+ h->reply_pool_wraparound = 1; /* spec: init to 1 */
+
+ /* Controller spec: zero out this buffer. */
+ memset(h->reply_pool, 0, h->max_commands * sizeof(__u64));
+ h->reply_pool_head = h->reply_pool;
+
+ trans_offset = readl(&(h->cfgtable->TransMethodOffset));
+ calc_bucket_map(bft, ARRAY_SIZE(bft), h->maxsgentries,
+ h->blockFetchTable);
+ writel(bft[0], &h->transtable->BlockFetch0);
+ writel(bft[1], &h->transtable->BlockFetch1);
+ writel(bft[2], &h->transtable->BlockFetch2);
+ writel(bft[3], &h->transtable->BlockFetch3);
+ writel(bft[4], &h->transtable->BlockFetch4);
+ writel(bft[5], &h->transtable->BlockFetch5);
+ writel(bft[6], &h->transtable->BlockFetch6);
+ writel(bft[7], &h->transtable->BlockFetch7);
+
+ /* size of controller ring buffer */
+ writel(h->max_commands, &h->transtable->RepQSize);
+ writel(1, &h->transtable->RepQCount);
+ writel(0, &h->transtable->RepQCtrAddrLow32);
+ writel(0, &h->transtable->RepQCtrAddrHigh32);
+ writel(h->reply_pool_dhandle, &h->transtable->RepQAddr0Low32);
+ writel(0, &h->transtable->RepQAddr0High32);
+ writel(CFGTBL_Trans_Performant,
+ &(h->cfgtable->HostWrite.TransportRequest));
+
+ writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL);
+ cciss_wait_for_mode_change_ack(h);
+ register_value = readl(&(h->cfgtable->TransportActive));
+ if (!(register_value & CFGTBL_Trans_Performant))
+ printk(KERN_WARNING "cciss: unable to get board into"
+ " performant mode\n");
+}
+
+static void __devinit cciss_put_controller_into_performant_mode(ctlr_info_t *h)
+{
+ __u32 trans_support;
+
+ dev_dbg(&h->pdev->dev, "Trying to put board into Performant mode\n");
+ /* Attempt to put controller into performant mode if supported */
+ /* Does board support performant mode? */
+ trans_support = readl(&(h->cfgtable->TransportSupport));
+ if (!(trans_support & PERFORMANT_MODE))
+ return;
+
+ printk(KERN_WARNING "cciss%d: Placing controller into "
+ "performant mode\n", h->ctlr);
+ /* Performant mode demands commands on a 32 byte boundary
+ * pci_alloc_consistent aligns on page boundarys already.
+ * Just need to check if divisible by 32
+ */
+ if ((sizeof(CommandList_struct) % 32) != 0) {
+ printk(KERN_WARNING "%s %d %s\n",
+ "cciss info: command size[",
+ (int)sizeof(CommandList_struct),
+ "] not divisible by 32, no performant mode..\n");
+ return;
+ }
+
+ /* Performant mode ring buffer and supporting data structures */
+ h->reply_pool = (__u64 *)pci_alloc_consistent(
+ h->pdev, h->max_commands * sizeof(__u64),
+ &(h->reply_pool_dhandle));
+
+ /* Need a block fetch table for performant mode */
+ h->blockFetchTable = kmalloc(((h->maxsgentries+1) *
+ sizeof(__u32)), GFP_KERNEL);
+
+ if ((h->reply_pool == NULL) || (h->blockFetchTable == NULL))
+ goto clean_up;
+
+ cciss_enter_performant_mode(h);
+
+ /* Change the access methods to the performant access methods */
+ h->access = SA5_performant_access;
+ h->transMethod = CFGTBL_Trans_Performant;
+
+ return;
+clean_up:
+ kfree(h->blockFetchTable);
+ if (h->reply_pool)
+ pci_free_consistent(h->pdev,
+ h->max_commands * sizeof(__u64),
+ h->reply_pool,
+ h->reply_pool_dhandle);
+ return;
+
+} /* cciss_put_controller_into_performant_mode */
+
/* If MSI/MSI-X is supported by the kernel we will try to enable it on
* controllers that are capable. If not, we use IO-APIC mode.
