diff --git a/drivers/scsi/hpsa.c b/drivers/scsi/hpsa.c index cef5d49b59cd24..abf5e1ef89e99a 100644 --- a/drivers/scsi/hpsa.c +++ b/drivers/scsi/hpsa.c @@ -54,12 +54,15 @@ #include "hpsa.h" /* HPSA_DRIVER_VERSION must be 3 byte values (0-255) separated by '.' */ -#define HPSA_DRIVER_VERSION "3.4.4-1" +#define HPSA_DRIVER_VERSION "3.4.9-0" #define DRIVER_NAME "HP HPSA Driver (v " HPSA_DRIVER_VERSION ")" #define HPSA "hpsa" -/* How long to wait (in milliseconds) for board to go into simple mode */ -#define MAX_CONFIG_WAIT 30000 +/* How long to wait for CISS doorbell communication */ +#define CLEAR_EVENT_WAIT_INTERVAL 20 /* ms for each msleep() call */ +#define MODE_CHANGE_WAIT_INTERVAL 10 /* ms for each msleep() call */ +#define MAX_CLEAR_EVENT_WAIT 30000 /* times 20 ms = 600 s */ +#define MAX_MODE_CHANGE_WAIT 100 /* times 10 ms = 1 s */ #define MAX_IOCTL_CONFIG_WAIT 1000 /*define how many times we will try a command because of bus resets */ @@ -103,7 +106,6 @@ static const struct pci_device_id hpsa_pci_device_id[] = { {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSH, 0x103C, 0x1922}, {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSH, 0x103C, 0x1923}, {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSH, 0x103C, 0x1924}, - {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSH, 0x103C, 0x1925}, {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSH, 0x103C, 0x1926}, {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSH, 0x103C, 0x1928}, {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSH, 0x103C, 0x1929}, @@ -149,6 +151,7 @@ static struct board_type products[] = { {0x3249103C, "Smart Array P812", &SA5_access}, {0x324A103C, "Smart Array P712m", &SA5_access}, {0x324B103C, "Smart Array P711m", &SA5_access}, + {0x3233103C, "HP StorageWorks 1210m", &SA5_access}, /* alias of 333f */ {0x3350103C, "Smart Array P222", &SA5_access}, {0x3351103C, "Smart Array P420", &SA5_access}, {0x3352103C, "Smart Array P421", &SA5_access}, @@ -194,17 +197,16 @@ static int number_of_controllers; static irqreturn_t do_hpsa_intr_intx(int irq, void *dev_id); static irqreturn_t do_hpsa_intr_msi(int irq, void *dev_id); static int hpsa_ioctl(struct scsi_device *dev, int cmd, void *arg); -static void lock_and_start_io(struct ctlr_info *h); -static void start_io(struct ctlr_info *h, unsigned long *flags); #ifdef CONFIG_COMPAT static int hpsa_compat_ioctl(struct scsi_device *dev, int cmd, void *arg); #endif static void cmd_free(struct ctlr_info *h, struct CommandList *c); -static void cmd_special_free(struct ctlr_info *h, struct CommandList *c); static struct CommandList *cmd_alloc(struct ctlr_info *h); -static struct CommandList *cmd_special_alloc(struct ctlr_info *h); +static void cmd_tagged_free(struct ctlr_info *h, struct CommandList *c); +static struct CommandList *cmd_tagged_alloc(struct ctlr_info *h, + struct scsi_cmnd *scmd); static int fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, void *buff, size_t size, u16 page_code, unsigned char *scsi3addr, int cmd_type); @@ -216,10 +218,12 @@ static int hpsa_scan_finished(struct Scsi_Host *sh, unsigned long elapsed_time); static int hpsa_change_queue_depth(struct scsi_device *sdev, int qdepth, int reason); +static int hpsa_change_queue_type(struct scsi_device *sdev, int type); static int hpsa_eh_device_reset_handler(struct scsi_cmnd *scsicmd); static int hpsa_eh_abort_handler(struct scsi_cmnd *scsicmd); static int hpsa_slave_alloc(struct scsi_device *sdev); +static int hpsa_slave_configure(struct scsi_device *sdev); static void hpsa_slave_destroy(struct scsi_device *sdev); static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno); @@ -230,7 +234,8 @@ static void check_ioctl_unit_attention(struct ctlr_info *h, /* performant mode helper functions */ static void calc_bucket_map(int *bucket, int num_buckets, int nsgs, int min_blocks, int *bucket_map); -static void hpsa_put_ctlr_into_performant_mode(struct ctlr_info *h); +static void hpsa_free_performant_mode(struct ctlr_info *h); +static int hpsa_put_ctlr_into_performant_mode(struct ctlr_info *h); static inline u32 next_command(struct ctlr_info *h, u8 q); static int hpsa_find_cfg_addrs(struct pci_dev *pdev, void __iomem *vaddr, u32 *cfg_base_addr, u64 *cfg_base_addr_index, @@ -241,14 +246,17 @@ static int hpsa_lookup_board_id(struct pci_dev *pdev, u32 *board_id); static int hpsa_wait_for_board_state(struct pci_dev *pdev, void __iomem *vaddr, int wait_for_ready); static inline void finish_cmd(struct CommandList *c); -static void hpsa_wait_for_mode_change_ack(struct ctlr_info *h); +static int hpsa_wait_for_mode_change_ack(struct ctlr_info *h); #define BOARD_NOT_READY 0 #define BOARD_READY 1 static void hpsa_drain_accel_commands(struct ctlr_info *h); static void hpsa_flush_cache(struct ctlr_info *h); static int hpsa_scsi_ioaccel_queue_command(struct ctlr_info *h, struct CommandList *c, u32 ioaccel_handle, u8 *cdb, int cdb_len, - u8 *scsi3addr); + u8 *scsi3addr, struct hpsa_scsi_dev_t *phys_disk); +static void hpsa_command_resubmit_worker(struct work_struct *work); +static void print_cfg_table(struct device *dev, struct CfgTable *tb); +static void detect_controller_lockup(struct ctlr_info *h); static inline struct ctlr_info *sdev_to_hba(struct scsi_device *sdev) { @@ -262,40 +270,84 @@ static inline struct ctlr_info *shost_to_hba(struct Scsi_Host *sh) return (struct ctlr_info *) *priv; } +/* extract sense key, asc, and ascq from sense data. -1 means invalid. */ +static void decode_sense_data(const u8 *sense_data, int sense_data_len, + int *sense_key, int *asc, int *ascq) +{ + if (sense_data_len < 1) { + *sense_key = -1; + *ascq = -1; + *asc = -1; + return; + } + + switch (sense_data[0]) { + case 0x70: /* old format sense data */ + *sense_key = (sense_data_len > 2) ? sense_data[2] & 0x0f : -1; + *ascq = (sense_data_len > 13) ? sense_data[13] : -1; + *asc = (sense_data_len > 12) ? sense_data[12] : -1; + break; + case 0x72: /* descriptor format sense data */ + *sense_key = (sense_data_len > 1) ? sense_data[1] & 0x0f : -1; + *ascq = (sense_data_len > 2) ? sense_data[2] : -1; + *asc = (sense_data_len > 3) ? sense_data[3] : -1; + break; + default: + *sense_key = -1; + *ascq = -1; + *asc = -1; + break; + } +} + static int check_for_unit_attention(struct ctlr_info *h, struct CommandList *c) { - if (c->err_info->SenseInfo[2] != UNIT_ATTENTION) + int sense_key, asc, ascq; + int sense_len; + + if (c->err_info->SenseLen > sizeof(c->err_info->SenseInfo)) + sense_len = sizeof(c->err_info->SenseInfo); + else + sense_len = c->err_info->SenseLen; + + decode_sense_data(c->err_info->SenseInfo, sense_len, + &sense_key, &asc, &ascq); + if (sense_key != UNIT_ATTENTION || asc == -1) return 0; - switch (c->err_info->SenseInfo[12]) { + switch (asc) { case STATE_CHANGED: - dev_warn(&h->pdev->dev, HPSA "%d: a state change " - "detected, command retried\n", h->ctlr); + dev_warn(&h->pdev->dev, + "%s: a state change detected, command retried\n", + h->devname); break; case LUN_FAILED: - dev_warn(&h->pdev->dev, HPSA "%d: LUN failure " - "detected, action required\n", h->ctlr); + dev_warn(&h->pdev->dev, + "%s: LUN failure detected\n", h->devname); break; case REPORT_LUNS_CHANGED: - dev_warn(&h->pdev->dev, HPSA "%d: report LUN data " - "changed, action required\n", h->ctlr); + dev_warn(&h->pdev->dev, + "%s: report LUN data changed\n", h->devname); /* * Note: this REPORT_LUNS_CHANGED condition only occurs on the external * target (array) devices. */ break; case POWER_OR_RESET: - dev_warn(&h->pdev->dev, HPSA "%d: a power on " - "or device reset detected\n", h->ctlr); + dev_warn(&h->pdev->dev, + "%s: a power on or device reset detected\n", + h->devname); break; case UNIT_ATTENTION_CLEARED: - dev_warn(&h->pdev->dev, HPSA "%d: unit attention " - "cleared by another initiator\n", h->ctlr); + dev_warn(&h->pdev->dev, + "%s: unit attention cleared by another initiator\n", + h->devname); break; default: - dev_warn(&h->pdev->dev, HPSA "%d: unknown " - "unit attention detected\n", h->ctlr); + dev_warn(&h->pdev->dev, + "%s: unknown unit attention detected\n", + h->devname); break; } return 1; @@ -311,6 +363,95 @@ static int check_for_busy(struct ctlr_info *h, struct CommandList *c) return 1; } +static ssize_t host_store_lockup_detector(struct device *dev, + struct device_attribute *attr, + const char *buf, size_t count) +{ + struct Scsi_Host *shost; + struct ctlr_info *h; + int len, enabled; + char tmpbuf[10]; + + if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SYS_RAWIO)) + return -EACCES; + len = count > sizeof(tmpbuf) - 1 ? sizeof(tmpbuf) - 1 : count; + strncpy(tmpbuf, buf, len); + tmpbuf[len] = '\0'; + if (sscanf(tmpbuf, "%d", &enabled) != 1) + return -EINVAL; + shost = class_to_shost(dev); + h = shost_to_hba(shost); + h->lockup_detector_enabled = !!enabled; + return count; +} + +static ssize_t host_show_lockup_detector(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct ctlr_info *h; + struct Scsi_Host *shost = class_to_shost(dev); + + h = shost_to_hba(shost); + return snprintf(buf, 20, "%d\n", h->lockup_detector_enabled); +} + +static ssize_t host_store_abort_test(struct device *dev, + struct device_attribute *attr, + const char *buf, size_t count) +{ + struct Scsi_Host *shost; + struct ctlr_info *h; + int len, value, timeout; + char tmpbuf[10]; + + if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SYS_RAWIO)) + return -EACCES; + len = count > sizeof(tmpbuf) - 1 ? sizeof(tmpbuf) - 1 : count; + strncpy(tmpbuf, buf, len); + tmpbuf[len] = '\0'; + if (sscanf(tmpbuf, "%d %d", &value, &timeout) != 2) + return -EINVAL; + shost = class_to_shost(dev); + h = shost_to_hba(shost); + h->abort_test = value; + h->abort_timeout = timeout; + return count; +} + +static ssize_t host_show_abort_test(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct ctlr_info *h; + struct Scsi_Host *shost = class_to_shost(dev); + + h = shost_to_hba(shost); + return snprintf(buf, 20, "%d %d\n", h->abort_test, h->abort_timeout); +} + +static u32 lockup_detected(struct ctlr_info *h); +static ssize_t host_show_lockup_detected(struct device *dev, + struct device_attribute *attr, char *buf) +{ + int i, ld, c, cpu; + struct ctlr_info *h; + struct Scsi_Host *shost = class_to_shost(dev); + + h = shost_to_hba(shost); + ld = lockup_detected(h); + + c = sprintf(buf, "ld=%d: ", ld); + cpu = cpumask_first(cpu_online_mask); + for (i = 0; i < num_online_cpus(); i++) { + u32 *lockup_detected; + lockup_detected = per_cpu_ptr(h->lockup_detected, cpu); + ld = *lockup_detected; + cpu = cpumask_next(cpu, cpu_online_mask); + c += sprintf(buf + c, "%d,", ld); + } + c += sprintf(buf + c, "\n"); + return c; +} + static ssize_t host_store_hp_ssd_smart_path_status(struct device *dev, struct device_attribute *attr, const char *buf, size_t count) @@ -371,6 +512,85 @@ static ssize_t host_store_rescan(struct device *dev, return count; } +static ssize_t host_store_hpsa_intcoal_delay(struct device *dev, + struct device_attribute *attr, + const char *buf, size_t count) +{ + int len; + unsigned int value; + char tmpbuf[8]; + struct ctlr_info *h; + struct Scsi_Host *shost = class_to_shost(dev); + + len = count > sizeof(tmpbuf) - 1 ? sizeof(tmpbuf) - 1 : count; + strncpy(tmpbuf, buf, len); + tmpbuf[len] = '\0'; + if (sscanf(tmpbuf, "%u", &value) != 1) + return -EINVAL; + /* the device policies the values written, so no need here */ + + h = shost_to_hba(shost); + print_cfg_table(&h->pdev->dev, h->cfgtable); + writel(value, &h->cfgtable->HostWrite.CoalIntDelay); + print_cfg_table(&h->pdev->dev, h->cfgtable); + writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL); + (void) hpsa_wait_for_mode_change_ack(h); + print_cfg_table(&h->pdev->dev, h->cfgtable); + return count; +} + +static ssize_t host_store_hpsa_intcoal_count(struct device *dev, + struct device_attribute *attr, + const char *buf, size_t count) +{ + int len; + unsigned int value; + char tmpbuf[8]; + struct ctlr_info *h; + struct Scsi_Host *shost = class_to_shost(dev); + + len = count > sizeof(tmpbuf) - 1 ? sizeof(tmpbuf) - 1 : count; + strncpy(tmpbuf, buf, len); + tmpbuf[len] = '\0'; + if (sscanf(tmpbuf, "%u", &value) != 1) + return -EINVAL; + /* the device policies the values written, so no need here */ + + h = shost_to_hba(shost); + print_cfg_table(&h->pdev->dev, h->cfgtable); + writel(value, &h->cfgtable->HostWrite.CoalIntCount); + print_cfg_table(&h->pdev->dev, h->cfgtable); + writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL); + (void) hpsa_wait_for_mode_change_ack(h); + print_cfg_table(&h->pdev->dev, h->cfgtable); + return count; +} + +static ssize_t host_show_hpsa_intcoal_delay(struct device *dev, + struct device_attribute *attr, char *buf) +{ + unsigned int value; + struct ctlr_info *h; + struct Scsi_Host *shost = class_to_shost(dev); + + h = shost_to_hba(shost); + value = readl(&h->cfgtable->HostWrite.CoalIntDelay); + return snprintf(buf, 20, "%u\n", value); +} + +static ssize_t host_show_hpsa_intcoal_count(struct device *dev, + struct device_attribute *attr, char *buf) +{ + unsigned int value; + struct ctlr_info *h; + struct Scsi_Host *shost = class_to_shost(dev); + + h = shost_to_hba(shost); + value = readl(&h->cfgtable->HostWrite.CoalIntCount); + return snprintf(buf, 20, "%u\n", value); +} + + static ssize_t host_show_firmware_revision(struct device *dev, struct device_attribute *attr, char *buf) { @@ -392,7 +612,8 @@ static ssize_t host_show_commands_outstanding(struct device *dev, struct Scsi_Host *shost = class_to_shost(dev); struct ctlr_info *h = shost_to_hba(shost); - return snprintf(buf, 20, "%d\n", h->commands_outstanding); + return snprintf(buf, 20, "%d\n", + atomic_read(&h->commands_outstanding)); } static ssize_t host_show_transport_mode(struct device *dev, @@ -463,24 +684,31 @@ static u32 soft_unresettable_controller[] = { 0x409D0E11, /* Smart Array 6400 EM */ }; -static int ctlr_is_hard_resettable(u32 board_id) +static u32 needs_abort_tags_swizzled[] = { + 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) @@ -489,6 +717,12 @@ 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) { @@ -504,8 +738,8 @@ static inline int is_logical_dev_addr_mode(unsigned char scsi3addr[]) return (scsi3addr[3] & 0xC0) == 0x40; } -static const char *raid_label[] = { "0", "4", "1(1+0)", "5", "5+1", "ADG", - "1(ADM)", "UNKNOWN" +static const char * const raid_label[] = { "0", "4", "1(+0)", "5", "5+1", "6", + "1(+0)ADM", "UNKNOWN" }; #define HPSA_RAID_0 0 #define HPSA_RAID_4 1 @@ -628,6 +862,8 @@ static DEVICE_ATTR(raid_level, S_IRUGO, raid_level_show, NULL); static DEVICE_ATTR(lunid, S_IRUGO, lunid_show, NULL); static DEVICE_ATTR(unique_id, S_IRUGO, unique_id_show, NULL); static DEVICE_ATTR(rescan, S_IWUSR, NULL, host_store_rescan); +static DEVICE_ATTR(hpsa_intcoal_delay, S_IWUSR|S_IRUGO, host_show_hpsa_intcoal_delay, host_store_hpsa_intcoal_delay); +static DEVICE_ATTR(hpsa_intcoal_count, S_IWUSR|S_IRUGO, host_show_hpsa_intcoal_count, host_store_hpsa_intcoal_count); static DEVICE_ATTR(hp_ssd_smart_path_enabled, S_IRUGO, host_show_hp_ssd_smart_path_enabled, NULL); static DEVICE_ATTR(hp_ssd_smart_path_status, S_IWUSR|S_IRUGO|S_IROTH, @@ -643,6 +879,12 @@ static DEVICE_ATTR(transport_mode, S_IRUGO, host_show_transport_mode, NULL); static DEVICE_ATTR(resettable, S_IRUGO, host_show_resettable, NULL); +static DEVICE_ATTR(lockup_detector, S_IWUSR|S_IRUGO, + host_show_lockup_detector, host_store_lockup_detector); +static DEVICE_ATTR(lockup_detected, S_IRUGO, + host_show_lockup_detected, NULL); +static DEVICE_ATTR(abort_test, S_IWUSR|S_IRUGO, + host_show_abort_test, host_store_abort_test); static struct device_attribute *hpsa_sdev_attrs[] = { &dev_attr_raid_level, @@ -657,12 +899,20 @@ static struct device_attribute *hpsa_shost_attrs[] = { &dev_attr_firmware_revision, &dev_attr_commands_outstanding, &dev_attr_transport_mode, + &dev_attr_hpsa_intcoal_delay, + &dev_attr_hpsa_intcoal_count, &dev_attr_resettable, &dev_attr_hp_ssd_smart_path_status, &dev_attr_raid_offload_debug, + &dev_attr_lockup_detector, + &dev_attr_lockup_detected, + &dev_attr_abort_test, NULL, }; +#define HPSA_NRESERVED_CMDS (HPSA_CMDS_RESERVED_FOR_ABORTS + \ + HPSA_CMDS_RESERVED_FOR_DRIVER + HPSA_MAX_CONCURRENT_PASSTHRUS) + static struct scsi_host_template hpsa_driver_template = { .module = THIS_MODULE, .name = HPSA, @@ -671,12 +921,14 @@ static struct scsi_host_template hpsa_driver_template = { .scan_start = hpsa_scan_start, .scan_finished = hpsa_scan_finished, .change_queue_depth = hpsa_change_queue_depth, + .change_queue_type = hpsa_change_queue_type, .this_id = -1, .use_clustering = ENABLE_CLUSTERING, .eh_abort_handler = hpsa_eh_abort_handler, .eh_device_reset_handler = hpsa_eh_device_reset_handler, .ioctl = hpsa_ioctl, .slave_alloc = hpsa_slave_alloc, + .slave_configure = hpsa_slave_configure, .slave_destroy = hpsa_slave_destroy, #ifdef CONFIG_COMPAT .compat_ioctl = hpsa_compat_ioctl, @@ -687,18 +939,10 @@ static struct scsi_host_template hpsa_driver_template = { .no_write_same = 1, }; - -/* Enqueuing and dequeuing functions for cmdlists. */ -static inline void addQ(struct list_head *list, struct CommandList *c) -{ - list_add_tail(&c->list, list); -} - static inline u32 next_command(struct ctlr_info *h, u8 q) { u32 a; struct reply_queue_buffer *rq = &h->reply_queue[q]; - unsigned long flags; if (h->transMethod & CFGTBL_Trans_io_accel1) return h->access.command_completed(h, q); @@ -709,9 +953,7 @@ static inline u32 next_command(struct ctlr_info *h, u8 q) if ((rq->head[rq->current_entry] & 1) == rq->wraparound) { a = rq->head[rq->current_entry]; rq->current_entry++; - spin_lock_irqsave(&h->lock, flags); - h->commands_outstanding--; - spin_unlock_irqrestore(&h->lock, flags); + atomic_dec(&h->commands_outstanding); } else { a = FIFO_EMPTY; } @@ -753,25 +995,35 @@ static inline u32 next_command(struct ctlr_info *h, u8 q) * 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 * processor. This seems to give the best I/O throughput. */ - cp->ReplyQueue = smp_processor_id() % h->nreply_queues; + 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) @@ -781,15 +1033,41 @@ static void set_ioaccel1_performant_mode(struct ctlr_info *h, IOACCEL1_BUSADDR_CMDTYPE; } +static void set_ioaccel2_tmf_performant_mode(struct ctlr_info *h, + struct CommandList *c, + int reply_queue) +{ + struct hpsa_tmf_struct *cp = (struct hpsa_tmf_struct *) + &h->ioaccel2_cmd_pool[c->cmdindex]; + + /* Tell the controller to post the reply to the queue for this + * processor. This seems to give the best I/O throughput. + */ + 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 + */ + c->busaddr |= (h->ioaccel2_blockFetchTable[0]); +} + 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 * processor. This seems to give the best I/O throughput. */ - cp->reply_queue = smp_processor_id() % h->nreply_queues; + 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) @@ -827,34 +1105,61 @@ static void dial_up_lockup_detection_on_fw_flash_complete(struct ctlr_info *h, 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); + +static void hpsa_abort_torture_worker(struct work_struct *work) { - unsigned long flags; + struct CommandList *c = container_of(to_delayed_work(work), + struct CommandList, abort_torture_work); + dev_warn(&c->h->pdev->dev, "Submitting delayed command\n"); + __enqueue_cmd_and_start_io(c->h, c, DEFAULT_REPLY_QUEUE); +} +static void __enqueue_cmd_and_start_io(struct ctlr_info *h, + struct CommandList *c, int reply_queue) +{ + if (h->abort_test > 0 && + (c->cmd_type == CMD_SCSI || + c->cmd_type == CMD_IOACCEL1|| + c->cmd_type == CMD_IOACCEL2)) { + h->abort_test = 0; + dev_warn(&h->pdev->dev, "delaying command for %d secs.\n", + h->abort_timeout); + INIT_DELAYED_WORK(&c->abort_torture_work, + hpsa_abort_torture_worker); + schedule_delayed_work(&c->abort_torture_work, h->abort_timeout * HZ); + atomic_inc(&h->cmds_sent); + return; + } + dial_down_lockup_detection_during_fw_flash(h, c); + atomic_inc(&h->commands_outstanding); + atomic_inc(&h->cmds_sent); 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; + case IOACCEL2_TMF: + set_ioaccel2_tmf_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); } - dial_down_lockup_detection_during_fw_flash(h, c); - spin_lock_irqsave(&h->lock, flags); - addQ(&h->reqQ, c); - h->Qdepth++; - start_io(h, &flags); - spin_unlock_irqrestore(&h->lock, flags); } -static inline void removeQ(struct CommandList *c) +static void enqueue_cmd_and_start_io(struct ctlr_info *h, struct CommandList *c) { - if (WARN_ON(list_empty(&c->list))) - return; - list_del_init(&c->list); + if (unlikely(c->abort_pending)) + return finish_cmd(c); + + __enqueue_cmd_and_start_io(h, c, DEFAULT_REPLY_QUEUE); } static inline int is_hba_lunid(unsigned char scsi3addr[]) @@ -962,17 +1267,23 @@ static int hpsa_scsi_add_entry(struct ctlr_info *h, int hostno, h->dev[n] = device; h->ndevices++; + device->offload_to_be_enabled = device->offload_enabled; + 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); + dev_info(&h->pdev->dev, + "%6s scsi %d:%d:%d:%d: %s %.8s %.16s RAID-%s SSDSmartPathCap%c En%c Exp=%d\n", + device->expose_state & HPSA_SCSI_ADD ? "added" : "masked", + hostno, device->bus, device->target, device->lun, + scsi_device_type(device->devtype), + device->vendor, + device->model, + device->raid_level > RAID_UNKNOWN ? + "RAID-?" : raid_label[device->raid_level], + device->offload_config ? '+' : '-', + device->offload_to_be_enabled ? '+' : '-', + device->expose_state); return 0; } @@ -986,16 +1297,47 @@ static void hpsa_scsi_update_entry(struct ctlr_info *h, int hostno, /* Raid level changed. */ h->dev[entry]->raid_level = new_entry->raid_level; - /* Raid offload parameters changed. */ + /* Raid offload parameters changed. Careful about the ordering. */ + if (new_entry->offload_config && new_entry->offload_enabled) { + /* if drive is newly offload_enabled, we want to copy the + * raid map data first. If previously offload_enabled and + * offload_config were set, raid map data had better be + * the same as it was before. if raid map data is changed + * then it had better be the case that + * h->dev[entry]->offload_enabled is currently 0. + */ + h->dev[entry]->raid_map = new_entry->raid_map; + h->dev[entry]->ioaccel_handle = new_entry->ioaccel_handle; + } + if (new_entry->hba_ioaccel_enabled) { + h->dev[entry]->ioaccel_handle = new_entry->ioaccel_handle; + wmb(); /* set ioaccel_handle *before* hba_ioaccel_eanbled */ + } + h->dev[entry]->hba_ioaccel_enabled = new_entry->hba_ioaccel_enabled; h->dev[entry]->offload_config = new_entry->offload_config; - h->dev[entry]->offload_enabled = new_entry->offload_enabled; - h->dev[entry]->ioaccel_handle = new_entry->ioaccel_handle; h->dev[entry]->offload_to_mirror = new_entry->offload_to_mirror; - h->dev[entry]->raid_map = new_entry->raid_map; - dev_info(&h->pdev->dev, "%s device c%db%dt%dl%d updated.\n", - scsi_device_type(new_entry->devtype), hostno, new_entry->bus, - new_entry->target, new_entry->lun); + /* We can turn off ioaccel offload now, but need to delay turning + * it on until we can update h->dev[entry]->phys_disk[], but we + * can't do that until all the devices are updated. + */ + h->dev[entry]->offload_to_be_enabled = new_entry->offload_enabled; + if (!new_entry->offload_enabled) + h->dev[entry]->offload_enabled = 0; + h->dev[entry]->queue_depth = new_entry->queue_depth; + + dev_info(&h->pdev->dev, + "updated scsi %d:%d:%d:%d: %s %.8s %.16s RAID-%s SSDSmartPathCap%c En%c Exp=%d\n", + hostno, h->dev[entry]->bus, + h->dev[entry]->target, h->dev[entry]->lun, + scsi_device_type(h->dev[entry]->devtype), + h->dev[entry]->vendor, + h->dev[entry]->model, + h->dev[entry]->raid_level > RAID_UNKNOWN ? + "RAID-?" : raid_label[h->dev[entry]->raid_level], + h->dev[entry]->offload_config ? '+' : '-', + h->dev[entry]->offload_to_be_enabled ? '+' : '-', + h->dev[entry]->expose_state); } /* Replace an entry from h->dev[] array. */ @@ -1018,12 +1360,22 @@ static void hpsa_scsi_replace_entry(struct ctlr_info *h, int hostno, new_entry->lun = h->dev[entry]->lun; } + new_entry->offload_to_be_enabled = new_entry->offload_enabled; + new_entry->offload_enabled = 0; 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); + dev_info(&h->pdev->dev, + "replaced scsi %d:%d:%d:%d: %s %.8s %.16s RAID-%s SSDSmartPathCap%c En%c Exp=%d\n", + hostno, new_entry->bus, new_entry->target, new_entry->lun, + scsi_device_type(new_entry->devtype), + new_entry->vendor, + new_entry->model, + new_entry->raid_level > RAID_UNKNOWN ? + "RAID-?" : raid_label[new_entry->raid_level], + new_entry->offload_config ? '+' : '-', + new_entry->offload_to_be_enabled ? '+' : '-', + new_entry->expose_state); } /* Remove an entry from h->dev[] array. */ @@ -1043,9 +1395,17 @@ static void hpsa_scsi_remove_entry(struct ctlr_info *h, int hostno, int entry, 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); + dev_info(&h->pdev->dev, + "removed scsi %d:%d:%d:%d: %s %.8s %.16s RAID-%s SSDSmartPathCap%c En%c Exp=%d\n", + hostno, sd->bus, sd->target, sd->lun, + scsi_device_type(sd->devtype), + sd->vendor, + sd->model, + sd->raid_level > RAID_UNKNOWN ? + "RAID-?" : raid_label[sd->raid_level], + sd->offload_config ? '+' : '-', + sd->offload_enabled ? '+' : '-', + sd->expose_state); } #define SCSI3ADDR_EQ(a, b) ( \ @@ -1117,6 +1477,8 @@ static inline int device_updated(struct hpsa_scsi_dev_t *dev1, return 1; if (dev1->offload_enabled != dev2->offload_enabled) return 1; + if (dev1->queue_depth != dev2->queue_depth) + return 1; return 0; } @@ -1262,6 +1624,91 @@ static void hpsa_show_volume_status(struct ctlr_info *h, } } +/* Figure the list of physical drive pointers for a logical drive with + * raid offload configured. + */ +static void hpsa_figure_phys_disk_ptrs(struct ctlr_info *h, + struct hpsa_scsi_dev_t *dev[], int ndevices, + struct hpsa_scsi_dev_t *logical_drive) +{ + struct raid_map_data *map = &logical_drive->raid_map; + struct raid_map_disk_data *dd = &map->data[0]; + int i, j; + int nraid_map_entries = map->row_cnt * map->layout_map_count * + (map->data_disks_per_row + map->metadata_disks_per_row); + int nphys_disk = map->layout_map_count * + (map->data_disks_per_row + map->metadata_disks_per_row); + int qdepth; + + if (nraid_map_entries > RAID_MAP_MAX_ENTRIES) + nraid_map_entries = RAID_MAP_MAX_ENTRIES; + + qdepth = 0; + for (i = 0; i < nraid_map_entries; i++) { + logical_drive->phys_disk[i] = NULL; + if (!logical_drive->offload_config) + continue; + for (j = 0; j < ndevices; j++) { + if (dev[j]->devtype != TYPE_DISK) + continue; + if (is_logical_dev_addr_mode(dev[j]->scsi3addr)) + continue; + if (dev[j]->ioaccel_handle != dd[i].ioaccel_handle) + continue; + + logical_drive->phys_disk[i] = dev[j]; + if (i < nphys_disk) + qdepth = min(h->nr_cmds, qdepth + + logical_drive->phys_disk[i]->queue_depth); + break; + } + + /* + * This can happen if a physical drive is removed and + * the logical drive is degraded. In that case, the RAID + * map data will refer to a physical disk which isn't actually + * present. And in that case offload_enabled should already + * be 0, but we'll turn it off here just in case + */ + if (!logical_drive->phys_disk[i]) { + logical_drive->offload_enabled = 0; + logical_drive->offload_to_be_enabled = 0; + logical_drive->queue_depth = h->nr_cmds; + } + } + if (nraid_map_entries) + /* + * This is correct for reads, too high for full stripe writes, + * way too high for partial stripe writes + */ + logical_drive->queue_depth = qdepth; + else + logical_drive->queue_depth = h->nr_cmds; +} + +static void hpsa_update_log_drive_phys_drive_ptrs(struct ctlr_info *h, + struct hpsa_scsi_dev_t *dev[], int ndevices) +{ + int i; + + for (i = 0; i < ndevices; i++) { + if (dev[i]->devtype != TYPE_DISK) + continue; + if (!is_logical_dev_addr_mode(dev[i]->scsi3addr)) + continue; + + /* If offload is currently enabled, the RAID map and + * phys_disk[] assignment *better* not be changing + * and since it isn't changing, we do not need to + * update it. + */ + if (dev[i]->offload_enabled) + continue; + + hpsa_figure_phys_disk_ptrs(h, dev, ndevices, dev[i]); + } +} + static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno, struct hpsa_scsi_dev_t *sd[], int nsds) { @@ -1334,9 +1781,18 @@ static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno, */ 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); + dev_info(&h->pdev->dev, + "offline scsi %d:%d:%d:%d: %s %.8s %.16s RAID-%s SSDSmartPathCap%c En%c Exp=%d\n", + hostno, sd[i]->bus, sd[i]->target, sd[i]->lun, + scsi_device_type(sd[i]->devtype), + sd[i]->vendor, + sd[i]->model, + sd[i]->raid_level > RAID_UNKNOWN ? + "RAID-?" : + raid_label[sd[i]->raid_level], + sd[i]->offload_config ? '+' : '-', + sd[i]->offload_enabled ? '+' : '-', + sd[i]->expose_state); continue; } @@ -1356,6 +1812,14 @@ static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno, /* but if it does happen, we just ignore that device */ } } + hpsa_update_log_drive_phys_drive_ptrs(h, h->dev, h->ndevices); + + /* Now that h->dev[]->phys_disk[] is coherent, we can enable + * any logical drives that need it enabled. + */ + for (i = 0; i < h->ndevices; i++) + h->dev[i]->offload_enabled = h->dev[i]->offload_to_be_enabled; + spin_unlock_irqrestore(&h->devlock, flags); /* Monitor devices which are in one of several NOT READY states to be @@ -1379,20 +1843,23 @@ static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno, sh = h->scsi_host; /* Notify scsi mid layer of any removed devices */ for (i = 0; i < nremoved; i++) { - struct scsi_device *sdev = - scsi_device_lookup(sh, removed[i]->bus, - removed[i]->target, removed[i]->lun); - if (sdev != NULL) { - scsi_remove_device(sdev); - scsi_device_put(sdev); - } else { - /* We don't expect to get here. - * 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.", hostno, removed[i]->bus, - removed[i]->target, removed[i]->lun); + if (removed[i]->expose_state & HPSA_SCSI_ADD) { + struct scsi_device *sdev = + scsi_device_lookup(sh, removed[i]->bus, + removed[i]->target, removed[i]->lun); + if (sdev != NULL) { + scsi_remove_device(sdev); + scsi_device_put(sdev); + } else { + /* We don't expect to get here. + * future cmds to this device will get selection + * timeout as if the device was gone. + */ + dev_warn(&h->pdev->dev, + "didn't find scsi %d:%d:%d:%d for removal.", + hostno, removed[i]->bus, + removed[i]->target, removed[i]->lun); + } } kfree(removed[i]); removed[i] = NULL; @@ -1400,16 +1867,20 @@ static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno, /* Notify scsi mid layer of any added devices */ for (i = 0; i < nadded; i++) { + if (!