diff options
Diffstat (limited to 'drivers/block/cciss.c')
| -rw-r--r-- | drivers/block/cciss.c | 2165 |
1 files changed, 1281 insertions, 884 deletions
diff --git a/drivers/block/cciss.c b/drivers/block/cciss.c index e1e7143ca1e3..31064df1370a 100644 --- a/drivers/block/cciss.c +++ b/drivers/block/cciss.c @@ -56,16 +56,14 @@ #include <linux/kthread.h> #define CCISS_DRIVER_VERSION(maj,min,submin) ((maj<<16)|(min<<8)|(submin)) -#define DRIVER_NAME "HP CISS Driver (v 3.6.20)" -#define DRIVER_VERSION CCISS_DRIVER_VERSION(3, 6, 20) +#define DRIVER_NAME "HP CISS Driver (v 3.6.26)" +#define DRIVER_VERSION CCISS_DRIVER_VERSION(3, 6, 26) /* Embedded module documentation macros - see modules.h */ MODULE_AUTHOR("Hewlett-Packard Company"); MODULE_DESCRIPTION("Driver for HP Smart Array Controllers"); -MODULE_SUPPORTED_DEVICE("HP SA5i SA5i+ SA532 SA5300 SA5312 SA641 SA642 SA6400" - " SA6i P600 P800 P400 P400i E200 E200i E500 P700m" - " Smart Array G2 Series SAS/SATA Controllers"); -MODULE_VERSION("3.6.20"); +MODULE_SUPPORTED_DEVICE("HP Smart Array Controllers"); +MODULE_VERSION("3.6.26"); MODULE_LICENSE("GPL"); static int cciss_allow_hpsa; @@ -107,6 +105,11 @@ static const struct pci_device_id cciss_pci_device_id[] = { {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x3249}, {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x324A}, {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x324B}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x3250}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x3251}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x3252}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x3253}, + {PCI_VENDOR_ID_HP, PCI_DEVICE_ID_HP_CISSE, 0x103C, 0x3254}, {0,} }; @@ -146,6 +149,11 @@ static struct board_type products[] = { {0x3249103C, "Smart Array P812", &SA5_access}, {0x324A103C, "Smart Array P712m", &SA5_access}, {0x324B103C, "Smart Array P711m", &SA5_access}, + {0x3250103C, "Smart Array", &SA5_access}, + {0x3251103C, "Smart Array", &SA5_access}, + {0x3252103C, "Smart Array", &SA5_access}, + {0x3253103C, "Smart Array", &SA5_access}, + {0x3254103C, "Smart Array", &SA5_access}, }; /* How long to wait (in milliseconds) for board to go into simple mode */ @@ -167,9 +175,13 @@ static DEFINE_MUTEX(scan_mutex); static LIST_HEAD(scan_q); static void do_cciss_request(struct request_queue *q); -static irqreturn_t do_cciss_intr(int irq, void *dev_id); +static irqreturn_t do_cciss_intx(int irq, void *dev_id); +static irqreturn_t do_cciss_msix_intr(int irq, void *dev_id); static int cciss_open(struct block_device *bdev, fmode_t mode); +static int cciss_unlocked_open(struct block_device *bdev, fmode_t mode); static int cciss_release(struct gendisk *disk, fmode_t mode); +static int do_ioctl(struct block_device *bdev, fmode_t mode, + unsigned int cmd, unsigned long arg); static int cciss_ioctl(struct block_device *bdev, fmode_t mode, unsigned int cmd, unsigned long arg); static int cciss_getgeo(struct block_device *bdev, struct hd_geometry *geo); @@ -179,25 +191,23 @@ static int rebuild_lun_table(ctlr_info_t *h, int first_time, int via_ioctl); static int deregister_disk(ctlr_info_t *h, int drv_index, int clear_all, int via_ioctl); -static void cciss_read_capacity(int ctlr, int logvol, +static void cciss_read_capacity(ctlr_info_t *h, int logvol, sector_t *total_size, unsigned int *block_size); -static void cciss_read_capacity_16(int ctlr, int logvol, +static void cciss_read_capacity_16(ctlr_info_t *h, int logvol, sector_t *total_size, unsigned int *block_size); -static void cciss_geometry_inquiry(int ctlr, int logvol, +static void cciss_geometry_inquiry(ctlr_info_t *h, int logvol, sector_t total_size, unsigned int block_size, InquiryData_struct *inq_buff, drive_info_struct *drv); -static void __devinit cciss_interrupt_mode(ctlr_info_t *, struct pci_dev *, - __u32); +static void __devinit cciss_interrupt_mode(ctlr_info_t *); static void start_io(ctlr_info_t *h); -static int sendcmd_withirq(__u8 cmd, int ctlr, void *buff, size_t