*/
-static void __devinit cciss_interrupt_mode(ctlr_info_t *c,
- struct pci_dev *pdev, __u32 board_id)
+static void __devinit cciss_interrupt_mode(ctlr_info_t *c)
{
#ifdef CONFIG_PCI_MSI
int err;
};
/* Some boards advertise MSI but don't really support it */
- if ((board_id == 0x40700E11) ||
- (board_id == 0x40800E11) ||
- (board_id == 0x40820E11) || (board_id == 0x40830E11))
+ if ((c->board_id == 0x40700E11) || (c->board_id == 0x40800E11) ||
+ (c->board_id == 0x40820E11) || (c->board_id == 0x40830E11))
goto default_int_mode;
- if (pci_find_capability(pdev, PCI_CAP_ID_MSIX)) {
- err = pci_enable_msix(pdev, cciss_msix_entries, 4);
+ if (pci_find_capability(c->pdev, PCI_CAP_ID_MSIX)) {
+ err = pci_enable_msix(c->pdev, cciss_msix_entries, 4);
if (!err) {
c->intr[0] = cciss_msix_entries[0].vector;
c->intr[1] = cciss_msix_entries[1].vector;
goto default_int_mode;
}
}
- if (pci_find_capability(pdev, PCI_CAP_ID_MSI)) {
- if (!pci_enable_msi(pdev)) {
+ if (pci_find_capability(c->pdev, PCI_CAP_ID_MSI)) {
+ if (!pci_enable_msi(c->pdev)) {
c->msi_vector = 1;
} else {
printk(KERN_WARNING "cciss: MSI init failed\n");
default_int_mode:
#endif /* CONFIG_PCI_MSI */
/* if we get here we're going to use the default interrupt mode */
- c->intr[SIMPLE_MODE_INT] = pdev->irq;
+ c->intr[PERF_MODE_INT] = c->pdev->irq;
return;
}
-static int __devinit cciss_pci_init(ctlr_info_t *c, struct pci_dev *pdev)
+static int __devinit cciss_lookup_board_id(struct pci_dev *pdev, u32 *board_id)
{
- ushort subsystem_vendor_id, subsystem_device_id, command;
- __u32 board_id, scratchpad = 0;
- __u64 cfg_offset;
- __u32 cfg_base_addr;
- __u64 cfg_base_addr_index;
- int i, prod_index, err;
+ int i;
+ u32 subsystem_vendor_id, subsystem_device_id;
subsystem_vendor_id = pdev->subsystem_vendor;
subsystem_device_id = pdev->subsystem_device;
- board_id = (((__u32) (subsystem_device_id << 16) & 0xffff0000) |
- subsystem_vendor_id);
+ *board_id = ((subsystem_device_id << 16) & 0xffff0000) |
+ subsystem_vendor_id;
for (i = 0; i < ARRAY_SIZE(products); i++) {
/* Stand aside for hpsa driver on request */
if (cciss_allow_hpsa && products[i].board_id == HPSA_BOUNDARY)
return -ENODEV;
- if (board_id == products[i].board_id)
- break;
- }
- prod_index = i;
- if (prod_index == ARRAY_SIZE(products)) {
- dev_warn(&pdev->dev,
- "unrecognized board ID: 0x%08lx, ignoring.\n",
- (unsigned long) board_id);
- return -ENODEV;
+ if (*board_id == products[i].board_id)
+ return i;
}
+ dev_warn(&pdev->dev, "unrecognized board ID: 0x%08x, ignoring.\n",
+ *board_id);
+ return -ENODEV;
+}
- /* check to see if controller has been disabled */
- /* BEFORE trying to enable it */
- (void)pci_read_config_word(pdev, PCI_COMMAND, &command);
- if (!(command & 0x02)) {
- printk(KERN_WARNING
- "cciss: controller appears to be disabled\n");
- return -ENODEV;
- }
+static inline bool cciss_board_disabled(ctlr_info_t *h)
+{
+ u16 command;
- err = pci_enable_device(pdev);
- if (err) {
- printk(KERN_ERR "cciss: Unable to Enable PCI device\n");
- return err;
- }
+ (void) pci_read_config_word(h->pdev, PCI_COMMAND, &command);
+ return ((command & PCI_COMMAND_MEMORY) == 0);
+}
- err = pci_request_regions(pdev, "cciss");
- if (err) {
- printk(KERN_ERR "cciss: Cannot obtain PCI resources, "
- "aborting\n");
- return err;
- }
+static int __devinit cciss_pci_find_memory_BAR(struct pci_dev *pdev,
+ unsigned long *memory_bar)
+{
+ int i;
-#ifdef CCISS_DEBUG
- printk("command = %x\n", command);
- printk("irq = %x\n", pdev->irq);
- printk("board_id = %x\n", board_id);
-#endif /* CCISS_DEBUG */
+ for (i = 0; i < DEVICE_COUNT_RESOURCE; i++)
+ if (pci_resource_flags(pdev, i) & IORESOURCE_MEM) {
+ /* addressing mode bits already removed */
+ *memory_bar = pci_resource_start(pdev, i);
+ dev_dbg(&pdev->dev, "memory BAR = %lx\n",
+ *memory_bar);
+ return 0;
+ }
+ dev_warn(&pdev->dev, "no memory BAR found\n");
+ return -ENODEV;
+}
-/* If the kernel supports MSI/MSI-X we will try to enable that functionality,
- * else we use the IO-APIC interrupt assigned to us by system ROM.