(added[i]->expose_state & HPSA_SCSI_ADD)) + continue; 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, + dev_warn(&h->pdev->dev, + "scsi %d:%d:%d:%d addition failed, device not added.\n", + hostno, added[i]->bus, added[i]->target, added[i]->lun); /* now we have to remove it from h->dev, * since it didn't get added to scsi mid layer */ fixup_botched_add(h, added[i]); + added[i] = NULL; } free_and_out: @@ -1446,19 +1917,95 @@ static int hpsa_slave_alloc(struct scsi_device *sdev) spin_lock_irqsave(&h->devlock, flags); sd = lookup_hpsa_scsi_dev(h, sdev_channel(sdev), sdev_id(sdev), sdev->lun); - if (sd != NULL) + if (sd && (sd->expose_state & HPSA_SCSI_ADD)) { + int queue_depth = sd->queue_depth; + + if (queue_depth == 0) + queue_depth = sdev->host->can_queue; + sdev->hostdata = sd; + + if (shost_use_blk_mq(sdev->host)) { + scsi_adjust_queue_depth(sdev, scsi_get_tag_type(sdev), + queue_depth); + } else { + /* We depend on tags for cmd allocation. */ + BUG_ON(!sdev->tagged_supported); + scsi_set_tag_type(sdev, MSG_SIMPLE_TAG); + scsi_activate_tcq(sdev, queue_depth); + } + atomic_set(&sd->ioaccel_cmds_out, 0); + } else { + sdev->hostdata = NULL; + } spin_unlock_irqrestore(&h->devlock, flags); return 0; } -static void hpsa_slave_destroy(struct scsi_device *sdev) +/* configure scsi device based on internal per-device structure */ +static int hpsa_slave_configure(struct scsi_device *sdev) { - /* nothing to do. */ -} + struct hpsa_scsi_dev_t *sd; + unsigned long flags; + struct ctlr_info *h; -static void hpsa_free_sg_chain_blocks(struct ctlr_info *h) -{ + h = sdev_to_hba(sdev); + spin_lock_irqsave(&h->devlock, flags); + sd = sdev->hostdata; + if (sd && !(sd->expose_state & HPSA_ULD_ATTACH)) + sdev->no_uld_attach = 1; + + spin_unlock_irqrestore(&h->devlock, flags); + return 0; +} + +static void hpsa_slave_destroy(struct scsi_device *sdev) +{ + /* nothing to do. */ +} + +static void hpsa_free_ioaccel2_sg_chain_blocks(struct ctlr_info *h) +{ + int i; + + if (!h->ioaccel2_cmd_sg_list) + return; + for (i = 0; i < h->nr_cmds; i++) { + kfree(h->ioaccel2_cmd_sg_list[i]); + h->ioaccel2_cmd_sg_list[i] = NULL; + } + kfree(h->ioaccel2_cmd_sg_list); + h->ioaccel2_cmd_sg_list = NULL; +} + +static int hpsa_allocate_ioaccel2_sg_chain_blocks(struct ctlr_info *h) +{ + int i; + + if (h->chainsize <= 0) + return 0; + + h->ioaccel2_cmd_sg_list = + kzalloc(sizeof(*h->ioaccel2_cmd_sg_list) * h->nr_cmds, + GFP_KERNEL); + if (!h->ioaccel2_cmd_sg_list) + return -ENOMEM; + for (i = 0; i < h->nr_cmds; i++) { + h->ioaccel2_cmd_sg_list[i] = + kmalloc(sizeof(*h->ioaccel2_cmd_sg_list[i]) * + h->maxsgentries, GFP_KERNEL); + if (!h->ioaccel2_cmd_sg_list[i]) + goto clean; + } + return 0; + +clean: + hpsa_free_ioaccel2_sg_chain_blocks(h); + return -ENOMEM; +} + +static void hpsa_free_sg_chain_blocks(struct ctlr_info *h) +{ int i; if (!h->cmd_sg_list) @@ -1471,7 +2018,7 @@ static void hpsa_free_sg_chain_blocks(struct ctlr_info *h) h->cmd_sg_list = NULL; } -static int hpsa_allocate_sg_chain_blocks(struct ctlr_info *h) +static int hpsa_alloc_sg_chain_blocks(struct ctlr_info *h) { int i; @@ -1480,13 +2027,17 @@ static int hpsa_allocate_sg_chain_blocks(struct ctlr_info *h) h->cmd_sg_list = kzalloc(sizeof(*h->cmd_sg_list) * h->nr_cmds, GFP_KERNEL); - if (!h->cmd_sg_list) + if (!h->cmd_sg_list) { + dev_err(&h->pdev->dev, "Failed to allocate SG list\n"); return -ENOMEM; + } for (i = 0; i < h->nr_cmds; i++) { h->cmd_sg_list[i] = kmalloc(sizeof(*h->cmd_sg_list[i]) * h->chainsize, GFP_KERNEL); - if (!h->cmd_sg_list[i]) + if (!h->cmd_sg_list[i]) { + dev_err(&h->pdev->dev, "Failed to allocate cmd SG\n"); goto clean; + } } return 0; @@ -1495,27 +2046,60 @@ static int hpsa_allocate_sg_chain_blocks(struct ctlr_info *h) return -ENOMEM; } +static int hpsa_map_ioaccel2_sg_chain_block(struct ctlr_info *h, + struct io_accel2_cmd *cp, struct CommandList *c) +{ + struct ioaccel2_sg_element *chain_block; + u64 temp64; + u32 chain_size; + + chain_block = h->ioaccel2_cmd_sg_list[c->cmdindex]; + chain_size = le32_to_cpu(cp->data_len); + temp64 = pci_map_single(h->pdev, chain_block, chain_size, + PCI_DMA_TODEVICE); + if (dma_mapping_error(&h->pdev->dev, temp64)) { + /* prevent subsequent unmapping */ + cp->sg->address = 0; + return -1; + } + cp->sg->address = (u64) cpu_to_le64(temp64); + return 0; +} + +static void hpsa_unmap_ioaccel2_sg_chain_block(struct ctlr_info *h, + struct io_accel2_cmd *cp) +{ + struct ioaccel2_sg_element *chain_sg; + u64 temp64; + u32 chain_size; + + chain_sg = cp->sg; + temp64 = le64_to_cpu(chain_sg->address); + chain_size = le32_to_cpu(cp->data_len); + pci_unmap_single(h->pdev, temp64, chain_size, PCI_DMA_TODEVICE); +} + static int hpsa_map_sg_chain_block(struct ctlr_info *h, struct CommandList *c) { struct SGDescriptor *chain_sg, *chain_block; u64 temp64; + u32 chain_len; chain_sg = &c->SG[h->max_cmd_sg_entries - 1]; chain_block = h->cmd_sg_list[c->cmdindex]; - chain_sg->Ext = HPSA_SG_CHAIN; - chain_sg->Len = sizeof(*chain_sg) * + chain_sg->Ext = cpu_to_le32(HPSA_SG_CHAIN); + chain_len = sizeof(*chain_sg) * (c->Header.SGTotal - h->max_cmd_sg_entries); - temp64 = pci_map_single(h->pdev, chain_block, chain_sg->Len, + chain_sg->Len = cpu_to_le32(chain_len); + temp64 = pci_map_single(h->pdev, chain_block, chain_len, PCI_DMA_TODEVICE); if (dma_mapping_error(&h->pdev->dev, temp64)) { /* prevent subsequent unmapping */ - chain_sg->Addr.lower = 0; - chain_sg->Addr.upper = 0; + chain_sg->Addr = 0; return -1; } - chain_sg->Addr.lower = (u32) (temp64 & 0x0FFFFFFFFULL); - chain_sg->Addr.upper = (u32) ((temp64 >> 32) & 0x0FFFFFFFFULL); + chain_sg->Addr = cpu_to_le64(temp64); return 0; } @@ -1523,15 +2107,13 @@ static void hpsa_unmap_sg_chain_block(struct ctlr_info *h, struct CommandList *c) { struct SGDescriptor *chain_sg; - union u64bit temp64; - if (c->Header.SGTotal <= h->max_cmd_sg_entries) + if (le16_to_cpu(c->Header.SGTotal) <= h->max_cmd_sg_entries) return; chain_sg = &c->SG[h->max_cmd_sg_entries - 1]; - temp64.val32.lower = chain_sg->Addr.lower; - temp64.val32.upper = chain_sg->Addr.upper; - pci_unmap_single(h->pdev, temp64.val, chain_sg->Len, PCI_DMA_TODEVICE); + pci_unmap_single(h->pdev, le64_to_cpu(chain_sg->Addr), + le32_to_cpu(chain_sg->Len), PCI_DMA_TODEVICE); } @@ -1546,6 +2128,7 @@ static int handle_ioaccel_mode2_error(struct ctlr_info *h, { int data_len; int retry = 0; + u32 ioaccel2_resid = 0; switch (c2->error_data.serv_response) { case IOACCEL2_SERV_RESPONSE_COMPLETE: @@ -1553,9 +2136,6 @@ static int handle_ioaccel_mode2_error(struct ctlr_info *h, case IOACCEL2_STATUS_SR_TASK_COMP_GOOD: break; case IOACCEL2_STATUS_SR_TASK_COMP_CHK_COND: - dev_warn(&h->pdev->dev, - "%s: task complete with check condition.\n", - "HP SSD Smart Path"); cmd->result |= SAM_STAT_CHECK_CONDITION; if (c2->error_data.data_present != IOACCEL2_SENSE_DATA_PRESENT) { @@ -1575,58 +2155,58 @@ static int handle_ioaccel_mode2_error(struct ctlr_info *h, retry = 1; break; case IOACCEL2_STATUS_SR_TASK_COMP_BUSY: - dev_warn(&h->pdev->dev, - "%s: task complete with BUSY status.\n", - "HP SSD Smart Path"); retry = 1; break; case IOACCEL2_STATUS_SR_TASK_COMP_RES_CON: - dev_warn(&h->pdev->dev, - "%s: task complete with reservation conflict.\n", - "HP SSD Smart Path"); retry = 1; break; case IOACCEL2_STATUS_SR_TASK_COMP_SET_FULL: - /* Make scsi midlayer do unlimited retries */ - cmd->result = DID_IMM_RETRY << 16; + retry = 1; break; case IOACCEL2_STATUS_SR_TASK_COMP_ABORTED: - dev_warn(&h->pdev->dev, - "%s: task complete with aborted status.\n", - "HP SSD Smart Path"); retry = 1; break; default: - dev_warn(&h->pdev->dev, - "%s: task complete with unrecognized status: 0x%02x\n", - "HP SSD Smart Path", c2->error_data.status); retry = 1; break; } break; case IOACCEL2_SERV_RESPONSE_FAILURE: - /* don't expect to get here. */ - dev_warn(&h->pdev->dev, - "unexpected delivery or target failure, status = 0x%02x\n", - c2->error_data.status); - retry = 1; + switch (c2->error_data.status) { + case IOACCEL2_STATUS_SR_IO_ERROR: + case IOACCEL2_STATUS_SR_IO_ABORTED: + case IOACCEL2_STATUS_SR_OVERRUN: + retry = 1; + break; + case IOACCEL2_STATUS_SR_UNDERRUN: + cmd->result = (DID_OK << 16); /* host byte */ + cmd->result |= (COMMAND_COMPLETE << 8); /* msg byte */ + ioaccel2_resid = c2->error_data.resid_cnt[3] << 24; + ioaccel2_resid |= c2->error_data.resid_cnt[2] << 16; + ioaccel2_resid |= c2->error_data.resid_cnt[1] << 8; + ioaccel2_resid |= c2->error_data.resid_cnt[0]; + scsi_set_resid(cmd, ioaccel2_resid); + break; + case IOACCEL2_STATUS_SR_NO_PATH_TO_DEVICE: + case IOACCEL2_STATUS_SR_INVALID_DEVICE: + case IOACCEL2_STATUS_SR_IOACCEL_DISABLED: + /* We will get an event from ctlr to trigger rescan */ + retry = 1; + break; + default: + retry = 1; + } break; case IOACCEL2_SERV_RESPONSE_TMF_COMPLETE: break; case IOACCEL2_SERV_RESPONSE_TMF_SUCCESS: break; case IOACCEL2_SERV_RESPONSE_TMF_REJECTED: - dev_warn(&h->pdev->dev, "task management function rejected.\n"); retry = 1; break; case IOACCEL2_SERV_RESPONSE_TMF_WRONG_LUN: - dev_warn(&h->pdev->dev, "task management function invalid LUN\n"); break; default: - dev_warn(&h->pdev->dev, - "%s: Unrecognized server response: 0x%02x\n", - "HP SSD Smart Path", - c2->error_data.serv_response); retry = 1; break; } @@ -1634,20 +2214,65 @@ static int handle_ioaccel_mode2_error(struct ctlr_info *h, return retry; /* retry on raid path? */ } +static void hpsa_cmd_free_and_done(struct ctlr_info *h, + struct CommandList *c, struct scsi_cmnd *cmd) +{ + /* Prevent the following race in the abort handler: + * + * 1. LLD is requested to abort a SCSI command + * 2. The SCSI command completes + * 3. The struct CommandList associated with step 2 is made available + * 4. New I/O request to LLD to another LUN re-uses struct CommandList + * 5. Abort handler follows scsi_cmnd->host_scribble and + * finds struct CommandList and tries to aborts it + * Now we have aborted the wrong command. + * + * Clear c->scsi_cmd here so that if this command gets re-used, the + * abort handler will know it's a different scsi_cmnd. + */ + c->scsi_cmd = NULL; + cmd_tagged_free(h, c); + cmd->scsi_done(cmd); +} + +static void hpsa_retry_cmd(struct ctlr_info *h, struct CommandList *c) +{ + INIT_WORK(&c->work, hpsa_command_resubmit_worker); + queue_work_on(raw_smp_processor_id(), h->resubmit_wq, &c->work); +} + +static void hpsa_set_scsi_cmd_aborted(struct scsi_cmnd *cmd) +{ + cmd->result = DID_ABORT << 16; +} + +static void hpsa_cmd_abort_and_free(struct ctlr_info *h, struct CommandList *c, + struct scsi_cmnd *cmd) +{ + hpsa_set_scsi_cmd_aborted(cmd); + dev_warn(&h->pdev->dev, "CDB %16phN was aborted with status 0x%x\n", + c->Request.CDB, c->err_info->ScsiStatus); + c->scsi_cmd = NULL; + cmd_tagged_free(h, c); + wake_up_all(&h->abort_sync_wait_queue); +} + static void process_ioaccel2_completion(struct ctlr_info *h, struct CommandList *c, struct scsi_cmnd *cmd, struct hpsa_scsi_dev_t *dev) { struct io_accel2_cmd *c2 = &h->ioaccel2_cmd_pool[c->cmdindex]; - int raid_retry = 0; + + atomic_dec(&c->phys_disk->ioaccel_cmds_out); /* check for good status */ if (likely(c2->error_data.serv_response == 0 && - c2->error_data.status == 0)) { - cmd_free(h, c); - cmd->scsi_done(cmd); - return; - } + c2->error_data.status == 0)) + return hpsa_cmd_free_and_done(h, c, cmd); + + /* don't requeue a command which is being aborted */ + if (unlikely(c->abort_pending)) + return hpsa_cmd_abort_and_free(h, c, cmd); /* Any RAID offload error results in retry which will use * the normal I/O path so the controller can handle whatever's @@ -1656,26 +2281,44 @@ static void process_ioaccel2_completion(struct ctlr_info *h, if (is_logical_dev_addr_mode(dev->scsi3addr) && c2->error_data.serv_response == IOACCEL2_SERV_RESPONSE_FAILURE) { - dev->offload_enabled = 0; - h->drv_req_rescan = 1; /* schedule controller for a rescan */ - cmd->result = DID_SOFT_ERROR << 16; - cmd_free(h, c); - cmd->scsi_done(cmd); - return; + if (c2->error_data.status == + IOACCEL2_STATUS_SR_IOACCEL_DISABLED) + dev->offload_enabled = 0; + + return hpsa_retry_cmd(h, c); } - raid_retry = handle_ioaccel_mode2_error(h, c, cmd, c2); - /* If error found, disable Smart Path, schedule a rescan, - * and force a retry on the standard path. - */ - if (raid_retry) { - dev_warn(&h->pdev->dev, "%s: Retrying on standard path.\n", - "HP SSD Smart Path"); - dev->offload_enabled = 0; /* Disable Smart Path */ - h->drv_req_rescan = 1; /* schedule controller rescan */ - cmd->result = DID_SOFT_ERROR << 16; + + if (handle_ioaccel_mode2_error(h, c, cmd, c2)) + return hpsa_retry_cmd(h, c); + + return hpsa_cmd_free_and_done(h, c, cmd); +} + +/* Returns 0 on success, < 0 otherwise. */ +static int hpsa_evaluate_tmf_status(struct ctlr_info *h, + struct CommandList *cp) +{ + u8 tmf_status = cp->err_info->ScsiStatus; + + switch (tmf_status) { + case CISS_TMF_COMPLETE: + /* CISS_TMF_COMPLETE never happens, instead, + * ei->CommandStatus == 0 for this case. + */ + case CISS_TMF_SUCCESS: + return 0; + case CISS_TMF_INVALID_FRAME: + case CISS_TMF_NOT_SUPPORTED: + case CISS_TMF_FAILED: + case CISS_TMF_WRONG_LUN: + case CISS_TMF_OVERLAPPED_TAG: + break; + default: + dev_warn(&h->pdev->dev, "Unknown TMF status: %02x\n", + tmf_status); + break; } - cmd_free(h, c); - cmd->scsi_done(cmd); + return -tmf_status; } static void complete_scsi_command(struct CommandList *cp) @@ -1684,56 +2327,60 @@ static void complete_scsi_command(struct CommandList *cp) struct ctlr_info *h; struct ErrorInfo *ei; struct hpsa_scsi_dev_t *dev; + struct io_accel2_cmd *c2; - unsigned char sense_key; - unsigned char asc; /* additional sense code */ - unsigned char ascq; /* additional sense code qualifier */ + int sense_key; + int asc; /* additional sense code */ + int ascq; /* additional sense code qualifier */ unsigned long sense_data_size; ei = cp->err_info; - cmd = (struct scsi_cmnd *) cp->scsi_cmd; + cmd = cp->scsi_cmd; h = cp->h; dev = cmd->device->hostdata; + c2 = &h->ioaccel2_cmd_pool[cp->cmdindex]; scsi_dma_unmap(cmd); /* undo the DMA mappings */ if ((cp->cmd_type == CMD_SCSI) && (cp->Header.SGTotal > h->max_cmd_sg_entries)) hpsa_unmap_sg_chain_block(h, cp); + if ((cp->cmd_type == CMD_IOACCEL2) && + (c2->sg[0].chain_indicator == IOACCEL2_CHAIN)) + hpsa_unmap_ioaccel2_sg_chain_block(h, c2); + cmd->result = (DID_OK << 16); /* host byte */ cmd->result |= (COMMAND_COMPLETE << 8); /* msg byte */ + /* 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; + return hpsa_cmd_free_and_done(h, cp, cmd); + } + if (cp->cmd_type == CMD_IOACCEL2) return process_ioaccel2_completion(h, cp, cmd, dev); - cmd->result |= ei->ScsiStatus; - scsi_set_resid(cmd, ei->ResidualCnt); if (ei->CommandStatus == 0) { - cmd_free(h, cp); - cmd->scsi_done(cmd); - return; + if (cp->cmd_type == CMD_IOACCEL1) + atomic_dec(&cp->phys_disk->ioaccel_cmds_out); + return hpsa_cmd_free_and_done(h, cp, cmd); } - /* copy the sense data */ - if (SCSI_SENSE_BUFFERSIZE < sizeof(ei->SenseInfo)) - sense_data_size = SCSI_SENSE_BUFFERSIZE; - else - sense_data_size = sizeof(ei->SenseInfo); - if (ei->SenseLen < sense_data_size) - sense_data_size = ei->SenseLen; - - memcpy(cmd->sense_buffer, ei->SenseInfo, sense_data_size); - /* For I/O accelerator commands, copy over some fields to the normal * CISS header used below for error handling. */ if (cp->cmd_type == CMD_IOACCEL1) { struct io_accel1_cmd *c = &h->ioaccel_cmd_pool[cp->cmdindex]; + atomic_dec(&cp->phys_disk->ioaccel_cmds_out); cp->Header.SGList = cp->Header.SGTotal = scsi_sg_count(cmd); cp->Request.CDBLen = c->io_flags & IOACCEL1_IOFLAGS_CDBLEN_MASK; - cp->Header.Tag.lower = c->Tag.lower; - cp->Header.Tag.upper = c->Tag.upper; + cp->Header.tag = c->tag; memcpy(cp->Header.LUN.LunAddrBytes, c->CISS_LUN, 8); memcpy(cp->Request.CDB, c->CDB, cp->Request.CDBLen); @@ -1744,91 +2391,37 @@ static void complete_scsi_command(struct CommandList *cp) if (is_logical_dev_addr_mode(dev->scsi3addr)) { if (ei->CommandStatus == CMD_IOACCEL_DISABLED) dev->offload_enabled = 0; - cmd->result = DID_SOFT_ERROR << 16; - cmd_free(h, cp); - cmd->scsi_done(cmd); - return; + if (!cp->abort_pending) + return hpsa_retry_cmd(h, cp); } } + if (cp->abort_pending) + ei->CommandStatus = CMD_ABORTED; + /* an error has occurred */ switch (ei->CommandStatus) { case CMD_TARGET_STATUS: - if (ei->ScsiStatus) { - /* Get sense key */ - sense_key = 0xf & ei->SenseInfo[2]; - /* Get additional sense code */ - asc = ei->SenseInfo[12]; - /* Get addition sense code qualifier */ - ascq = ei->SenseInfo[13]; - } - + cmd->result |= ei->ScsiStatus; + /* copy the sense data */ + if (SCSI_SENSE_BUFFERSIZE < sizeof(ei->SenseInfo)) + sense_data_size = SCSI_SENSE_BUFFERSIZE; + else + sense_data_size = sizeof(ei->SenseInfo); + if (ei->SenseLen < sense_data_size) + sense_data_size = ei->SenseLen; + memcpy(cmd->sense_buffer, ei->SenseInfo, sense_data_size); + if (ei->ScsiStatus) + decode_sense_data(ei->SenseInfo, sense_data_size, + &sense_key, &asc, &ascq); if (ei->ScsiStatus == SAM_STAT_CHECK_CONDITION) { - if (check_for_unit_attention(h, cp)) - break; - if (sense_key == ILLEGAL_REQUEST) { - /* - * SCSI REPORT_LUNS is commonly unsupported on - * Smart Array. Suppress noisy complaint. - */ - if (cp->Request.CDB[0] == REPORT_LUNS) - break; - - /* If ASC/ASCQ indicate Logical Unit - * Not Supported condition, - */ - if ((asc == 0x25) && (ascq == 0x0)) { - dev_warn(&h->pdev->dev, "cp %p " - "has check condition\n", cp); - break; - } - } - - if (sense_key == NOT_READY) { - /* If Sense is Not Ready, Logical Unit - * Not ready, Manual Intervention - * required - */ - if ((asc == 0x04) && (ascq == 0x03)) { - dev_warn(&h->pdev->dev, "cp %p " - "has check condition: unit " - "not ready, manual " - "intervention required\n", cp); - break; - } - } if (sense_key == ABORTED_COMMAND) { - /* Aborted command is retryable */ - dev_warn(&h->pdev->dev, "cp %p " - "has check condition: aborted command: " - "ASC: 0x%x, ASCQ: 0x%x\n", - cp, asc, ascq); cmd->result |= DID_SOFT_ERROR << 16; break; } - /* Must be some other type of check condition */ - dev_dbg(&h->pdev->dev, "cp %p has check condition: " - "unknown type: " - "Sense: 0x%x, ASC: 0x%x, ASCQ: 0x%x, " - "Returning result: 0x%x, " - "cmd=[%02x %02x %02x %02x %02x " - "%02x %02x %02x %02x %02x %02x " - "%02x %02x %02x %02x %02x]\n", - cp, sense_key, asc, ascq, - cmd->result, - cmd->cmnd[0], cmd->cmnd[1], - cmd->cmnd[2], cmd->cmnd[3], - cmd->cmnd[4], cmd->cmnd[5], - cmd->cmnd[6], cmd->cmnd[7], - cmd->cmnd[8], cmd->cmnd[9], - cmd->cmnd[10], cmd->cmnd[11], - cmd->cmnd[12], cmd->cmnd[13], - cmd->cmnd[14], cmd->cmnd[15]); break; } - - /* Problem was not a check condition * Pass it up to the upper layers... */ @@ -1862,9 +2455,8 @@ static void complete_scsi_command(struct CommandList *cp) case CMD_DATA_UNDERRUN: /* let mid layer handle it. */ break; case CMD_DATA_OVERRUN: - dev_warn(&h->pdev->dev, "cp %p has" - " completed with data overrun " - "reported\n", cp); + dev_warn(&h->pdev->dev, + "CDB %16phN data overrun\n", cp->Request.CDB); break; case CMD_INVALID: { /* print_bytes(cp, sizeof(*cp), 1, 0); @@ -1880,39 +2472,45 @@ static void complete_scsi_command(struct CommandList *cp) break; case CMD_PROTOCOL_ERR: cmd->result = DID_ERROR << 16; - dev_warn(&h->pdev->dev, "cp %p has " - "protocol error\n", cp); + dev_warn(&h->pdev->dev, "CDB %16phN : protocol error\n", + cp->Request.CDB); break; case CMD_HARDWARE_ERR: cmd->result = DID_ERROR << 16; - dev_warn(&h->pdev->dev, "cp %p had hardware error\n", cp); + dev_warn(&h->pdev->dev, "CDB %16phN : hardware error\n", + cp->Request.CDB); break; case CMD_CONNECTION_LOST: cmd->result = DID_ERROR << 16; - dev_warn(&h->pdev->dev, "cp %p had connection lost\n", cp); + dev_warn(&h->pdev->dev, "CDB %16phN : connection lost\n", + cp->Request.CDB); break; case CMD_ABORTED: - cmd->result = DID_ABORT << 16; - dev_warn(&h->pdev->dev, "cp %p was aborted with status 0x%x\n", - cp, ei->ScsiStatus); - break; + /* Return now to avoid calling scsi_done(). */ + return hpsa_cmd_abort_and_free(h, cp, cmd); case CMD_ABORT_FAILED: cmd->result = DID_ERROR << 16; - dev_warn(&h->pdev->dev, "cp %p reports abort failed\n", cp); + dev_warn(&h->pdev->dev, "CDB %16phN : abort failed\n", + cp->Request.CDB); break; case CMD_UNSOLICITED_ABORT: cmd->result = DID_SOFT_ERROR << 16; /* retry the command */ - dev_warn(&h->pdev->dev, "cp %p aborted due to an unsolicited " - "abort\n", cp); + dev_warn(&h->pdev->dev, "CDB %16phN : unsolicited abort\n", + cp->Request.CDB); break; case CMD_TIMEOUT: cmd->result = DID_TIME_OUT << 16; - dev_warn(&h->pdev->dev, "cp %p timedout\n", cp); + dev_warn(&h->pdev->dev, "CDB %16phN timedout\n", + cp->Request.CDB); break; case CMD_UNABORTABLE: cmd->result = DID_ERROR << 16; dev_warn(&h->pdev->dev, "Command unabortable\n"); break; + case CMD_TMF_STATUS: + if (hpsa_evaluate_tmf_status(h, cp)) /* TMF failed? */ + cmd->result = DID_ERROR << 16; + break; case CMD_IOACCEL_DISABLED: /* This only handles the direct pass-through case since RAID * offload is handled above. Just attempt a retry. @@ -1926,22 +2524,19 @@ static void complete_scsi_command(struct CommandList *cp) dev_warn(&h->pdev->dev, "cp %p returned unknown status %x\n", cp, ei->CommandStatus); } - cmd_free(h, cp); - cmd->scsi_done(cmd); + + return hpsa_cmd_free_and_done(h, cp, cmd); } static void hpsa_pci_unmap(struct pci_dev *pdev, struct CommandList *c, int sg_used, int data_direction) { int i; - union u64bit addr64; - for (i = 0; i < sg_used; i++) { - addr64.val32.lower = c->SG[i].Addr.lower; - addr64.val32.upper = c->SG[i].Addr.upper; - pci_unmap_single(pdev, (dma_addr_t) addr64.val, c->SG[i].Len, - data_direction); - } + for (i = 0; i < sg_used; i++) + pci_unmap_single(pdev, (dma_addr_t) le64_to_cpu(c->SG[i].Addr), + le32_to_cpu(c->SG[i].Len), + data_direction); } static int hpsa_map_one(struct pci_dev *pdev, @@ -1954,7 +2549,7 @@ static int hpsa_map_one(struct pci_dev *pdev, if (buflen == 0 || data_direction == PCI_DMA_NONE) { cp->Header.SGList = 0; - cp->Header.SGTotal = 0; + cp->Header.SGTotal = cpu_to_le16(0); return 0; } @@ -1962,28 +2557,47 @@ static int hpsa_map_one(struct pci_dev *pdev, if (dma_mapping_error(&pdev->dev, addr64)) { /* Prevent subsequent unmap of something never mapped */ cp->Header.SGList = 0; - cp->Header.SGTotal = 0; + cp->Header.SGTotal = cpu_to_le16(0); return -1; } - cp->SG[0].Addr.lower = - (u32) (addr64 & (u64) 0x00000000FFFFFFFF); - cp->SG[0].Addr.upper = - (u32) ((addr64 >> 32) & (u64) 0x00000000FFFFFFFF); - cp->SG[0].Len = buflen; - cp->SG[0].Ext = HPSA_SG_LAST; /* we are not chaining */ + cp->SG[0].Addr = cpu_to_le64(addr64); + cp->SG[0].Len = cpu_to_le32(buflen); + cp->SG[0].Ext = cpu_to_le32(HPSA_SG_LAST); /* we are not chaining */ cp->Header.SGList = (u8) 1; /* no. SGs contig in this cmd */ - cp->Header.SGTotal = (u16) 1; /* total sgs in this cmd list */ + cp->Header.SGTotal = (u16) cpu_to_le16(1); /* total sgs in cmd list */ 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 0; + } + if (!wait_for_completion_io_timeout(&wait, + msecs_to_jiffies(timeout_msecs))) { + dev_warn(&h->pdev->dev, "Command timed out.\n"); + return -ETIMEDOUT; + } + return 0; +} + +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 0; + } + return hpsa_scsi_do_simple_cmd_core(h, c, reply_queue, timeout_msecs); } static u32 lockup_detected(struct ctlr_info *h) @@ -1998,25 +2612,19 @@ 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); @@ -2027,6 +2635,9 @@ static void hpsa_scsi_do_simple_cmd_with_retry(struct ctlr_info *h, 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 = -1; /* FIXME do something better? */ + return rc; } static void hpsa_print_cmd(struct ctlr_info *h, char *txt, @@ -2050,14 +2661,20 @@ static void hpsa_scsi_interpret_error(struct ctlr_info *h, { const struct ErrorInfo *ei = cp->err_info; struct device *d = &cp->h->pdev->dev; - const u8 *sd = ei->SenseInfo; + int sense_key, asc, ascq, sense_len; switch (ei->CommandStatus) { case CMD_TARGET_STATUS: + if (ei->SenseLen > sizeof(ei->SenseInfo)) + sense_len = sizeof(ei->SenseInfo); + else + sense_len = ei->SenseLen; + decode_sense_data(ei->SenseInfo, sense_len, + &sense_key, &asc, &ascq); hpsa_print_cmd(h, "SCSI status", cp); if (ei->ScsiStatus == SAM_STAT_CHECK_CONDITION) dev_warn(d, "SCSI Status = 02, Sense key = %02x, ASC = %02x, ASCQ = %02x\n", - sd[2] & 0x0f, sd[12], sd[13]); + sense_key, asc, ascq); else dev_warn(d, "SCSI Status = %02x\n", ei->ScsiStatus); if (ei->ScsiStatus == 0) @@ -2103,6 +2720,9 @@ static void hpsa_scsi_interpret_error(struct ctlr_info *h, 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", @@ -2118,26 +2738,24 @@ static int hpsa_scsi_do_inquiry(struct ctlr_info *h, unsigned char *scsi3addr, struct CommandList *c; struct ErrorInfo *ei; - c = cmd_special_alloc(h); - - if (c == NULL) { /* trouble... */ - dev_warn(&h->pdev->dev, "cmd_special_alloc returned NULL!