- */
- cciss_interrupt_mode(c, pdev, board_id);
+static int __devinit cciss_wait_for_board_ready(ctlr_info_t *h)
+{
+ int i;
+ u32 scratchpad;
- /* find the memory BAR */
- for (i = 0; i < DEVICE_COUNT_RESOURCE; i++) {
- if (pci_resource_flags(pdev, i) & IORESOURCE_MEM)
- break;
+ for (i = 0; i < CCISS_BOARD_READY_ITERATIONS; i++) {
+ scratchpad = readl(h->vaddr + SA5_SCRATCHPAD_OFFSET);
+ if (scratchpad == CCISS_FIRMWARE_READY)
+ return 0;
+ msleep(CCISS_BOARD_READY_POLL_INTERVAL_MSECS);
}
- if (i == DEVICE_COUNT_RESOURCE) {
- printk(KERN_WARNING "cciss: No memory BAR found\n");
- err = -ENODEV;
- goto err_out_free_res;
+ dev_warn(&h->pdev->dev, "board not ready, timed out.\n");
+ return -ENODEV;
+}
+
+static int __devinit cciss_find_cfg_addrs(struct pci_dev *pdev,
+ void __iomem *vaddr, u32 *cfg_base_addr, u64 *cfg_base_addr_index,
+ u64 *cfg_offset)
+{
+ *cfg_base_addr = readl(vaddr + SA5_CTCFG_OFFSET);
+ *cfg_offset = readl(vaddr + SA5_CTMEM_OFFSET);
+ *cfg_base_addr &= (u32) 0x0000ffff;
+ *cfg_base_addr_index = find_PCI_BAR_index(pdev, *cfg_base_addr);
+ if (*cfg_base_addr_index == -1) {
+ dev_warn(&pdev->dev, "cannot find cfg_base_addr_index, "
+ "*cfg_base_addr = 0x%08x\n", *cfg_base_addr);
+ return -ENODEV;
}
+ return 0;
+}
- c->paddr = pci_resource_start(pdev, i); /* addressing mode bits
- * already removed
- */
+static int __devinit cciss_find_cfgtables(ctlr_info_t *h)
+{
+ u64 cfg_offset;
+ u32 cfg_base_addr;
+ u64 cfg_base_addr_index;
+ u32 trans_offset;
+ int rc;
-#ifdef CCISS_DEBUG
- printk("address 0 = %lx\n", c->paddr);
-#endif /* CCISS_DEBUG */
- c->vaddr = remap_pci_mem(c->paddr, 0x250);
+ rc = cciss_find_cfg_addrs(h->pdev, h->vaddr, &cfg_base_addr,
+ &cfg_base_addr_index, &cfg_offset);
+ if (rc)
+ return rc;
+ h->cfgtable = remap_pci_mem(pci_resource_start(h->pdev,
+ cfg_base_addr_index) + cfg_offset, sizeof(h->cfgtable));
+ if (!h->cfgtable)
+ return -ENOMEM;
+ /* Find performant mode table. */
+ trans_offset = readl(&h->cfgtable->TransMethodOffset);
+ h->transtable = remap_pci_mem(pci_resource_start(h->pdev,
+ cfg_base_addr_index)+cfg_offset+trans_offset,
+ sizeof(*h->transtable));
+ if (!h->transtable)
+ return -ENOMEM;
+ return 0;
+}
- /* Wait for the board to become ready. (PCI hotplug needs this.)
- * We poll for up to 120 secs, once per 100ms. */
- for (i = 0; i < 1200; i++) {
- scratchpad = readl(c->vaddr + SA5_SCRATCHPAD_OFFSET);
- if (scratchpad == CCISS_FIRMWARE_READY)
- break;
- set_current_state(TASK_INTERRUPTIBLE);
- schedule_timeout(msecs_to_jiffies(100)); /* wait 100ms */
+static void __devinit cciss_get_max_perf_mode_cmds(struct ctlr_info *h)
+{
+ h->max_commands = readl(&(h->cfgtable->MaxPerformantModeCommands));
+ if (h->max_commands < 16) {
+ dev_warn(&h->pdev->dev, "Controller reports "
+ "max supported commands of %d, an obvious lie. "
+ "Using 16. Ensure that firmware is up to date.\n",
+ h->max_commands);
+ h->max_commands = 16;
}
- if (scratchpad != CCISS_FIRMWARE_READY) {
- printk(KERN_WARNING "cciss: Board not ready. Timed out.\n");
- err = -ENODEV;
- goto err_out_free_res;
+}
+
+/* Interrogate the hardware for some limits:
+ * max commands, max SG elements without chaining, and with chaining,
+ * SG chain block size, etc.
+ */
+static void __devinit cciss_find_board_params(ctlr_info_t *h)
+{
+ cciss_get_max_perf_mode_cmds(h);
+ h->nr_cmds = h->max_commands - 4; /* Allow room for some ioctls */
+ h->maxsgentries = readl(&(h->cfgtable->MaxSGElements));
+ /*
+ * Limit in-command s/g elements to 32 save dma'able memory.
+ * Howvever spec says if 0, use 31
+ */
+ h->max_cmd_sgentries = 31;
+ if (h->maxsgentries > 512) {
+ h->max_cmd_sgentries = 32;
+ h->chainsize = h->maxsgentries - h->max_cmd_sgentries + 1;
+ h->maxsgentries--; /* save one for chain pointer */
+ } else {
+ h->maxsgentries = 31; /* default to traditional values */
+ h->chainsize = 0;
}
+}
- /* get the address index number */
- cfg_base_addr = readl(c->vaddr + SA5_CTCFG_OFFSET);
- cfg_base_addr &= (__u32) 0x0000ffff;
-#ifdef CCISS_DEBUG
- printk("cfg base address = %x\n", cfg_base_addr);
-#endif /* CCISS_DEBUG */
- cfg_base_addr_index = find_PCI_BAR_index(pdev, cfg_base_addr);
-#ifdef CCISS_DEBUG
- printk("cfg base address index = %llx\n",
- (unsigned long long)cfg_base_addr_index);
-#endif /* CCISS_DEBUG */
- if (cfg_base_addr_index == -1) {
- printk(KERN_WARNING "cciss: Cannot find cfg_base_addr_index\n");
- err = -ENODEV;
- goto err_out_free_res;
+static inline bool CISS_signature_present(ctlr_info_t *h)
+{
+ if ((readb(&h->cfgtable->Signature[0]) != 'C') ||
+ (readb(&h->cfgtable->Signature[1]) != 'I') ||
+ (readb(&h->cfgtable->Signature[2]) != 'S') ||
+ (readb(&h->cfgtable->Signature[3]) != 'S')) {
+ dev_warn(&h->pdev->dev, "not a valid CISS config table\n");
+ return false;
}
+ return true;
+}
- cfg_offset = readl(c->vaddr + SA5_CTMEM_OFFSET);
-#ifdef CCISS_DEBUG
- printk("cfg offset = %llx\n", (unsigned long long)cfg_offset);
-#endif /* CCISS_DEBUG */
- c->cfgtable = remap_pci_mem(pci_resource_start(pdev,
- cfg_base_addr_index) +
- cfg_offset, sizeof(CfgTable_struct));
- c->board_id = board_id;
+/* Need to enable prefetch in the SCSI core for 6400 in x86 */
+static inline void cciss_enable_scsi_prefetch(ctlr_info_t *h)
+{
+#ifdef CONFIG_X86
+ u32 prefetch;
-#ifdef CCISS_DEBUG
- print_cfg_table(c->cfgtable);
-#endif /* CCISS_DEBUG */
+ prefetch = readl(&(h->cfgtable->SCSI_Prefetch));
+ prefetch |= 0x100;
+ writel(prefetch, &(h->cfgtable->SCSI_Prefetch));
+#endif
+}
- /* Some controllers support Zero Memory Raid (ZMR).
- * When configured in ZMR mode the number of supported
- * commands drops to 64. So instead of just setting an
- * arbitrary value we make the driver a little smarter.
- * We read the config table to tell us how many commands
- * are supported on the controller then subtract 4 to
- * leave a little room for ioctl calls.
- */
- c->max_commands = readl(&(c->cfgtable->CmdsOutMax));
- c->maxsgentries = readl(&(c->cfgtable->MaxSGElements));
+/* Disable DMA prefetch for the P600. Otherwise an ASIC bug may result
+ * in a prefetch beyond physical memory.
+ */
+static inline void cciss_p600_dma_prefetch_quirk(ctlr_info_t *h)
+{
+ u32 dma_prefetch;
+ __u32 dma_refetch;
- /*
- * Limit native command to 32 s/g elements to save dma'able memory.
- * Howvever spec says if 0, use 31
- */
+ if (h->board_id != 0x3225103C)
+ return;
+ dma_prefetch = readl(h->vaddr + I2O_DMA1_CFG);
+ dma_prefetch |= 0x8000;
+ writel(dma_prefetch, h->vaddr + I2O_DMA1_CFG);
+ pci_read_config_dword(h->pdev, PCI_COMMAND_PARITY, &dma_refetch);
+ dma_refetch |= 0x1;
+ pci_write_config_dword(h->pdev, PCI_COMMAND_PARITY, dma_refetch);
+}
- c->max_cmd_sgentries = 31;
- if (c->maxsgentries > 512) {
- c->max_cmd_sgentries = 32;
- c->chainsize = c->maxsgentries - c->max_cmd_sgentries + 1;
- c->maxsgentries -= 1; /* account for chain pointer */
- } else {
- c->maxsgentries = 31; /* Default to traditional value */
- c->chainsize = 0; /* traditional */
- }
+static int __devinit cciss_pci_init(ctlr_info_t *c)
+{
+ int prod_index, err;
+ prod_index = cciss_lookup_board_id(c->pdev, &c->board_id);
+ if (prod_index < 0)
+ return -ENODEV;
c->product_name = products[prod_index].product_name;
c->access = *(products[prod_index].access);
- c->nr_cmds = c->max_commands - 4;
- if ((readb(&c->cfgtable->Signature[0]) != 'C') ||
- (readb(&c->cfgtable->Signature[1]) != 'I') ||
- (readb(&c->cfgtable->Signature[2]) != 'S') ||
- (readb(&c->cfgtable->Signature[3]) != 'S')) {
- printk("Does not appear to be a valid CISS config table\n");
- err = -ENODEV;
- goto err_out_free_res;
+
+ if (cciss_board_disabled(c)) {
+ printk(KERN_WARNING
+ "cciss: controller appears to be disabled\n");
+ return -ENODEV;
}
-#ifdef CONFIG_X86
- {
- /* Need to enable prefetch in the SCSI core for 6400 in x86 */
- __u32 prefetch;
- prefetch = readl(&(c->cfgtable->SCSI_Prefetch));
- prefetch |= 0x100;
- writel(prefetch, &(c->cfgtable->SCSI_Prefetch));
+ err = pci_enable_device(c->pdev);
+ if (err) {
+ printk(KERN_ERR "cciss: Unable to Enable PCI device\n");
+ return err;
}
-#endif
- /* Disabling DMA prefetch and refetch for the P600.
- * An ASIC bug may result in accesses to invalid memory addresses.