\n"); - return -ENOMEM; - } + c = cmd_alloc(h); if (fill_cmd(c, HPSA_INQUIRY, h, buf, bufsize, page, scsi3addr, TYPE_CMD)) { 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; } out: - cmd_special_free(h, c); + cmd_free(h, c); return rc; } @@ -2149,48 +2767,45 @@ static int hpsa_bmic_ctrl_mode_sense(struct ctlr_info *h, struct CommandList *c; struct ErrorInfo *ei; - c = cmd_special_alloc(h); - - if (c == NULL) { /* trouble... */ - dev_warn(&h->pdev->dev, "cmd_special_alloc returned NULL!\n"); - return -ENOMEM; - } + c = cmd_alloc(h); if (fill_cmd(c, BMIC_SENSE_CONTROLLER_PARAMETERS, h, buf, bufsize, page, scsi3addr, TYPE_CMD)) { 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; } out: - cmd_special_free(h, c); + cmd_free(h, c); return rc; - } +} 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; struct ErrorInfo *ei; - c = cmd_special_alloc(h); - - if (c == NULL) { /* trouble... */ - dev_warn(&h->pdev->dev, "cmd_special_alloc returned NULL!\n"); - return -ENOMEM; - } + c = cmd_alloc(h); /* fill_cmd can't fail here, no data buffer to map. */ (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; @@ -2198,7 +2813,8 @@ static int hpsa_send_reset(struct ctlr_info *h, unsigned char *scsi3addr, hpsa_scsi_interpret_error(h, c); rc = -1; } - cmd_special_free(h, c); +out: + cmd_free(h, c); return rc; } @@ -2308,26 +2924,26 @@ static int hpsa_get_raid_map(struct ctlr_info *h, struct CommandList *c; struct ErrorInfo *ei; - c = cmd_special_alloc(h); - if (c == NULL) { - dev_warn(&h->pdev->dev, "cmd_special_alloc returned NULL!\n"); - return -ENOMEM; - } + c = cmd_alloc(h); + if (fill_cmd(c, HPSA_GET_RAID_MAP, h, &this_device->raid_map, sizeof(this_device->raid_map), 0, scsi3addr, TYPE_CMD)) { - dev_warn(&h->pdev->dev, "Out of memory in hpsa_get_raid_map()\n"); - cmd_special_free(h, c); - return -ENOMEM; + dev_warn(&h->pdev->dev, "hpsa_get_raid_map fill_cmd failed\n"); + cmd_free(h, c); + return -1; } - 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_special_free(h, c); - return -1; + rc = -1; + goto out; } - cmd_special_free(h, c); + cmd_free(h, c); /* @todo in the future, dynamically allocate RAID map memory */ if (le32_to_cpu(this_device->raid_map.structure_size) > @@ -2337,6 +2953,38 @@ static int hpsa_get_raid_map(struct ctlr_info *h, } 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, + unsigned char scsi3addr[], u16 bmic_device_index, + struct bmic_identify_physical_device *buf, size_t bufsize) +{ + int rc = IO_OK; + struct CommandList *c; + struct ErrorInfo *ei; + + c = cmd_alloc(h); + rc = fill_cmd(c, BMIC_IDENTIFY_PHYSICAL_DEVICE, h, buf, bufsize, + 0, RAID_CTLR_LUNID, TYPE_CMD); + if (rc) + goto out; + + 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, + NO_TIMEOUT); + ei = c->err_info; + if (ei->CommandStatus != 0 && ei->CommandStatus != CMD_DATA_UNDERRUN) { + hpsa_scsi_interpret_error(h, c); + rc = -1; + } +out: + cmd_free(h, c); + return rc; } static int hpsa_vpd_page_supported(struct ctlr_info *h, @@ -2391,6 +3039,7 @@ static void hpsa_get_ioaccel_status(struct ctlr_info *h, this_device->offload_config = 0; this_device->offload_enabled = 0; + this_device->offload_to_be_enabled = 0; buf = kzalloc(64, GFP_KERNEL); if (!buf) @@ -2414,6 +3063,7 @@ static void hpsa_get_ioaccel_status(struct ctlr_info *h, if (hpsa_get_raid_map(h, scsi3addr, this_device)) this_device->offload_enabled = 0; } + this_device->offload_to_be_enabled = this_device->offload_enabled; out: kfree(buf); return; @@ -2439,7 +3089,7 @@ static int hpsa_get_device_id(struct ctlr_info *h, unsigned char *scsi3addr, } static int hpsa_scsi_do_report_luns(struct ctlr_info *h, int logical, - struct ReportLUNdata *buf, int bufsize, + void *buf, int bufsize, int extended_response) { int rc = IO_OK; @@ -2447,11 +3097,8 @@ static int hpsa_scsi_do_report_luns(struct ctlr_info *h, int logical, unsigned char scsi3addr[8]; struct ErrorInfo *ei; - c = cmd_special_alloc(h); - if (c == NULL) { /* trouble... */ - dev_err(&h->pdev->dev, "cmd_special_alloc returned NULL!\n"); - return -1; - } + c = cmd_alloc(h); + /* address the controller */ memset(scsi3addr, 0, sizeof(scsi3addr)); if (fill_cmd(c, logical ? HPSA_REPORT_LOG : HPSA_REPORT_PHYS, h, @@ -2461,31 +3108,35 @@ static int hpsa_scsi_do_report_luns(struct ctlr_info *h, int logical, } 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) { hpsa_scsi_interpret_error(h, c); rc = -1; } else { - if (buf->extended_response_flag != extended_response) { + struct ReportLUNdata *rld = buf; + if (rld->extended_response_flag != extended_response) { dev_err(&h->pdev->dev, "report luns requested format %u, got %u\n", extended_response, - buf->extended_response_flag); + rld->extended_response_flag); rc = -1; } } out: - cmd_special_free(h, c); + cmd_free(h, c); return rc; } static inline int hpsa_scsi_do_report_phys_luns(struct ctlr_info *h, - struct ReportLUNdata *buf, - int bufsize, int extended_response) + struct ReportExtendedLUNdata *buf, int bufsize) { - return hpsa_scsi_do_report_luns(h, 0, buf, bufsize, extended_response); + return hpsa_scsi_do_report_luns(h, 0, buf, bufsize, + HPSA_REPORT_PHYS_EXTENDED); } static inline int hpsa_scsi_do_report_log_luns(struct ctlr_info *h, @@ -2551,8 +3202,9 @@ static int hpsa_volume_offline(struct ctlr_info *h, unsigned char scsi3addr[]) { struct CommandList *c; - unsigned char *sense, sense_key, asc, ascq; - int ldstat = 0; + unsigned char *sense; + int sense_key, asc, ascq, sense_len; + int rc, ldstat = 0; u16 cmd_status; u8 scsi_status; #define ASC_LUN_NOT_READY 0x04 @@ -2560,16 +3212,21 @@ static int hpsa_volume_offline(struct ctlr_info *h, #define ASCQ_LUN_NOT_READY_INITIALIZING_CMD_REQ 0x02 c = cmd_alloc(h); - 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); - sense = c->err_info->SenseInfo; - sense_key = sense[2]; - asc = sense[12]; - ascq = sense[13]; - cmd_status = c->err_info->CommandStatus; - scsi_status = c->err_info->ScsiStatus; + 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; + if (c->err_info->SenseLen > sizeof(c->err_info->SenseInfo)) + sense_len = sizeof(c->err_info->SenseInfo); + else + sense_len = c->err_info->SenseLen; + decode_sense_data(sense, sense_len, &sense_key, &asc, &ascq); + cmd_status = c->err_info->CommandStatus; + scsi_status = c->err_info->ScsiStatus; cmd_free(h, c); /* Is the volume 'not ready'? */ if (cmd_status != CMD_TARGET_STATUS || @@ -2607,6 +3264,51 @@ static int hpsa_volume_offline(struct ctlr_info *h, 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); + + (void) fill_cmd(c, HPSA_ABORT_MSG, h, &tag, 0, 0, scsi3addr, TYPE_MSG); + (void) hpsa_scsi_do_simple_cmd(h, c, 0, 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; + case CMD_TMF_STATUS: + rc = hpsa_evaluate_tmf_status(h, c); + 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) @@ -2659,7 +3361,10 @@ static int hpsa_update_device_info(struct ctlr_info *h, this_device->raid_level = RAID_UNKNOWN; this_device->offload_config = 0; this_device->offload_enabled = 0; + this_device->offload_to_be_enabled = 0; + this_device->hba_ioaccel_enabled = 0; this_device->volume_offline = 0; + this_device->queue_depth = h->nr_cmds; } if (is_OBDR_device) { @@ -2671,7 +3376,6 @@ static int hpsa_update_device_info(struct ctlr_info *h, strncmp(obdr_sig, OBDR_TAPE_SIG, OBDR_SIG_LEN) == 0); } - kfree(inq_buff); return 0; @@ -2680,6 +3384,31 @@ static int hpsa_update_device_info(struct ctlr_info *h, 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", @@ -2785,6 +3514,7 @@ static int add_ext_target_dev(struct ctlr_info *h, (*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; } @@ -2800,90 +3530,23 @@ static int add_ext_target_dev(struct ctlr_info *h, static int hpsa_get_pdisk_of_ioaccel2(struct ctlr_info *h, struct CommandList *ioaccel2_cmd_to_abort, unsigned char *scsi3addr) { - struct ReportExtendedLUNdata *physicals = NULL; - int responsesize = 24; /* size of physical extended response */ - int extended = 2; /* flag forces reporting 'other dev info'. */ - int reportsize = sizeof(*physicals) + HPSA_MAX_PHYS_LUN * responsesize; - u32 nphysicals = 0; /* number of reported physical devs */ - int found = 0; /* found match (1) or not (0) */ - u32 find; /* handle we need to match */ + struct io_accel2_cmd *c2 = + &h->ioaccel2_cmd_pool[ioaccel2_cmd_to_abort->cmdindex]; + unsigned long flags; int i; - struct scsi_cmnd *scmd; /* scsi command within request being aborted */ - struct hpsa_scsi_dev_t *d; /* device of request being aborted */ - struct io_accel2_cmd *c2a; /* ioaccel2 command to abort */ - u32 it_nexus; /* 4 byte device handle for the ioaccel2 cmd */ - u32 scsi_nexus; /* 4 byte device handle for the ioaccel2 cmd */ - - if (ioaccel2_cmd_to_abort->cmd_type != CMD_IOACCEL2) - return 0; /* no match */ - - /* point to the ioaccel2 device handle */ - c2a = &h->ioaccel2_cmd_pool[ioaccel2_cmd_to_abort->cmdindex]; - if (c2a == NULL) - return 0; /* no match */ - - scmd = (struct scsi_cmnd *) ioaccel2_cmd_to_abort->scsi_cmd; - if (scmd == NULL) - return 0; /* no match */ - - d = scmd->device->hostdata; - if (d == NULL) - return 0; /* no match */ - - it_nexus = cpu_to_le32((u32) d->ioaccel_handle); - scsi_nexus = cpu_to_le32((u32) c2a->scsi_nexus); - find = c2a->scsi_nexus; - - if (h->raid_offload_debug > 0) - dev_info(&h->pdev->dev, - "%s: scsi_nexus:0x%08x device id: 0x%02x%02x%02x%02x %02x%02x%02x%02x %02x%02x%02x%02x %02x%02x%02x%02x\n", - __func__, scsi_nexus, - d->device_id[0], d->device_id[1], d->device_id[2], - d->device_id[3], d->device_id[4], d->device_id[5], - d->device_id[6], d->device_id[7], d->device_id[8], - d->device_id[9], d->device_id[10], d->device_id[11], - d->device_id[12], d->device_id[13], d->device_id[14], - d->device_id[15]); - - /* Get the list of physical devices */ - physicals = kzalloc(reportsize, GFP_KERNEL); - if (physicals == NULL) - return 0; - if (hpsa_scsi_do_report_phys_luns(h, (struct ReportLUNdata *) physicals, - reportsize, extended)) { - dev_err(&h->pdev->dev, - "Can't lookup %s device handle: report physical LUNs failed.\n", - "HP SSD Smart Path"); - kfree(physicals); - return 0; - } - nphysicals = be32_to_cpu(*((__be32 *)physicals->LUNListLength)) / - responsesize; - - /* find ioaccel2 handle in list of physicals: */ - for (i = 0; i < nphysicals; i++) { - struct ext_report_lun_entry *entry = &physicals->LUN[i]; - - /* handle is in bytes 28-31 of each lun */ - if (entry->ioaccel_handle != find) - continue; /* didn't match */ - found = 1; - memcpy(scsi3addr, entry->lunid, 8); - if (h->raid_offload_debug > 0) - dev_info(&h->pdev->dev, - "%s: Searched h=0x%08x, Found h=0x%08x, scsiaddr 0x%8phN\n", - __func__, find, - entry->ioaccel_handle, scsi3addr); - break; /* found it */ - } - - kfree(physicals); - if (found) - return 1; - else - return 0; + spin_lock_irqsave(&h->devlock, flags); + for (i = 0; i < h->ndevices; i++) + if (h->dev[i]->ioaccel_handle == c2->scsi_nexus) { + memcpy(scsi3addr, h->dev[i]->scsi3addr, + sizeof(h->dev[i]->scsi3addr)); + spin_unlock_irqrestore(&h->devlock, flags); + return 1; + } + spin_unlock_irqrestore(&h->devlock, flags); + return 0; } + /* * Do CISS_REPORT_PHYS and CISS_REPORT_LOG. Data is returned in physdev, * logdev. The number of luns in physdev and logdev are returned in @@ -2891,34 +3554,21 @@ static int hpsa_get_pdisk_of_ioaccel2(struct ctlr_info *h, * Returns 0 on success, -1 otherwise. */ static int hpsa_gather_lun_info(struct ctlr_info *h, - int reportlunsize, - struct ReportLUNdata *physdev, u32 *nphysicals, int *physical_mode, + struct ReportExtendedLUNdata *physdev, u32 *nphysicals, struct ReportLUNdata *logdev, u32 *nlogicals) { - int physical_entry_size = 8; - - *physical_mode = 0; - - /* For I/O accelerator mode we need to read physical device handles */ - if (h->transMethod & CFGTBL_Trans_io_accel1 || - h->transMethod & CFGTBL_Trans_io_accel2) { - *physical_mode = HPSA_REPORT_PHYS_EXTENDED; - physical_entry_size = 24; - } - if (hpsa_scsi_do_report_phys_luns(h, physdev, reportlunsize, - *physical_mode)) { + if (hpsa_scsi_do_report_phys_luns(h, physdev, sizeof(*physdev))) { dev_err(&h->pdev->dev, "report physical LUNs failed.\n"); return -1; } - *nphysicals = be32_to_cpu(*((__be32 *)physdev->LUNListLength)) / - physical_entry_size; + *nphysicals = be32_to_cpu(*((__be32 *)physdev->LUNListLength)) / 24; if (*nphysicals > HPSA_MAX_PHYS_LUN) { dev_warn(&h->pdev->dev, "maximum physical LUNs (%d) exceeded." " %d LUNs ignored.\n", HPSA_MAX_PHYS_LUN, *nphysicals - HPSA_MAX_PHYS_LUN); *nphysicals = HPSA_MAX_PHYS_LUN; } - if (hpsa_scsi_do_report_log_luns(h, logdev, reportlunsize)) { + if (hpsa_scsi_do_report_log_luns(h, logdev, sizeof(*logdev))) { dev_err(&h->pdev->dev, "report logical LUNs failed.\n"); return -1; } @@ -2991,6 +3641,34 @@ static int hpsa_hba_mode_enabled(struct ctlr_info *h) return hba_mode_enabled; } +/* get physical drive ioaccel handle and queue depth */ +void hpsa_get_ioaccel_drive_info(struct ctlr_info *h, + struct hpsa_scsi_dev_t *dev, + u8 *lunaddrbytes, + struct bmic_identify_physical_device *id_phys) +{ + int rc; + struct ext_report_lun_entry *rle = + (struct ext_report_lun_entry *) lunaddrbytes; + + dev->ioaccel_handle = rle->ioaccel_handle; + if (PHYS_IOACCEL(lunaddrbytes) && dev->ioaccel_handle) + dev->hba_ioaccel_enabled = 1; + memset(id_phys, 0, sizeof(*id_phys)); + rc = hpsa_bmic_id_physical_device(h, lunaddrbytes, + GET_BMIC_DRIVE_NUMBER(lunaddrbytes), id_phys, + sizeof(*id_phys)); + if (!rc) + /* Reserve space for FW operations */ +#define DRIVE_CMDS_RESERVED_FOR_FW 2 + dev->queue_depth = + le16_to_cpu(id_phys->current_queue_depth_limit) - + DRIVE_CMDS_RESERVED_FOR_FW; + else + dev->queue_depth = 7; /* conservative */ + atomic_set(&dev->ioaccel_cmds_out, 0); +} + static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) { /* the idea here is we could get notified @@ -3005,24 +3683,25 @@ static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) */ struct ReportExtendedLUNdata *physdev_list = NULL; struct ReportLUNdata *logdev_list = NULL; + struct bmic_identify_physical_device *id_phys = NULL; u32 nphysicals = 0; u32 nlogicals = 0; - int physical_mode = 0; u32 ndev_allocated = 0; struct hpsa_scsi_dev_t **currentsd, *this_device, *tmpdevice; int ncurrent = 0; - int reportlunsize = sizeof(*physdev_list) + HPSA_MAX_PHYS_LUN * 24; int i, n_ext_target_devs, ndevs_to_allocate; int raid_ctlr_position; int rescan_hba_mode; DECLARE_BITMAP(lunzerobits, MAX_EXT_TARGETS); currentsd = kzalloc(sizeof(*currentsd) * HPSA_MAX_DEVICES, GFP_KERNEL); - physdev_list = kzalloc(reportlunsize, GFP_KERNEL); - logdev_list = kzalloc(reportlunsize, GFP_KERNEL); + physdev_list = kzalloc(sizeof(*physdev_list), GFP_KERNEL); + logdev_list = kzalloc(sizeof(*logdev_list), GFP_KERNEL); tmpdevice = kzalloc(sizeof(*tmpdevice), GFP_KERNEL); + id_phys = kzalloc(sizeof(*id_phys), GFP_KERNEL); - if (!currentsd || !physdev_list || !logdev_list || !tmpdevice) { + if (!currentsd || !physdev_list || !logdev_list || + !tmpdevice || !id_phys) { dev_err(&h->pdev->dev, "out of memory\n"); goto out; } @@ -3039,9 +3718,8 @@ static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) h->hba_mode_enabled = rescan_hba_mode; - if (hpsa_gather_lun_info(h, reportlunsize, - (struct ReportLUNdata *) physdev_list, &nphysicals, - &physical_mode, logdev_list, &nlogicals)) + if (hpsa_gather_lun_info(h, physdev_list, &nphysicals, + logdev_list, &nlogicals)) goto out; /* We might see up to the maximum number of logical and physical disks @@ -3081,16 +3759,19 @@ static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) /* Figure out where the LUN ID info is coming from */ lunaddrbytes = figure_lunaddrbytes(h, raid_ctlr_position, i, nphysicals, nlogicals, physdev_list, logdev_list); - /* skip masked physical devices. */ - if (lunaddrbytes[3] & 0xC0 && - i < nphysicals + (raid_ctlr_position == 0)) - continue; + + /* skip masked non-disk devices */ + if (MASKED_DEVICE(lunaddrbytes)) + if (i < nphysicals + (raid_ctlr_position == 0) && + NON_DISK_PHYS_DEV(lunaddrbytes)) + continue; /* Get device type, vendor, model, device id */ if (hpsa_update_device_info(h, lunaddrbytes, tmpdevice, &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]; /* @@ -3109,6 +3790,18 @@ static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) *this_device = *tmpdevice; + /* do not expose masked devices */ + if (MASKED_DEVICE(lunaddrbytes) && + i < nphysicals + (raid_ctlr_position == 0)) { + if (h->hba_mode_enabled) + dev_warn(&h->pdev->dev, + "Masked physical device detected\n"); + this_device->expose_state = HPSA_DO_NOT_EXPOSE; + } else { + this_device->expose_state = + HPSA_SG_ATTACH | HPSA_ULD_ATTACH; + } + switch (this_device->devtype) { case TYPE_ROM: /* We don't *really* support actual CD-ROM devices, @@ -3122,33 +3815,30 @@ static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) ncurrent++; break; case TYPE_DISK: - if (h->hba_mode_enabled) { - /* never use raid mapper in HBA mode */ - this_device->offload_enabled = 0; - ncurrent++; - break; - } else if (h->acciopath_status) { - if (i >= nphysicals) { - ncurrent++; - break; - } - } else { - if (i < nphysicals) - break; + if (i >= nphysicals) { ncurrent++; break; } - if (physical_mode == HPSA_REPORT_PHYS_EXTENDED) { - memcpy(&this_device->ioaccel_handle, - &lunaddrbytes[20], - sizeof(this_device->ioaccel_handle)); - ncurrent++; - } + + if (h->hba_mode_enabled) + /* never use raid mapper in HBA mode */ + this_device->offload_enabled = 0; + else if (!(h->transMethod & CFGTBL_Trans_io_accel1 || + h->transMethod & CFGTBL_Trans_io_accel2)) + break; + + hpsa_get_ioaccel_drive_info(h, this_device, + lunaddrbytes, id_phys); + ncurrent++; break; case TYPE_TAPE: case TYPE_MEDIUM_CHANGER: ncurrent++; break; + case TYPE_ENCLOSURE: + if (h->hba_mode_enabled) + ncurrent++; + break; case TYPE_RAID: /* Only present the Smartarray HBA as a RAID controller. * If it's a RAID controller other than the HBA itself @@ -3173,6 +3863,18 @@ static void hpsa_update_scsi_devices(struct ctlr_info *h, int hostno) kfree(currentsd); kfree(physdev_list); kfree(logdev_list); + kfree(id_phys); +} + +static void hpsa_set_sg_descriptor(struct SGDescriptor *desc, + struct scatterlist *sg) +{ + u64 addr64 = (u64) sg_dma_address(sg); + unsigned int len = sg_dma_len(sg); + + desc->Addr = cpu_to_le64(addr64); + desc->Len = cpu_to_le32(len); + desc->Ext = 0; } /* hpsa_scatter_gather takes a struct scsi_cmnd, (cmd), and does the pci @@ -3183,10 +3885,8 @@ static int hpsa_scatter_gather(struct ctlr_info *h, struct CommandList *cp, struct scsi_cmnd *cmd) { - unsigned int len; struct scatterlist *sg; - u64 addr64; - int use_sg, i, sg_index, chained; + int use_sg, i, sg_limit, chained, last_sg; struct SGDescriptor *curr_sg; BUG_ON(scsi_sg_count(cmd) > h->maxsgentries); @@ -3198,25 +3898,36 @@ static int hpsa_scatter_gather(struct ctlr_info *h, if (!use_sg) goto sglist_finished; + /* If the number of entries is greater than the max for a single list, + * then we have a chained list; we will set up all but one entry in the + * first list (the last entry is saved for link information); + * otherwise, we don't have a chained list and we'll set up at each of + * the entries in the one list. */ curr_sg = cp->SG; - chained = 0; - sg_index = 0; - scsi_for_each_sg(cmd, sg, use_sg, i) { - if (i == h->max_cmd_sg_entries - 1 && - use_sg > h->max_cmd_sg_entries) { - chained = 1; - curr_sg = h->cmd_sg_list[cp->cmdindex]; - sg_index = 0; - } - addr64 = (u64) sg_dma_address(sg); - len = sg_dma_len(sg); - curr_sg->Addr.lower = (u32) (addr64 & 0x0FFFFFFFFULL); - curr_sg->Addr.upper = (u32) ((addr64 >> 32) & 0x0FFFFFFFFULL); - curr_sg->Len = len; - curr_sg->Ext = (i < scsi_sg_count(cmd) - 1) ? 0 : HPSA_SG_LAST; + chained = use_sg > h->max_cmd_sg_entries; + sg_limit = chained ? h->max_cmd_sg_entries - 1 : use_sg; + last_sg = scsi_sg_count(cmd) - 1; + scsi_for_each_sg(cmd, sg, sg_limit, i) { + hpsa_set_sg_descriptor(curr_sg, sg); curr_sg++; } + if (chained) { + /* Continue with the chained list. Set curr_sg to the chained + * list. Modify the limit to the total count less the entries + * we've already set up. Resume the scan at the list entry + * where the previous loop left off. */ + curr_sg = h->cmd_sg_list[cp->cmdindex]; + sg_limit = use_sg - sg_limit; + for_each_sg(sg, sg, sg_limit, i) { + hpsa_set_sg_descriptor(curr_sg, sg); + curr_sg++; + } + } + + /* Back the pointer up to the last entry and mark it as "last". */ + (curr_sg - 1)->Ext = cpu_to_le32(HPSA_SG_LAST); + if (use_sg + chained > h->maxSG) h->maxSG = use_sg + chained; @@ -3287,7 +3998,7 @@ static int fixup_ioaccel_cdb(u8 *cdb, int *cdb_len) static int hpsa_scsi_ioaccel1_queue_command(struct ctlr_info *h, struct CommandList *c, u32 ioaccel_handle, u8 *cdb, int cdb_len, - u8 *scsi3addr) + u8 *scsi3addr, struct hpsa_scsi_dev_t *phys_disk) { struct scsi_cmnd *cmd = c->scsi_cmd; struct io_accel1_cmd *cp = &h->ioaccel_cmd_pool[c->cmdindex]; @@ -3295,18 +4006,22 @@ static int hpsa_scsi_ioaccel1_queue_command(struct ctlr_info *h, unsigned int total_len = 0; struct scatterlist *sg; u64 addr64; - int use_sg, i; + int use_sg, i, last_sg; struct SGDescriptor *curr_sg; u32 control = IOACCEL1_CONTROL_SIMPLEQUEUE; /* TODO: implement chaining support */ - if (scsi_sg_count(cmd) > h->ioaccel_maxsg) + if (scsi_sg_count(cmd) > h->ioaccel_maxsg) { + atomic_dec(&phys_disk->ioaccel_cmds_out); return IO_ACCEL_INELIGIBLE; + } BUG_ON(cmd->cmd_len > IOACCEL1_IOFLAGS_CDBLEN_MAX); - if (fixup_ioaccel_cdb(cdb, &cdb_len)) + if (fixup_ioaccel_cdb(cdb, &cdb_len)) { + atomic_dec(&phys_disk->ioaccel_cmds_out); return IO_ACCEL_INELIGIBLE; + } c->cmd_type = CMD_IOACCEL1; @@ -3315,25 +4030,30 @@ static int hpsa_scsi_ioaccel1_queue_command(struct ctlr_info *h, (c->cmdindex * sizeof(*cp)); BUG_ON(c->busaddr & 0x0000007F); + /* Try to honor the device's queue depth */ + if (atomic_inc_return(&phys_disk->ioaccel_cmds_out) > + phys_disk->queue_depth) { + atomic_dec(&phys_disk->ioaccel_cmds_out); + return IO_ACCEL_INELIGIBLE; + } + use_sg = scsi_dma_map(cmd); - if (use_sg < 0) + if (use_sg < 0) { + atomic_dec(&phys_disk->ioaccel_cmds_out); return use_sg; + } if (use_sg) { + last_sg = scsi_sg_count(cmd) - 1; curr_sg = cp->SG; scsi_for_each_sg(cmd, sg, use_sg, i) { addr64 = (u64) sg_dma_address(sg); len = sg_dma_len(sg); total_len += len; - curr_sg->Addr.lower = (u32) (addr64 & 0x0FFFFFFFFULL); - curr_sg->Addr.upper = - (u32) ((addr64 >> 32) & 0x0FFFFFFFFULL); - curr_sg->Len = len; - - if (i == (scsi_sg_count(cmd) - 1)) - curr_sg->Ext = HPSA_SG_LAST; - else - curr_sg->Ext = 0; /* we are not chaining */ + curr_sg->Addr = cpu_to_le64(addr64); + curr_sg->Len = cpu_to_le32(len); + curr_sg->Ext = + cpu_to_le32((i == last_sg) * HPSA_SG_LAST); curr_sg++; } @@ -3381,28 +4101,24 @@ static int hpsa_scsi_ioaccel_direct_map(struct ctlr_info *h, struct scsi_cmnd *cmd = c->scsi_cmd; struct hpsa_scsi_dev_t *dev = cmd->device->hostdata; + c->phys_disk = dev; + return hpsa_scsi_ioaccel_queue_command(h, c, dev->ioaccel_handle, - cmd->cmnd, cmd->cmd_len, dev->scsi3addr); + cmd->cmnd, cmd->cmd_len, dev->scsi3addr, dev); } /* * Set encryption parameters for the ioaccel2 request */ static void set_encrypt_ioaccel2(struct ctlr_info *h, - struct CommandList *c, struct io_accel2_cmd *cp) + struct CommandList *c, struct io_accel2_cmd *cp, u16 dekindex) { struct scsi_cmnd *cmd = c->scsi_cmd; struct hpsa_scsi_dev_t *dev = cmd->device->hostdata; struct raid_map_data *map = &dev->raid_map; u64 first_block; - BUG_ON(!(dev->offload_config && dev->offload_enabled)); - - /* Are we doing encryption on this device */ - if (!(map->flags & RAID_MAP_FLAG_ENCRYPT_ON)) - return; - /* Set the data encryption key index. */ - cp->dekindex = map->dekindex; + cp->dekindex = dekindex; /* Set the encryption enable flag, encoded into direction field. */ cp->direction |= IOACCEL2_DIRECTION_ENCRYPT_MASK; @@ -3509,7 +4225,7 @@ static void set_encrypt_ioaccel2(struct ctlr_info *h, static int hpsa_scsi_ioaccel2_queue_command(struct ctlr_info *h, struct CommandList *c, u32 ioaccel_handle, u8 *cdb, int cdb_len, - u8 *scsi3addr) + u8 *scsi3addr, struct hpsa_scsi_dev_t *phys_disk) { struct scsi_cmnd *cmd = c->scsi_cmd; struct io_accel2_cmd *cp = &h->ioaccel2_cmd_pool[c->cmdindex]; @@ -3520,11 +4236,13 @@ static int hpsa_scsi_ioaccel2_queue_command(struct ctlr_info *h, u32 len; u32 total_len = 0; - if (scsi_sg_count(cmd) > h->ioaccel_maxsg) - return IO_ACCEL_INELIGIBLE; + BUG_ON(scsi_sg_count(cmd) > h->maxsgentries); - if (fixup_ioaccel_cdb(cdb, &cdb_len)) + if (fixup_ioaccel_cdb(cdb, &cdb_len)) { + atomic_dec(&phys_disk->ioaccel_cmds_out); return IO_ACCEL_INELIGIBLE; + } + c->cmd_type = CMD_IOACCEL2; /* Adjust the DMA address to point to the accelerated command buffer */ c->busaddr = (u32) h->ioaccel2_cmd_pool_dhandle + @@ -3534,13 +4252,32 @@ static int hpsa_scsi_ioaccel2_queue_command(struct ctlr_info *h, memset(cp, 0, sizeof(*cp)); cp->IU_type = IOACCEL2_IU_TYPE; + /* Try to honor the device's queue depth */ + if (atomic_inc_return(&phys_disk->ioaccel_cmds_out) > + phys_disk->queue_depth) { + atomic_dec(&phys_disk->ioaccel_cmds_out); + return IO_ACCEL_INELIGIBLE; + } + use_sg = scsi_dma_map(cmd); - if (use_sg < 0) + if (use_sg < 0) { + atomic_dec(&phys_disk->ioaccel_cmds_out); return use_sg; + } if (use_sg) { - BUG_ON(use_sg > IOACCEL2_MAXSGENTRIES); curr_sg = cp->sg; + if (use_sg > h->ioaccel_maxsg) { + addr64 = h->ioaccel2_cmd_sg_list[c->cmdindex]->address; + curr_sg->address = cpu_to_le64(addr64); + curr_sg->length = 0; + curr_sg->reserved[0] = 0; + curr_sg->reserved[1] = 0; + curr_sg->reserved[2] = 0; + curr_sg->chain_indicator = 0x80; + + curr_sg = h->ioaccel2_cmd_sg_list[c->cmdindex]; + } scsi_for_each_sg(cmd, sg, use_sg, i) { addr64 = (u64) sg_dma_address(sg); len = sg_dma_len(sg); @@ -3579,21 +4316,30 @@ static int hpsa_scsi_ioaccel2_queue_command(struct ctlr_info *h, } /* Set encryption parameters, if necessary */ - set_encrypt_ioaccel2(h, c, cp); + if (phys_disk->raid_map.flags & RAID_MAP_FLAG_ENCRYPT_ON) + set_encrypt_ioaccel2(h, c, cp, phys_disk->raid_map.dekindex); cp->scsi_nexus = ioaccel_handle; - cp->Tag = (c->cmdindex << DIRECT_LOOKUP_SHIFT) | - DIRECT_LOOKUP_BIT; + cp->Tag = c->cmdindex << DIRECT_LOOKUP_SHIFT; memcpy(cp->cdb, cdb, sizeof(cp->cdb)); - /* fill in sg elements */ - cp->sg_count = (u8) use_sg; - cp->data_len = cpu_to_le32(total_len); cp->err_ptr = cpu_to_le64(c->busaddr + offsetof(struct io_accel2_cmd, error_data)); cp->err_len = cpu_to_le32((u32) sizeof(cp->error_data)); + /* fill in sg elements */ + if (use_sg > h->ioaccel_maxsg) { + cp->sg_count = 1; + if (hpsa_map_ioaccel2_sg_chain_block(h, cp, c)) { + atomic_dec(&phys_disk->ioaccel_cmds_out); + scsi_dma_unmap(cmd); + return -1; + } + } else { + cp->sg_count = (u8) use_sg; + } + enqueue_cmd_and_start_io(h, c); return 0; } @@ -3603,14 +4349,22 @@ static int hpsa_scsi_ioaccel2_queue_command(struct ctlr_info *h, */ static int hpsa_scsi_ioaccel_queue_command(struct ctlr_info *h, struct CommandList *c, u32 ioaccel_handle, u8 *cdb, int cdb_len, - u8 *scsi3addr) + u8 *scsi3addr, struct hpsa_scsi_dev_t *phys_disk) { + /* Try to honor the device's queue depth */ + if (atomic_inc_return(&phys_disk->ioaccel_cmds_out) > + phys_disk->queue_depth) { + atomic_dec(&phys_disk->ioaccel_cmds_out); + return IO_ACCEL_INELIGIBLE; + } if (h->transMethod & CFGTBL_Trans_io_accel1) return hpsa_scsi_ioaccel1_queue_command(h, c, ioaccel_handle, - cdb, cdb_len, scsi3addr); + cdb, cdb_len, scsi3addr, + phys_disk); else return hpsa_scsi_ioaccel2_queue_command(h, c, ioaccel_handle, - cdb, cdb_len, scsi3addr); + cdb, cdb_len, scsi3addr, + phys_disk); } static void raid_map_helper(struct raid_map_data *map, @@ -3675,8 +4429,6 @@ static int hpsa_scsi_ioaccel_raid_map(struct ctlr_info *h, #endif int offload_to_mirror; - BUG_ON(!(dev->offload_config && dev->offload_enabled)); - /* check for valid opcode, get LBA and block count */ switch (cmd->cmnd[0]) { case WRITE_6: @@ -3809,11 +4561,6 @@ static int hpsa_scsi_ioaccel_raid_map(struct ctlr_info *h, offload_to_mirror = (offload_to_mirror >= map->layout_map_count - 1) ? 0 : offload_to_mirror + 1; - /* FIXME: remove after debug/dev */ - BUG_ON(offload_to_mirror >= map->layout_map_count); - dev_warn(&h->pdev->dev, - "DEBUG: Using physical disk map index %d from mirror group %d\n", - map_index, offload_to_mirror); dev->offload_to_mirror = offload_to_mirror; /* Avoid direct use of dev->offload_to_mirror within this * function since multiple threads might simultaneously @@ -3911,6 +4658,11 @@ static int hpsa_scsi_ioaccel_raid_map(struct ctlr_info *h, return IO_ACCEL_INELIGIBLE; } + if (unlikely(map_index >= RAID_MAP_MAX_ENTRIES)) + return IO_ACCEL_INELIGIBLE; + + c->phys_disk = dev->phys_disk[map_index]; + disk_handle = dd[map_index].ioaccel_handle; disk_block = map->disk_starting_blk + (first_row * map->strip_size) + (first_row_offset - (first_column * map->strip_size)); @@ -3956,97 +4708,43 @@ static int hpsa_scsi_ioaccel_raid_map(struct ctlr_info *h, cdb_len = 10; } return hpsa_scsi_ioaccel_queue_command(h, c, disk_handle, cdb, cdb_len, - dev->scsi3addr); + dev->scsi3addr, + dev->phys_disk[map_index]); } -static int hpsa_scsi_queue_command_lck(struct scsi_cmnd *cmd, - void (*done)(struct scsi_cmnd *)) +/* 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[]) { - struct ctlr_info *h; - struct hpsa_scsi_dev_t *dev; - unsigned char scsi3addr[8]; - struct CommandList *c; - int rc = 0; - - /* Get the ptr to our adapter structure out of cmd->host. */ - h = sdev_to_hba(cmd->device); - dev = cmd->device->hostdata; - if (!dev) { - cmd->result = DID_NO_CONNECT << 16; - done(cmd); - return 0; - } - memcpy(scsi3addr, dev->scsi3addr, sizeof(scsi3addr)); - - if (unlikely(lockup_detected(h))) { - cmd->result = DID_ERROR << 16; - done(cmd); - return 0; - } - c = cmd_alloc(h); - if (c == NULL) { /* trouble... */ - dev_err(&h->pdev->dev, "cmd_alloc returned NULL!\n"); - return SCSI_MLQUEUE_HOST_BUSY; - } - - /* Fill in the command list header */ - - cmd->scsi_done = done; /* save this for use by completion code */ - - /* save c in case we have to abort it */ cmd->host_scribble = (unsigned char *) c; - c->cmd_type = CMD_SCSI; c->scsi_cmd = cmd; - - /* Call alternate submit routine for I/O accelerated commands. - * Retries always go down the normal I/O path. - */ - if (likely(cmd->retries == 0 && - cmd->request->cmd_type == REQ_TYPE_FS && - h->acciopath_status)) { - if (dev->offload_enabled) { - rc = hpsa_scsi_ioaccel_raid_map(h, c); - if (rc == 0) - return 0; /* Sent on ioaccel path */ - if (rc < 0) { /* scsi_dma_map failed. */ - cmd_free(h, c); - return SCSI_MLQUEUE_HOST_BUSY; - } - } else if (dev->ioaccel_handle) { - rc = hpsa_scsi_ioaccel_direct_map(h, c); - if (rc == 0) - return 0; /* Sent on direct map path */ - if (rc < 0) { /* scsi_dma_map failed. */ - cmd_free(h, c); - return SCSI_MLQUEUE_HOST_BUSY; - } - } - } - c->Header.ReplyQueue = 0; /* unused in simple mode */ memcpy(&c->Header.LUN.LunAddrBytes[0], &scsi3addr[0], 8); - c->Header.Tag.lower = (c->cmdindex << DIRECT_LOOKUP_SHIFT); - c->Header.Tag.lower |= DIRECT_LOOKUP_BIT; + c->Header.tag = cpu_to_le64((u64) c->cmdindex << DIRECT_LOOKUP_SHIFT); /* Fill in the request block... */ c->Request.Timeout = 0; - memset(c->Request.CDB, 0, sizeof(c->Request.CDB)); BUG_ON(cmd->cmd_len > sizeof(c->Request.CDB)); c->Request.CDBLen = cmd->cmd_len; memcpy(c->Request.CDB, cmd->cmnd, cmd->cmd_len); - c->Request.Type.Type = TYPE_CMD; - c->Request.Type.Attribute = ATTR_SIMPLE; switch (cmd->sc_data_direction) { case DMA_TO_DEVICE: - c->Request.Type.Direction = XFER_WRITE; + c->Request.type_attr_dir = + TYPE_ATTR_DIR(TYPE_CMD, ATTR_SIMPLE, XFER_WRITE); break; case DMA_FROM_DEVICE: - c->Request.Type.Direction = XFER_READ; + c->Request.type_attr_dir = + TYPE_ATTR_DIR(TYPE_CMD, ATTR_SIMPLE, XFER_READ); break; case DMA_NONE: - c->Request.Type.Direction = XFER_NONE; + c->Request.type_attr_dir = + TYPE_ATTR_DIR(TYPE_CMD, ATTR_SIMPLE, XFER_NONE); break; case DMA_BIDIRECTIONAL: /* This can happen if a buggy application does a scsi passthru @@ -4054,7 +4752,8 @@ static int hpsa_scsi_queue_command_lck(struct scsi_cmnd *cmd, * ../scsi/scsi_ioctl.c:scsi_ioctl_send_command() ) */ - c->Request.Type.Direction = XFER_RSVD; + c->Request.type_attr_dir = + TYPE_ATTR_DIR(TYPE_CMD, ATTR_SIMPLE, XFER_RSVD); /* This is technically wrong, and hpsa controllers should * reject it with CMD_INVALID, which is the most correct * response, but non-fibre backends appear to let it @@ -4073,7 +4772,7 @@ static int hpsa_scsi_queue_command_lck(struct scsi_cmnd *cmd, } if (hpsa_scatter_gather(h, c, cmd) < 0) { /* Fill SG list */ - cmd_free(h, c); + cmd_tagged_free(h, c); return SCSI_MLQUEUE_HOST_BUSY; } enqueue_cmd_and_start_io(h, c); @@ -4081,50 +4780,293 @@ static int hpsa_scsi_queue_command_lck(struct scsi_cmnd *cmd, return 0; } -static DEF_SCSI_QCMD(hpsa_scsi_queue_command) +static void hpsa_cmd_init(struct ctlr_info *h, int index, + struct CommandList *c) +{ + dma_addr_t cmd_dma_handle, err_dma_handle; + + /* Zero out all of commandlist except the last field, refcount */ + memset(c, 0, offsetof(struct CommandList, refcount)); + c->Header.tag = cpu_to_le64((u64) (index << DIRECT_LOOKUP_SHIFT)); + cmd_dma_handle = h->cmd_pool_dhandle + index * sizeof(*c); + c->err_info = h->errinfo_pool + index; + memset(c->err_info, 0, sizeof(*c->err_info)); + err_dma_handle = h->errinfo_pool_dhandle + + index * sizeof(*c->err_info); + c->cmdindex = index; + c->busaddr = (u32) cmd_dma_handle; + c->ErrDesc.Addr = cpu_to_le64((u64) err_dma_handle); + c->ErrDesc.Len = cpu_to_le32((u32) sizeof(*c->err_info)); + c->h = h; +} -static int do_not_scan_if_controller_locked_up(struct ctlr_info *h) +static void hpsa_preinitialize_commands(struct ctlr_info *h) { - unsigned long flags; + int i; - /* - * Don't let rescans be initiated on a controller known - * to be locked up. If the controller locks up *during* - * a rescan, that thread is probably hosed, but at least - * we can prevent new rescan threads from piling up on a - * locked up controller. - */ - if (unlikely(lockup_detected(h))) { - spin_lock_irqsave(&h->scan_lock, flags); - h->scan_finished = 1; - wake_up_all(&h->scan_wait_queue); - spin_unlock_irqrestore(&h->scan_lock, flags); - return 1; + for (i = 0; i < h->nr_cmds; i++) { + struct CommandList *c = h->cmd_pool + i; + hpsa_cmd_init(h, i, c); + atomic_set(&c->refcount, 0); } - return 0; } -static void hpsa_scan_start(struct Scsi_Host *sh) +static inline void hpsa_cmd_partial_init(struct ctlr_info *h, int index, + struct CommandList *c) { - struct ctlr_info *h = shost_to_hba(sh); - unsigned long flags; + dma_addr_t cmd_dma_handle = h->cmd_pool_dhandle + index * sizeof(*c); - if (do_not_scan_if_controller_locked_up(h)) - return; + memset(c->Request.CDB, 0, sizeof(c->Request.CDB)); + memset(c->err_info, 0, sizeof(*c->err_info)); + c->busaddr = (u32) cmd_dma_handle; +} - /* wait until any scan already in progress is finished. */ - while (1) { - spin_lock_irqsave(&h->scan_lock, flags); - if (h->scan_finished) - break; - spin_unlock_irqrestore(&h->scan_lock, flags); - wait_event(h->scan_wait_queue, h->scan_finished); - /* Note: We don't need to worry about a race between this - * thread and driver unload because the midlayer will - * have incremented the reference count, so unload won't - * happen if we're in here. - */ - } +static int hpsa_ioaccel_submit(struct ctlr_info *h, + struct CommandList *c, struct scsi_cmnd *cmd, + unsigned char *scsi3addr) +{ + struct hpsa_scsi_dev_t *dev = cmd->device->hostdata; + int rc = IO_ACCEL_INELIGIBLE; + + cmd->host_scribble = (unsigned char *) c; + + if (dev->offload_enabled) { + hpsa_cmd_init(h, c->cmdindex, c); + c->cmd_type = CMD_SCSI; + c->scsi_cmd = cmd; + rc = hpsa_scsi_ioaccel_raid_map(h, c); + if (rc == 0) + return 0; /* Sent on ioaccel path */ + if (rc < 0) { /* scsi_dma_map failed. */ + cmd_tagged_free(h, c); + return SCSI_MLQUEUE_HOST_BUSY; + } + } else if (dev->hba_ioaccel_enabled) { + hpsa_cmd_init(h, c->cmdindex, c); + c->cmd_type = CMD_SCSI; + c->scsi_cmd = cmd; + rc = hpsa_scsi_ioaccel_direct_map(h, c); + if (rc == 0) + return 0; /* Sent on direct map path */ + if (rc < 0) { /* scsi_dma_map failed. */ + cmd_tagged_free(h, c); + return SCSI_MLQUEUE_HOST_BUSY; + } + } + return rc; +} + +static void hpsa_command_resubmit_worker(struct work_struct *work) +{ + struct scsi_cmnd *cmd; + struct hpsa_scsi_dev_t *dev; + struct CommandList *c = container_of(work, struct CommandList, work); + + cmd = c->scsi_cmd; + dev = cmd->device->hostdata; + if (!dev) { + cmd->result = DID_NO_CONNECT << 16; + return hpsa_cmd_free_and_done(c->h, c, cmd); + } + if (c->abort_pending) + return hpsa_cmd_abort_and_free(c->h, c, cmd); + if (c->cmd_type == CMD_IOACCEL2) { + struct ctlr_info *h = c->h; + struct io_accel2_cmd *c2 = &h->ioaccel2_cmd_pool[c->cmdindex]; + int rc; + + if (c2->error_data.serv_response == + IOACCEL2_STATUS_SR_TASK_COMP_SET_FULL) { + rc = hpsa_ioaccel_submit(h, c, cmd, dev->scsi3addr); + if (rc == 0) + return; + if (rc == SCSI_MLQUEUE_HOST_BUSY) { + /* + * If we get here, it means dma mapping failed. + * Try again via scsi mid layer, which will + * then get SCSI_MLQUEUE_HOST_BUSY. + * + * hpsa_ioaccel_submit will have already freed c + * if it encountered a dma mapping failure. + */ + cmd->result = DID_IMM_RETRY << 16; + cmd->scsi_done(cmd); + return; + } + /* else, fall thru and resubmit down CISS path */ + } + } + hpsa_cmd_partial_init(c->h, c->cmdindex, c); + if (hpsa_ciss_submit(c->h, c, cmd, dev->scsi3addr)) { + /* + * If we get here, it means dma mapping failed. Try + * again via scsi mid layer, which will then get + * SCSI_MLQUEUE_HOST_BUSY. + * + * hpsa_ciss_submit will have already freed c + * if it encountered a dma mapping failure. + */ + cmd->result = DID_IMM_RETRY << 16; + cmd->scsi_done(cmd); + } +} + +/* Running in struct Scsi_Host->host_lock less mode */ +static int hpsa_scsi_queue_command(struct Scsi_Host *sh, struct scsi_cmnd *cmd) +{ + struct ctlr_info *h; + struct hpsa_scsi_dev_t *dev; + unsigned char scsi3addr[8]; + struct CommandList *c; + int rc = 0; + + /* Get the ptr to our adapter structure out of cmd->host. */ + h = sdev_to_hba(cmd->device); + dev = cmd->device->hostdata; + if (!dev) { + cmd->result = DID_NO_CONNECT << 16; + cmd->scsi_done(cmd); + return 0; + } + memcpy(scsi3addr, dev->scsi3addr, sizeof(scsi3addr)); + + if (unlikely(lockup_detected(h))) { + cmd->result = DID_NO_CONNECT << 16; + cmd->scsi_done(cmd); + return 0; + } + c = cmd_tagged_alloc(h, cmd); + + if (unlikely(lockup_detected(h))) { + cmd->result = DID_NO_CONNECT << 16; + cmd_tagged_free(h, c); /* FIXME may not be necessary, as lockup detector also frees everything */ + cmd->scsi_done(cmd); + return 0; + } + + if (hpsa_scatter_gather(h, c, cmd) < 0) { /* Fill SG list */ + cmd_free(h, c); + return SCSI_MLQUEUE_HOST_BUSY; + } + enqueue_cmd_and_start_io(h, c); + /* the cmd'll come back via intr handler in complete_scsi_command() */ + return 0; +} + +static void hpsa_command_resubmit_worker(struct work_struct *work) +{ + struct scsi_cmnd *cmd; + struct hpsa_scsi_dev_t *dev; + struct CommandList *c = + container_of(work, struct CommandList, work); + + cmd = c->scsi_cmd; + dev = cmd->device->hostdata; + if (!dev) { + cmd->result = DID_NO_CONNECT << 16; + cmd->scsi_done(cmd); + return; + } + if (hpsa_ciss_submit(c->h, c, cmd, dev->scsi3addr)) { + /* + * If we get here, it means dma mapping failed. Try + * again via scsi mid layer, which will then get + * SCSI_MLQUEUE_HOST_BUSY. + */ + cmd->result = DID_IMM_RETRY << 16; + cmd->scsi_done(cmd); + } +} + +/* Running in struct Scsi_Host->host_lock less mode */ +static int hpsa_scsi_queue_command(struct Scsi_Host *sh, struct scsi_cmnd *cmd) +{ + struct ctlr_info *h; + struct hpsa_scsi_dev_t *dev; + unsigned char scsi3addr[8]; + struct CommandList *c; + int rc = 0; + + /* Get the ptr to our adapter structure out of cmd->host. */ + h = sdev_to_hba(cmd->device); + dev = cmd->device->hostdata; + if (!dev) { + cmd->result = DID_NO_CONNECT << 16; + cmd->scsi_done(cmd); + return 0; + } + memcpy(scsi3addr, dev->scsi3addr, sizeof(scsi3addr)); + + if (unlikely(lockup_detected(h))) { + cmd->result = DID_ERROR << 16; + cmd->scsi_done(cmd); + return 0; + } + c = cmd_alloc(h); + if (c == NULL) { /* trouble... */ + dev_err(&h->pdev->dev, "cmd_alloc returned NULL!\n"); + return SCSI_MLQUEUE_HOST_BUSY; + } + +======= +>>>>>>> patched + /* Call alternate submit routine for I/O accelerated commands. + * Retries always go down the normal I/O path. + */ + if (likely(cmd->retries == 0 && + cmd->request->cmd_type == REQ_TYPE_FS && + h->acciopath_status)) { + rc = hpsa_ioaccel_submit(h, c, cmd, scsi3addr); + if (rc == 0) + return 0; + if (rc == SCSI_MLQUEUE_HOST_BUSY) + return SCSI_MLQUEUE_HOST_BUSY; + } + return hpsa_ciss_submit(h, c, cmd, scsi3addr); +} + +static int do_not_scan_if_controller_locked_up(struct ctlr_info *h) +{ + unsigned long flags; + + /* + * Don't let rescans be initiated on a controller known + * to be locked up. If the controller locks up *during* + * a rescan, that thread is probably hosed, but at least + * we can prevent new rescan threads from piling up on a + * locked up controller. + */ + if (unlikely(lockup_detected(h))) { + spin_lock_irqsave(&h->scan_lock, flags); + h->scan_finished = 1; + wake_up_all(&h->scan_wait_queue); + spin_unlock_irqrestore(&h->scan_lock, flags); + return 1; + } + return 0; +} + +static void hpsa_scan_start(struct Scsi_Host *sh) +{ + struct ctlr_info *h = shost_to_hba(sh); + unsigned long flags; + + if (do_not_scan_if_controller_locked_up(h)) + return; + + /* wait until any scan already in progress is finished. */ + while (1) { + spin_lock_irqsave(&h->scan_lock, flags); + if (h->scan_finished) + break; + spin_unlock_irqrestore(&h->scan_lock, flags); + wait_event(h->scan_wait_queue, h->scan_finished); + /* Note: We don't need to worry about a race between this + * thread and driver unload because the midlayer will + * have incremented the reference count, so unload won't + * happen if we're in here. + */ + } h->scan_finished = 0; /* mark scan as in progress */ spin_unlock_irqrestore(&h->scan_lock, flags); @@ -4152,39 +5094,65 @@ static int hpsa_scan_finished(struct Scsi_Host *sh, return finished; } +/* scsi host template change_queue_depth function */ static int hpsa_change_queue_depth(struct scsi_device *sdev, int qdepth, int reason) { - struct ctlr_info *h = sdev_to_hba(sdev); + if (reason == SCSI_QDEPTH_DEFAULT || reason == SCSI_QDEPTH_RAMP_UP) { + struct hpsa_scsi_dev_t *logical_drive = sdev->hostdata; - if (reason != SCSI_QDEPTH_DEFAULT) - return -ENOTSUPP; + if (!logical_drive) + return -ENODEV; + + if (qdepth < 1) + qdepth = 1; + else if (qdepth > logical_drive->queue_depth) + qdepth = logical_drive->queue_depth; - if (qdepth < 1) - qdepth = 1; + scsi_adjust_queue_depth(sdev, scsi_get_tag_type(sdev), qdepth); + } else if (reason == SCSI_QDEPTH_QFULL) + scsi_track_queue_full(sdev, qdepth); else - if (qdepth > h->nr_cmds) - qdepth = h->nr_cmds; - scsi_adjust_queue_depth(sdev, scsi_get_tag_type(sdev), qdepth); + return -ENOTSUPP; + return sdev->queue_depth; } -static void hpsa_unregister_scsi(struct ctlr_info *h) +static int hpsa_change_queue_type(struct scsi_device *sdev, int tag_type) { - /* we are being forcibly unloaded, and may not refuse. */ - scsi_remove_host(h->scsi_host); - scsi_host_put(h->scsi_host); - h->scsi_host = NULL; + if (sdev->tagged_supported) { + if (shost_use_blk_mq(sdev->host)) { + scsi_set_tag_type(sdev, tag_type); + if (tag_type) + scsi_activate_tcq(sdev, sdev->queue_depth); + else + scsi_deactivate_tcq(sdev, sdev->queue_depth); + } else { + /* We require tags for our internal cmd allocation; if + * the caller wants to switch tag types, that's fine, + * but don't let them be disabled. */ + if (tag_type) + scsi_set_tag_type(sdev, tag_type); + else + tag_type = scsi_get_tag_type(sdev); + } + } else { + BUG_ON(!shost_use_blk_mq(sdev->host)); + tag_type = 0; + } + + return tag_type; } -static int hpsa_register_scsi(struct ctlr_info *h) +static int hpsa_scsi_host_alloc(struct ctlr_info *h) { struct Scsi_Host *sh; - int error; sh = scsi_host_alloc(&hpsa_driver_template, sizeof(h)); - if (sh == NULL) - goto fail; + if (sh == NULL) { + dev_err(&h->pdev->dev, "scsi_host_alloc failed\n"); + return -ENOMEM; + } sh->io_port = 0; sh->n_io_port = 0; @@ -4193,80 +5161,133 @@ static int hpsa_register_scsi(struct ctlr_info *h) sh->max_cmd_len = MAX_COMMAND_SIZE; sh->max_lun = HPSA_MAX_LUN; sh->max_id = HPSA_MAX_LUN; - sh->can_queue = h->nr_cmds; - if (h->hba_mode_enabled) - sh->cmd_per_lun = 7; - else - sh->cmd_per_lun = h->nr_cmds; + sh->can_queue = h->nr_cmds - HPSA_NRESERVED_CMDS; + sh->cmd_per_lun = sh->can_queue; sh->sg_tablesize = h->maxsgentries; - h->scsi_host = sh; sh->hostdata[0] = (unsigned long) h; sh->irq = h->intr[h->intr_mode]; sh->unique_id = sh->irq; - error = scsi_add_host(sh, &h->pdev->dev); - if (error) - goto fail_host_put; - scsi_scan_host(sh); + if (!shost_use_blk_mq(sh)) { + int error = scsi_init_shared_tag_map(sh, sh->can_queue); + + if (error) { + dev_err(&h->pdev->dev, + "%s: scs_init_shared_tag_map failed for controller %d\n", + __func__, h->ctlr); + scsi_host_put(sh); + return error; + } + } + h->scsi_host = sh; + return 0; +} + +static int hpsa_scsi_add_host(struct ctlr_info *h) +{ + int rv; + + rv = scsi_add_host(h->scsi_host, &h->pdev->dev); + if (rv) { + dev_err(&h->pdev->dev, "scsi_add_host failed\n"); + return rv; + } + scsi_scan_host(h->scsi_host); return 0; +} - fail_host_put: - dev_err(&h->pdev->dev, "%s: scsi_add_host" - " failed for controller %d\n", __func__, h->ctlr); - scsi_host_put(sh); - return error; - fail: - dev_err(&h->pdev->dev, "%s: scsi_host_alloc" - " failed for controller %d\n", __func__, h->ctlr); - return -ENOMEM; +/* Send a TEST_UNIT_READY command to the specified LUN using the specified + * reply queue; returns zero if the unit is ready, and non-zero otherwise. */ +static int hpsa_send_test_unit_ready(struct ctlr_info *h, + struct CommandList *c, unsigned char lunaddr[], + int reply_queue) +{ + int rc; + + /* 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); + rc = hpsa_scsi_do_simple_cmd(h, c, reply_queue, NO_TIMEOUT); + if (rc) + return rc; + /* no unmap needed here because no data xfer. */ + + /* Check if the unit is already ready. */ + if (c->err_info->CommandStatus == CMD_SUCCESS) + return 0; + + /* The first command sent after reset will receive "unit attention" to + * indicate that the LUN has been reset...this is actually what we're + * looking for (but, success is good too). */ + if (c->err_info->CommandStatus == CMD_TARGET_STATUS && + c->err_info->ScsiStatus == SAM_STAT_CHECK_CONDITION && + (c->err_info->SenseInfo[2] == NO_SENSE || + c->err_info->SenseInfo[2] == UNIT_ATTENTION)) + return 0; + + return 1; } -static int wait_for_device_to_become_ready(struct ctlr_info *h, - unsigned char lunaddr[]) +/* Wait for a TEST_UNIT_READY command to complete, retrying as necessary; + * returns zero when the unit is ready, and non-zero when giving up. */ +static int hpsa_wait_for_test_unit_ready(struct ctlr_info *h, struct CommandList *c, + unsigned char lunaddr[], int reply_queue) { int rc; int count = 0; int waittime = 1; /* seconds */ - struct CommandList *c; - - c = cmd_special_alloc(h); - if (!c) { - dev_warn(&h->pdev->dev, "out of memory in " - "wait_for_device_to_become_ready.\n"); - return IO_ERROR; - } /* Send test unit ready until device ready, or give up. */ - while (count < HPSA_TUR_RETRY_LIMIT) { + for (count = 0; count < HPSA_TUR_RETRY_LIMIT; count++) { /* Wait for a bit. do this first, because if we send * the TUR right away, the reset will just abort it. */ msleep(1000 * waittime); - count++; - rc = 0; /* Device ready. */ + + rc = hpsa_send_test_unit_ready(h, c, lunaddr, reply_queue); + if (!rc) + break; /* Increase wait time with each try, up to a point. */ if (waittime < HPSA_MAX_WAIT_INTERVAL_SECS) waittime = waittime * 2; - /* 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); - /* no unmap needed here because no data xfer. */ + dev_warn(&h->pdev->dev, + "waiting %d secs for device to become ready.\n", + waittime); + } - if (c->err_info->CommandStatus == CMD_SUCCESS) - break; + return rc; +} - if (c->err_info->CommandStatus == CMD_TARGET_STATUS && - c->err_info->ScsiStatus == SAM_STAT_CHECK_CONDITION && - (c->err_info->SenseInfo[2] == NO_SENSE || - c->err_info->SenseInfo[2] == UNIT_ATTENTION)) - break; +static int wait_for_device_to_become_ready(struct ctlr_info *h, + unsigned char lunaddr[], + int reply_queue) +{ + int first_queue; + int last_queue; + int rq; + int rc = 0; + struct CommandList *c; + + c = cmd_alloc(h); + + /* If no specific reply queue was requested, then send the TUR + * repeatedly, requesting a reply on each reply queue; otherwise execute + * the loop exactly once using only the specified queue. */ + if (likely(reply_queue == DEFAULT_REPLY_QUEUE)) { + first_queue = 0; + last_queue = h->nreply_queues - 1; + } else { + first_queue = reply_queue; + last_queue = reply_queue; + } - dev_warn(&h->pdev->dev, "waiting %d secs " - "for device to become ready.\n", waittime); - rc = 1; /* device not ready. */ + for (rq = first_queue; rq <= last_queue; rq++) { + rc = hpsa_wait_for_test_unit_ready(h, c, lunaddr, + reply_queue == DEFAULT_REPLY_QUEUE ? rq : reply_queue); + if (rc) + break; } if (rc) @@ -4274,7 +5295,7 @@ static int wait_for_device_to_become_ready(struct ctlr_info *h, else dev_warn(&h->pdev->dev, "device is ready.\n"); - cmd_special_free(h, c); + cmd_free(h, c); return rc; } @@ -4291,20 +5312,37 @@ static int hpsa_eh_device_reset_handler(struct scsi_cmnd *scsicmd) h = sdev_to_hba(scsicmd->device); if (h == NULL) /* paranoia */ return FAILED; + + if (lockup_detected(h)) + return FAILED; + dev = scsicmd->device->hostdata; if (!dev) { dev_err(&h->pdev->dev, "hpsa_eh_device_reset_handler: " "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); + dev_warn(&h->pdev->dev, + "resetting scsi %d:%d:%d:%d: %s %.8s %.16s RAID-%s SSDSmartPathCap%c En%c Exp=%d\n", + h->scsi_host->host_no, dev->bus, dev->target, dev->lun, + 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); + /* send a reset to the SCSI LUN which the command was sent to */ - rc = hpsa_send_reset(h, dev->scsi3addr, HPSA_RESET_TYPE_LUN); - if (rc == 0 && wait_for_device_to_become_ready(h, dev->scsi3addr) == 0) + rc = hpsa_send_reset(h, dev->scsi3addr, HPSA_RESET_TYPE_LUN, + DEFAULT_REPLY_QUEUE); + if (rc == 0) return SUCCESS; - dev_warn(&h->pdev->dev, "resetting device failed.\n"); + dev_warn(&h->pdev->dev, + "resetting scsi %d:%d:%d:%d failed\n", + h->scsi_host->host_no, dev->bus, dev->target, dev->lun); return FAILED; } @@ -4329,8 +5367,8 @@ static void hpsa_get_tag(struct ctlr_info *h, if (c->cmd_type == CMD_IOACCEL1) { struct io_accel1_cmd *cm1 = (struct io_accel1_cmd *) &h->ioaccel_cmd_pool[c->cmdindex]; - *tagupper = cm1->Tag.upper; - *taglower = cm1->Tag.lower; + *tagupper = (u32) (cm1->tag >> 32); + *taglower = (u32) (cm1->tag & 0x0ffffffffULL); return; } if (c->cmd_type == CMD_IOACCEL2) { @@ -4341,33 +5379,28 @@ static void hpsa_get_tag(struct ctlr_info *h, *taglower = cm2->Tag; return; } - *tagupper = c->Header.Tag.upper; - *taglower = c->Header.Tag.lower; + *tagupper = (u32) (c->Header.tag >> 32); + *taglower = (u32) (c->Header.tag & 0x0ffffffffULL); } - 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; struct ErrorInfo *ei; u32 tagupper, taglower; - c = cmd_special_alloc(h); - if (c == NULL) { /* trouble... */ - dev_warn(&h->pdev->dev, "cmd_special_alloc returned NULL!\n"); - return -ENOMEM; - } + c = cmd_alloc(h); /* 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. */ @@ -4375,6 +5408,9 @@ static int hpsa_send_abort(struct ctlr_info *h, unsigned char *scsi3addr, switch (ei->CommandStatus) { case CMD_SUCCESS: break; + case CMD_TMF_STATUS: + rc = hpsa_evaluate_tmf_status(h, c); + break; case CMD_UNABORTABLE: /* Very common, don't make noise. */ rc = -1; break; @@ -4385,60 +5421,51 @@ static int hpsa_send_abort(struct ctlr_info *h, unsigned char *scsi3addr, rc = -1; break; } - cmd_special_free(h, c); + cmd_free(h, c); dev_dbg(&h->pdev->dev, "%s: Tag:0x%08x:%08x: Finished.