- * We've disabled prefetch for some time now. Testing with XEN
- * kernels revealed a bug in the refetch if dom0 resides on a P600.
- */
- if(board_id == 0x3225103C) {
- __u32 dma_prefetch;
- __u32 dma_refetch;
- dma_prefetch = readl(c->vaddr + I2O_DMA1_CFG);
- dma_prefetch |= 0x8000;
- writel(dma_prefetch, c->vaddr + I2O_DMA1_CFG);
- pci_read_config_dword(pdev, PCI_COMMAND_PARITY, &dma_refetch);
- dma_refetch |= 0x1;
- pci_write_config_dword(pdev, PCI_COMMAND_PARITY, dma_refetch);
+ err = pci_request_regions(c->pdev, "cciss");
+ if (err) {
+ printk(KERN_ERR "cciss: Cannot obtain PCI resources, "
+ "aborting\n");
+ return err;
}
#ifdef CCISS_DEBUG
- printk("Trying to put board into Simple mode\n");
+ printk(KERN_INFO "command = %x\n", command);
+ printk(KERN_INFO "irq = %x\n", c->pdev->irq);
+ printk(KERN_INFO "board_id = %x\n", c->board_id);
#endif /* CCISS_DEBUG */
- c->max_commands = readl(&(c->cfgtable->CmdsOutMax));
- /* Update the field, and then ring the doorbell */
- writel(CFGTBL_Trans_Simple, &(c->cfgtable->HostWrite.TransportRequest));
- writel(CFGTBL_ChangeReq, c->vaddr + SA5_DOORBELL);
- /* under certain very rare conditions, this can take awhile.
- * (e.g.: hot replace a failed 144GB drive in a RAID 5 set right
- * as we enter this code.) */
- for (i = 0; i < MAX_CONFIG_WAIT; i++) {
- if (!(readl(c->vaddr + SA5_DOORBELL) & CFGTBL_ChangeReq))
- break;
- /* delay and try again */
- set_current_state(TASK_INTERRUPTIBLE);
- schedule_timeout(msecs_to_jiffies(1));
+/* If the kernel supports MSI/MSI-X we will try to enable that functionality,
+ * else we use the IO-APIC interrupt assigned to us by system ROM.
+ */
+ cciss_interrupt_mode(c);
+ err = cciss_pci_find_memory_BAR(c->pdev, &c->paddr);
+ if (err)
+ goto err_out_free_res;
+ c->vaddr = remap_pci_mem(c->paddr, 0x250);
+ if (!c->vaddr) {
+ err = -ENOMEM;
+ goto err_out_free_res;
}
-
-#ifdef CCISS_DEBUG
- printk(KERN_DEBUG "I counter got to %d %x\n", i,
- readl(c->vaddr + SA5_DOORBELL));
-#endif /* CCISS_DEBUG */
-#ifdef CCISS_DEBUG
+ err = cciss_wait_for_board_ready(c);
+ if (err)
+ goto err_out_free_res;
+ err = cciss_find_cfgtables(c);
+ if (err)
+ goto err_out_free_res;
print_cfg_table(c->cfgtable);
-#endif /* CCISS_DEBUG */
+ cciss_find_board_params(c);
- if (!(readl(&(c->cfgtable->TransportActive)) & CFGTBL_Trans_Simple)) {
- printk(KERN_WARNING "cciss: unable to get board into"
- " simple mode\n");
+ if (!CISS_signature_present(c)) {
err = -ENODEV;
goto err_out_free_res;
}
+ cciss_enable_scsi_prefetch(c);
+ cciss_p600_dma_prefetch_quirk(c);
+ cciss_put_controller_into_performant_mode(c);
return 0;
err_out_free_res:
* Deliberately omit pci_disable_device(): it does something nasty to
* Smart Array controllers that pci_enable_device does not undo
*/
- pci_release_regions(pdev);
+ if (c->transtable)
+ iounmap(c->transtable);
+ if (c->cfgtable)
+ iounmap(c->cfgtable);
+ if (c->vaddr)
+ iounmap(c->vaddr);
+ pci_release_regions(c->pdev);
return err;
}
return 0;
}
-/* This does a hard reset of the controller using PCI power management
- * states. */
-static __devinit int cciss_hard_reset_controller(struct pci_dev *pdev)
+static int cciss_controller_hard_reset(struct pci_dev *pdev,
+ void * __iomem vaddr, bool use_doorbell)
{
- u16 pmcsr, saved_config_space[32];
- int i, pos;
+ u16 pmcsr;
+ int pos;
- printk(KERN_INFO "cciss: using PCI PM to reset controller\n");
+ if (use_doorbell) {
+ /* For everything after the P600, the PCI power state method
+ * of resetting the controller doesn't work, so we have this
+ * other way using the doorbell register.