\n", __func__, tagupper, taglower); return rc; } -/* - * hpsa_find_cmd_in_queue - * - * Used to determine whether a command (find) is still present - * in queue_head. Optionally excludes the last element of queue_head. - * - * This is used to avoid unnecessary aborts. Commands in h->reqQ have - * not yet been submitted, and so can be aborted by the driver without - * sending an abort to the hardware. - * - * Returns pointer to command if found in queue, NULL otherwise. - */ -static struct CommandList *hpsa_find_cmd_in_queue(struct ctlr_info *h, - struct scsi_cmnd *find, struct list_head *queue_head) +static void setup_ioaccel2_abort_cmd(struct CommandList *c, struct ctlr_info *h, + struct CommandList *command_to_abort, int reply_queue) { - unsigned long flags; - struct CommandList *c = NULL; /* ptr into cmpQ */ + struct io_accel2_cmd *c2 = &h->ioaccel2_cmd_pool[c->cmdindex]; + struct hpsa_tmf_struct *ac = (struct hpsa_tmf_struct *) c2; + struct io_accel2_cmd *c2a = + &h->ioaccel2_cmd_pool[command_to_abort->cmdindex]; + struct scsi_cmnd *scmd = + (struct scsi_cmnd *) command_to_abort->scsi_cmd; + struct hpsa_scsi_dev_t *dev = scmd->device->hostdata; - if (!find) - return 0; - spin_lock_irqsave(&h->lock, flags); - list_for_each_entry(c, queue_head, list) { - if (c->scsi_cmd == NULL) /* e.g.: passthru ioctl */ - continue; - if (c->scsi_cmd == find) { - spin_unlock_irqrestore(&h->lock, flags); - return c; - } - } - spin_unlock_irqrestore(&h->lock, flags); - return NULL; -} + /* + * We're overlaying struct hpsa_tmf_struct on top of something which + * was allocated as a struct io_accel2_cmd, so we better be sure it + * actually fits, and doesn't overrun the error info space. + */ + BUILD_BUG_ON(sizeof(struct hpsa_tmf_struct) > + sizeof(struct io_accel2_cmd)); + BUG_ON(offsetof(struct io_accel2_cmd, error_data) < + offsetof(struct hpsa_tmf_struct, error_len) + + sizeof(ac->error_len)); -static struct CommandList *hpsa_find_cmd_in_queue_by_tag(struct ctlr_info *h, - u8 *tag, struct list_head *queue_head) -{ - unsigned long flags; - struct CommandList *c; + c->cmd_type = IOACCEL2_TMF; + /* Adjust the DMA address to point to the accelerated command buffer */ + c->busaddr = (u32) h->ioaccel2_cmd_pool_dhandle + + (c->cmdindex * sizeof(struct io_accel2_cmd)); + BUG_ON(c->busaddr & 0x0000007F); - spin_lock_irqsave(&h->lock, flags); - list_for_each_entry(c, queue_head, list) { - if (memcmp(&c->Header.Tag, tag, 8) != 0) - continue; - spin_unlock_irqrestore(&h->lock, flags); - return c; - } - spin_unlock_irqrestore(&h->lock, flags); - return NULL; + memset(ac, 0, sizeof(*c2)); /* yes this is correct */ + ac->iu_type = IOACCEL2_IU_TMF_TYPE; + ac->reply_queue = reply_queue; + ac->tmf = IOACCEL2_TMF_ABORT; + ac->it_nexus = cpu_to_le32((u32) dev->ioaccel_handle); + memset(ac->lun_id, 0, sizeof(ac->lun_id)); + ac->tag = c->cmdindex << DIRECT_LOOKUP_SHIFT; + ac->abort_tag = c2a->Tag; + ac->error_ptr = cpu_to_le64((u64) c->busaddr + + offsetof(struct io_accel2_cmd, error_data)); + ac->error_len = cpu_to_le32((u32) sizeof(c2->error_data)); } /* ioaccel2 path firmware cannot handle abort task requests. @@ -4449,7 +5476,7 @@ static struct CommandList *hpsa_find_cmd_in_queue_by_tag(struct ctlr_info *h, */ 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 */ @@ -4458,7 +5485,7 @@ static int hpsa_send_reset_as_abort_ioaccel2(struct ctlr_info *h, unsigned char *psa = &phys_scsi3addr[0]; /* Get a pointer to the hpsa logical device. */ - scmd = (struct scsi_cmnd *) abort->scsi_cmd; + scmd = abort->scsi_cmd; dev = (struct hpsa_scsi_dev_t *)(scmd->device->hostdata); if (dev == NULL) { dev_warn(&h->pdev->dev, @@ -4468,7 +5495,7 @@ static int hpsa_send_reset_as_abort_ioaccel2(struct ctlr_info *h, 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", + "Reset as abort: scsi %d:%d:%d:%d scsi3addr 0x%02x%02x%02x%02x%02x%02x%02x%02x\n", h->scsi_host->host_no, dev->bus, dev->target, dev->lun, scsi3addr[0], scsi3addr[1], scsi3addr[2], scsi3addr[3], scsi3addr[4], scsi3addr[5], scsi3addr[6], scsi3addr[7]); @@ -4491,7 +5518,7 @@ static int hpsa_send_reset_as_abort_ioaccel2(struct ctlr_info *h, "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", @@ -4501,7 +5528,7 @@ static int hpsa_send_reset_as_abort_ioaccel2(struct ctlr_info *h, } /* wait for device to recover */ - if (wait_for_device_to_become_ready(h, psa) != 0) { + if (wait_for_device_to_become_ready(h, psa, reply_queue) != 0) { dev_warn(&h->pdev->dev, "Reset as abort: Failed: Device never recovered from reset: 0x%02x%02x%02x%02x%02x%02x%02x%02x\n", psa[0], psa[1], psa[2], psa[3], @@ -4518,48 +5545,93 @@ static int hpsa_send_reset_as_abort_ioaccel2(struct ctlr_info *h, 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) +static int hpsa_send_abort_ioaccel2(struct ctlr_info *h, + struct CommandList *abort, int reply_queue) { - u8 swizzled_tag[8]; + int rc = IO_OK; struct CommandList *c; - int rc = 0, rc2 = 0; + u32 taglower, tagupper; + struct hpsa_scsi_dev_t *dev; + struct io_accel2_cmd *c2; + dev = abort->scsi_cmd->device->hostdata; + if (!dev->offload_enabled && !dev->hba_ioaccel_enabled) + return -1; + + c = cmd_alloc(h); + setup_ioaccel2_abort_cmd(c, h, abort, reply_queue); + c2 = &h->ioaccel2_cmd_pool[c->cmdindex]; + (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(ioaccel2 abort) completed.\n", + __func__, tagupper, taglower); + /* no unmap needed here because no data xfer. */ + + dev_dbg(&h->pdev->dev, + "%s: Tag:0x%08x:%08x: abort service response = 0x%02x.\n", + __func__, tagupper, taglower, c2->error_data.serv_response); + switch (c2->error_data.serv_response) { + case IOACCEL2_SERV_RESPONSE_TMF_COMPLETE: + case IOACCEL2_SERV_RESPONSE_TMF_SUCCESS: + rc = 0; + break; + case IOACCEL2_SERV_RESPONSE_TMF_REJECTED: + case IOACCEL2_SERV_RESPONSE_FAILURE: + case IOACCEL2_SERV_RESPONSE_TMF_WRONG_LUN: + rc = -1; + break; + default: + dev_warn(&h->pdev->dev, + "%s: Tag:0x%08x:%08x: unknown abort service response x0%02x\n", + __func__, tagupper, taglower, + c2->error_data.serv_response); + rc = -1; + } + cmd_free(h, c); + dev_dbg(&h->pdev->dev, "%s: Tag:0x%08x:%08x: Finished.\n", __func__, + tagupper, taglower); + return rc; +} + +static int hpsa_send_abort_both_ways(struct ctlr_info *h, + 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, - * but underlying firmware can't handle abort TMF. - * Change abort to physical device reset. + * but not all underlying firmware can handle abort TMF. + * Change abort to physical device reset when abort TMF is unsupported. */ - if (abort->cmd_type == CMD_IOACCEL2) - return hpsa_send_reset_as_abort_ioaccel2(h, scsi3addr, abort); + if (abort->cmd_type == CMD_IOACCEL2) { + if (HPSATMF_IOACCEL_ENABLED & h->TMFSupportFlags) + return hpsa_send_abort_ioaccel2(h, abort, + reply_queue); + else + return hpsa_send_reset_as_abort_ioaccel2(h, scsi3addr, + abort, reply_queue); + } + return hpsa_send_abort(h, scsi3addr, abort, reply_queue); +} - /* we do not expect to find the swizzled tag in our queue, but - * check anyway just to be sure the assumptions which make this - * the case haven't become wrong. - */ - memcpy(swizzled_tag, &abort->Request.CDB[4], 8); - swizzle_abort_tag(swizzled_tag); - c = hpsa_find_cmd_in_queue_by_tag(h, swizzled_tag, &h->cmpQ); - if (c != NULL) { - dev_warn(&h->pdev->dev, "Unexpectedly found byte-swapped tag in completion queue.\n"); - return hpsa_send_abort(h, scsi3addr, abort, 0); - } - rc = hpsa_send_abort(h, scsi3addr, abort, 0); - - /* if the command is still in our queue, we can't conclude that it was - * aborted (it might have just completed normally) but in any case - * we don't need to try to abort it another way. - */ - c = hpsa_find_cmd_in_queue(h, abort->scsi_cmd, &h->cmpQ); - if (c) - rc2 = hpsa_send_abort(h, scsi3addr, abort, 1); - return rc && rc2; +/* 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. @@ -4569,31 +5641,64 @@ static int hpsa_send_abort_both_ways(struct ctlr_info *h, static int hpsa_eh_abort_handler(struct scsi_cmnd *sc) { - int i, rc; + int rc; struct ctlr_info *h; struct hpsa_scsi_dev_t *dev; struct CommandList *abort; /* pointer to command to be aborted */ - struct CommandList *found; struct scsi_cmnd *as; /* ptr to scsi cmd inside aborted command. */ char msg[256]; /* For debug messaging. */ int ml = 0; u32 tagupper, taglower; + 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")) + "scsi ?:?:?:? scmd %p ABORT FAILED, Controller lookup failed.\n", + sc)) return FAILED; + dev_warn(&h->pdev->dev, "abort 0\n"); + /* if controller locked up, we can guarantee command won't complete */ + if (lockup_detected(h)) { + dev_warn(&h->pdev->dev, "abort 0.1\n"); + dev_warn(&h->pdev->dev, + "scsi %d:%d:%d:%llu scmd %p ABORT FAILED, lockup detected\n", + h->scsi_host->host_no, sc->device->channel, + sc->device->id, sc->device->lun, sc); + return FAILED; + } else { + /* good time to check if controller lockup has occurred */ + /* FIXME eh_timeout_handler would be even better. + * for testing, abort is just being used for timeouts, + * so is equivalent */ + dev_warn(&h->pdev->dev, "abort 0.2\n"); + detect_controller_lockup(h); + dev_warn(&h->pdev->dev, "abort 0.3\n"); + + /* check again in case one just occurred */ + if (lockup_detected(h)) { + dev_warn(&h->pdev->dev, "abort 0.4\n"); + dev_warn(&h->pdev->dev, + "scsi %d:%d:%d:%llu scmd %p ABORT FAILED, lockup detected\n", + h->scsi_host->host_no, sc->device->channel, + sc->device->id, sc->device->lun, sc); + return FAILED; + } + } + + dev_warn(&h->pdev->dev, "abort 1\n"); /* Check that controller supports some kind of task abort */ if (!(HPSATMF_PHYS_TASK_ABORT & h->TMFSupportFlags) && !(HPSATMF_LOG_TASK_ABORT & h->TMFSupportFlags)) return FAILED; + dev_warn(&h->pdev->dev, "abort 2\n"); memset(msg, 0, sizeof(msg)); - ml += sprintf(msg+ml, "ABORT REQUEST on C%d:B%d:T%d:L%llu ", + ml += sprintf(msg+ml, "scsi %d:%d:%d:%llu scmd %p ABORT ", h->scsi_host->host_no, sc->device->channel, - sc->device->id, sc->device->lun); + sc->device->id, sc->device->lun, sc); /* Find the device of the command to be aborted */ dev = sc->device->hostdata; @@ -4603,184 +5708,229 @@ static int hpsa_eh_abort_handler(struct scsi_cmnd *sc) return FAILED; } + dev_warn(&h->pdev->dev, "abort 3\n"); /* Get SCSI command to be aborted */ abort = (struct CommandList *) sc->host_scribble; if (abort == NULL) { - dev_err(&h->pdev->dev, "%s FAILED, Command to abort is NULL.\n", - msg); + /* This can happen if the command already completed. */ + return SUCCESS; + } + dev_warn(&h->pdev->dev, "abort 4\n"); + refcount = atomic_inc_return(&abort->refcount); + if (refcount == 1) { /* Command is done already. */ + cmd_free(h, abort); + return SUCCESS; + } + + dev_warn(&h->pdev->dev, "abort 5\n"); + /* 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); - ml += sprintf(msg+ml, "Tag:0x%08x:%08x ", tagupper, taglower); - as = (struct scsi_cmnd *) 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); - /* Search reqQ to See if command is queued but not submitted, - * if so, complete the command with aborted status and remove - * it from the reqQ. + dev_warn(&h->pdev->dev, "abort 6\n"); + /* Check that we're aborting the right command. + * It's possible the CommandList already completed and got re-used. */ - found = hpsa_find_cmd_in_queue(h, sc, &h->reqQ); - if (found) { - found->err_info->CommandStatus = CMD_ABORTED; - finish_cmd(found); - dev_info(&h->pdev->dev, "%s Request SUCCEEDED (driver queue).\n", - msg); + if (abort->scsi_cmd != sc) { + cmd_free(h, abort); return SUCCESS; } + dev_warn(&h->pdev->dev, "abort 7\n"); - /* not in reqQ, if also not in cmpQ, must have already completed */ - found = hpsa_find_cmd_in_queue(h, sc, &h->cmpQ); - if (!found) { - dev_dbg(&h->pdev->dev, "%s Request SUCCEEDED (not known to driver).\n", - msg); - return SUCCESS; - } + abort->abort_pending = true; + 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, + "CDBLen: %d CDB: 0x%02x%02x... SN: 0x%lx ", + as->cmd_len, as->cmnd[0], as->cmnd[1], + as->serial_number); + dev_warn(&h->pdev->dev, "%s BEING SENT\n", msg); + dev_warn(&h->pdev->dev, "abort 8 (waiting for abort command to be free)\n"); /* * 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, + "%s FAILED, timeout waiting for an abort command to become available.\n", + msg); + cmd_free(h, abort); + return FAILED; + } + dev_warn(&h->pdev->dev, "abort 9 (got abort command)\n"); + 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); + dev_warn(&h->pdev->dev, "%s SENT, FAILED\n", msg); + cmd_free(h, abort); return FAILED; } - dev_info(&h->pdev->dev, "%s REQUEST SUCCEEDED.\n", msg); + dev_warn(&h->pdev->dev, "abort 10\n"); + dev_info(&h->pdev->dev, "%s SENT, SUCCESS\n", msg); + wait_event(h->abort_sync_wait_queue, atomic_read(&abort->refcount) == 1); + cmd_free(h, abort); + dev_warn(&h->pdev->dev, "abort 11\n"); + return SUCCESS; +} + +/* + * One of the upper layers has already gone to the trouble of picking out a + * unique, small-integer tag for this request. If the blk-mq support is not + * available, we use the SCSI tag in the SCSI command; otherwise, we use the + * block layer tag in the embedded request structure. We use an offset from + * that value as an index to select our command block. (The offset allows us to + * reserve the low-numbered entries for our own uses.) + */ +static int hpsa_get_cmd_index(struct scsi_cmnd *scmd) +{ + struct scsi_device *dev = scmd->device; + int idx = shost_use_blk_mq(dev->host) ? scmd->request->tag : scmd->tag; + + if (idx < 0) { + dev_err(&dev->sdev_dev , "Invalid block tag: %d\n", idx); + /* This value comes from an upper layer...it's not our bug. */ + BUG_ON(idx < 0); + } + + /* Offset to leave space for internal cmds. */ + idx += HPSA_NRESERVED_CMDS; + + return idx; +} + +static void print_scsi_cmd(struct ctlr_info *h, struct scsi_cmnd *scmd, int idx) +{ + char format[50]; + + sprintf(format, "Tag %d: CDB: %dph\n", idx, scmd->cmd_len); + dev_warn(&h->pdev->dev, format, scmd->cmnd); +} + +/* + * For operations with an associated SCSI command, a command block is allocated + * at init, and managed by cmd_tagged_alloc() and cmd_tagged_free() using the + * block request tag as an index into a table of entries. cmd_tagged_free() is + * the complement, although cmd_free() may be called instead. + */ +static struct CommandList *cmd_tagged_alloc(struct ctlr_info *h, + struct scsi_cmnd *scmd) +{ + int idx = hpsa_get_cmd_index(scmd); + struct CommandList *c = h->cmd_pool + idx; + int refcount = 0; + + if (idx < HPSA_NRESERVED_CMDS || idx >= h->nr_cmds) { + dev_err(&h->pdev->dev, "Bad block tag: %d\n", idx); + /* The index value comes from the block layer, so if it's out of + * bounds, it's probably not our bug. + */ + BUG_ON(idx < HPSA_NRESERVED_CMDS || idx >= h->nr_cmds); + } + + refcount = atomic_inc_return(&c->refcount); + if (unlikely(refcount > 1)) { + /* + * We expect that the SCSI layer will hand us a unique tag + * value. Thus, there should never be a collision here between + * two requests; however, it's possible that a lingering abort + * might have an outstanding reference when the next request + * comes in (the block layer tends to reuse the last-used tag + * first). Hopefully, that won't be a problem. + */ + dev_warn(&h->pdev->dev, + "tag collision (tag=%d) in cmd_tagged_alloc().\n", + idx); + print_scsi_cmd(h, scmd, idx); + } - /* If the abort(s) above completed and actually aborted the - * command, then the command to be aborted should already be - * completed. If not, wait around a bit more to see if they - * manage to complete normally. - */ -#define ABORT_COMPLETE_WAIT_SECS 30 - for (i = 0; i < ABORT_COMPLETE_WAIT_SECS * 10; i++) { - found = hpsa_find_cmd_in_queue(h, sc, &h->cmpQ); - if (!found) - return SUCCESS; - msleep(100); - } - dev_warn(&h->pdev->dev, "%s FAILED. Aborted command has not completed after %d seconds.\n", - msg, ABORT_COMPLETE_WAIT_SECS); - return FAILED; + hpsa_cmd_partial_init(h, idx, c); + return c; } +static void cmd_tagged_free(struct ctlr_info *h, struct CommandList *c) +{ + /* + * Release our reference to the block. We don't need to do anything + * else to free it, because it is accessed by index. (There's no point + * in checking the result of the decrement, since we cannot guarantee + * that there isn't a concurrent abort which is also accessing it.) + */ + (void)atomic_dec_and_test(&c->refcount); +} /* * For operations that cannot sleep, a command block is allocated at init, * and managed by cmd_alloc() and cmd_free() using a simple bitmap to track * which ones are free or in use. Lock must be held when calling this. * cmd_free() is the complement. + * This function never gives up and returns NULL. If it hangs, + * another thread must call cmd_free() to free some tags. */ static struct CommandList *cmd_alloc(struct ctlr_info *h) { struct CommandList *c; - int i; - union u64bit temp64; - dma_addr_t cmd_dma_handle, err_dma_handle; - unsigned long flags; + int refcount, i; + + /* There is some *extremely* small but non-zero chance that that + * multiple threads could get in here, and one thread could + * be scanning through the list of bits looking for a free + * one, but the free ones are always behind him, and other + * threads sneak in behind him and eat them before he can + * get to them, so that while there is always a free one, a + * very unlucky thread might be starved anyway, never able to + * beat the other threads. In reality, this happens so + * infrequently as to be indistinguishable from never. + * + * Note that we start allocating commands before the SCSI host structure + * is initialized. Since the search starts at bit zero, this + * all works, since we have at least one command structure available; + * however, it means that the structures with the low indexes have to be + * reserved for driver-initiated requests, while requests from the block + * layer will use the higher indexes. + */ - spin_lock_irqsave(&h->lock, flags); - do { - i = find_first_zero_bit(h->cmd_pool_bits, h->nr_cmds); - if (i == h->nr_cmds) { - spin_unlock_irqrestore(&h->lock, flags); - return NULL; + for (;;) { + i = find_first_zero_bit(h->cmd_pool_bits, HPSA_NRESERVED_CMDS); + if (unlikely(i >= HPSA_NRESERVED_CMDS)) + continue; + c = h->cmd_pool + i; + refcount = atomic_inc_return(&c->refcount); + if (unlikely(refcount > 1)) { + cmd_free(h, c); /* already in use */ + continue; } - } while (test_and_set_bit - (i & (BITS_PER_LONG - 1), - h->cmd_pool_bits + (i / BITS_PER_LONG)) != 0); - spin_unlock_irqrestore(&h->lock, flags); - - c = h->cmd_pool + i; - memset(c, 0, sizeof(*c)); - cmd_dma_handle = h->cmd_pool_dhandle - + i * sizeof(*c); - c->err_info = h->errinfo_pool + i; - memset(c->err_info, 0, sizeof(*c->err_info)); - err_dma_handle = h->errinfo_pool_dhandle - + i * sizeof(*c->err_info); - - c->cmdindex = i; - - INIT_LIST_HEAD(&c->list); - c->busaddr = (u32) cmd_dma_handle; - temp64.val = (u64) err_dma_handle; - c->ErrDesc.Addr.lower = temp64.val32.lower; - c->ErrDesc.Addr.upper = temp64.val32.upper; - c->ErrDesc.Len = sizeof(*c->err_info); - - c->h = h; - return c; -} - -/* For operations that can wait for kmalloc to possibly sleep, - * this routine can be called. Lock need not be held to call - * cmd_special_alloc. cmd_special_free() is the complement. - */ -static struct CommandList *cmd_special_alloc(struct ctlr_info *h) -{ - struct CommandList *c; - union u64bit temp64; - dma_addr_t cmd_dma_handle, err_dma_handle; - - c = pci_zalloc_consistent(h->pdev, sizeof(*c), &cmd_dma_handle); - if (c == NULL) - return NULL; - - c->cmd_type = CMD_SCSI; - c->cmdindex = -1; - - c->err_info = pci_zalloc_consistent(h->pdev, sizeof(*c->err_info), - &err_dma_handle); - - if (c->err_info == NULL) { - pci_free_consistent(h->pdev, - sizeof(*c), c, cmd_dma_handle); - return NULL; + set_bit(i & (BITS_PER_LONG - 1), + h->cmd_pool_bits + (i / BITS_PER_LONG)); + break; /* it's ours now. */ } - - INIT_LIST_HEAD(&c->list); - c->busaddr = (u32) cmd_dma_handle; - temp64.val = (u64) err_dma_handle; - c->ErrDesc.Addr.lower = temp64.val32.lower; - c->ErrDesc.Addr.upper = temp64.val32.upper; - c->ErrDesc.Len = sizeof(*c->err_info); - - c->h = h; + hpsa_cmd_partial_init(h, i, c); return c; } +/* + * This is the complementary operation to cmd_alloc(). Note, however, in some + * corner cases it may also be used to free blocks allocated by + * cmd_tagged_alloc() in which case the ref-count decrement does the trick and + * the clear-bit is harmless. + */ static void cmd_free(struct ctlr_info *h, struct CommandList *c) { - int i; - unsigned long flags; - - i = c - h->cmd_pool; - spin_lock_irqsave(&h->lock, flags); - clear_bit(i & (BITS_PER_LONG - 1), - h->cmd_pool_bits + (i / BITS_PER_LONG)); - spin_unlock_irqrestore(&h->lock, flags); -} - -static void cmd_special_free(struct ctlr_info *h, struct CommandList *c) -{ - union u64bit temp64; + if (atomic_dec_and_test(&c->refcount)) { + int i; - temp64.val32.lower = c->ErrDesc.Addr.lower; - temp64.val32.upper = c->ErrDesc.Addr.upper; - pci_free_consistent(h->pdev, sizeof(*c->err_info), - c->err_info, (dma_addr_t) temp64.val); - pci_free_consistent(h->pdev, sizeof(*c), - c, (dma_addr_t) (c->busaddr & DIRECT_LOOKUP_MASK)); + i = c - h->cmd_pool; + clear_bit(i & (BITS_PER_LONG - 1), + h->cmd_pool_bits + (i / BITS_PER_LONG)); + } } #ifdef CONFIG_COMPAT @@ -4932,7 +6082,7 @@ static int hpsa_passthru_ioctl(struct ctlr_info *h, void __user *argp) IOCTL_Command_struct iocommand; struct CommandList *c; char *buff = NULL; - union u64bit temp64; + u64 temp64; int rc = 0; if (!argp) @@ -4948,7 +6098,7 @@ static int hpsa_passthru_ioctl(struct ctlr_info *h, void __user *argp) if (iocommand.buf_size > 0) { buff = kmalloc(iocommand.buf_size, GFP_KERNEL); if (buff == NULL) - return -EFAULT; + return -ENOMEM; if (iocommand.Request.Type.Direction & XFER_WRITE) { /* Copy the data into the buffer we created */ if (copy_from_user(buff, iocommand.buf, @@ -4960,25 +6110,20 @@ static int hpsa_passthru_ioctl(struct ctlr_info *h, void __user *argp) memset(buff, 0, iocommand.buf_size); } } - c = cmd_special_alloc(h); - if (c == NULL) { - rc = -ENOMEM; - goto out_kfree; - } + c = cmd_alloc(h); + /* Fill in the command type */ c->cmd_type = CMD_IOCTL_PEND; /* Fill in Command Header */ c->Header.ReplyQueue = 0; /* unused in simple mode */ if (iocommand.buf_size > 0) { /* buffer to fill */ - c->Header.SGList = 1; - c->Header.SGTotal = 1; + c->Header.SGList = (u8) 1; + c->Header.SGTotal = cpu_to_le16(1); } else { /* no buffers to fill */ c->Header.SGList = 0; - c->Header.SGTotal = 0; + c->Header.SGTotal = cpu_to_le16(0); } memcpy(&c->Header.LUN, &iocommand.LUN_info, sizeof(c->Header.LUN)); - /* use the kernel address the cmd block for tag */ - c->Header.Tag.lower = c->busaddr; /* Fill in Request block */ memcpy(&c->Request, &iocommand.Request, @@ -4986,21 +6131,21 @@ static int hpsa_passthru_ioctl(struct ctlr_info *h, void __user *argp) /* Fill in the scatter gather information */ if (iocommand.buf_size > 0) { - temp64.val = pci_map_single(h->pdev, buff, + temp64 = (u64) pci_map_single(h->pdev, buff, iocommand.buf_size, PCI_DMA_BIDIRECTIONAL); - if (dma_mapping_error(&h->pdev->dev, temp64.val)) { - c->SG[0].Addr.lower = 0; - c->SG[0].Addr.upper = 0; - c->SG[0].Len = 0; + if (dma_mapping_error(&h->pdev->dev, (dma_addr_t) temp64)) { + c->SG[0].Addr = cpu_to_le64(0); + c->SG[0].Len = cpu_to_le32(0); rc = -ENOMEM; goto out; } - c->SG[0].Addr.lower = temp64.val32.lower; - c->SG[0].Addr.upper = temp64.val32.upper; - c->SG[0].Len = iocommand.buf_size; - c->SG[0].Ext = HPSA_SG_LAST; /* we are not chaining*/ + c->SG[0].Addr = cpu_to_le64(temp64); + 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 (rc) + rc = -EIO; if (iocommand.buf_size > 0) hpsa_pci_unmap(h->pdev, c, 1, PCI_DMA_BIDIRECTIONAL); check_ioctl_unit_attention(h, c); @@ -5021,7 +6166,7 @@ static int hpsa_passthru_ioctl(struct ctlr_info *h, void __user *argp) } } out: - cmd_special_free(h, c); + cmd_free(h, c); out_kfree: kfree(buff); return rc; @@ -5033,7 +6178,7 @@ static int hpsa_big_passthru_ioctl(struct ctlr_info *h, void __user *argp) struct CommandList *c; unsigned char **buff = NULL; int *buff_size = NULL; - union u64bit temp64; + u64 temp64; BYTE sg_used = 0; int status = 0; int i; @@ -5100,38 +6245,39 @@ static int hpsa_big_passthru_ioctl(struct ctlr_info *h, void __user *argp) data_ptr += sz; sg_used++; } - c = cmd_special_alloc(h); - if (c == NULL) { - status = -ENOMEM; - goto cleanup1; - } + c = cmd_alloc(h); + c->cmd_type = CMD_IOCTL_PEND; c->Header.ReplyQueue = 0; - c->Header.SGList = c->Header.SGTotal = sg_used; + c->Header.SGList = (u8) sg_used; + c->Header.SGTotal = cpu_to_le16(sg_used); memcpy(&c->Header.LUN, &ioc->LUN_info, sizeof(c->Header.LUN)); - c->Header.Tag.lower = c->busaddr; memcpy(&c->Request, &ioc->Request, sizeof(c->Request)); if (ioc->buf_size > 0) { int i; for (i = 0; i < sg_used; i++) { - temp64.val = pci_map_single(h->pdev, buff[i], + temp64 = (u64) pci_map_single(h->pdev, buff[i], buff_size[i], PCI_DMA_BIDIRECTIONAL); - if (dma_mapping_error(&h->pdev->dev, temp64.val)) { - c->SG[i].Addr.lower = 0; - c->SG[i].Addr.upper = 0; + if (dma_mapping_error(&h->pdev->dev, + (dma_addr_t) temp64)) { + c->SG[i].Addr = 0; c->SG[i].Len = 0; hpsa_pci_unmap(h->pdev, c, i, PCI_DMA_BIDIRECTIONAL); status = -ENOMEM; goto cleanup0; } - c->SG[i].Addr.lower = temp64.val32.lower; - c->SG[i].Addr.upper = temp64.val32.upper; - c->SG[i].Len = buff_size[i]; - c->SG[i].Ext = i < sg_used - 1 ? 0 : HPSA_SG_LAST; + c->SG[i].Addr = cpu_to_le64(temp64); + c->SG[i].Len = cpu_to_le32(buff_size[i]); + c->SG[i].Ext = + cpu_to_le32((i == sg_used) * 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 (status) { + status = -EIO; + goto cleanup0; + } if (sg_used) hpsa_pci_unmap(h->pdev, c, sg_used, PCI_DMA_BIDIRECTIONAL); check_ioctl_unit_attention(h, c); @@ -5154,7 +6300,7 @@ static int hpsa_big_passthru_ioctl(struct ctlr_info *h, void __user *argp) } status = 0; cleanup0: - cmd_special_free(h, c); + cmd_free(h, c); cleanup1: if (buff) { for (i = 0; i < sg_used; i++) @@ -5174,35 +6320,6 @@ static void check_ioctl_unit_attention(struct ctlr_info *h, (void) check_for_unit_attention(h, c); } -static int increment_passthru_count(struct ctlr_info *h) -{ - unsigned long flags; - - spin_lock_irqsave(&h->passthru_count_lock, flags); - if (h->passthru_count >= HPSA_MAX_CONCURRENT_PASSTHRUS) { - spin_unlock_irqrestore(&h->passthru_count_lock, flags); - return -1; - } - h->passthru_count++; - spin_unlock_irqrestore(&h->passthru_count_lock, flags); - return 0; -} - -static void decrement_passthru_count(struct ctlr_info *h) -{ - unsigned long flags; - - spin_lock_irqsave(&h->passthru_count_lock, flags); - if (h->passthru_count <= 0) { - spin_unlock_irqrestore(&h->passthru_count_lock, flags); - /* not expecting to get here. */ - dev_warn(&h->pdev->dev, "Bug detected, passthru_count seems to be incorrect.\n"); - return; - } - h->passthru_count--; - spin_unlock_irqrestore(&h->passthru_count_lock, flags); -} - /* * ioctl */ @@ -5225,30 +6342,29 @@ static int hpsa_ioctl(struct scsi_device *dev, int cmd, void *arg) case CCISS_GETDRIVVER: return hpsa_getdrivver_ioctl(h, argp); case CCISS_PASSTHRU: - if (increment_passthru_count(h)) + if (atomic_dec_if_positive(&h->passthru_cmds_avail) < 0) return -EAGAIN; rc = hpsa_passthru_ioctl(h, argp); - decrement_passthru_count(h); + atomic_inc(&h->passthru_cmds_avail); return rc; case CCISS_BIG_PASSTHRU: - if (increment_passthru_count(h)) + if (atomic_dec_if_positive(&h->passthru_cmds_avail) < 0) return -EAGAIN; rc = hpsa_big_passthru_ioctl(h, argp); - decrement_passthru_count(h); + atomic_inc(&h->passthru_cmds_avail); return rc; default: return -ENOTTY; } } -static int hpsa_send_host_reset(struct ctlr_info *h, unsigned char *scsi3addr, +static void hpsa_send_host_reset(struct ctlr_info *h, unsigned char *scsi3addr, u8 reset_type) { struct CommandList *c; c = cmd_alloc(h); - if (!c) - return -ENOMEM; + /* fill_cmd can't fail here, no data buffer to map */ (void) fill_cmd(c, HPSA_DEVICE_RESET_MSG, h, NULL, 0, 0, RAID_CTLR_LUNID, TYPE_MSG); @@ -5259,7 +6375,7 @@ static int hpsa_send_host_reset(struct ctlr_info *h, unsigned char *scsi3addr, * the command either. This is the last command we will send before * re-initializing everything, so it doesn't matter and won't leak. */ - return 0; + return; } static int fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, @@ -5267,21 +6383,20 @@ static int fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, int cmd_type) { int pci_dir = XFER_NONE; - struct CommandList *a; /* for commands to be aborted */ + u64 tag; /* for commands to be aborted */ + u32 tupper, tlower; c->cmd_type = CMD_IOCTL_PEND; c->Header.ReplyQueue = 0; if (buff != NULL && size > 0) { - c->Header.SGList = 1; - c->Header.SGTotal = 1; + c->Header.SGList = (u8) 1; + c->Header.SGTotal = cpu_to_le32(1); } else { c->Header.SGList = 0; c->Header.SGTotal = 0; } - c->Header.Tag.lower = c->busaddr; memcpy(c->Header.LUN.LunAddrBytes, scsi3addr, 8); - c->Request.Type.Type = cmd_type; if (cmd_type == TYPE_CMD) { switch (cmd) { case HPSA_INQUIRY: @@ -5291,8 +6406,8 @@ static int fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, c->Request.CDB[2] = (page_code & 0xff); } c->Request.CDBLen = 6; - c->Request.Type.Attribute = ATTR_SIMPLE; - c->Request.Type.Direction = XFER_READ; + c->Request.type_attr_dir = + TYPE_ATTR_DIR(cmd_type, ATTR_SIMPLE, XFER_READ); c->Request.Timeout = 0; c->Request.CDB[0] = HPSA_INQUIRY; c->Request.CDB[4] = size & 0xFF; @@ -5303,8 +6418,8 @@ static int fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, mode = 00 target = 0. Nothing to write. */ c->Request.CDBLen = 12; - c->Request.Type.Attribute = ATTR_SIMPLE; - c->Request.Type.Direction = XFER_READ; + c->Request.type_attr_dir = + TYPE_ATTR_DIR(cmd_type, ATTR_SIMPLE, XFER_READ); c->Request.Timeout = 0; c->Request.CDB[0] = cmd; c->Request.CDB[6] = (size >> 24) & 0xFF; /* MSB */ @@ -5314,8 +6429,9 @@ static int fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, break; case HPSA_CACHE_FLUSH: c->Request.CDBLen = 12; - c->Request.Type.Attribute = ATTR_SIMPLE; - c->Request.Type.Direction = XFER_WRITE; + c->Request.type_attr_dir = + TYPE_ATTR_DIR(cmd_type, + ATTR_SIMPLE, XFER_WRITE); c->Request.Timeout = 0; c->Request.CDB[0] = BMIC_WRITE; c->Request.CDB[6] = BMIC_CACHE_FLUSH; @@ -5324,14 +6440,14 @@ static int fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, break; case TEST_UNIT_READY: c->Request.CDBLen = 6; - c->Request.Type.Attribute = ATTR_SIMPLE; - c->Request.Type.Direction = XFER_NONE; + c->Request.type_attr_dir = + TYPE_ATTR_DIR(cmd_type, ATTR_SIMPLE, XFER_NONE); c->Request.Timeout = 0; break; case HPSA_GET_RAID_MAP: c->Request.CDBLen = 12; - c->Request.Type.Attribute = ATTR_SIMPLE; - c->Request.Type.Direction = XFER_READ; + c->Request.type_attr_dir = + TYPE_ATTR_DIR(cmd_type, ATTR_SIMPLE, XFER_READ); c->Request.Timeout = 0; c->Request.CDB[0] = HPSA_CISS_READ; c->Request.CDB[1] = cmd; @@ -5342,14 +6458,24 @@ static int fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, break; case BMIC_SENSE_CONTROLLER_PARAMETERS: c->Request.CDBLen = 10; - c->Request.Type.Attribute = ATTR_SIMPLE; - c->Request.Type.Direction = XFER_READ; + c->Request.type_attr_dir = + TYPE_ATTR_DIR(cmd_type, ATTR_SIMPLE, XFER_READ); c->Request.Timeout = 0; c->Request.CDB[0] = BMIC_READ; c->Request.CDB[6] = BMIC_SENSE_CONTROLLER_PARAMETERS; c->Request.CDB[7] = (size >> 16) & 0xFF; c->Request.CDB[8] = (size >> 8) & 0xFF; break; + case BMIC_IDENTIFY_PHYSICAL_DEVICE: + c->Request.CDBLen = 10; + c->Request.type_attr_dir = + TYPE_ATTR_DIR(cmd_type, ATTR_SIMPLE, XFER_READ); + c->Request.Timeout = 0; + c->Request.CDB[0] = BMIC_READ; + c->Request.CDB[6] = BMIC_IDENTIFY_PHYSICAL_DEVICE; + c->Request.CDB[7] = (size >> 16) & 0xFF; + c->Request.CDB[8] = (size >> 8) & 0XFF; + break; default: dev_warn(&h->pdev->dev, "unknown command 0x%c\n", cmd); BUG(); @@ -5360,9 +6486,8 @@ static int fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, case HPSA_DEVICE_RESET_MSG: c->Request.CDBLen = 16; - c->Request.Type.Type = 1; /* It is a MSG not a CMD */ - c->Request.Type.Attribute = ATTR_SIMPLE; - c->Request.Type.Direction = XFER_NONE; + c->Request.type_attr_dir = + TYPE_ATTR_DIR(cmd_type, ATTR_SIMPLE, XFER_NONE); c->Request.Timeout = 0; /* Don't time out */ memset(&c->Request.CDB[0], 0, sizeof(c->Request.CDB)); c->Request.CDB[0] = cmd; @@ -5375,28 +6500,29 @@ static int fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, c->Request.CDB[7] = 0x00; break; case HPSA_ABORT_MSG: - a = buff; /* point to command to be aborted */ - dev_dbg(&h->pdev->dev, "Abort Tag:0x%08x:%08x using request Tag:0x%08x:%08x\n", - a->Header.Tag.upper, a->Header.Tag.lower, - c->Header.Tag.upper, c->Header.Tag.lower); + memcpy(&tag, buff, sizeof(tag)); + dev_dbg(&h->pdev->dev, "Abort Tag:0x%016llx using request Tag:0x%016llx", + tag, c->Header.tag); + tlower = (u32) (tag >> 32); + tupper = (u32) (tag & 0x0ffffffffULL); c->Request.CDBLen = 16; - c->Request.Type.Type = TYPE_MSG; - c->Request.Type.Attribute = ATTR_SIMPLE; - c->Request.Type.Direction = XFER_WRITE; + c->Request.type_attr_dir = + TYPE_ATTR_DIR(cmd_type, + ATTR_SIMPLE, XFER_WRITE); c->Request.Timeout = 0; /* Don't time out */ c->Request.CDB[0] = HPSA_TASK_MANAGEMENT; c->Request.CDB[1] = HPSA_TMF_ABORT_TASK; c->Request.CDB[2] = 0x00; /* reserved */ c->Request.CDB[3] = 0x00; /* reserved */ /* Tag to abort goes in CDB[4]-CDB[11] */ - c->Request.CDB[4] = a->Header.Tag.lower & 0xFF; - c->Request.CDB[5] = (a->Header.Tag.lower >> 8) & 0xFF; - c->Request.CDB[6] = (a->Header.Tag.lower >> 16) & 0xFF; - c->Request.CDB[7] = (a->Header.Tag.lower >> 24) & 0xFF; - c->Request.CDB[8] = a->Header.Tag.upper & 0xFF; - c->Request.CDB[9] = (a->Header.Tag.upper >> 8) & 0xFF; - c->Request.CDB[10] = (a->Header.Tag.upper >> 16) & 0xFF; - c->Request.CDB[11] = (a->Header.Tag.upper >> 24) & 0xFF; + c->Request.CDB[4] = tlower & 0xFF; + c->Request.CDB[5] = (tlower >> 8) & 0xFF; + c->Request.CDB[6] = (tlower >> 16) & 0xFF; + c->Request.CDB[7] = (tlower >> 24) & 0xFF; + c->Request.CDB[8] = tupper & 0xFF; + c->Request.CDB[9] = (tupper >> 8) & 0xFF; + c->Request.CDB[10] = (tupper >> 16) & 0xFF; + c->Request.CDB[11] = (tupper >> 24) & 0xFF; c->Request.CDB[12] = 0x00; /* reserved */ c->Request.CDB[13] = 0x00; /* reserved */ c->Request.CDB[14] = 0x00; /* reserved */ @@ -5412,7 +6538,7 @@ static int fill_cmd(struct CommandList *c, u8 cmd, struct ctlr_info *h, BUG(); } - switch (c->Request.Type.Direction) { + switch (GET_DIR(c->Request.type_attr_dir)) { case XFER_READ: pci_dir = PCI_DMA_FROMDEVICE; break; @@ -5443,53 +6569,6 @@ static void __iomem *remap_pci_mem(ulong base, ulong size) return page_remapped ? (page_remapped + page_offs) : NULL; } -/* Takes cmds off the submission queue and sends them to the hardware, - * then puts them on the queue of cmds waiting for completion. - * Assumes h->lock is held - */ -static void start_io(struct ctlr_info *h, unsigned long *flags) -{ - struct CommandList *c; - - while (!list_empty(&h->reqQ)) { - c = list_entry(h->reqQ.next, struct CommandList, list); - /* can't do anything if fifo is full */ - if ((h->access.fifo_full(h))) { - h->fifo_recently_full = 1; - dev_warn(&h->pdev->dev, "fifo full\n"); - break; - } - h->fifo_recently_full = 0; - - /* Get the first entry from the Request Q */ - removeQ(c); - h->Qdepth--; - - /* Put job onto the completed Q */ - addQ(&h->cmpQ, c); - - /* Must increment commands_outstanding before unlocking - * and submitting to avoid race checking for fifo full - * condition. - */ - h->commands_outstanding++; - - /* Tell the controller execute command */ - spin_unlock_irqrestore(&h->lock, *flags); - h->access.submit_command(h, c); - spin_lock_irqsave(&h->lock, *flags); - } -} - -static void lock_and_start_io(struct ctlr_info *h) -{ - unsigned long flags; - - spin_lock_irqsave(&h->lock, flags); - start_io(h, &flags); - spin_unlock_irqrestore(&h->lock, flags); -} - static inline unsigned long get_next_completion(struct ctlr_info *h, u8 q) { return h->access.command_completed(h, q); @@ -5518,53 +6597,12 @@ static inline int bad_tag(struct ctlr_info *h, u32 tag_index, static inline void finish_cmd(struct CommandList *c) { - unsigned long flags; - int io_may_be_stalled = 0; - struct ctlr_info *h = c->h; - - spin_lock_irqsave(&h->lock, flags); - removeQ(c); - - /* - * Check for possibly stalled i/o. - * - * If a fifo_full condition is encountered, requests will back up - * in h->reqQ. This queue is only emptied out by start_io which is - * only called when a new i/o request comes in. If no i/o's are - * forthcoming, the i/o's in h->reqQ can get stuck. So we call - * start_io from here if we detect such a danger. - * - * Normally, we shouldn't hit this case, but pounding on the - * CCISS_PASSTHRU ioctl can provoke it. Only call start_io if - * commands_outstanding is low. We want to avoid calling - * start_io from in here as much as possible, and esp. don't - * want to get in a cycle where we call start_io every time - * through here. - */ - if (unlikely(h->fifo_recently_full) && - h->commands_outstanding < 5) - io_may_be_stalled = 1; - - spin_unlock_irqrestore(&h->lock, flags); - dial_up_lockup_detection_on_fw_flash_complete(c->h, c); if (likely(c->cmd_type == CMD_IOACCEL1 || c->cmd_type == CMD_SCSI || c->cmd_type == CMD_IOACCEL2)) complete_scsi_command(c); - else if (c->cmd_type == CMD_IOCTL_PEND) + else if (c->cmd_type == CMD_IOCTL_PEND || c->cmd_type == IOACCEL2_TMF) complete(c->waiting); - if (unlikely(io_may_be_stalled)) - lock_and_start_io(h); -} - -static inline u32 hpsa_tag_contains_index(u32 tag) -{ - return tag & DIRECT_LOOKUP_BIT; -} - -static inline u32 hpsa_tag_to_index(u32 tag) -{ - return tag >> DIRECT_LOOKUP_SHIFT; } @@ -5584,34 +6622,13 @@ static inline void process_indexed_cmd(struct ctlr_info *h, u32 tag_index; struct CommandList *c; - tag_index = hpsa_tag_to_index(raw_tag); + tag_index = raw_tag >> DIRECT_LOOKUP_SHIFT; if (!bad_tag(h, tag_index, raw_tag)) { c = h->cmd_pool + tag_index; finish_cmd(c); } } -/* process completion of a non-indexed command */ -static inline void process_nonindexed_cmd(struct ctlr_info *h, - u32 raw_tag) -{ - u32 tag; - struct CommandList *c = NULL; - unsigned long flags; - - tag = hpsa_tag_discard_error_bits(h, raw_tag); - spin_lock_irqsave(&h->lock, flags); - list_for_each_entry(c, &h->cmpQ, list) { - if ((c->busaddr & 0xFFFFFFE0) == (tag & 0xFFFFFFE0)) { - spin_unlock_irqrestore(&h->lock, flags); - finish_cmd(c); - return; - } - } - spin_unlock_irqrestore(&h->lock, flags); - bad_tag(h, h->nr_cmds + 1, raw_tag); -} - /* Some controllers, like p400, will give us one interrupt * after a soft reset, even if we turned interrupts off. * Only need to check for this in the hpsa_xxx_discard_completions @@ -5689,10 +6706,7 @@ static irqreturn_t do_hpsa_intr_intx(int irq, void *queue) while (interrupt_pending(h)) { raw_tag = get_next_completion(h, q); while (raw_tag != FIFO_EMPTY) { - if (likely(hpsa_tag_contains_index(raw_tag))) - process_indexed_cmd(h, raw_tag); - else - process_nonindexed_cmd(h, raw_tag); + process_indexed_cmd(h, raw_tag); raw_tag = next_command(h, q); } } @@ -5708,10 +6722,7 @@ static irqreturn_t do_hpsa_intr_msi(int irq, void *queue) h->last_intr_timestamp = get_jiffies_64(); raw_tag = get_next_completion(h, q); while (raw_tag != FIFO_EMPTY) { - if (likely(hpsa_tag_contains_index(raw_tag))) - process_indexed_cmd(h, raw_tag); - else - process_nonindexed_cmd(h, raw_tag); + process_indexed_cmd(h, raw_tag); raw_tag = next_command(h, q); } return IRQ_HANDLED; @@ -5748,7 +6759,7 @@ static int hpsa_message(struct pci_dev *pdev, unsigned char opcode, err = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32)); if (err) { iounmap(vaddr); - return -ENOMEM; + return err; } cmd = pci_alloc_consistent(pdev, cmd_sz, &paddr64); @@ -5766,21 +6777,19 @@ static int hpsa_message(struct pci_dev *pdev, unsigned char opcode, cmd->CommandHeader.ReplyQueue = 0; cmd->CommandHeader.SGList = 0; cmd->CommandHeader.SGTotal = 0; - cmd->CommandHeader.Tag.lower = paddr32; - cmd->CommandHeader.Tag.upper = 0; + cmd->CommandHeader.tag = (u64) paddr32; memset(&cmd->CommandHeader.LUN.LunAddrBytes, 0, 8); cmd->Request.CDBLen = 16; - cmd->Request.Type.Type = TYPE_MSG; - cmd->Request.Type.Attribute = ATTR_HEADOFQUEUE; - cmd->Request.Type.Direction = XFER_NONE; + cmd->Request.type_attr_dir = + TYPE_ATTR_DIR(TYPE_MSG, ATTR_HEADOFQUEUE, XFER_NONE); cmd->Request.Timeout = 0; /* Don't time out */ cmd->Request.CDB[0] = opcode; cmd->Request.CDB[1] = type; memset(&cmd->Request.CDB[2], 0, 14); /* rest of the CDB is reserved */ - cmd->ErrorDescriptor.Addr.lower = paddr32 + sizeof(*cmd); - cmd->ErrorDescriptor.Addr.upper = 0; - cmd->ErrorDescriptor.Len = sizeof(struct ErrorInfo); + cmd->ErrorDescriptor.Addr = + cpu_to_le64((u64) (paddr32 + sizeof(*cmd))); + cmd->ErrorDescriptor.Len = cpu_to_le32(sizeof(struct ErrorInfo)); writel(paddr32, vaddr + SA5_REQUEST_PORT_OFFSET); @@ -5960,8 +6969,12 @@ static int hpsa_kdump_hard_reset_controller(struct pci_dev *pdev) */ rc = hpsa_lookup_board_id(pdev, &board_id); - if (rc < 0 || !ctlr_is_resettable(board_id)) { - dev_warn(&pdev->dev, "Not resetting device.\n"); + if (rc < 0) { + dev_warn(&pdev->dev, "Board ID not found\n"); + return rc; + } + if (!ctlr_is_resettable(board_id)) { + dev_warn(&pdev->dev, "Controller not resettable\n"); return -ENODEV; } @@ -6027,7 +7040,7 @@ static int hpsa_kdump_hard_reset_controller(struct pci_dev *pdev) rc = hpsa_wait_for_board_state(pdev, vaddr, BOARD_READY); if (rc) { dev_warn(&pdev->dev, - "failed waiting for board to become ready " + "Failed waiting for board to become ready " "after hard reset\n"); goto unmap_cfgtable; } @@ -6058,7 +7071,6 @@ static int hpsa_kdump_hard_reset_controller(struct pci_dev *pdev) */ static void print_cfg_table(struct device *dev, struct CfgTable *tb) { -#ifdef HPSA_DEBUG int i; char temp_name[17]; @@ -6079,7 +7091,7 @@ static void print_cfg_table(struct device *dev, struct CfgTable *tb) readl(&(tb->HostWrite.CoalIntDelay))); dev_info(dev, " Coalesce Interrupt Count = 0x%x\n", readl(&(tb->HostWrite.CoalIntCount))); - dev_info(dev, " Max outstanding commands = 0x%d\n", + dev_info(dev, " Max outstanding commands = %d\n", readl(&(tb->CmdsOutMax))); dev_info(dev, " Bus Types = 0x%x\n", readl(&(tb->BusTypes))); for (i = 0; i < 16; i++) @@ -6088,7 +7100,6 @@ static void print_cfg_table(struct device *dev, struct CfgTable *tb) dev_info(dev, " Server Name = %s\n", temp_name); dev_info(dev, " Heartbeat Counter = 0x%x\n\n\n", readl(&(tb->HeartBeat))); -#endif /* HPSA_DEBUG */ } static int find_PCI_BAR_index(struct pci_dev *pdev, unsigned long pci_bar_addr) @@ -6126,10 +7137,24 @@ static int find_PCI_BAR_index(struct pci_dev *pdev, unsigned long pci_bar_addr) return -1; } +static void hpsa_disable_interrupt_mode(struct ctlr_info *h) +{ +#ifdef CONFIG_PCI_MSI + if (h->msix_vector) { + if (h->pdev->msix_enabled) + pci_disable_msix(h->pdev); + h->msix_vector = 0; + } else if (h->msi_vector) { + if (h->pdev->msi_enabled) + pci_disable_msi(h->pdev); + h->msi_vector = 0; + } +#endif /* CONFIG_PCI_MSI */ +} + /* 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. + * controllers that are capable. If not, we use legacy INTx mode. */ - static void hpsa_interrupt_mode(struct ctlr_info *h) { #ifdef CONFIG_PCI_MSI @@ -6146,7 +7171,7 @@ static void hpsa_interrupt_mode(struct ctlr_info *h) (h->board_id == 0x40820E11) || (h->board_id == 0x40830E11)) goto default_int_mode; if (pci_find_capability(h->pdev, PCI_CAP_ID_MSIX)) { - dev_info(&h->pdev->dev, "MSIX\n"); + dev_info(&h->pdev->dev, "MSI-X capable controller\n"); h->msix_vector = MAX_REPLY_QUEUES; if (h->msix_vector > num_online_cpus()) h->msix_vector = num_online_cpus(); @@ -6167,7 +7192,7 @@ static void hpsa_interrupt_mode(struct ctlr_info *h) } single_msi_mode: if (pci_find_capability(h->pdev, PCI_CAP_ID_MSI)) { - dev_info(&h->pdev->dev, "MSI\n"); + dev_info(&h->pdev->dev, "MSI capable controller\n"); if (!pci_enable_msi(h->pdev)) h->msi_vector = 1; else @@ -6260,6 +7285,21 @@ static int hpsa_find_cfg_addrs(struct pci_dev *pdev, void __iomem *vaddr, return 0; } +static void hpsa_free_cfgtables(struct ctlr_info *h) +{ + if (h->transtable) { + iounmap(h->transtable); + h->transtable = NULL; + } + if (h->cfgtable) { + iounmap(h->cfgtable); + h->cfgtable = NULL; + } +} + +/* Find and map CISS config table and transfer table ++ * several items must be unmapped (freed) later ++ * */ static int hpsa_find_cfgtables(struct ctlr_info *h) { u64 cfg_offset; @@ -6274,8 +7314,10 @@ static int hpsa_find_cfgtables(struct ctlr_info *h) return rc; h->cfgtable = remap_pci_mem(pci_resource_start(h->pdev, cfg_base_addr_index) + cfg_offset, sizeof(*h->cfgtable)); - if (!h->cfgtable) + if (!h->cfgtable) { + dev_err(&h->pdev->dev, "Failed mapping cfgtable\n"); return -ENOMEM; + } rc = write_driver_ver_to_cfgtable(h->cfgtable); if (rc) return rc; @@ -6284,28 +7326,43 @@ static int hpsa_find_cfgtables(struct ctlr_info *h) h->transtable = remap_pci_mem(pci_resource_start(h->pdev, cfg_base_addr_index)+cfg_offset+trans_offset, sizeof(*h->transtable)); - if (!h->transtable) + if (!h->transtable) { + dev_err(&h->pdev->dev, "Failed mapping transfer table\n"); + hpsa_free_cfgtables(h); return -ENOMEM; + } return 0; } static void hpsa_get_max_perf_mode_cmds(struct ctlr_info *h) { +#define MIN_MAX_COMMANDS 16 + BUILD_BUG_ON(MIN_MAX_COMMANDS <= HPSA_NRESERVED_CMDS); + h->max_commands = readl(&(h->cfgtable->MaxPerformantModeCommands)); /* Limit commands in memory limited kdump scenario. */ if (reset_devices && h->max_commands > 32) h->max_commands = 32; - if (h->max_commands < 16) { + if (h->max_commands < MIN_MAX_COMMANDS) { 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; + h->max_commands = MIN_MAX_COMMANDS; } } +/* If the controller reports that the total max sg entries is greater than 512, + * then we know that chained SG blocks work. (Original smart arrays did not + * support chained SG blocks and would return zero for max sg entries.) + */ +static int hpsa_supports_chained_sg_blocks(struct ctlr_info *h) +{ + return h->maxsgentries > 512; +} + /* Interrogate the hardware for some limits: * max commands, max SG elements without chaining, and with chaining, * SG chain block size, etc. @@ -6313,19 +7370,21 @@ static void hpsa_get_max_perf_mode_cmds(struct ctlr_info *h) static void hpsa_find_board_params(struct ctlr_info *h) { hpsa_get_max_perf_mode_cmds(h); - h->nr_cmds = h->max_commands - 4; /* Allow room for some ioctls */ + h->nr_cmds = h->max_commands; h->maxsgentries = readl(&(h->cfgtable->MaxScatterGatherElements)); h->fw_support = readl(&(h->cfgtable->misc_fw_support)); - /* - * Limit in-command s/g elements to 32 save dma'able memory. - * Howvever spec says if 0, use 31 - */ - h->max_cmd_sg_entries = 31; - if (h->maxsgentries > 512) { + if (hpsa_supports_chained_sg_blocks(h)) { + /* Limit in-command s/g elements to 32 save dma'able memory. */ h->max_cmd_sg_entries = 32; - h->chainsize = h->maxsgentries - h->max_cmd_sg_entries + 1; + h->chainsize = h->maxsgentries - h->max_cmd_sg_entries; h->maxsgentries--; /* save one for chain pointer */ } else { + /* Original smart arrays supported at most 31 scatter gather + * entries embedded inline in the command (trying to use more + * would lock up the controller, see + * https://lkml.org/lkml/2001/12/4/139 ). + */ + h->max_cmd_sg_entries = 31; h->maxsgentries = 31; /* default to traditional values */ h->chainsize = 0; } @@ -6336,12 +7395,14 @@ static void hpsa_find_board_params(struct ctlr_info *h) dev_warn(&h->pdev->dev, "Physical aborts not supported\n"); if (!(HPSATMF_LOG_TASK_ABORT & h->TMFSupportFlags)) dev_warn(&h->pdev->dev, "Logical aborts not supported\n"); + if (!(HPSATMF_IOACCEL_ENABLED & h->TMFSupportFlags)) + dev_warn(&h->pdev->dev, "HP SSD Smart Path aborts supported\n"); } static inline bool hpsa_CISS_signature_present(struct ctlr_info *h) { if (!check_signature(h->cfgtable->Signature, "CISS", 4)) { - dev_warn(&h->pdev->dev, "not a valid CISS config table\n"); + dev_err(&h->pdev->dev, "not a valid CISS config table\n"); return false; } return true; @@ -6374,24 +7435,27 @@ static inline void hpsa_p600_dma_prefetch_quirk(struct ctlr_info *h) writel(dma_prefetch, h->vaddr + I2O_DMA1_CFG); } -static void hpsa_wait_for_clear_event_notify_ack(struct ctlr_info *h) +static int hpsa_wait_for_clear_event_notify_ack(struct ctlr_info *h) { int i; u32 doorbell_value; unsigned long flags; /* wait until the clear_event_notify bit 6 is cleared by controller. */ - for (i = 0; i < MAX_CONFIG_WAIT; i++) { + for (i = 0; i < MAX_CLEAR_EVENT_WAIT; i++) { spin_lock_irqsave(&h->lock, flags); doorbell_value = readl(h->vaddr + SA5_DOORBELL); spin_unlock_irqrestore(&h->lock, flags); if (!(doorbell_value & DOORBELL_CLEAR_EVENTS)) - break; + goto done; /* delay and try again */ - msleep(20); + msleep(CLEAR_EVENT_WAIT_INTERVAL); } + return -ENODEV; +done: + return 0; } -static void hpsa_wait_for_mode_change_ack(struct ctlr_info *h) +static int hpsa_wait_for_mode_change_ack(struct ctlr_info *h) { int i; u32 doorbell_value; @@ -6401,17 +7465,21 @@ static void hpsa_wait_for_mode_change_ack(struct ctlr_info *h) * (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++) { + for (i = 0; i < MAX_MODE_CHANGE_WAIT; i++) { spin_lock_irqsave(&h->lock, flags); doorbell_value = readl(h->vaddr + SA5_DOORBELL); spin_unlock_irqrestore(&h->lock, flags); if (!(doorbell_value & CFGTBL_ChangeReq)) - break; + goto done; /* delay and try again */ - usleep_range(10000, 20000); + msleep(MODE_CHANGE_WAIT_INTERVAL); } + return -ENODEV; +done: + return 0; } +/* return -ENODEV or other reason on error, 0 on success */ static int hpsa_enter_simple_mode(struct ctlr_info *h) { u32 trans_support; @@ -6426,33 +7494,54 @@ static int hpsa_enter_simple_mode(struct ctlr_info *h) writel(CFGTBL_Trans_Simple, &(h->cfgtable->HostWrite.TransportRequest)); writel(0, &h->cfgtable->HostWrite.command_pool_addr_hi); writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL); - hpsa_wait_for_mode_change_ack(h); + if (hpsa_wait_for_mode_change_ack(h)) + goto error; print_cfg_table(&h->pdev->dev, h->cfgtable); if (!(readl(&(h->cfgtable->TransportActive)) & CFGTBL_Trans_Simple)) goto error; h->transMethod = CFGTBL_Trans_Simple; return 0; error: - dev_warn(&h->pdev->dev, "unable to get board into simple mode\n"); + dev_err(&h->pdev->dev, "failed to enter simple mode\n"); return -ENODEV; } +/* free items allocated or mapped by hpsa_pci_init */ +static void hpsa_free_pci_init(struct ctlr_info *h) +{ + hpsa_free_cfgtables(h); /* pci_init 4 */ + iounmap(h->vaddr); /* pci_init 3 */ + h->vaddr = NULL; + hpsa_disable_interrupt_mode(h); /* pci_init 2 */ + /* + * call pci_disable_device before pci_release_regions per + * Documentation/PCI/pci.txt + */ + pci_disable_device(h->pdev); /* pci_init 1 */ + pci_release_regions(h->pdev); /* pci_init 2 */ +} + +/* several items must be freed later */ static int hpsa_pci_init(struct ctlr_info *h) { int prod_index, err; prod_index = hpsa_lookup_board_id(h->pdev, &h->board_id); if (prod_index < 0) - return -ENODEV; + return prod_index; 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); err = pci_enable_device(h->pdev); if (err) { - dev_warn(&h->pdev->dev, "unable to enable PCI device\n"); + dev_err(&h->pdev->dev, "failed to enable PCI device\n"); + pci_disable_device(h->pdev); return err; } @@ -6462,44 +7551,50 @@ static int hpsa_pci_init(struct ctlr_info *h) err = pci_request_regions(h->pdev, HPSA); if (err) { dev_err(&h->pdev->dev, - "cannot obtain PCI resources, aborting\n"); + "failed to obtain PCI resources\n"); + pci_disable_device(h->pdev); return err; } hpsa_interrupt_mode(h); err = hpsa_pci_find_memory_BAR(h->pdev, &h->paddr); if (err) - goto err_out_free_res; + goto clean2; /* intmode+region, pci */ h->vaddr = remap_pci_mem(h->paddr, 0x250); if (!h->vaddr) { + dev_err(&h->pdev->dev, "failed to remap PCI mem\n"); err = -ENOMEM; - goto err_out_free_res; + goto clean2; /* intmode+region, pci */ } err = hpsa_wait_for_board_state(h->pdev, h->vaddr, BOARD_READY); if (err) - goto err_out_free_res; + goto clean3; /* vaddr, intmode+region, pci */ err = hpsa_find_cfgtables(h); if (err) - goto err_out_free_res; + goto clean3; /* vaddr, intmode+region, pci */ hpsa_find_board_params(h); if (!hpsa_CISS_signature_present(h)) { err = -ENODEV; - goto err_out_free_res; + goto clean4; /* cfgtables, vaddr, intmode+region, pci */ } hpsa_set_driver_support_bits(h); hpsa_p600_dma_prefetch_quirk(h); err = hpsa_enter_simple_mode(h); if (err) - goto err_out_free_res; + goto clean4; /* cfgtables, vaddr, intmode+region, pci */ return 0; -err_out_free_res: - if (h->transtable) - iounmap(h->transtable); - if (h->cfgtable) - iounmap(h->cfgtable); - if (h->vaddr) - iounmap(h->vaddr); +clean4: /* cfgtables, vaddr, intmode+region, pci */ + hpsa_free_cfgtables(h); +clean3: /* vaddr, intmode+region, pci */ + iounmap(h->vaddr); + h->vaddr = NULL; +clean2: /* intmode+region, pci */ + hpsa_disable_interrupt_mode(h); + /* + * call pci_disable_device before pci_release_regions per + * Documentation/PCI/pci.txt + */ pci_disable_device(h->pdev); pci_release_regions(h->pdev); return err; @@ -6553,14 +7648,11 @@ static int hpsa_init_reset_devices(struct pci_dev *pdev) * "performant mode". Or, it might be 640x, which can't reset * due to concerns about shared bbwc between 6402/6404 pair. */ - if (rc) { - if (rc != -ENOTSUPP) /* just try to do the kdump anyhow. */ - rc = -ENODEV; + if (rc) goto out_disable; - } /* Now try to get the controller to respond to a no-op */ - dev_warn(&pdev->dev, "Waiting for controller to respond to no-op\n"); + dev_info(&pdev->dev, "Waiting for controller to respond to no-op\n"); for (i = 0; i < HPSA_POST_RESET_NOOP_RETRIES; i++) { if (hpsa_noop(pdev) == 0) break; @@ -6575,7 +7667,28 @@ static int hpsa_init_reset_devices(struct pci_dev *pdev) return rc; } -static int hpsa_allocate_cmd_pool(struct ctlr_info *h) +static void hpsa_free_cmd_pool(struct ctlr_info *h) +{ + kfree(h->cmd_pool_bits); + h->cmd_pool_bits = NULL; + if (h->cmd_pool) { + pci_free_consistent(h->pdev, + h->nr_cmds * sizeof(struct CommandList), + h->cmd_pool, h->cmd_pool_dhandle); + h->cmd_pool = NULL; + h->cmd_pool_dhandle = 0; + } + if (h->errinfo_pool) { + pci_free_consistent(h->pdev, + h->nr_cmds * sizeof(struct ErrorInfo), + h->errinfo_pool, + h->errinfo_pool_dhandle); + h->errinfo_pool = NULL; + h->errinfo_pool_dhandle = 0; + } +} + +static int hpsa_alloc_cmd_pool(struct ctlr_info *h) { h->cmd_pool_bits = kzalloc( DIV_ROUND_UP(h->nr_cmds, BITS_PER_LONG) * @@ -6590,45 +7703,51 @@ static int hpsa_allocate_cmd_pool(struct ctlr_info *h) || (h->cmd_pool == NULL) || (h->errinfo_pool == NULL)) { dev_err(&h->pdev->dev, "out of memory in %s", __func__); - return -ENOMEM; + goto clean_up; } + hpsa_preinitialize_commands(h); return 0; +clean_up: + hpsa_free_cmd_pool(h); + return -ENOMEM; } -static void hpsa_free_cmd_pool(struct ctlr_info *h) +static void hpsa_irq_affinity_hints(struct ctlr_info *h) { - kfree(h->cmd_pool_bits); - if (h->cmd_pool) - pci_free_consistent(h->pdev, - h->nr_cmds * sizeof(struct CommandList), - h->cmd_pool, h->cmd_pool_dhandle); - if (h->ioaccel2_cmd_pool) - pci_free_consistent(h->pdev, - h->nr_cmds * sizeof(*h->ioaccel2_cmd_pool), - h->ioaccel2_cmd_pool, h->ioaccel2_cmd_pool_dhandle); - if (h->errinfo_pool) - pci_free_consistent(h->pdev, - h->nr_cmds * sizeof(struct ErrorInfo), - h->errinfo_pool, - h->errinfo_pool_dhandle); - if (h->ioaccel_cmd_pool) - pci_free_consistent(h->pdev, - h->nr_cmds * sizeof(struct io_accel1_cmd), - h->ioaccel_cmd_pool, h->ioaccel_cmd_pool_dhandle); + int i, cpu, rc; + + cpu = cpumask_first(cpu_online_mask); + for (i = 0; i < h->msix_vector; i++) { + rc = irq_set_affinity_hint(h->intr[i], get_cpu_mask(cpu)); + cpu = cpumask_next(cpu, cpu_online_mask); + } } -static void hpsa_irq_affinity_hints(struct ctlr_info *h) +/* clear affinity hints and free MSI-X, MSI, or legacy INTx vectors */ +static void hpsa_free_irqs(struct ctlr_info *h) { - int i, cpu, rc; + int i; + + if (!h->msix_vector || h->intr_mode != PERF_MODE_INT) { + /* Single reply queue, only one irq to free */ + i = h->intr_mode; + irq_set_affinity_hint(h->intr[i], NULL); + free_irq(h->intr[i], &h->q[i]); + h->q[i] = 0; + return; + } - cpu = cpumask_first(cpu_online_mask); for (i = 0; i < h->msix_vector; i++) { - rc = irq_set_affinity_hint(h->intr[i], get_cpu_mask(cpu)); - cpu = cpumask_next(cpu, cpu_online_mask); + irq_set_affinity_hint(h->intr[i], NULL); + free_irq(h->intr[i], &h->q[i]); + h->q[i] = 0; } + for (; i < MAX_REPLY_QUEUES; i++) + h->q[i] = 0; } -static int hpsa_request_irq(struct ctlr_info *h, +/* returns 0 on success; cleans up and returns -Enn on error */ +static int hpsa_request_irqs(struct ctlr_info *h, irqreturn_t (*msixhandler)(int, void *), irqreturn_t (*intxhandler)(int, void *)) { @@ -6643,26 +7762,54 @@ static int hpsa_request_irq(struct ctlr_info *h, if (h->intr_mode == PERF_MODE_INT && h->msix_vector > 0) { /* If performant mode and MSI-X, use multiple reply queues */ - for (i = 0; i < h->msix_vector; i++) + for (i = 0; i < h->msix_vector; i++) { + sprintf(h->intrname[i], "%s-msix%d", h->devname, i); rc = request_irq(h->intr[i], msixhandler, - 0, h->devname, + 0, h->intrname[i], &h->q[i]); + if (rc) { + int j; + + dev_err(&h->pdev->dev, + "failed to get irq %d for %s\n", + h->intr[i], h->devname); + for (j = 0; j < i; j++) { + free_irq(h->intr[j], &h->q[j]); + h->q[j] = 0; + } + for (; j < MAX_REPLY_QUEUES; j++) + h->q[j] = 0; + return rc; + } + } hpsa_irq_affinity_hints(h); } else { /* Use single reply pool */ if (h->msix_vector > 0 || h->msi_vector) { + if (h->msix_vector) + sprintf(h->intrname[h->intr_mode], + "%s-msix", h->devname); + else + sprintf(h->intrname[h->intr_mode], + "%s-msi", h->devname); rc = request_irq(h->intr[h->intr_mode], - msixhandler, 0, h->devname, + msixhandler, 0, + h->intrname[h->intr_mode], &h->q[h->intr_mode]); } else { + sprintf(h->intrname[h->intr_mode], + "%s-intx", h->devname); rc = request_irq(h->intr[h->intr_mode], - intxhandler, IRQF_SHARED, h->devname, + intxhandler, IRQF_SHARED, + h->intrname[h->intr_mode], &h->q[h->intr_mode]); } + irq_set_affinity_hint(h->intr[h->intr_mode], NULL); } if (rc) { - dev_err(&h->pdev->dev, "unable to get irq %d for %s\n", + dev_err(&h->pdev->dev, "failed to get irq %d for %s\n", h->intr[h->intr_mode], h->devname); + hpsa_free_irqs(h); return -ENODEV; } return 0; @@ -6670,60 +7817,27 @@ static int hpsa_request_irq(struct ctlr_info *h, static int hpsa_kdump_soft_reset(struct ctlr_info *h) { - if (hpsa_send_host_reset(h, RAID_CTLR_LUNID, - HPSA_RESET_TYPE_CONTROLLER)) { - dev_warn(&h->pdev->dev, "Resetting array controller failed.