+ */
+ dev_info(&pdev->dev, "using doorbell to reset controller\n");
+ writel(DOORBELL_CTLR_RESET, vaddr + SA5_DOORBELL);
+ msleep(1000);
+ } else { /* Try to do it the PCI power state way */
+
+ /* Quoting from the Open CISS Specification: "The Power
+ * Management Control/Status Register (CSR) controls the power
+ * state of the device. The normal operating state is D0,
+ * CSR=00h. The software off state is D3, CSR=03h. To reset
+ * the controller, place the interface device in D3 then to D0,
+ * this causes a secondary PCI reset which will reset the
+ * controller." */
+
+ pos = pci_find_capability(pdev, PCI_CAP_ID_PM);
+ if (pos == 0) {
+ dev_err(&pdev->dev,
+ "cciss_controller_hard_reset: "
+ "PCI PM not supported\n");
+ return -ENODEV;
+ }
+ dev_info(&pdev->dev, "using PCI PM to reset controller\n");
+ /* enter the D3hot power management state */
+ pci_read_config_word(pdev, pos + PCI_PM_CTRL, &pmcsr);
+ pmcsr &= ~PCI_PM_CTRL_STATE_MASK;
+ pmcsr |= PCI_D3hot;
+ pci_write_config_word(pdev, pos + PCI_PM_CTRL, pmcsr);
- /* This is very nearly the same thing as
+ msleep(500);
- pci_save_state(pci_dev);
- pci_set_power_state(pci_dev, PCI_D3hot);
- pci_set_power_state(pci_dev, PCI_D0);
- pci_restore_state(pci_dev);
+ /* enter the D0 power management state */
+ pmcsr &= ~PCI_PM_CTRL_STATE_MASK;
+ pmcsr |= PCI_D0;
+ pci_write_config_word(pdev, pos + PCI_PM_CTRL, pmcsr);
- but we can't use these nice canned kernel routines on
- kexec, because they also check the MSI/MSI-X state in PCI
- configuration space and do the wrong thing when it is
- set/cleared. Also, the pci_save/restore_state functions
- violate the ordering requirements for restoring the
- configuration space from the CCISS document (see the
- comment below). So we roll our own .... */
+ msleep(500);
+ }
+ return 0;
+}
- for (i = 0; i < 32; i++)
- pci_read_config_word(pdev, 2*i, &saved_config_space[i]);
+/* This does a hard reset of the controller using PCI power management
+ * states or using the doorbell register. */
+static __devinit int cciss_kdump_hard_reset_controller(struct pci_dev *pdev)
+{
+ u16 saved_config_space[32];
+ u64 cfg_offset;
+ u32 cfg_base_addr;
+ u64 cfg_base_addr_index;
+ void __iomem *vaddr;
+ unsigned long paddr;
+ u32 misc_fw_support, active_transport;
+ int rc, i;
+ CfgTable_struct __iomem *cfgtable;
+ bool use_doorbell;
+ u32 board_id;
+
+ /* For controllers as old a the p600, this is very nearly
+ * the same thing as
+ *
+ * pci_save_state(pci_dev);
+ * pci_set_power_state(pci_dev, PCI_D3hot);
+ * pci_set_power_state(pci_dev, PCI_D0);
+ * pci_restore_state(pci_dev);
+ *
+ * but we can't use these nice canned kernel routines on
+ * kexec, because they also check the MSI/MSI-X state in PCI
+ * configuration space and do the wrong thing when it is
+ * set/cleared. Also, the pci_save/restore_state functions
+ * violate the ordering requirements for restoring the
+ * configuration space from the CCISS document (see the
+ * comment below). So we roll our own ....
+ *
+ * For controllers newer than the P600, the pci power state
+ * method of resetting doesn't work so we have another way
+ * using the doorbell register.
+ */
- pos = pci_find_capability(pdev, PCI_CAP_ID_PM);
- if (pos == 0) {
- printk(KERN_ERR "cciss_reset_controller: PCI PM not supported\n");
+ /* Exclude 640x boards. These are two pci devices in one slot
+ * which share a battery backed cache module. One controls the
+ * cache, the other accesses the cache through the one that controls
+ * it. If we reset the one controlling the cache, the other will
+ * likely not be happy. Just forbid resetting this conjoined mess.