\n"); - return -EIO; - } + int rc; + hpsa_send_host_reset(h, RAID_CTLR_LUNID, HPSA_RESET_TYPE_CONTROLLER); dev_info(&h->pdev->dev, "Waiting for board to soft reset.\n"); - if (hpsa_wait_for_board_state(h->pdev, h->vaddr, BOARD_NOT_READY)) { + rc = hpsa_wait_for_board_state(h->pdev, h->vaddr, BOARD_NOT_READY); + if (rc) { dev_warn(&h->pdev->dev, "Soft reset had no effect.\n"); - return -1; + return rc; } dev_info(&h->pdev->dev, "Board reset, awaiting READY status.\n"); - if (hpsa_wait_for_board_state(h->pdev, h->vaddr, BOARD_READY)) { + rc = hpsa_wait_for_board_state(h->pdev, h->vaddr, BOARD_READY); + if (rc) { dev_warn(&h->pdev->dev, "Board failed to become ready " "after soft reset.\n"); - return -1; + return rc; } return 0; } -static void free_irqs(struct ctlr_info *h) -{ - int i; - - if (!h->msix_vector || h->intr_mode != PERF_MODE_INT) { - /* Single reply queue, only one irq to free */ - i = h->intr_mode; - irq_set_affinity_hint(h->intr[i], NULL); - free_irq(h->intr[i], &h->q[i]); - return; - } - - for (i = 0; i < h->msix_vector; i++) { - irq_set_affinity_hint(h->intr[i], NULL); - free_irq(h->intr[i], &h->q[i]); - } -} - -static void hpsa_free_irqs_and_disable_msix(struct ctlr_info *h) -{ - free_irqs(h); -#ifdef CONFIG_PCI_MSI - if (h->msix_vector) { - if (h->pdev->msix_enabled) - pci_disable_msix(h->pdev); - } else if (h->msi_vector) { - if (h->pdev->msi_enabled) - pci_disable_msi(h->pdev); - } -#endif /* CONFIG_PCI_MSI */ -} - static void hpsa_free_reply_queues(struct ctlr_info *h) { int i; @@ -6735,40 +7849,50 @@ static void hpsa_free_reply_queues(struct ctlr_info *h) h->reply_queue[i].head, h->reply_queue[i].busaddr); h->reply_queue[i].head = NULL; h->reply_queue[i].busaddr = 0; + h->reply_queue_size = 0; } } static void hpsa_undo_allocations_after_kdump_soft_reset(struct ctlr_info *h) { - hpsa_free_irqs_and_disable_msix(h); - hpsa_free_sg_chain_blocks(h); - hpsa_free_cmd_pool(h); - kfree(h->ioaccel1_blockFetchTable); - kfree(h->blockFetchTable); - hpsa_free_reply_queues(h); - if (h->vaddr) - iounmap(h->vaddr); - if (h->transtable) - iounmap(h->transtable); - if (h->cfgtable) - iounmap(h->cfgtable); - pci_disable_device(h->pdev); - pci_release_regions(h->pdev); - kfree(h); + hpsa_free_performant_mode(h); /* init_one 7 */ + hpsa_free_sg_chain_blocks(h); /* init_one 6 */ + hpsa_free_cmd_pool(h); /* init_one 5 */ + hpsa_free_irqs(h); /* init_one 4 */ + scsi_host_put(h->scsi_host); /* init_one 3 */ + h->scsi_host = NULL; /* init_one 3 */ + hpsa_free_pci_init(h); /* init_one 2_5 */ + free_percpu(h->lockup_detected); /* init_one 2 */ + h->lockup_detected = NULL; /* init_one 2 */ + if (h->resubmit_wq) + destroy_workqueue(h->resubmit_wq); /* init_one 1 */ + kfree(h); /* init_one 1 */ } /* Called when controller lockup detected. */ -static void fail_all_cmds_on_list(struct ctlr_info *h, struct list_head *list) +static void fail_all_outstanding_cmds(struct ctlr_info *h) { - struct CommandList *c = NULL; + int i, refcount; + struct CommandList *c; + int failcount = 0; - assert_spin_locked(&h->lock); - /* Mark all outstanding commands as failed and complete them. */ - while (!list_empty(list)) { - c = list_entry(list->next, struct CommandList, list); - c->err_info->CommandStatus = CMD_HARDWARE_ERR; - finish_cmd(c); + 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_CTLR_LOCKUP; + /* CMD_CTLR_LOCKUP gets finish_cmd to return + * DID_NO_CONNECT which doesn't get retried + */ + finish_cmd(c); + atomic_dec(&h->commands_outstanding); + 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) @@ -6796,18 +7920,16 @@ static void controller_lockup_detected(struct ctlr_info *h) 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); - spin_lock_irqsave(&h->lock, flags); - fail_all_cmds_on_list(h, &h->cmpQ); - fail_all_cmds_on_list(h, &h->reqQ); - spin_unlock_irqrestore(&h->lock, flags); + fail_all_outstanding_cmds(h); } static void detect_controller_lockup(struct ctlr_info *h) @@ -6816,6 +7938,9 @@ static void detect_controller_lockup(struct ctlr_info *h) u32 heartbeat; unsigned long flags; + if (!h->lockup_detector_enabled) + return; + now = get_jiffies_64(); /* If we've received an interrupt recently, we're ok. */ if (time_after64(h->last_intr_timestamp + @@ -6850,8 +7975,8 @@ static void hpsa_ack_ctlr_events(struct ctlr_info *h) int i; char *event_type; - /* Clear the driver-requested rescan flag */ - h->drv_req_rescan = 0; + if (!(h->fw_support & MISC_FW_EVENT_NOTIFY)) + return; /* Ask the controller to clear the events we're handling. */ if ((h->transMethod & (CFGTBL_Trans_io_accel1 @@ -6898,9 +8023,6 @@ static void hpsa_ack_ctlr_events(struct ctlr_info *h) */ static int hpsa_ctlr_needs_rescan(struct ctlr_info *h) { - if (h->drv_req_rescan) - return 1; - if (!(h->fw_support & MISC_FW_EVENT_NOTIFY)) return 0; @@ -6934,32 +8056,117 @@ static int hpsa_offline_devices_ready(struct ctlr_info *h) return 0; } - -static void hpsa_monitor_ctlr_worker(struct work_struct *work) +static void hpsa_requeue_worker(struct ctlr_info *h, struct delayed_work *wi) { unsigned long flags; + + spin_lock_irqsave(&h->lock, flags); + if (!h->remove_in_progress) + schedule_delayed_work(wi, h->heartbeat_sample_interval); + spin_unlock_irqrestore(&h->lock, flags); +} + +static void hpsa_rescan_ctlr_worker(struct work_struct *work) +{ struct ctlr_info *h = container_of(to_delayed_work(work), - struct ctlr_info, monitor_ctlr_work); - detect_controller_lockup(h); - if (lockup_detected(h)) - return; + struct ctlr_info, rescan_ctlr_work); if (hpsa_ctlr_needs_rescan(h) || hpsa_offline_devices_ready(h)) { scsi_host_get(h->scsi_host); - h->drv_req_rescan = 0; hpsa_ack_ctlr_events(h); hpsa_scan_start(h->scsi_host); scsi_host_put(h->scsi_host); } - spin_lock_irqsave(&h->lock, flags); - if (h->remove_in_progress) { - spin_unlock_irqrestore(&h->lock, flags); + hpsa_requeue_worker(h, &h->rescan_ctlr_work); +} + +static void hpsa_monitor_ctlr_worker(struct work_struct *work) +{ + struct ctlr_info *h = container_of(to_delayed_work(work), + struct ctlr_info, monitor_ctlr_work); + detect_controller_lockup(h); + if (lockup_detected(h)) return; + + hpsa_requeue_worker(h, &h->monitor_ctlr_work); +} + +static void check_board_id_tables(void) +{ + int i, j, found; + uint32_t board_id; + int count; + + for (i = 0; i < ARRAY_SIZE(products); i++) { + board_id = products[i].board_id; + if (board_id == 0xFFFF103C) /* sentinel entry */ + continue; + found = 0; + count = 0; + for (j = 0; j < ARRAY_SIZE(hpsa_pci_device_id); j++) { + uint32_t bid; + + bid = (hpsa_pci_device_id[j].subdevice << 16) | + hpsa_pci_device_id[j].subvendor; + if (bid == board_id) { + found = 1; + if (i != j) { + printk(KERN_WARNING HPSA + ": products[%d] (%s) found at pci table entry %d\n", + i, products[i].product_name, j); + } + count++; + } + } + if (!found) + printk(KERN_WARNING HPSA + ": products[%d] (%s) not found in pci table\n", + i, products[i].product_name); + if (count > 1) + printk(KERN_WARNING HPSA + ": products[%d] (%s) has duplicate entries in pci table\n", + i, products[i].product_name); + } + + for (i = 0; i < ARRAY_SIZE(hpsa_pci_device_id); i++) { + board_id = (hpsa_pci_device_id[i].subdevice << 16) | + hpsa_pci_device_id[i].subvendor; + if (hpsa_pci_device_id[i].vendor == 0) /* sentinel */ + continue; + if (hpsa_pci_device_id[i].subvendor == PCI_ANY_ID && + hpsa_pci_device_id[i].subdevice == PCI_ANY_ID) + continue; + count = 0; + found = 0; + for (j = 0; j < ARRAY_SIZE(products); j++) { + if (board_id == products[j].board_id) { + if (j != i) { + printk(KERN_WARNING HPSA + ": pci table entry %d found at product entry %d (%s)\n", + i, j, products[j].product_name); + } + found = 1; + count++; + } + } + if (!found) + printk(KERN_WARNING HPSA + ": pci table entry %d (%04x:%04x) not found in product table\n", + i, hpsa_pci_device_id[i].subvendor, + hpsa_pci_device_id[i].subdevice); + if (count > 1) + printk(KERN_WARNING HPSA + ": pci table entry %d (%04x:%04x) has duplicate product[] entries\n", + i, hpsa_pci_device_id[i].subvendor, + hpsa_pci_device_id[i].subdevice); + } + if (ARRAY_SIZE(products) != ARRAY_SIZE(hpsa_pci_device_id) - 1) { + printk(KERN_WARNING HPSA + ": suspicious relative cardinality of products vs hpsa_pci_device_id (%lu/%lu)\n", + (unsigned long) ARRAY_SIZE(products), + (unsigned long) ARRAY_SIZE(hpsa_pci_device_id)); } - schedule_delayed_work(&h->monitor_ctlr_work, - h->heartbeat_sample_interval); - spin_unlock_irqrestore(&h->lock, flags); } static int hpsa_init_one(struct pci_dev *pdev, const struct pci_device_id *ent) @@ -6969,6 +8176,7 @@ static int hpsa_init_one(struct pci_dev *pdev, const struct pci_device_id *ent) int try_soft_reset = 0; unsigned long flags; + check_board_id_tables(); if (number_of_controllers == 0) printk(KERN_INFO DRIVER_NAME "\n"); @@ -6993,33 +8201,42 @@ static int hpsa_init_one(struct pci_dev *pdev, const struct pci_device_id *ent) */ BUILD_BUG_ON(sizeof(struct CommandList) % COMMANDLIST_ALIGNMENT); h = kzalloc(sizeof(*h), GFP_KERNEL); - if (!h) + if (!h) { + dev_err(&pdev->dev, "Failed to allocate controller head\n"); return -ENOMEM; + } h->pdev = pdev; + h->intr_mode = hpsa_simple_mode ? SIMPLE_MODE_INT : PERF_MODE_INT; - INIT_LIST_HEAD(&h->cmpQ); - INIT_LIST_HEAD(&h->reqQ); INIT_LIST_HEAD(&h->offline_device_list); spin_lock_init(&h->lock); spin_lock_init(&h->offline_device_lock); spin_lock_init(&h->scan_lock); - spin_lock_init(&h->passthru_count_lock); + atomic_set(&h->passthru_cmds_avail, HPSA_MAX_CONCURRENT_PASSTHRUS); + atomic_set(&h->abort_cmds_available, HPSA_CMDS_RESERVED_FOR_ABORTS); + atomic_set(&h->cmds_sent, 0); /* Allocate and clear per-cpu variable lockup_detected */ h->lockup_detected = alloc_percpu(u32); if (!h->lockup_detected) { + dev_err(&h->pdev->dev, "Failed to allocate lockup detector\n"); rc = -ENOMEM; - goto clean1; + goto clean1; /* aer/h */ } set_lockup_detected_for_all_cpus(h, 0); rc = hpsa_pci_init(h); - if (rc != 0) - goto clean1; + if (rc) + goto clean2; /* lu, aer/h */ + + /* relies on h-> settings made by hpsa_pci_init, including + * interrupt_mode h->intr */ + rc = hpsa_scsi_host_alloc(h); + if (rc) + goto clean2_5; /* pci, lu, aer/h */ - sprintf(h->devname, HPSA "%d", number_of_controllers); - h->ctlr = number_of_controllers; + sprintf(h->devname, HPSA "%d", h->scsi_host->host_no); number_of_controllers++; /* configure PCI DMA stuff */ @@ -7032,31 +8249,48 @@ static int hpsa_init_one(struct pci_dev *pdev, const struct pci_device_id *ent) dac = 0; } else { dev_err(&pdev->dev, "no suitable DMA available\n"); - goto clean1; + goto clean3; /* shost, pci, lu, aer/h */ } } /* make sure the board interrupts are off */ h->access.set_intr_mask(h, HPSA_INTR_OFF); - if (hpsa_request_irq(h, do_hpsa_intr_msi, do_hpsa_intr_intx)) - goto clean2; - dev_info(&pdev->dev, "%s: <0x%x> at IRQ %d%s using DAC\n", - h->devname, pdev->device, - h->intr[h->intr_mode], dac ? "" : " not"); - if (hpsa_allocate_cmd_pool(h)) - goto clean4; - if (hpsa_allocate_sg_chain_blocks(h)) - goto clean4; + rc = hpsa_request_irqs(h, do_hpsa_intr_msi, do_hpsa_intr_intx); + if (rc) + goto clean3; /* shost, pci, lu, aer/h */ + rc = hpsa_alloc_cmd_pool(h); + if (rc) + goto clean4; /* irq, shost, pci, lu, aer/h */ + rc = hpsa_alloc_sg_chain_blocks(h); + if (rc) + goto clean5; /* cmd, irq, shost, pci, lu, aer/h */ init_waitqueue_head(&h->scan_wait_queue); + init_waitqueue_head(&h->abort_cmd_wait_queue); + init_waitqueue_head(&h->abort_sync_wait_queue); h->scan_finished = 1; /* no scan currently in progress */ pci_set_drvdata(pdev, h); h->ndevices = 0; h->hba_mode_enabled = 0; - h->scsi_host = NULL; + spin_lock_init(&h->devlock); - hpsa_put_ctlr_into_performant_mode(h); + rc = hpsa_put_ctlr_into_performant_mode(h); + if (rc) + goto clean6; /* sg, cmd, irq, shost, pci, lu, aer/h */ + + /* hook into SCSI subsystem */ + rc = hpsa_scsi_add_host(h); + if (rc) + goto clean7; /* perf, sg, cmd, irq, shost, pci, lu, aer/h */ + + /* create the resubmit workqueue */ + h->resubmit_wq = alloc_workqueue("hpsa", WQ_MEM_RECLAIM, 0); + if (!h->resubmit_wq) { + dev_err(&h->pdev->dev, "Failed to allocate work queue\n"); + rc = -ENOMEM; + goto clean8; /* sh, perf, sg, cmd, irq, shost, pci, lu, aer/h */ + } /* At this point, the controller is ready to take commands. * Now, if reset_devices and the hard reset didn't work, try @@ -7074,19 +8308,28 @@ static int hpsa_init_one(struct pci_dev *pdev, const struct pci_device_id *ent) spin_lock_irqsave(&h->lock, flags); h->access.set_intr_mask(h, HPSA_INTR_OFF); spin_unlock_irqrestore(&h->lock, flags); - free_irqs(h); - rc = hpsa_request_irq(h, hpsa_msix_discard_completions, + hpsa_free_irqs(h); + rc = hpsa_request_irqs(h, hpsa_msix_discard_completions, hpsa_intx_discard_completions); if (rc) { - dev_warn(&h->pdev->dev, "Failed to request_irq after " - "soft reset.\n"); - goto clean4; + dev_warn(&h->pdev->dev, + "Failed to request_irq after soft reset.\n"); + /* cannot goto clean9 or free_irqs will be called + * again. Instead, do its work */ + if (h->resubmit_wq) + destroy_workqueue(h->resubmit_wq); /* clean9 */ + hpsa_free_performant_mode(h); /* clean7 */ + hpsa_free_sg_chain_blocks(h); /* clean6 */ + hpsa_free_cmd_pool(h); /* clean5 */ + /* skip hpsa_free_irqs(h) clean4 since that + * was just called before request_irqs failed */ + goto clean3; } rc = hpsa_kdump_soft_reset(h); if (rc) /* Neither hard nor soft reset worked, we're hosed. */ - goto clean4; + goto clean9; dev_info(&h->pdev->dev, "Board READY.\n"); dev_info(&h->pdev->dev, @@ -7107,38 +8350,58 @@ static int hpsa_init_one(struct pci_dev *pdev, const struct pci_device_id *ent) hpsa_undo_allocations_after_kdump_soft_reset(h); try_soft_reset = 0; if (rc) - /* don't go to clean4, we already unallocated */ + /* don't goto clean, we already unallocated */ return -ENODEV; goto reinit_after_soft_reset; } - /* Enable Accelerated IO path at driver layer */ - h->acciopath_status = 1; + /* Enable Accelerated IO path at driver layer */ + h->acciopath_status = 1; - h->drv_req_rescan = 0; + h->lockup_detector_enabled = 1; /* Turn the interrupts on so we can service requests */ h->access.set_intr_mask(h, HPSA_INTR_ON); hpsa_hba_inquiry(h); - hpsa_register_scsi(h); /* hook ourselves into SCSI subsystem */ /* Monitor the controller for firmware lockups */ h->heartbeat_sample_interval = HEARTBEAT_SAMPLE_INTERVAL; INIT_DELAYED_WORK(&h->monitor_ctlr_work, hpsa_monitor_ctlr_worker); schedule_delayed_work(&h->monitor_ctlr_work, h->heartbeat_sample_interval); + INIT_DELAYED_WORK(&h->rescan_ctlr_work, hpsa_rescan_ctlr_worker); + schedule_delayed_work(&h->rescan_ctlr_work, + h->heartbeat_sample_interval); return 0; -clean4: +clean9: /* wq, sh, perf, sg, cmd, irq, shost, pci, lu, aer/h */ + if (h->resubmit_wq) + destroy_workqueue(h->resubmit_wq); +clean8: /* sh, perf, sg, cmd, irq, shost, pci, lu, aer/h */ + /* nothing to clean from scsi_add_host... + * scsi_host_put will do so below */ +clean7: /* perf, sg, cmd, irq, shost, pci, lu, aer/h */ + hpsa_free_performant_mode(h); +clean6: /* sg, cmd, irq, shost, pci, lu, aer/h */ hpsa_free_sg_chain_blocks(h); +clean5: /* cmd, irq, shost, pci, lu, aer/h */ hpsa_free_cmd_pool(h); - free_irqs(h); -clean2: -clean1: - if (h->lockup_detected) +clean4: /* irq, shost, pci, lu, aer/h */ + hpsa_free_irqs(h); +clean3: /* shost, pci, lu, aer/h */ + scsi_host_put(h->scsi_host); + h->scsi_host = NULL; +clean2_5: /* pci, lu, aer/h */ + hpsa_free_pci_init(h); +clean2: /* lu, aer/h */ + if (h->lockup_detected) { free_percpu(h->lockup_detected); + h->lockup_detected = NULL; + } +clean1: /* aer/h */ + /* (void) pci_disable_pcie_error_reporting(pdev); */ kfree(h); return rc; } @@ -7147,30 +8410,31 @@ static void hpsa_flush_cache(struct ctlr_info *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); if (!flush_buf) return; - c = cmd_special_alloc(h); - if (!c) { - dev_warn(&h->pdev->dev, "cmd_special_alloc returned NULL!\n"); - goto out_of_memory; - } + c = cmd_alloc(h); + if (fill_cmd(c, HPSA_CACHE_FLUSH, h, flush_buf, 4, 0, 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, "error flushing cache on controller\n"); - cmd_special_free(h, c); -out_of_memory: + cmd_free(h, c); kfree(flush_buf); } @@ -7185,15 +8449,18 @@ static void hpsa_shutdown(struct pci_dev *pdev) */ hpsa_flush_cache(h); h->access.set_intr_mask(h, HPSA_INTR_OFF); - hpsa_free_irqs_and_disable_msix(h); + hpsa_free_irqs(h); /* init_one 4 */ + hpsa_disable_interrupt_mode(h); /* pci_init 2 */ } static void hpsa_free_device_info(struct ctlr_info *h) { int i; - for (i = 0; i < h->ndevices; i++) + for (i = 0; i < h->ndevices; i++) { kfree(h->dev[i]); + h->dev[i] = NULL; + } } static void hpsa_remove_one(struct pci_dev *pdev) @@ -7210,32 +8477,38 @@ static void hpsa_remove_one(struct pci_dev *pdev) /* Get rid of any controller monitoring work items */ spin_lock_irqsave(&h->lock, flags); h->remove_in_progress = 1; - cancel_delayed_work(&h->monitor_ctlr_work); + cancel_delayed_work_sync(&h->monitor_ctlr_work); + cancel_delayed_work_sync(&h->rescan_ctlr_work); spin_unlock_irqrestore(&h->lock, flags); - hpsa_unregister_scsi(h); /* unhook from SCSI subsystem */ + /* includes hpsa_free_irqs - init_one 4 */ + /* includes hpsa_disable_interrupt_mode - pci_init 2 */ hpsa_shutdown(pdev); - iounmap(h->vaddr); - iounmap(h->transtable); - iounmap(h->cfgtable); - hpsa_free_device_info(h); - hpsa_free_sg_chain_blocks(h); - pci_free_consistent(h->pdev, - h->nr_cmds * sizeof(struct CommandList), - h->cmd_pool, h->cmd_pool_dhandle); - pci_free_consistent(h->pdev, - h->nr_cmds * sizeof(struct ErrorInfo), - h->errinfo_pool, h->errinfo_pool_dhandle); - hpsa_free_reply_queues(h); - kfree(h->cmd_pool_bits); - kfree(h->blockFetchTable); - kfree(h->ioaccel1_blockFetchTable); - kfree(h->ioaccel2_blockFetchTable); - kfree(h->hba_inquiry_data); - pci_disable_device(pdev); - pci_release_regions(pdev); - free_percpu(h->lockup_detected); - kfree(h); + + hpsa_free_device_info(h); /* scan */ + + kfree(h->hba_inquiry_data); /* init_one 10 */ + h->hba_inquiry_data = NULL; /* init_one 10 */ + if (h->resubmit_wq) + destroy_workqueue(h->resubmit_wq); /* init_one 9 */ + scsi_remove_host(h->scsi_host); /* init_one 8 */ + hpsa_free_ioaccel2_sg_chain_blocks(h); + hpsa_free_performant_mode(h); /* init_one 7 */ + hpsa_free_sg_chain_blocks(h); /* init_one 6 */ + hpsa_free_cmd_pool(h); /* init_one 5 */ + + /* hpsa_free_irqs already called via hpsa_shutdown init_one 4 */ + + scsi_host_put(h->scsi_host); /* init_one 3 */ + h->scsi_host = NULL; /* init_one 3 */ + + /* includes hpsa_disable_interrupt_mode - pci_init 2 */ + hpsa_free_pci_init(h); /* init_one 2.5 */ + + free_percpu(h->lockup_detected); /* init_one 2 */ + h->lockup_detected = NULL; /* init_one 2 */ + /* (void) pci_disable_pcie_error_reporting(pdev); */ /* init_one 1 */ + kfree(h); /* init_one 1 */ } static int hpsa_suspend(__attribute__((unused)) struct pci_dev *pdev, @@ -7293,7 +8566,8 @@ static void calc_bucket_map(int bucket[], int num_buckets, } } -static void hpsa_enter_performant_mode(struct ctlr_info *h, u32 trans_support) +/* return -ENODEV or other reason on error, 0 on success */ +static int hpsa_enter_performant_mode(struct ctlr_info *h, u32 trans_support) { int i; unsigned long register_value; @@ -7385,12 +8659,17 @@ static void hpsa_enter_performant_mode(struct ctlr_info *h, u32 trans_support) } } writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL); - hpsa_wait_for_mode_change_ack(h); + if (hpsa_wait_for_mode_change_ack(h)) { + dev_err(&h->pdev->dev, + "performant mode problem - doorbell timeout\n"); + return -ENODEV; + } + print_cfg_table(&h->pdev->dev, h->cfgtable); register_value = readl(&(h->cfgtable->TransportActive)); if (!(register_value & CFGTBL_Trans_Performant)) { - dev_warn(&h->pdev->dev, "unable to get board into" - " performant mode\n"); - return; + dev_err(&h->pdev->dev, + "performant mode problem - transport not active\n"); + return -ENODEV; } /* Change the access methods to the performant access methods */ h->access = access; @@ -7398,7 +8677,7 @@ static void hpsa_enter_performant_mode(struct ctlr_info *h, u32 trans_support) if (!((trans_support & CFGTBL_Trans_io_accel1) || (trans_support & CFGTBL_Trans_io_accel2))) - return; + return 0; if (trans_support & CFGTBL_Trans_io_accel1) { /* Set up I/O accelerator mode */ @@ -7431,13 +8710,10 @@ static void hpsa_enter_performant_mode(struct ctlr_info *h, u32 trans_support) cp->host_context_flags = IOACCEL1_HCFLAGS_CISS_FORMAT; cp->timeout_sec = 0; cp->ReplyQueue = 0; - cp->Tag.lower = (i << DIRECT_LOOKUP_SHIFT) | - DIRECT_LOOKUP_BIT; - cp->Tag.upper = 0; - cp->host_addr.lower = - (u32) (h->ioaccel_cmd_pool_dhandle + - (i * sizeof(struct io_accel1_cmd))); - cp->host_addr.upper = 0; + cp->tag = cpu_to_le64((u64) i << DIRECT_LOOKUP_SHIFT); + cp->host_addr = + cpu_to_le64((u64) (h->ioaccel_cmd_pool_dhandle + + (i * sizeof(struct io_accel1_cmd)))); } } else if (trans_support & CFGTBL_Trans_io_accel2) { u64 cfg_offset, cfg_base_addr_index; @@ -7463,10 +8739,31 @@ static void hpsa_enter_performant_mode(struct ctlr_info *h, u32 trans_support) writel(bft2[i], &h->ioaccel2_bft2_regs[i]); } writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL); - hpsa_wait_for_mode_change_ack(h); + if (hpsa_wait_for_mode_change_ack(h)) { + dev_err(&h->pdev->dev, + "performant mode problem - enabling ioaccel mode\n"); + return -ENODEV; + } + print_cfg_table(&h->pdev->dev, h->cfgtable); + return 0; +} + +/* Free ioaccel1 mode command blocks and block fetch table */ +static void hpsa_free_ioaccel1_cmd_and_bft(struct ctlr_info *h) +{ + if (h->ioaccel_cmd_pool) { + pci_free_consistent(h->pdev, + h->nr_cmds * sizeof(*h->ioaccel_cmd_pool), + h->ioaccel_cmd_pool, h->ioaccel_cmd_pool_dhandle); + h->ioaccel_cmd_pool = NULL; + h->ioaccel_cmd_pool_dhandle = 0; + } + kfree(h->ioaccel1_blockFetchTable); + h->ioaccel1_blockFetchTable = NULL; } -static int hpsa_alloc_ioaccel_cmd_and_bft(struct ctlr_info *h) +/* Allocate ioaccel1 mode command blocks and block fetch table */ +static int hpsa_alloc_ioaccel1_cmd_and_bft(struct ctlr_info *h) { h->ioaccel_maxsg = readl(&(h->cfgtable->io_accel_max_embedded_sg_count)); @@ -7497,16 +8794,31 @@ static int hpsa_alloc_ioaccel_cmd_and_bft(struct ctlr_info *h) return 0; clean_up: - if (h->ioaccel_cmd_pool) + hpsa_free_ioaccel1_cmd_and_bft(h); + return -ENOMEM; +} + +/* Free ioaccel2 mode command blocks and block fetch table */ +static void hpsa_free_ioaccel2_cmd_and_bft(struct ctlr_info *h) +{ + hpsa_free_ioaccel2_sg_chain_blocks(h); + + if (h->ioaccel2_cmd_pool) { pci_free_consistent(h->pdev, - h->nr_cmds * sizeof(*h->ioaccel_cmd_pool), - h->ioaccel_cmd_pool, h->ioaccel_cmd_pool_dhandle); - kfree(h->ioaccel1_blockFetchTable); - return 1; + h->nr_cmds * sizeof(*h->ioaccel2_cmd_pool), + h->ioaccel2_cmd_pool, h->ioaccel2_cmd_pool_dhandle); + h->ioaccel2_cmd_pool = NULL; + h->ioaccel2_cmd_pool_dhandle = 0; + } + kfree(h->ioaccel2_blockFetchTable); + h->ioaccel2_blockFetchTable = NULL; } -static int ioaccel2_alloc_cmds_and_bft(struct ctlr_info *h) +/* Allocate ioaccel2 mode command blocks and block fetch table */ +static int hpsa_alloc_ioaccel2_cmd_and_bft(struct ctlr_info *h) { + int rc; + /* Allocate ioaccel2 mode command blocks and block fetch table */ h->ioaccel_maxsg = @@ -7526,7 +8838,13 @@ static int ioaccel2_alloc_cmds_and_bft(struct ctlr_info *h) sizeof(u32)), GFP_KERNEL); if ((h->ioaccel2_cmd_pool == NULL) || - (h->ioaccel2_blockFetchTable == NULL)) + (h->ioaccel2_blockFetchTable == NULL)) { + rc = -ENOMEM; + goto clean_up; + } + + rc = hpsa_allocate_ioaccel2_sg_chain_blocks(h); + if (rc) goto clean_up; memset(h->ioaccel2_cmd_pool, 0, @@ -7534,41 +8852,50 @@ static int ioaccel2_alloc_cmds_and_bft(struct ctlr_info *h) return 0; clean_up: - if (h->ioaccel2_cmd_pool) - pci_free_consistent(h->pdev, - h->nr_cmds * sizeof(*h->ioaccel2_cmd_pool), - h->ioaccel2_cmd_pool, h->ioaccel2_cmd_pool_dhandle); - kfree(h->ioaccel2_blockFetchTable); - return 1; + hpsa_free_ioaccel2_cmd_and_bft(h); + return rc; +} + +/* Free items allocated by hpsa_put_ctlr_into_performant_mode */ +static void hpsa_free_performant_mode(struct ctlr_info *h) +{ + kfree(h->blockFetchTable); + h->blockFetchTable = NULL; + hpsa_free_reply_queues(h); + hpsa_free_ioaccel1_cmd_and_bft(h); + hpsa_free_ioaccel2_cmd_and_bft(h); } -static void hpsa_put_ctlr_into_performant_mode(struct ctlr_info *h) +/* return -ENODEV on error, 0 on success (or no action) + * allocates numerous items that must be freed later + */ +static int hpsa_put_ctlr_into_performant_mode(struct ctlr_info *h) { u32 trans_support; unsigned long transMethod = CFGTBL_Trans_Performant | CFGTBL_Trans_use_short_tags; - int i; + int i, rc; if (hpsa_simple_mode) - return; + return 0; trans_support = readl(&(h->cfgtable->TransportSupport)); if (!