+ */
+ cciss_lookup_board_id(pdev, &board_id);
+ if (board_id == 0x409C0E11 || board_id == 0x409D0E11) {
+ dev_warn(&pdev->dev, "Cannot reset Smart Array 640x "
+ "due to shared cache module.");
return -ENODEV;
}
- /* Quoting from the Open CISS Specification: "The Power
- * Management Control/Status Register (CSR) controls the power
- * state of the device. The normal operating state is D0,
- * CSR=00h. The software off state is D3, CSR=03h. To reset
- * the controller, place the interface device in D3 then to
- * D0, this causes a secondary PCI reset which will reset the
- * controller." */
+ for (i = 0; i < 32; i++)
+ pci_read_config_word(pdev, 2*i, &saved_config_space[i]);
- /* enter the D3hot power management state */
- pci_read_config_word(pdev, pos + PCI_PM_CTRL, &pmcsr);
- pmcsr &= ~PCI_PM_CTRL_STATE_MASK;
- pmcsr |= PCI_D3hot;
- pci_write_config_word(pdev, pos + PCI_PM_CTRL, pmcsr);
+ /* find the first memory BAR, so we can find the cfg table */
+ rc = cciss_pci_find_memory_BAR(pdev, &paddr);
+ if (rc)
+ return rc;
+ vaddr = remap_pci_mem(paddr, 0x250);
+ if (!vaddr)
+ return -ENOMEM;
- schedule_timeout_uninterruptible(HZ >> 1);
+ /* find cfgtable in order to check if reset via doorbell is supported */
+ rc = cciss_find_cfg_addrs(pdev, vaddr, &cfg_base_addr,
+ &cfg_base_addr_index, &cfg_offset);
+ if (rc)
+ goto unmap_vaddr;
+ cfgtable = remap_pci_mem(pci_resource_start(pdev,
+ cfg_base_addr_index) + cfg_offset, sizeof(*cfgtable));
+ if (!cfgtable) {
+ rc = -ENOMEM;
+ goto unmap_vaddr;
+ }
- /* enter the D0 power management state */
- pmcsr &= ~PCI_PM_CTRL_STATE_MASK;
- pmcsr |= PCI_D0;
- pci_write_config_word(pdev, pos + PCI_PM_CTRL, pmcsr);
+ /* If reset via doorbell register is supported, use that. */
+ misc_fw_support = readl(&cfgtable->misc_fw_support);
+ use_doorbell = misc_fw_support & MISC_FW_DOORBELL_RESET;
- schedule_timeout_uninterruptible(HZ >> 1);
+ rc = cciss_controller_hard_reset(pdev, vaddr, use_doorbell);
+ if (rc)
+ goto unmap_cfgtable;
/* Restore the PCI configuration space. The Open CISS
* Specification says, "Restore the PCI Configuration
* Registers, offsets 00h through 60h. It is important to
* restore the command register, 16-bits at offset 04h,
* last. Do not restore the configuration status register,
- * 16-bits at offset 06h." Note that the offset is 2*i. */
+ * 16-bits at offset 06h." Note that the offset is 2*i.
+ */
for (i = 0; i < 32; i++) {
if (i == 2 || i == 3)
continue;
wmb();
pci_write_config_word(pdev, 4, saved_config_space[2]);
+ /* Some devices (notably the HP Smart Array 5i Controller)
+ need a little pause here */
+ msleep(CCISS_POST_RESET_PAUSE_MSECS);
+
+ /* Controller should be in simple mode at this point. If it's not,
+ * It means we're on one of those controllers which doesn't support
+ * the doorbell reset method and on which the PCI power management reset
+ * method doesn't work (P800, for example.)
+ * In those cases, don't try to proceed, as it generally doesn't work.
+ */
+ active_transport = readl(&cfgtable->TransportActive);
+ if (active_transport & PERFORMANT_MODE) {
+ dev_warn(&pdev->dev, "Unable to successfully reset controller,"
+ " Ignoring controller.\n");
+ rc = -ENODEV;
+ }
+
+unmap_cfgtable:
+ iounmap(cfgtable);
+
+unmap_vaddr:
+ iounmap(vaddr);
+ return rc;
+}
+
+static __devinit int cciss_init_reset_devices(struct pci_dev *pdev)
+{
+ int rc, i;
+
+ if (!reset_devices)
+ return 0;
+
+ /* Reset the controller with a PCI power-cycle or via doorbell */
+ rc = cciss_kdump_hard_reset_controller(pdev);
+
+ /* -ENOTSUPP here means we cannot reset the controller
+ * but it's already (and still) up and running in
+ * "performant mode". Or, it might be 640x, which can't reset
+ * due to concerns about shared bbwc between 6402/6404 pair.
+ */
+ if (rc == -ENOTSUPP)
+ return 0; /* just try to do the kdump anyhow. */
+ if (rc)
+ return -ENODEV;
+ if (cciss_reset_msi(pdev))
+ return -ENODEV;
+
+ /* Now try to get the controller to respond to a no-op */
+ for (i = 0; i < CCISS_POST_RESET_NOOP_RETRIES; i++) {
+ if (cciss_noop(pdev) == 0)
+ break;
+ else
+ dev_warn(&pdev->dev, "no-op failed%s\n",
+ (i < CCISS_POST_RESET_NOOP_RETRIES - 1 ?