(trans_support & PERFORMANT_MODE)) - return; + return 0; /* Check for I/O accelerator mode support */ if (trans_support & CFGTBL_Trans_io_accel1) { transMethod |= CFGTBL_Trans_io_accel1 | CFGTBL_Trans_enable_directed_msix; - if (hpsa_alloc_ioaccel_cmd_and_bft(h)) - goto clean_up; - } else { - if (trans_support & CFGTBL_Trans_io_accel2) { - transMethod |= CFGTBL_Trans_io_accel2 | + rc = hpsa_alloc_ioaccel1_cmd_and_bft(h); + if (rc) + return rc; + } else if (trans_support & CFGTBL_Trans_io_accel2) { + transMethod |= CFGTBL_Trans_io_accel2 | CFGTBL_Trans_enable_directed_msix; - if (ioaccel2_alloc_cmds_and_bft(h)) - goto clean_up; - } + rc = hpsa_alloc_ioaccel2_cmd_and_bft(h); + if (rc) + return rc; } h->nreply_queues = h->msix_vector > 0 ? h->msix_vector : 1; @@ -7580,8 +8907,10 @@ static void hpsa_put_ctlr_into_performant_mode(struct ctlr_info *h) h->reply_queue[i].head = pci_alloc_consistent(h->pdev, h->reply_queue_size, &(h->reply_queue[i].busaddr)); - if (!h->reply_queue[i].head) - goto clean_up; + if (!h->reply_queue[i].head) { + rc = -ENOMEM; + goto clean1; /* rq, ioaccel */ + } h->reply_queue[i].size = h->max_commands; h->reply_queue[i].wraparound = 1; /* spec: init to 1 */ h->reply_queue[i].current_entry = 0; @@ -7590,15 +8919,24 @@ static void hpsa_put_ctlr_into_performant_mode(struct ctlr_info *h) /* Need a block fetch table for performant mode */ h->blockFetchTable = kmalloc(((SG_ENTRIES_IN_CMD + 1) * sizeof(u32)), GFP_KERNEL); - if (!h->blockFetchTable) - goto clean_up; + if (!h->blockFetchTable) { + rc = -ENOMEM; + goto clean1; /* rq, ioaccel */ + } - hpsa_enter_performant_mode(h, trans_support); - return; + rc = hpsa_enter_performant_mode(h, trans_support); + if (rc) + goto clean2; /* bft, rq, ioaccel */ + return 0; -clean_up: - hpsa_free_reply_queues(h); +clean2: /* bft, rq, ioaccel */ kfree(h->blockFetchTable); + h->blockFetchTable = NULL; +clean1: /* rq, ioaccel */ + hpsa_free_reply_queues(h); + hpsa_free_ioaccel1_cmd_and_bft(h); + hpsa_free_ioaccel2_cmd_and_bft(h); + return rc; } static int is_accelerated_cmd(struct CommandList *c) @@ -7609,17 +8947,18 @@ static int is_accelerated_cmd(struct CommandList *c) static void hpsa_drain_accel_commands(struct ctlr_info *h) { struct CommandList *c = NULL; - unsigned long flags; - int accel_cmds_out; + int i, accel_cmds_out; + int refcount; - do { /* wait for all outstanding commands to drain out */ + do { /* wait for all outstanding ioaccel commands to drain out */ accel_cmds_out = 0; - spin_lock_irqsave(&h->lock, flags); - list_for_each_entry(c, &h->cmpQ, list) - accel_cmds_out += is_accelerated_cmd(c); - list_for_each_entry(c, &h->reqQ, list) - accel_cmds_out += is_accelerated_cmd(c); - spin_unlock_irqrestore(&h->lock, flags); + for (i = 0; i < h->nr_cmds; i++) { + c = h->cmd_pool + i; + refcount = atomic_inc_return(&c->refcount); + if (refcount > 1) /* Command is allocated */ + accel_cmds_out += is_accelerated_cmd(c); + cmd_free(h, c); + } if (accel_cmds_out <= 0) break; msleep(100); @@ -7711,7 +9050,7 @@ static void __attribute__((unused)) verify_offsets(void) VERIFY_OFFSET(timeout_sec, 0x62); VERIFY_OFFSET(ReplyQueue, 0x64); VERIFY_OFFSET(reserved9, 0x65); - VERIFY_OFFSET(Tag, 0x68); + VERIFY_OFFSET(tag, 0x68); VERIFY_OFFSET(host_addr, 0x70); VERIFY_OFFSET(CISS_LUN, 0x78); VERIFY_OFFSET(SG, 0x78 + 8); diff --git a/drivers/scsi/hpsa.h b/drivers/scsi/hpsa.h index 24472cec7de34d..5643d55ac5a805 100644 --- a/drivers/scsi/hpsa.h +++ b/drivers/scsi/hpsa.h @@ -32,7 +32,6 @@ struct access_method { void (*submit_command)(struct ctlr_info *h, struct CommandList *c); void (*set_intr_mask)(struct ctlr_info *h, unsigned long val); - unsigned long (*fifo_full)(struct ctlr_info *h); bool (*intr_pending)(struct ctlr_info *h); unsigned long (*command_completed)(struct ctlr_info *h, u8 q); }; @@ -47,14 +46,36 @@ struct hpsa_scsi_dev_t { unsigned char model[16]; /* bytes 16-31 of inquiry data */ unsigned char raid_level; /* from inquiry page 0xC1 */ unsigned char volume_offline; /* discovered via TUR or VPD */ + u16 queue_depth; /* max queue_depth for this device */ + atomic_t ioaccel_cmds_out; /* Only used for physical devices + * counts commands sent to physical + * device via "ioaccel" path. + */ u32 ioaccel_handle; int offload_config; /* I/O accel RAID offload configured */ int offload_enabled; /* I/O accel RAID offload enabled */ + int offload_to_be_enabled; + int hba_ioaccel_enabled; int offload_to_mirror; /* Send next I/O accelerator RAID * offload request to mirror drive */ struct raid_map_data raid_map; /* I/O accelerator RAID map */ - + int supports_aborts; +#define HPSA_DO_NOT_EXPOSE 0x0 +#define HPSA_SG_ATTACH 0x1 +#define HPSA_ULD_ATTACH 0x2 +#define HPSA_SCSI_ADD (HPSA_SG_ATTACH | HPSA_ULD_ATTACH) + u8 expose_state; + + /* Pointers from logical drive map indices to the phys drives that + * make those logical drives. Note, multiple logical drives may + * share physical drives. You can have for instance 5 physical + * drives with 3 logical drives each using those same 5 physical + * disks. We need these pointers for counting i/o's out to physical + * devices in order to honor physical device queue depth limits. + */ + struct hpsa_scsi_dev_t *phys_disk[RAID_MAP_MAX_ENTRIES]; + int nphysical_disks; }; struct reply_queue_buffer { @@ -115,10 +136,12 @@ struct ctlr_info { void __iomem *vaddr; unsigned long paddr; int nr_cmds; /* Number of commands allowed on this controller */ +#define HPSA_CMDS_RESERVED_FOR_ABORTS 2 +#define HPSA_CMDS_RESERVED_FOR_DRIVER 1 struct CfgTable __iomem *cfgtable; int interrupts_enabled; int max_commands; - int commands_outstanding; + atomic_t commands_outstanding; # define PERF_MODE_INT 0 # define DOORBELL_INT 1 # define SIMPLE_MODE_INT 2 @@ -131,8 +154,6 @@ struct ctlr_info { char hba_mode_enabled; /* queue and queue Info */ - struct list_head reqQ; - struct list_head cmpQ; unsigned int Qdepth; unsigned int maxSG; spinlock_t lock; @@ -140,6 +161,7 @@ struct ctlr_info { u8 max_cmd_sg_entries; int chainsize; struct SGDescriptor **cmd_sg_list; + struct ioaccel2_sg_element **ioaccel2_cmd_sg_list; /* pointers to command and error info pool */ struct CommandList *cmd_pool; @@ -168,9 +190,8 @@ struct ctlr_info { unsigned long transMethod; /* cap concurrent passthrus at some reasonable maximum */ -#define HPSA_MAX_CONCURRENT_PASSTHRUS (20) - spinlock_t passthru_count_lock; /* protects passthru_count */ - int passthru_count; +#define HPSA_MAX_CONCURRENT_PASSTHRUS (10) + atomic_t passthru_cmds_avail; /* * Performant mode completion buffers @@ -194,10 +215,11 @@ struct ctlr_info { atomic_t firmware_flash_in_progress; u32 *lockup_detected; struct delayed_work monitor_ctlr_work; + struct delayed_work rescan_ctlr_work; int remove_in_progress; - u32 fifo_recently_full; /* Address of h->q[x] is passed to intr handler to know which queue */ u8 q[MAX_REPLY_QUEUES]; + char intrname[MAX_REPLY_QUEUES][16]; /* "hpsa0-msix00" names */ u32 TMFSupportFlags; /* cache what task mgmt funcs are supported. */ #define HPSATMF_BITS_SUPPORTED (1 << 0) #define HPSATMF_PHYS_LUN_RESET (1 << 1) @@ -209,6 +231,7 @@ struct ctlr_info { #define HPSATMF_PHYS_QRY_TASK (1 << 7) #define HPSATMF_PHYS_QRY_TSET (1 << 8) #define HPSATMF_PHYS_QRY_ASYNC (1 << 9) +#define HPSATMF_IOACCEL_ENABLED (1 << 15) #define HPSATMF_MASK_SUPPORTED (1 << 16) #define HPSATMF_LOG_LUN_RESET (1 << 17) #define HPSATMF_LOG_NEX_RESET (1 << 18) @@ -237,8 +260,16 @@ struct ctlr_info { spinlock_t offline_device_lock; struct list_head offline_device_list; int acciopath_status; - int drv_req_rescan; /* flag for driver to request rescan event */ int raid_offload_debug; + int lockup_detector_enabled; + int needs_abort_tags_swizzled; + struct workqueue_struct *resubmit_wq; + atomic_t abort_cmds_available; + wait_queue_head_t abort_cmd_wait_queue; + wait_queue_head_t abort_sync_wait_queue; + atomic_t cmds_sent; + int abort_test; + int abort_timeout; }; struct offline_device_entry { @@ -297,6 +328,8 @@ struct offline_device_entry { */ #define SA5_DOORBELL 0x20 #define SA5_REQUEST_PORT_OFFSET 0x40 +#define SA5_REQUEST_PORT64_LO_OFFSET 0xC0 +#define SA5_REQUEST_PORT64_HI_OFFSET 0xC4 #define SA5_REPLY_INTR_MASK_OFFSET 0x34 #define SA5_REPLY_PORT_OFFSET 0x44 #define SA5_INTR_STATUS 0x30 @@ -353,10 +386,7 @@ static void SA5_submit_command_no_read(struct ctlr_info *h, static void SA5_submit_command_ioaccel2(struct ctlr_info *h, struct CommandList *c) { - if (c->cmd_type == CMD_IOACCEL2) - writel(c->busaddr, h->vaddr + IOACCEL2_INBOUND_POSTQ_32); - else - writel(c->busaddr, h->vaddr + SA5_REQUEST_PORT_OFFSET); + writel(c->busaddr, h->vaddr + SA5_REQUEST_PORT_OFFSET); } /* @@ -395,27 +425,25 @@ static void SA5_performant_intr_mask(struct ctlr_info *h, unsigned long val) static unsigned long SA5_performant_completed(struct ctlr_info *h, u8 q) { struct reply_queue_buffer *rq = &h->reply_queue[q]; - unsigned long flags, register_value = FIFO_EMPTY; + unsigned long register_value = FIFO_EMPTY; /* msi auto clears the interrupt pending bit. */ - if (!(h->msi_vector || h->msix_vector)) { + if (unlikely(!(h->msi_vector || h->msix_vector))) { /* flush the controller write of the reply queue by reading * outbound doorbell status register. */ - register_value = readl(h->vaddr + SA5_OUTDB_STATUS); + (void) readl(h->vaddr + SA5_OUTDB_STATUS); writel(SA5_OUTDB_CLEAR_PERF_BIT, h->vaddr + SA5_OUTDB_CLEAR); /* Do a read in order to flush the write to the controller * (as per spec.) */ - register_value = readl(h->vaddr + SA5_OUTDB_STATUS); + (void) readl(h->vaddr + SA5_OUTDB_STATUS); } - if ((rq->head[rq->current_entry] & 1) == rq->wraparound) { + if ((((u32) rq->head[rq->current_entry]) & 1) == rq->wraparound) { register_value = rq->head[rq->current_entry]; rq->current_entry++; - spin_lock_irqsave(&h->lock, flags); - h->commands_outstanding--; - spin_unlock_irqrestore(&h->lock, flags); + atomic_dec(&h->commands_outstanding); } else { register_value = FIFO_EMPTY; } @@ -427,18 +455,6 @@ static unsigned long SA5_performant_completed(struct ctlr_info *h, u8 q) return register_value; } -/* - * Returns true if fifo is full. - * - */ -static unsigned long SA5_fifo_full(struct ctlr_info *h) -{ - if (h->commands_outstanding >= h->max_commands) - return 1; - else - return 0; - -} /* * returns value read from hardware. * returns FIFO_EMPTY if there is nothing to read @@ -448,13 +464,9 @@ static unsigned long SA5_completed(struct ctlr_info *h, { unsigned long register_value = readl(h->vaddr + SA5_REPLY_PORT_OFFSET); - unsigned long flags; - if (register_value != FIFO_EMPTY) { - spin_lock_irqsave(&h->lock, flags); - h->commands_outstanding--; - spin_unlock_irqrestore(&h->lock, flags); - } + if (register_value != FIFO_EMPTY) + atomic_dec(&h->commands_outstanding); #ifdef HPSA_DEBUG if (register_value != FIFO_EMPTY) @@ -483,9 +495,6 @@ static bool SA5_performant_intr_pending(struct ctlr_info *h) if (!register_value) return false; - if (h->msi_vector || h->msix_vector) - return true; - /* Read outbound doorbell to flush */ register_value = readl(h->vaddr + SA5_OUTDB_STATUS); return register_value & SA5_OUTDB_STATUS_PERF_BIT; @@ -510,7 +519,6 @@ static unsigned long SA5_ioaccel_mode1_completed(struct ctlr_info *h, u8 q) { u64 register_value; struct reply_queue_buffer *rq = &h->reply_queue[q]; - unsigned long flags; BUG_ON(q >= h->nreply_queues); @@ -528,9 +536,7 @@ static unsigned long SA5_ioaccel_mode1_completed(struct ctlr_info *h, u8 q) wmb(); writel((q << 24) | rq->current_entry, h->vaddr + IOACCEL_MODE1_CONSUMER_INDEX); - spin_lock_irqsave(&h->lock, flags); - h->commands_outstanding--; - spin_unlock_irqrestore(&h->lock, flags); + atomic_dec(&h->commands_outstanding); } return (unsigned long) register_value; } @@ -538,7 +544,6 @@ static unsigned long SA5_ioaccel_mode1_completed(struct ctlr_info *h, u8 q) static struct access_method SA5_access = { SA5_submit_command, SA5_intr_mask, - SA5_fifo_full, SA5_intr_pending, SA5_completed, }; @@ -546,7 +551,6 @@ static struct access_method SA5_access = { static struct access_method SA5_ioaccel_mode1_access = { SA5_submit_command, SA5_performant_intr_mask, - SA5_fifo_full, SA5_ioaccel_mode1_intr_pending, SA5_ioaccel_mode1_completed, }; @@ -554,7 +558,6 @@ static struct access_method SA5_ioaccel_mode1_access = { static struct access_method SA5_ioaccel_mode2_access = { SA5_submit_command_ioaccel2, SA5_performant_intr_mask, - SA5_fifo_full, SA5_performant_intr_pending, SA5_performant_completed, }; @@ -562,7 +565,6 @@ static struct access_method SA5_ioaccel_mode2_access = { static struct access_method SA5_performant_access = { SA5_submit_command, SA5_performant_intr_mask, - SA5_fifo_full, SA5_performant_intr_pending, SA5_performant_completed, }; @@ -570,7 +572,6 @@ static struct access_method SA5_performant_access = { static struct access_method SA5_performant_access_no_read = { SA5_submit_command_no_read, SA5_performant_intr_mask, - SA5_fifo_full, SA5_performant_intr_pending, SA5_performant_completed, }; diff --git a/drivers/scsi/hpsa_cmd.h b/drivers/scsi/hpsa_cmd.h index b5125dc3143912..5ab1b13f996f83 100644 --- a/drivers/scsi/hpsa_cmd.h +++ b/drivers/scsi/hpsa_cmd.h @@ -42,8 +42,22 @@ #define CMD_UNSOLICITED_ABORT 0x000A #define CMD_TIMEOUT 0x000B #define CMD_UNABORTABLE 0x000C +#define CMD_TMF_STATUS 0x000D #define CMD_IOACCEL_DISABLED 0x000E +#define CMD_CTLR_LOCKUP 0xffff +/* Note: CMD_CTLR_LOCKUP is not a value defined by the CISS spec + * it is a value defined by the driver that commands can be marked + * with when a controller lockup has been detected by the driver + */ +/* TMF function status values */ +#define CISS_TMF_COMPLETE 0x00 +#define CISS_TMF_INVALID_FRAME 0x02 +#define CISS_TMF_NOT_SUPPORTED 0x04 +#define CISS_TMF_FAILED 0x05 +#define CISS_TMF_SUCCESS 0x08 +#define CISS_TMF_WRONG_LUN 0x09 +#define CISS_TMF_OVERLAPPED_TAG 0x0a /* Unit Attentions ASC's as defined for the MSA2012sa */ #define POWER_OR_RESET 0x29 @@ -240,9 +254,16 @@ struct ReportLUNdata { struct ext_report_lun_entry { u8 lunid[8]; +#define MASKED_DEVICE(x) ((x)[3] & 0xC0) +#define GET_BMIC_BUS(lunid) ((lunid)[7] & 0x3F) +#define GET_BMIC_LEVEL_TWO_TARGET(lunid) ((lunid)[6]) +#define GET_BMIC_DRIVE_NUMBER(lunid) (((GET_BMIC_BUS((lunid)) - 1) << 8) + \ + GET_BMIC_LEVEL_TWO_TARGET((lunid))) u8 wwid[8]; u8 device_type; u8 device_flags; +#define NON_DISK_PHYS_DEV(x) ((x)[17] & 0x01) +#define PHYS_IOACCEL(x) ((x)[17] & 0x08) u8 lun_count; /* multi-lun device, how many luns */ u8 redundant_paths; u32 ioaccel_handle; /* ioaccel1 only uses lower 16 bits */ @@ -252,7 +273,7 @@ struct ReportExtendedLUNdata { u8 LUNListLength[4]; u8 extended_response_flag; u8 reserved[3]; - struct ext_report_lun_entry LUN[HPSA_MAX_LUN]; + struct ext_report_lun_entry LUN[HPSA_MAX_PHYS_LUN]; }; struct SenseSubsystem_info { @@ -268,6 +289,7 @@ struct SenseSubsystem_info { #define HPSA_CACHE_FLUSH 0x01 /* C2 was already being used by HPSA */ #define BMIC_FLASH_FIRMWARE 0xF7 #define BMIC_SENSE_CONTROLLER_PARAMETERS 0x64 +#define BMIC_IDENTIFY_PHYSICAL_DEVICE 0x15 /* Command List Structure */ union SCSI3Addr { @@ -314,28 +336,36 @@ struct CommandListHeader { u8 ReplyQueue; u8 SGList; u16 SGTotal; - struct vals32 Tag; + u64 tag; union LUNAddr LUN; }; struct RequestBlock { u8 CDBLen; - struct { - u8 Type:3; - u8 Attribute:3; - u8 Direction:2; - } Type; + /* + * type_attr_dir: + * type: low 3 bits + * attr: middle 3 bits + * dir: high 2 bits + */ + u8 type_attr_dir; +#define TYPE_ATTR_DIR(t, a, d) ((((d) & 0x03) << 6) |\ + (((a) & 0x07) << 3) |\ + ((t) & 0x07)) +#define GET_TYPE(tad) ((tad) & 0x07) +#define GET_ATTR(tad) (((tad) >> 3) & 0x07) +#define GET_DIR(tad) (((tad) >> 6) & 0x03) u16 Timeout; u8 CDB[16]; }; struct ErrDescriptor { - struct vals32 Addr; + u64 Addr; u32 Len; }; struct SGDescriptor { - struct vals32 Addr; + u64 Addr; u32 Len; u32 Ext; }; @@ -366,23 +396,21 @@ struct ErrorInfo { #define CMD_SCSI 0x03 #define CMD_IOACCEL1 0x04 #define CMD_IOACCEL2 0x05 +#define IOACCEL2_TMF 0x06 -#define DIRECT_LOOKUP_SHIFT 5 -#define DIRECT_LOOKUP_BIT 0x10 +#define DIRECT_LOOKUP_SHIFT 4 #define DIRECT_LOOKUP_MASK (~((1 << DIRECT_LOOKUP_SHIFT) - 1)) #define HPSA_ERROR_BIT 0x02 struct ctlr_info; /* defined in hpsa.h */ -/* The size of this structure needs to be divisible by 32 - * on all architectures because low 5 bits of the addresses +/* The size of this structure needs to be divisible by 128 + * on all architectures. The low 4 bits of the addresses * are used as follows: * * bit 0: to device, used to indicate "performant mode" command * from device, indidcates error status. * bit 1-3: to device, indicates block fetch table entry for * reducing DMA in fetching commands from host memory. - * bit 4: used to indicate whether tag is "direct lookup" (index), - * or a bus address. */ #define COMMANDLIST_ALIGNMENT 128 @@ -397,9 +425,23 @@ struct CommandList { struct ctlr_info *h; int cmd_type; long cmdindex; - struct list_head list; struct completion *waiting; - void *scsi_cmd; + struct scsi_cmnd *scsi_cmd; + struct work_struct work; + struct delayed_work abort_torture_work; + + /* For commands using either of the two "ioaccel" paths to + * bypass the RAID stack and go directly to the physical disk + * phys_disk is a pointer to the hpsa_scsi_dev_t to which the + * i/o is destined. We need to store that here because the command + * may potentially encounter TASK SET FULL and need to be resubmitted + * For "normal" i/o's not using the "ioaccel" paths, phys_disk is + * not used. + */ + struct hpsa_scsi_dev_t *phys_disk; + + int abort_pending; + atomic_t refcount; /* Must be last to avoid memset in cmd_alloc */ } __aligned(COMMANDLIST_ALIGNMENT); /* Max S/G elements in I/O accelerator command */ @@ -434,8 +476,8 @@ struct io_accel1_cmd { u16 timeout_sec; /* 0x62 - 0x63 */ u8 ReplyQueue; /* 0x64 */ u8 reserved9[3]; /* 0x65 - 0x67 */ - struct vals32 Tag; /* 0x68 - 0x6F */ - struct vals32 host_addr; /* 0x70 - 0x77 */ + u64 tag; /* 0x68 - 0x6F */ + u64 host_addr; /* 0x70 - 0x77 */ u8 CISS_LUN[8]; /* 0x78 - 0x7F */ struct SGDescriptor SG[IOACCEL1_MAXSGENTRIES]; } __aligned(IOACCEL1_COMMANDLIST_ALIGNMENT); @@ -493,6 +535,12 @@ struct io_accel2_scsi_response { #define IOACCEL2_STATUS_SR_TASK_COMP_SET_FULL 0x28 #define IOACCEL2_STATUS_SR_TASK_COMP_ABORTED 0x40 #define IOACCEL2_STATUS_SR_IOACCEL_DISABLED 0x0E +#define IOACCEL2_STATUS_SR_IO_ERROR 0x01 +#define IOACCEL2_STATUS_SR_IO_ABORTED 0x02 +#define IOACCEL2_STATUS_SR_NO_PATH_TO_DEVICE 0x03 +#define IOACCEL2_STATUS_SR_INVALID_DEVICE 0x04 +#define IOACCEL2_STATUS_SR_UNDERRUN 0x51 +#define IOACCEL2_STATUS_SR_OVERRUN 0x75 u8 data_present; /* low 2 bits */ #define IOACCEL2_NO_DATAPRESENT 0x000 #define IOACCEL2_RESPONSE_DATAPRESENT 0x001 @@ -545,6 +593,7 @@ struct io_accel2_cmd { #define IOACCEL2_DIR_NO_DATA 0x00 #define IOACCEL2_DIR_DATA_IN 0x01 #define IOACCEL2_DIR_DATA_OUT 0x02 +#define IOACCEL2_TMF_ABORT 0x01 /* * SCSI Task Management Request format for Accelerator Mode 2 */ @@ -555,11 +604,11 @@ struct hpsa_tmf_struct { u8 reserved1; /* byte 3 Reserved */ u32 it_nexus; /* SCSI I-T Nexus */ u8 lun_id[8]; /* LUN ID for TMF request */ - struct vals32 Tag; /* cciss tag associated w/ request */ - struct vals32 abort_tag;/* cciss tag of SCSI cmd or task to abort */ + u64 tag; /* cciss tag associated w/ request */ + u64 abort_tag; /* cciss tag of SCSI cmd or task to abort */ u64 error_ptr; /* Error Pointer */ u32 error_len; /* Error Length */ -}; +} __aligned(IOACCEL2_COMMANDLIST_ALIGNMENT); /* Configuration Table Structure */ struct HostWrite { @@ -636,5 +685,137 @@ struct hpsa_pci_info { u32 board_id; }; +struct bmic_identify_physical_device { + u8 scsi_bus; /* SCSI Bus number on controller */ + u8 scsi_id; /* SCSI ID on this bus */ + u16 block_size; /* sector size in bytes */ + u32 total_blocks; /* number for sectors on drive */ + u32 reserved_blocks; /* controller reserved (RIS) */ + u8 model[40]; /* Physical Drive Model */ + u8 serial_number[40]; /* Drive Serial Number */ + u8 firmware_revision[8]; /* drive firmware revision */ + u8 scsi_inquiry_bits; /* inquiry byte 7 bits */ + u8 compaq_drive_stamp; /* 0 means drive not stamped */ + u8 last_failure_reason; +#define BMIC_LAST_FAILURE_TOO_SMALL_IN_LOAD_CONFIG 0x01 +#define BMIC_LAST_FAILURE_ERROR_ERASING_RIS 0x02 +#define BMIC_LAST_FAILURE_ERROR_SAVING_RIS 0x03 +#define BMIC_LAST_FAILURE_FAIL_DRIVE_COMMAND 0x04 +#define BMIC_LAST_FAILURE_MARK_BAD_FAILED 0x05 +#define BMIC_LAST_FAILURE_MARK_BAD_FAILED_IN_FINISH_REMAP 0x06 +#define BMIC_LAST_FAILURE_TIMEOUT 0x07 +#define BMIC_LAST_FAILURE_AUTOSENSE_FAILED 0x08 +#define BMIC_LAST_FAILURE_MEDIUM_ERROR_1 0x09 +#define BMIC_LAST_FAILURE_MEDIUM_ERROR_2 0x0a +#define BMIC_LAST_FAILURE_NOT_READY_BAD_SENSE 0x0b +#define BMIC_LAST_FAILURE_NOT_READY 0x0c +#define BMIC_LAST_FAILURE_HARDWARE_ERROR 0x0d +#define BMIC_LAST_FAILURE_ABORTED_COMMAND 0x0e +#define BMIC_LAST_FAILURE_WRITE_PROTECTED 0x0f +#define BMIC_LAST_FAILURE_SPIN_UP_FAILURE_IN_RECOVER 0x10 +#define BMIC_LAST_FAILURE_REBUILD_WRITE_ERROR 0x11 +#define BMIC_LAST_FAILURE_TOO_SMALL_IN_HOT_PLUG 0x12 +#define BMIC_LAST_FAILURE_BUS_RESET_RECOVERY_ABORTED 0x13 +#define BMIC_LAST_FAILURE_REMOVED_IN_HOT_PLUG 0x14 +#define BMIC_LAST_FAILURE_INIT_REQUEST_SENSE_FAILED 0x15 +#define BMIC_LAST_FAILURE_INIT_START_UNIT_FAILED 0x16 +#define BMIC_LAST_FAILURE_INQUIRY_FAILED 0x17 +#define BMIC_LAST_FAILURE_NON_DISK_DEVICE 0x18 +#define BMIC_LAST_FAILURE_READ_CAPACITY_FAILED 0x19 +#define BMIC_LAST_FAILURE_INVALID_BLOCK_SIZE 0x1a +#define BMIC_LAST_FAILURE_HOT_PLUG_REQUEST_SENSE_FAILED 0x1b +#define BMIC_LAST_FAILURE_HOT_PLUG_START_UNIT_FAILED 0x1c +#define BMIC_LAST_FAILURE_WRITE_ERROR_AFTER_REMAP 0x1d +#define BMIC_LAST_FAILURE_INIT_RESET_RECOVERY_ABORTED 0x1e +#define BMIC_LAST_FAILURE_DEFERRED_WRITE_ERROR 0x1f +#define BMIC_LAST_FAILURE_MISSING_IN_SAVE_RIS 0x20 +#define BMIC_LAST_FAILURE_WRONG_REPLACE 0x21 +#define BMIC_LAST_FAILURE_GDP_VPD_INQUIRY_FAILED 0x22 +#define BMIC_LAST_FAILURE_GDP_MODE_SENSE_FAILED 0x23 +#define BMIC_LAST_FAILURE_DRIVE_NOT_IN_48BIT_MODE 0x24 +#define BMIC_LAST_FAILURE_DRIVE_TYPE_MIX_IN_HOT_PLUG 0x25 +#define BMIC_LAST_FAILURE_DRIVE_TYPE_MIX_IN_LOAD_CFG 0x26 +#define BMIC_LAST_FAILURE_PROTOCOL_ADAPTER_FAILED 0x27 +#define BMIC_LAST_FAILURE_FAULTY_ID_BAY_EMPTY 0x28 +#define BMIC_LAST_FAILURE_FAULTY_ID_BAY_OCCUPIED 0x29 +#define BMIC_LAST_FAILURE_FAULTY_ID_INVALID_BAY 0x2a +#define BMIC_LAST_FAILURE_WRITE_RETRIES_FAILED 0x2b + +#define BMIC_LAST_FAILURE_SMART_ERROR_REPORTED 0x37 +#define BMIC_LAST_FAILURE_PHY_RESET_FAILED 0x38 +#define BMIC_LAST_FAILURE_ONLY_ONE_CTLR_CAN_SEE_DRIVE 0x40 +#define BMIC_LAST_FAILURE_KC_VOLUME_FAILED 0x41 +#define BMIC_LAST_FAILURE_UNEXPECTED_REPLACEMENT 0x42 +#define BMIC_LAST_FAILURE_OFFLINE_ERASE 0x80 +#define BMIC_LAST_FAILURE_OFFLINE_TOO_SMALL 0x81 +#define BMIC_LAST_FAILURE_OFFLINE_DRIVE_TYPE_MIX 0x82 +#define BMIC_LAST_FAILURE_OFFLINE_ERASE_COMPLETE 0x83 + + u8 flags; + u8 more_flags; + u8 scsi_lun; /* SCSI LUN for phys drive */ + u8 yet_more_flags; + u8 even_more_flags; + u32 spi_speed_rules;/* SPI Speed data:Ultra disable diagnose */ + u8 phys_connector[2]; /* connector number on controller */ + u8 phys_box_on_bus; /* phys enclosure this drive resides */ + u8 phys_bay_in_box; /* phys drv bay this drive resides */ + u32 rpm; /* Drive rotational speed in rpm */ + u8 device_type; /* type of drive */ + u8 sata_version; /* only valid when drive_type is SATA */ + u64 big_total_block_count; + u64 ris_starting_lba; + u32 ris_size; + u8 wwid[20]; + u8 controller_phy_map[32]; + u16 phy_count; + u8 phy_connected_dev_type[256]; + u8 phy_to_drive_bay_num[256]; + u16 phy_to_attached_dev_index[256]; + u8 box_index; + u8 reserved; + u16 extra_physical_drive_flags; +#define BMIC_PHYS_DRIVE_SUPPORTS_GAS_GAUGE(idphydrv) \ + (idphydrv->extra_physical_drive_flags & (1 << 10)) + u8 negotiated_link_rate[256]; + u8 phy_to_phy_map[256]; + u8 redundant_path_present_map; + u8 redundant_path_failure_map; + u8 active_path_number; + u16 alternate_paths_phys_connector[8]; + u8 alternate_paths_phys_box_on_port[8]; + u8 multi_lun_device_lun_count; + u8 minimum_good_fw_revision[8]; + u8 unique_inquiry_bytes[20]; + u8 current_temperature_degreesC; + u8 temperature_threshold_degreesC; + u8 max_temperature_degreesC; + u8 logical_blocks_per_phys_block_exp; /* phyblocksize = 512 * 2^exp */ + u16 current_queue_depth_limit; + u8 switch_name[10]; + u16 switch_port; + u8 alternate_paths_switch_name[40]; + u8 alternate_paths_switch_port[8]; + u16 power_on_hours; /* valid only if gas gauge supported */ + u16 percent_endurance_used; /* valid only if gas gauge supported. */ +#define BMIC_PHYS_DRIVE_SSD_WEAROUT(idphydrv) \ + ((idphydrv->percent_endurance_used & 0x80) || \ + (idphydrv->percent_endurance_used > 10000)) + u8 drive_authentication; +#define BMIC_PHYS_DRIVE_AUTHENTICATED(idphydrv) \ + (idphydrv->drive_authentication == 0x80) + u8 smart_carrier_authentication; +#define BMIC_SMART_CARRIER_AUTHENTICATION_SUPPORTED(idphydrv) \ + (idphydrv->smart_carrier_authentication != 0x0) +#define BMIC_SMART_CARRIER_AUTHENTICATED(idphydrv) \ + (idphydrv->smart_carrier_authentication == 0x01) + u8 smart_carrier_app_fw_version; + u8 smart_carrier_bootloader_fw_version; + u8 encryption_key_name[64]; + u32 misc_drive_flags; + u16 dek_index; + u8 padding[112]; +}; + #pragma pack() #endif /* HPSA_CMD_H */ diff --git a/drivers/scsi/scsi.c b/drivers/scsi/scsi.c index 79c77b485a6729..362e443e7d5eb1 100644 --- a/drivers/scsi/scsi.c +++ b/drivers/scsi/scsi.c @@ -80,7 +80,10 @@ * Note - the initial logging level can be set here to log events at boot time. * After the system is up, you may enable logging via the /proc interface. */ -unsigned int scsi_logging_level; +unsigned int scsi_logging_level = + (0x5 & ((1 << SCSI_LOG_ERROR_BITS) - 1)) << SCSI_LOG_ERROR_SHIFT | + (0x1 & ((1 << SCSI_LOG_MLCOMPLETE_BITS) - 1)) << SCSI_LOG_MLCOMPLETE_SHIFT; + #if defined(CONFIG_SCSI_LOGGING) EXPORT_SYMBOL(scsi_logging_level); #endif @@ -553,6 +556,23 @@ void scsi_log_send(struct scsi_cmnd *cmd) } } +/* Strings for internal return values in scsi.h */ +/* NEEDS_RETRY must be the lowest numbered value */ +static const char * const disposition_label[] = { + "NEEDS_RETRY", + "SUCCESS", + "FAILED", + "QUEUED", + "SOFT_ERROR", + "ADD_TO_MLQUEUE", + "TIMEOUT_ERROR", + "SCSI_RETURN_NOT_HANDLED", + "FAST_IO_FAIL", + "UNKNOWN", +}; +#define DISPOSITION_BASE NEEDS_RETRY +#define DISPOSITION_UNKNOWN (ARRAY_SIZE(disposition_label) - 1) + void scsi_log_completion(struct scsi_cmnd *cmd, int disposition) { unsigned int level; @@ -574,35 +594,21 @@ void scsi_log_completion(struct scsi_cmnd *cmd, int disposition) SCSI_LOG_MLCOMPLETE_BITS); if (((level > 0) && (cmd->result || disposition != SUCCESS)) || (level > 1)) { - scmd_printk(KERN_INFO, cmd, "Done: "); + int dindex; + + if (disposition >= DISPOSITION_BASE && + disposition <= DISPOSITION_BASE + + DISPOSITION_UNKNOWN) + dindex = disposition - DISPOSITION_BASE; + else + dindex = DISPOSITION_UNKNOWN; if (level > 2) - printk("0x%p ", cmd); - /* - * Dump truncated values, so we usually fit within - * 80 chars. - */ - switch (disposition) { - case SUCCESS: - printk("SUCCESS\n"); - break; - case NEEDS_RETRY: - printk("RETRY\n"); - break; - case ADD_TO_MLQUEUE: - printk("MLQUEUE\n"); - break; - case FAILED: - printk("FAILED\n"); - break; - case TIMEOUT_ERROR: - /* - * If called via scsi_times_out. - */ - printk("TIMEOUT\n"); - break; - default: - printk("UNKNOWN\n"); - } + scmd_printk(KERN_INFO, cmd, "Done: 0x%p %s\n", + cmd, + disposition_label[dindex]); + else + scmd_printk(KERN_INFO, cmd, "Done: %s\n", + disposition_label[dindex]); scsi_print_result(cmd); scsi_print_command(cmd); if (status_byte(cmd->result) & CHECK_CONDITION)