+ "; re-trying" : ""));
+ msleep(CCISS_POST_RESET_NOOP_INTERVAL_MSECS);
+ }
return 0;
}
int dac, return_code;
InquiryData_struct *inq_buff;
- if (reset_devices) {
- /* Reset the controller with a PCI power-cycle */
- if (cciss_hard_reset_controller(pdev) || cciss_reset_msi(pdev))
- return -ENODEV;
-
- /* Now try to get the controller to respond to a no-op. Some
- devices (notably the HP Smart Array 5i Controller) need
- up to 30 seconds to respond. */
- for (i=0; i<30; i++) {
- if (cciss_noop(pdev) == 0)
- break;
-
- schedule_timeout_uninterruptible(HZ);
- }
- if (i == 30) {
- printk(KERN_ERR "cciss: controller seems dead\n");
- return -EBUSY;
- }
- }
-
+ rc = cciss_init_reset_devices(pdev);
+ if (rc)
+ return rc;
i = alloc_cciss_hba();
if (i < 0)
return -1;
+ hba[i]->pdev = pdev;
hba[i]->busy_initializing = 1;
INIT_HLIST_HEAD(&hba[i]->cmpQ);
INIT_HLIST_HEAD(&hba[i]->reqQ);
mutex_init(&hba[i]->busy_shutting_down);
- if (cciss_pci_init(hba[i], pdev) != 0)
+ if (cciss_pci_init(hba[i]) != 0)
goto clean_no_release_regions;
sprintf(hba[i]->devname, "cciss%d", i);
hba[i]->ctlr = i;
- hba[i]->pdev = pdev;
init_completion(&hba[i]->scan_wait);
/* make sure the board interrupts are off */
hba[i]->access.set_intr_mask(hba[i], CCISS_INTR_OFF);
- if (request_irq(hba[i]->intr[SIMPLE_MODE_INT], do_cciss_intr,
- IRQF_DISABLED | IRQF_SHARED, hba[i]->devname, hba[i])) {
- printk(KERN_ERR "cciss: Unable to get irq %d for %s\n",
- hba[i]->intr[SIMPLE_MODE_INT], hba[i]->devname);
- goto clean2;
+ if (hba[i]->msi_vector || hba[i]->msix_vector) {
+ if (request_irq(hba[i]->intr[PERF_MODE_INT],
+ do_cciss_msix_intr,
+ IRQF_DISABLED, hba[i]->devname, hba[i])) {
+ printk(KERN_ERR "cciss: Unable to get irq %d for %s\n",
+ hba[i]->intr[PERF_MODE_INT], hba[i]->devname);
+ goto clean2;
+ }
+ } else {
+ if (request_irq(hba[i]->intr[PERF_MODE_INT], do_cciss_intx,
+ IRQF_DISABLED, hba[i]->devname, hba[i])) {
+ printk(KERN_ERR "cciss: Unable to get irq %d for %s\n",
+ hba[i]->intr[PERF_MODE_INT], hba[i]->devname);
+ goto clean2;
+ }
}
printk(KERN_INFO "%s: <0x%x> at PCI %s IRQ %d%s using DAC\n",
hba[i]->devname, pdev->device, pci_name(pdev),
- hba[i]->intr[SIMPLE_MODE_INT], dac ? "" : " not");
+ hba[i]->intr[PERF_MODE_INT], dac ? "" : " not");
hba[i]->cmd_pool_bits =
kmalloc(DIV_ROUND_UP(hba[i]->nr_cmds, BITS_PER_LONG)
hba[i]->nr_cmds * sizeof(ErrorInfo_struct),
hba[i]->errinfo_pool,
hba[i]->errinfo_pool_dhandle);
- free_irq(hba[i]->intr[SIMPLE_MODE_INT], hba[i]);
+ free_irq(hba[i]->intr[PERF_MODE_INT], hba[i]);
clean2:
unregister_blkdev(hba[i]->major, hba[i]->devname);
clean1:
printk(KERN_WARNING "cciss%d: Error flushing cache\n",
h->ctlr);
h->access.set_intr_mask(h, CCISS_INTR_OFF);
- free_irq(h->intr[2], h);
+ free_irq(h->intr[PERF_MODE_INT], h);
}
static void __devexit cciss_remove_one(struct pci_dev *pdev)
pci_disable_msi(hba[i]->pdev);
#endif /* CONFIG_PCI_MSI */
+ iounmap(hba[i]->transtable);
+ iounmap(hba[i]->cfgtable);
iounmap(hba[i]->vaddr);
pci_free_consistent(hba[i]->pdev, hba[i]->nr_cmds * sizeof(CommandList_struct),
* array of them, the size must be a multiple of 8 bytes.
*/
BUILD_BUG_ON(sizeof(CommandList_struct) % COMMANDLIST_ALIGNMENT);
-
printk(KERN_INFO DRIVER_NAME "\n");
err = bus_register(&cciss_bus_type);
bus_unregister(&cciss_bus_type);
}
-static void fail_all_cmds(unsigned long ctlr)
-{
- /* If we get here, the board is apparently dead. */
- ctlr_info_t *h = hba[ctlr];
- CommandList_struct *c;
- unsigned long flags;
-
- printk(KERN_WARNING "cciss%d: controller not responding.\n", h->ctlr);
- h->alive = 0; /* the controller apparently died... */
-
- spin_lock_irqsave(CCISS_LOCK(ctlr), flags);
-
- pci_disable_device(h->pdev); /* Make sure it is really dead. */
-
- /* move everything off the request queue onto the completed queue */
- while (!hlist_empty(&h->reqQ)) {
- c = hlist_entry(h->reqQ.first, CommandList_struct, list);
- removeQ(c);
- h->Qdepth--;
- addQ(&h->cmpQ, c);
- }
-
- /* Now, fail everything on the completed queue with a HW error */
- while (!hlist_empty(&h->cmpQ)) {
- c = hlist_entry(h->cmpQ.first, CommandList_struct, list);
- removeQ(c);
- if (c->cmd_type != CMD_MSG_STALE)
- c->err_info->CommandStatus = CMD_HARDWARE_ERR;
- if (c->cmd_type == CMD_RWREQ) {
- complete_command(h, c, 0);
- } else if (c->cmd_type == CMD_IOCTL_PEND)
- complete(c->waiting);
-#ifdef CONFIG_CISS_SCSI_TAPE
- else if (c->cmd_type == CMD_SCSI)
- complete_scsi_command(c, 0, 0);
-#endif
- }
- spin_unlock_irqrestore(CCISS_LOCK(ctlr), flags);
- return;
-}
-
module_init(cciss_init);
module_exit(cciss_cleanup);