size, +static int sendcmd_withirq(ctlr_info_t *h, __u8 cmd, void *buff, size_t size, __u8 page_code, unsigned char scsi3addr[], int cmd_type); static int sendcmd_withirq_core(ctlr_info_t *h, CommandList_struct *c, int attempt_retry); static int process_sendcmd_error(ctlr_info_t *h, CommandList_struct *c); -static void fail_all_cmds(unsigned long ctlr); static int add_to_scan_list(struct ctlr_info *h); static int scan_thread(void *data); static int check_for_unit_attention(ctlr_info_t *h, CommandList_struct *c); @@ -205,11 +215,23 @@ static void cciss_hba_release(struct device *dev); static void cciss_device_release(struct device *dev); static void cciss_free_gendisk(ctlr_info_t *h, int drv_index); static void cciss_free_drive_info(ctlr_info_t *h, int drv_index); +static inline u32 next_command(ctlr_info_t *h); +static int __devinit cciss_find_cfg_addrs(struct pci_dev *pdev, + void __iomem *vaddr, u32 *cfg_base_addr, u64 *cfg_base_addr_index, + u64 *cfg_offset); +static int __devinit cciss_pci_find_memory_BAR(struct pci_dev *pdev, + unsigned long *memory_bar); + + +/* performant mode helper functions */ +static void calc_bucket_map(int *bucket, int num_buckets, int nsgs, + int *bucket_map); +static void cciss_put_controller_into_performant_mode(ctlr_info_t *h); #ifdef CONFIG_PROC_FS -static void cciss_procinit(int i); +static void cciss_procinit(ctlr_info_t *h); #else -static void cciss_procinit(int i) +static void cciss_procinit(ctlr_info_t *h) { } #endif /* CONFIG_PROC_FS */ @@ -221,9 +243,9 @@ static int cciss_compat_ioctl(struct block_device *, fmode_t, static const struct block_device_operations cciss_fops = { .owner = THIS_MODULE, - .open = cciss_open, + .open = cciss_unlocked_open, .release = cciss_release, - .locked_ioctl = cciss_ioctl, + .ioctl = do_ioctl, .getgeo = cciss_getgeo, #ifdef CONFIG_COMPAT .compat_ioctl = cciss_compat_ioctl, @@ -231,6 +253,16 @@ static const struct block_device_operations cciss_fops = { .revalidate_disk = cciss_revalidate, }; +/* set_performant_mode: Modify the tag for cciss performant + * set bit 0 for pull model, bits 3-1 for block fetch + * register number + */ +static void set_performant_mode(ctlr_info_t *h, CommandList_struct *c) +{ + if (likely(h->transMethod == CFGTBL_Trans_Performant)) + c->busaddr |= 1 | (h->blockFetchTable[c->Header.SGList] << 1); +} + /* * Enqueuing and dequeuing functions for cmdlists. */ @@ -257,6 +289,18 @@ static inline void removeQ(CommandList_struct *c) hlist_del_init(&c->list); } +static void enqueue_cmd_and_start_io(ctlr_info_t *h, + CommandList_struct *c) +{ + unsigned long flags; + set_performant_mode(h, c); + spin_lock_irqsave(&h->lock, flags); + addQ(&h->reqQ, c); + h->Qdepth++; + start_io(h); + spin_unlock_irqrestore(&h->lock, flags); +} + static void cciss_free_sg_chain_blocks(SGDescriptor_struct **cmd_sg_list, int nr_cmds) { @@ -366,32 +410,31 @@ static void cciss_seq_show_header(struct seq_file *seq) h->product_name, (unsigned long)h->board_id, h->firm_ver[0], h->firm_ver[1], h->firm_ver[2], - h->firm_ver[3], (unsigned int)h->intr[SIMPLE_MODE_INT], + h->firm_ver[3], (unsigned int)h->intr[PERF_MODE_INT], h->num_luns, h->Qdepth, h->commands_outstanding, h->maxQsinceinit, h->max_outstanding, h->maxSG); #ifdef CONFIG_CISS_SCSI_TAPE - cciss_seq_tape_report(seq, h->ctlr); + cciss_seq_tape_report(seq, h); #endif /* CONFIG_CISS_SCSI_TAPE */ } static void *cciss_seq_start(struct seq_file *seq, loff_t *pos) { ctlr_info_t *h = seq->private; - unsigned ctlr = h->ctlr; unsigned long flags; /* prevent displaying bogus info during configuration * or deconfiguration of a logical volume */ - spin_lock_irqsave(CCISS_LOCK(ctlr), flags); + spin_lock_irqsave(&h->lock, flags); if (h->busy_configuring) { - spin_unlock_irqrestore(CCISS_LOCK(ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); return ERR_PTR(-EBUSY); } h->busy_configuring = 1; - spin_unlock_irqrestore(CCISS_LOCK(ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); if (*pos == 0) cciss_seq_show_header(seq); @@ -499,7 +542,7 @@ cciss_proc_write(struct file *file, const char __user *buf, struct seq_file *seq = file->private_data; ctlr_info_t *h = seq->private; - err = cciss_engage_scsi(h->ctlr); + err = cciss_engage_scsi(h); if (err == 0) err = length; } else @@ -522,7 +565,7 @@ static const struct file_operations cciss_proc_fops = { .write = cciss_proc_write, }; -static void __devinit cciss_procinit(int i) +static void __devinit cciss_procinit(ctlr_info_t *h) { struct proc_dir_entry *pde; @@ -530,9 +573,9 @@ static void __devinit cciss_procinit(int i) proc_cciss = proc_mkdir("driver/cciss", NULL); if (!proc_cciss) return; - pde = proc_create_data(hba[i]->devname, S_IWUSR | S_IRUSR | S_IRGRP | + pde = proc_create_data(h->devname, S_IWUSR | S_IRUSR | S_IRGRP | S_IROTH, proc_cciss, - &cciss_proc_fops, hba[i]); + &cciss_proc_fops, h); } #endif /* CONFIG_PROC_FS */ @@ -565,12 +608,12 @@ static ssize_t dev_show_unique_id(struct device *dev, unsigned long flags; int ret = 0; - spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags); + spin_lock_irqsave(&h->lock, flags); if (h->busy_configuring) ret = -EBUSY; else memcpy(sn, drv->serial_no, sizeof(sn)); - spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); if (ret) return ret; @@ -595,12 +638,12 @@ static ssize_t dev_show_vendor(struct device *dev, unsigned long flags; int ret = 0; - spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags); + spin_lock_irqsave(&h->lock, flags); if (h->busy_configuring) ret = -EBUSY; else memcpy(vendor, drv->vendor, VENDOR_LEN + 1); - spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); if (ret) return ret; @@ -619,12 +662,12 @@ static ssize_t dev_show_model(struct device *dev, unsigned long flags; int ret = 0; - spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags); + spin_lock_irqsave(&h->lock, flags); if (h->busy_configuring) ret = -EBUSY; else memcpy(model, drv->model, MODEL_LEN + 1); - spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); if (ret) return ret; @@ -643,12 +686,12 @@ static ssize_t dev_show_rev(struct device *dev, unsigned long flags; int ret = 0; - spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags); + spin_lock_irqsave(&h->lock, flags); if (h->busy_configuring) ret = -EBUSY; else memcpy(rev, drv->rev, REV_LEN + 1); - spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); if (ret) return ret; @@ -665,17 +708,17 @@ static ssize_t cciss_show_lunid(struct device *dev, unsigned long flags; unsigned char lunid[8]; - spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags); + spin_lock_irqsave(&h->lock, flags); if (h->busy_configuring) { - spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); return -EBUSY; } if (!drv->heads) { - spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); return -ENOTTY; } memcpy(lunid, drv->LunID, sizeof(lunid)); - spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); return snprintf(buf, 20, "0x%02x%02x%02x%02x%02x%02x%02x%02x\n", lunid[0], lunid[1], lunid[2], lunid[3], lunid[4], lunid[5], lunid[6], lunid[7]); @@ -690,13 +733,13 @@ static ssize_t cciss_show_raid_level(struct device *dev, int raid; unsigned long flags; - spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags); + spin_lock_irqsave(&h->lock, flags); if (h->busy_configuring) { - spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); return -EBUSY; } raid = drv->raid_level; - spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); if (raid < 0 || raid > RAID_UNKNOWN) raid = RAID_UNKNOWN; @@ -713,13 +756,13 @@ static ssize_t cciss_show_usage_count(struct device *dev, unsigned long flags; int count; - spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags); + spin_lock_irqsave(&h->lock, flags); if (h->busy_configuring) { - spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); return -EBUSY; } count = drv->usage_count; - spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); return snprintf(buf, 20, "%d\n", count); } static DEVICE_ATTR(usage_count, S_IRUGO, cciss_show_usage_count, NULL); @@ -864,60 +907,70 @@ static void cciss_destroy_ld_sysfs_entry(struct ctlr_info *h, int drv_index, /* * 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. For operations that can wait for kmalloc - * to possible sleep, this routine can be called with get_from_pool set to 0. - * cmd_free() MUST be called with a got_from_pool set to 0 if cmd_alloc was. + * which ones are free or in use. */ -static CommandList_struct *cmd_alloc(ctlr_info_t *h, int get_from_pool) +static CommandList_struct *cmd_alloc(ctlr_info_t *h) { CommandList_struct *c; int i; u64bit temp64; dma_addr_t cmd_dma_handle, err_dma_handle; - if (!get_from_pool) { - c = (CommandList_struct *) pci_alloc_consistent(h->pdev, - sizeof(CommandList_struct), &cmd_dma_handle); - if (c == NULL) + do { + i = find_first_zero_bit(h->cmd_pool_bits, h->nr_cmds); + if (i == h->nr_cmds) return NULL; - memset(c, 0, sizeof(CommandList_struct)); + } while (test_and_set_bit(i & (BITS_PER_LONG - 1), + h->cmd_pool_bits + (i / BITS_PER_LONG)) != 0); + c = h->cmd_pool + i; + memset(c, 0, sizeof(CommandList_struct)); + cmd_dma_handle = h->cmd_pool_dhandle + i * sizeof(CommandList_struct); + c->err_info = h->errinfo_pool + i; + memset(c->err_info, 0, sizeof(ErrorInfo_struct)); + err_dma_handle = h->errinfo_pool_dhandle + + i * sizeof(ErrorInfo_struct); + h->nr_allocs++; - c->cmdindex = -1; + c->cmdindex = i; - c->err_info = (ErrorInfo_struct *) - pci_alloc_consistent(h->pdev, sizeof(ErrorInfo_struct), - &err_dma_handle); + INIT_HLIST_NODE(&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(ErrorInfo_struct); - if (c->err_info == NULL) { - pci_free_consistent(h->pdev, - sizeof(CommandList_struct), c, cmd_dma_handle); - return NULL; - } - memset(c->err_info, 0, sizeof(ErrorInfo_struct)); - } else { /* get it out of the controllers pool */ - - do { - i = find_first_zero_bit(h->cmd_pool_bits, h->nr_cmds); - if (i == h->nr_cmds) - return NULL; - } while (test_and_set_bit - (i & (BITS_PER_LONG - 1), - h->cmd_pool_bits + (i / BITS_PER_LONG)) != 0); -#ifdef CCISS_DEBUG - printk(KERN_DEBUG "cciss: using command buffer %d\n", i); -#endif - c = h->cmd_pool + i; - memset(c, 0, sizeof(CommandList_struct)); - cmd_dma_handle = h->cmd_pool_dhandle - + i * sizeof(CommandList_struct); - c->err_info = h->errinfo_pool + i; - memset(c->err_info, 0, sizeof(ErrorInfo_struct)); - err_dma_handle = h->errinfo_pool_dhandle - + i * sizeof(ErrorInfo_struct); - h->nr_allocs++; + c->ctlr = h->ctlr; + return c; +} - c->cmdindex = i; +/* allocate a command using pci_alloc_consistent, used for ioctls, + * etc., not for the main i/o path. + */ +static CommandList_struct *cmd_special_alloc(ctlr_info_t *h) +{ + CommandList_struct *c; + u64bit temp64; + dma_addr_t cmd_dma_handle, err_dma_handle; + + c = (CommandList_struct *) pci_alloc_consistent(h->pdev, + sizeof(CommandList_struct), &cmd_dma_handle); + if (c == NULL) + return NULL; + memset(c, 0, sizeof(CommandList_struct)); + + c->cmdindex = -1; + + c->err_info = (ErrorInfo_struct *) + pci_alloc_consistent(h->pdev, sizeof(ErrorInfo_struct), + &err_dma_handle); + + if (c->err_info == NULL) { + pci_free_consistent(h->pdev, + sizeof(CommandList_struct), c, cmd_dma_handle); + return NULL; } + memset(c->err_info, 0, sizeof(ErrorInfo_struct)); INIT_HLIST_NODE(&c->list); c->busaddr = (__u32) cmd_dma_handle; @@ -930,27 +983,26 @@ static CommandList_struct *cmd_alloc(ctlr_info_t *h, int get_from_pool) return c; } -/* - * Frees a command block that was previously allocated with cmd_alloc(). - */ -static void cmd_free(ctlr_info_t *h, CommandList_struct *c, int got_from_pool) +static void cmd_free(ctlr_info_t *h, CommandList_struct *c) { int i; + + i = c - h->cmd_pool; + clear_bit(i & (BITS_PER_LONG - 1), + h->cmd_pool_bits + (i / BITS_PER_LONG)); + h->nr_frees++; +} + +static void cmd_special_free(ctlr_info_t *h, CommandList_struct *c) +{ u64bit temp64; - if (!got_from_pool) { - temp64.val32.lower = c->ErrDesc.Addr.lower; - temp64.val32.upper = c->ErrDesc.Addr.upper; - pci_free_consistent(h->pdev, sizeof(ErrorInfo_struct), - c->err_info, (dma_addr_t) temp64.val); - pci_free_consistent(h->pdev, sizeof(CommandList_struct), - c, (dma_addr_t) c->busaddr); - } else { - i = c - h->cmd_pool; - clear_bit(i & (BITS_PER_LONG - 1), - h->cmd_pool_bits + (i / BITS_PER_LONG)); - h->nr_frees++; - } + temp64.val32.lower = c->ErrDesc.Addr.lower; + temp64.val32.upper = c->ErrDesc.Addr.upper; + pci_free_consistent(h->pdev, sizeof(ErrorInfo_struct), + c->err_info, (dma_addr_t) temp64.val); + pci_free_consistent(h->pdev, sizeof(CommandList_struct), + c, (dma_addr_t) c->busaddr); } static inline ctlr_info_t *get_host(struct gendisk *disk) @@ -968,13 +1020,10 @@ static inline drive_info_struct *get_drv(struct gendisk *disk) */ static int cciss_open(struct block_device *bdev, fmode_t mode) { - ctlr_info_t *host = get_host(bdev->bd_disk); + ctlr_info_t *h = get_host(bdev->bd_disk); drive_info_struct *drv = get_drv(bdev->bd_disk); -#ifdef CCISS_DEBUG - printk(KERN_DEBUG "cciss_open %s\n", bdev->bd_disk->disk_name); -#endif /* CCISS_DEBUG */ - + dev_dbg(&h->pdev->dev, "cciss_open %s\n", bdev->bd_disk->disk_name); if (drv->busy_configuring) return -EBUSY; /* @@ -1000,29 +1049,39 @@ static int cciss_open(struct block_device *bdev, fmode_t mode) return -EPERM; } drv->usage_count++; - host->usage_count++; + h->usage_count++; return 0; } +static int cciss_unlocked_open(struct block_device *bdev, fmode_t mode) +{ + int ret; + + lock_kernel(); + ret = cciss_open(bdev, mode); + unlock_kernel(); + + return ret; +} + /* * Close. Sync first. */ static int cciss_release(struct gendisk *disk, fmode_t mode) { - ctlr_info_t *host = get_host(disk); - drive_info_struct *drv = get_drv(disk); - -#ifdef CCISS_DEBUG - printk(KERN_DEBUG "cciss_release %s\n", disk->disk_name); -#endif /* CCISS_DEBUG */ + ctlr_info_t *h; + drive_info_struct *drv; + lock_kernel(); + h = get_host(disk); + drv = get_drv(disk); + dev_dbg(&h->pdev->dev, "cciss_release %s\n", disk->disk_name); drv->usage_count--; - host->usage_count--; + h->usage_count--; + unlock_kernel(); return 0; } -#ifdef CONFIG_COMPAT - static int do_ioctl(struct block_device *bdev, fmode_t mode, unsigned cmd, unsigned long arg) { @@ -1033,6 +1092,8 @@ static int do_ioctl(struct block_device *bdev, fmode_t mode, return ret; } +#ifdef CONFIG_COMPAT + static int cciss_ioctl32_passthru(struct block_device *bdev, fmode_t mode, unsigned cmd, unsigned long arg); static int cciss_ioctl32_big_passthru(struct block_device *bdev, fmode_t mode, @@ -1163,11 +1224,11 @@ static int cciss_getgeo(struct block_device *bdev, struct hd_geometry *geo) return 0; } -static void check_ioctl_unit_attention(ctlr_info_t *host, CommandList_struct *c) +static void check_ioctl_unit_attention(ctlr_info_t *h, CommandList_struct *c) { if (c->err_info->CommandStatus == CMD_TARGET_STATUS && c->err_info->ScsiStatus != SAM_STAT_CHECK_CONDITION) - (void)check_for_unit_attention(host, c); + (void)check_for_unit_attention(h, c); } /* * ioctl @@ -1176,15 +1237,12 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, unsigned int cmd, unsigned long arg) { struct gendisk *disk = bdev->bd_disk; - ctlr_info_t *host = get_host(disk); + ctlr_info_t *h = get_host(disk); drive_info_struct *drv = get_drv(disk); - int ctlr = host->ctlr; void __user *argp = (void __user *)arg; -#ifdef CCISS_DEBUG - printk(KERN_DEBUG "cciss_ioctl: Called with cmd=%x %lx\n", cmd, arg); -#endif /* CCISS_DEBUG */ - + dev_dbg(&h->pdev->dev, "cciss_ioctl: Called with cmd=%x %lx\n", + cmd, arg); switch (cmd) { case CCISS_GETPCIINFO: { @@ -1192,10 +1250,10 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, if (!arg) return -EINVAL; - pciinfo.domain = pci_domain_nr(host->pdev->bus); - pciinfo.bus = host->pdev->bus->number; - pciinfo.dev_fn = host->pdev->devfn; - pciinfo.board_id = host->board_id; + pciinfo.domain = pci_domain_nr(h->pdev->bus); + pciinfo.bus = h->pdev->bus->number; + pciinfo.dev_fn = h->pdev->devfn; + pciinfo.board_id = h->board_id; if (copy_to_user (argp, &pciinfo, sizeof(cciss_pci_info_struct))) return -EFAULT; @@ -1207,9 +1265,9 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, if (!arg) return -EINVAL; intinfo.delay = - readl(&host->cfgtable->HostWrite.CoalIntDelay); + readl(&h->cfgtable->HostWrite.CoalIntDelay); intinfo.count = - readl(&host->cfgtable->HostWrite.CoalIntCount); + readl(&h->cfgtable->HostWrite.CoalIntCount); if (copy_to_user (argp, &intinfo, sizeof(cciss_coalint_struct))) return -EFAULT; @@ -1229,26 +1287,23 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, (&intinfo, argp, sizeof(cciss_coalint_struct))) return -EFAULT; if ((intinfo.delay == 0) && (intinfo.count == 0)) - { -// printk("cciss_ioctl: delay and count cannot be 0\n"); return -EINVAL; - } - spin_lock_irqsave(CCISS_LOCK(ctlr), flags); + spin_lock_irqsave(&h->lock, flags); /* Update the field, and then ring the doorbell */ writel(intinfo.delay, - &(host->cfgtable->HostWrite.CoalIntDelay)); + &(h->cfgtable->HostWrite.CoalIntDelay)); writel(intinfo.count, - &(host->cfgtable->HostWrite.CoalIntCount)); - writel(CFGTBL_ChangeReq, host->vaddr + SA5_DOORBELL); + &(h->cfgtable->HostWrite.CoalIntCount)); + writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL); for (i = 0; i < MAX_IOCTL_CONFIG_WAIT; i++) { - if (!(readl(host->vaddr + SA5_DOORBELL) + if (!(readl(h->vaddr + SA5_DOORBELL) & CFGTBL_ChangeReq)) break; /* delay and try again */ udelay(1000); } - spin_unlock_irqrestore(CCISS_LOCK(ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); if (i >= MAX_IOCTL_CONFIG_WAIT) return -EAGAIN; return 0; @@ -1262,7 +1317,7 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, return -EINVAL; for (i = 0; i < 16; i++) NodeName[i] = - readb(&host->cfgtable->ServerName[i]); + readb(&h->cfgtable->ServerName[i]); if (copy_to_user(argp, NodeName, sizeof(NodeName_type))) return -EFAULT; return 0; @@ -1282,23 +1337,23 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, (NodeName, argp, sizeof(NodeName_type))) return -EFAULT; - spin_lock_irqsave(CCISS_LOCK(ctlr), flags); + spin_lock_irqsave(&h->lock, flags); /* Update the field, and then ring the doorbell */ for (i = 0; i < 16; i++) writeb(NodeName[i], - &host->cfgtable->ServerName[i]); + &h->cfgtable->ServerName[i]); - writel(CFGTBL_ChangeReq, host->vaddr + SA5_DOORBELL); + writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL); for (i = 0; i < MAX_IOCTL_CONFIG_WAIT; i++) { - if (!(readl(host->vaddr + SA5_DOORBELL) + if (!(readl(h->vaddr + SA5_DOORBELL) & CFGTBL_ChangeReq)) break; /* delay and try again */ udelay(1000); } - spin_unlock_irqrestore(CCISS_LOCK(ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); if (i >= MAX_IOCTL_CONFIG_WAIT) return -EAGAIN; return 0; @@ -1310,7 +1365,7 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, if (!arg) return -EINVAL; - heartbeat = readl(&host->cfgtable->HeartBeat); + heartbeat = readl(&h->cfgtable->HeartBeat); if (copy_to_user (argp, &heartbeat, sizeof(Heartbeat_type))) return -EFAULT; @@ -1322,7 +1377,7 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, if (!arg) return -EINVAL; - BusTypes = readl(&host->cfgtable->BusTypes); + BusTypes = readl(&h->cfgtable->BusTypes); if (copy_to_user (argp, &BusTypes, sizeof(BusTypes_type))) return -EFAULT; @@ -1334,7 +1389,7 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, if (!arg) return -EINVAL; - memcpy(firmware, host->firm_ver, 4); + memcpy(firmware, h->firm_ver, 4); if (copy_to_user (argp, firmware, sizeof(FirmwareVer_type))) @@ -1357,7 +1412,7 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, case CCISS_DEREGDISK: case CCISS_REGNEWD: case CCISS_REVALIDVOLS: - return rebuild_lun_table(host, 0, 1); + return rebuild_lun_table(h, 0, 1); case CCISS_GETLUNINFO:{ LogvolInfo_struct luninfo; @@ -1377,7 +1432,6 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, CommandList_struct *c; char *buff = NULL; u64bit temp64; - unsigned long flags; DECLARE_COMPLETION_ONSTACK(wait); if (!arg) @@ -1413,7 +1467,8 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, } else { memset(buff, 0, iocommand.buf_size); } - if ((c = cmd_alloc(host, 0)) == NULL) { + c = cmd_special_alloc(h); + if (!c) { kfree(buff); return -ENOMEM; } @@ -1439,7 +1494,7 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, /* Fill in the scatter gather information */ if (iocommand.buf_size > 0) { - temp64.val = pci_map_single(host->pdev, buff, + temp64.val = pci_map_single(h->pdev, buff, iocommand.buf_size, PCI_DMA_BIDIRECTIONAL); c->SG[0].Addr.lower = temp64.val32.lower; @@ -1449,30 +1504,24 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, } c->waiting = &wait; - /* Put the request on the tail of the request queue */ - spin_lock_irqsave(CCISS_LOCK(ctlr), flags); - addQ(&host->reqQ, c); - host->Qdepth++; - start_io(host); - spin_unlock_irqrestore(CCISS_LOCK(ctlr), flags); - + enqueue_cmd_and_start_io(h, c); wait_for_completion(&wait); /* unlock the buffers from DMA */ temp64.val32.lower = c->SG[0].Addr.lower; temp64.val32.upper = c->SG[0].Addr.upper; - pci_unmap_single(host->pdev, (dma_addr_t) temp64.val, + pci_unmap_single(h->pdev, (dma_addr_t) temp64.val, iocommand.buf_size, PCI_DMA_BIDIRECTIONAL); - check_ioctl_unit_attention(host, c); + check_ioctl_unit_attention(h, c); /* Copy the error information out */ iocommand.error_info = *(c->err_info); if (copy_to_user (argp, &iocommand, sizeof(IOCTL_Command_struct))) { kfree(buff); - cmd_free(host, c, 0); + cmd_special_free(h, c); return -EFAULT; } @@ -1481,12 +1530,12 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, if (copy_to_user (iocommand.buf, buff, iocommand.buf_size)) { kfree(buff); - cmd_free(host, c, 0); + cmd_special_free(h, c); return -EFAULT; } } kfree(buff); - cmd_free(host, c, 0); + cmd_special_free(h, c); return 0; } case CCISS_BIG_PASSTHRU:{ @@ -1495,7 +1544,6 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, unsigned char **buff = NULL; int *buff_size = NULL; u64bit temp64; - unsigned long flags; BYTE sg_used = 0; int status = 0; int i; @@ -1569,7 +1617,8 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, data_ptr += sz; sg_used++; } - if ((c = cmd_alloc(host, 0)) == NULL) { + c = cmd_special_alloc(h); + if (!c) { status = -ENOMEM; goto cleanup1; } @@ -1590,7 +1639,7 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, if (ioc->buf_size > 0) { for (i = 0; i < sg_used; i++) { temp64.val = - pci_map_single(host->pdev, buff[i], + pci_map_single(h->pdev, buff[i], buff_size[i], PCI_DMA_BIDIRECTIONAL); c->SG[i].Addr.lower = @@ -1602,26 +1651,21 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, } } c->waiting = &wait; - /* Put the request on the tail of the request queue */ - spin_lock_irqsave(CCISS_LOCK(ctlr), flags); - addQ(&host->reqQ, c); - host->Qdepth++; - start_io(host); - spin_unlock_irqrestore(CCISS_LOCK(ctlr), flags); + enqueue_cmd_and_start_io(h, c); wait_for_completion(&wait); /* unlock the buffers from DMA */ for (i = 0; i < sg_used; i++) { temp64.val32.lower = c->SG[i].Addr.lower; temp64.val32.upper = c->SG[i].Addr.upper; - pci_unmap_single(host->pdev, + pci_unmap_single(h->pdev, (dma_addr_t) temp64.val, buff_size[i], PCI_DMA_BIDIRECTIONAL); } - check_ioctl_unit_attention(host, c); + check_ioctl_unit_attention(h, c); /* Copy the error information out */ ioc->error_info = *(c->err_info); if (copy_to_user(argp, ioc, sizeof(*ioc))) { - cmd_free(host, c, 0); + cmd_special_free(h, c); status = -EFAULT; goto cleanup1; } @@ -1631,14 +1675,14 @@ static int cciss_ioctl(struct block_device *bdev, fmode_t mode, for (i = 0; i < sg_used; i++) { if (copy_to_user (ptr, buff[i], buff_size[i])) { - cmd_free(host, c, 0); + cmd_special_free(h, c); status = -EFAULT; goto cleanup1; } ptr += buff_size[i]; } } - cmd_free(host, c, 0); + cmd_special_free(h, c); status = 0; cleanup1: if (buff) { @@ -1726,26 +1770,26 @@ static void cciss_check_queues(ctlr_info_t *h) static void cciss_softirq_done(struct request *rq) { - CommandList_struct *cmd = rq->completion_data; - ctlr_info_t *h = hba[cmd->ctlr]; - SGDescriptor_struct *curr_sg = cmd->SG; - unsigned long flags; + CommandList_struct *c = rq->completion_data; + ctlr_info_t *h = hba[c->ctlr]; + SGDescriptor_struct *curr_sg = c->SG; u64bit temp64; + unsigned long flags; int i, ddir; int sg_index = 0; - if (cmd->Request.Type.Direction == XFER_READ) + if (c->Request.Type.Direction == XFER_READ) ddir = PCI_DMA_FROMDEVICE; else ddir = PCI_DMA_TODEVICE; /* command did not need to be retried */ /* unmap the DMA mapping for all the scatter gather elements */ - for (i = 0; i < cmd->Header.SGList; i++) { + for (i = 0; i < c->Header.SGList; i++) { if (curr_sg[sg_index].Ext == CCISS_SG_CHAIN) { - cciss_unmap_sg_chain_block(h, cmd); + cciss_unmap_sg_chain_block(h, c); /* Point to the next block */ - curr_sg = h->cmd_sg_list[cmd->cmdindex]; + curr_sg = h->cmd_sg_list[c->cmdindex]; sg_index = 0; } temp64.val32.lower = curr_sg[sg_index].Addr.lower; @@ -1755,18 +1799,16 @@ static void cciss_softirq_done(struct request *rq) ++sg_index; } -#ifdef CCISS_DEBUG - printk("Done with %p\n", rq); -#endif /* CCISS_DEBUG */ + dev_dbg(&h->pdev->dev, "Done with %p\n", rq); /* set the residual count for pc requests */ - if (blk_pc_request(rq)) - rq->resid_len = cmd->err_info->ResidualCnt; + if (rq->cmd_type == REQ_TYPE_BLOCK_PC) + rq->resid_len = c->err_info->ResidualCnt; blk_end_request_all(rq, (rq->errors == 0) ? 0 : -EIO); spin_lock_irqsave(&h->lock, flags); - cmd_free(h, cmd, 1); + cmd_free(h, c); cciss_check_queues(h); spin_unlock_irqrestore(&h->lock, flags); } @@ -1782,7 +1824,7 @@ static inline void log_unit_to_scsi3addr(ctlr_info_t *h, * via the inquiry page 0. Model, vendor, and rev are set to empty strings if * they cannot be read. */ -static void cciss_get_device_descr(int ctlr, int logvol, +static void cciss_get_device_descr(ctlr_info_t *h, int logvol, char *vendor, char *model, char *rev) { int rc; @@ -1797,8 +1839,8 @@ static void cciss_get_device_descr(int ctlr, int logvol, if (!inq_buf) return; - log_unit_to_scsi3addr(hba[ctlr], scsi3addr, logvol); - rc = sendcmd_withirq(CISS_INQUIRY, ctlr, inq_buf, sizeof(*inq_buf), 0, + log_unit_to_scsi3addr(h, scsi3addr, logvol); + rc = sendcmd_withirq(h, CISS_INQUIRY, inq_buf, sizeof(*inq_buf), 0, scsi3addr, TYPE_CMD); if (rc == IO_OK) { memcpy(vendor, &inq_buf->data_byte[8], VENDOR_LEN); @@ -1818,7 +1860,7 @@ static void cciss_get_device_descr(int ctlr, int logvol, * number cannot be had, for whatever reason, 16 bytes of 0xff * are returned instead. */ -static void cciss_get_serial_no(int ctlr, int logvol, +static void cciss_get_serial_no(ctlr_info_t *h, int logvol, unsigned char *serial_no, int buflen) { #define PAGE_83_INQ_BYTES 64 @@ -1833,8 +1875,8 @@ static void cciss_get_serial_no(int ctlr, int logvol, if (!buf) return; memset(serial_no, 0, buflen); - log_unit_to_scsi3addr(hba[ctlr], scsi3addr, logvol); - rc = sendcmd_withirq(CISS_INQUIRY, ctlr, buf, + log_unit_to_scsi3addr(h, scsi3addr, logvol); + rc = sendcmd_withirq(h, CISS_INQUIRY, buf, PAGE_83_INQ_BYTES, 0x83, scsi3addr, TYPE_CMD); if (rc == IO_OK) memcpy(serial_no, &buf[8], buflen); @@ -1900,10 +1942,9 @@ init_queue_failure: * is also the controller node. Any changes to disk 0 will show up on * the next reboot. */ -static void cciss_update_drive_info(int ctlr, int drv_index, int first_time, - int via_ioctl) +static void cciss_update_drive_info(ctlr_info_t *h, int drv_index, + int first_time, int via_ioctl) { - ctlr_info_t *h = hba[ctlr]; struct gendisk *disk; InquiryData_struct *inq_buff = NULL; unsigned int block_size; @@ -1920,16 +1961,16 @@ static void cciss_update_drive_info(int ctlr, int drv_index, int first_time, /* testing to see if 16-byte CDBs are already being used */ if (h->cciss_read == CCISS_READ_16) { - cciss_read_capacity_16(h->ctlr, drv_index, + cciss_read_capacity_16(h, drv_index, &total_size, &block_size); } else { - cciss_read_capacity(ctlr, drv_index, &total_size, &block_size); + cciss_read_capacity(h, drv_index, &total_size, &block_size); /* if read_capacity returns all F's this volume is >2TB */ /* in size so we switch to 16-byte CDB's for all */ /* read/write ops */ if (total_size == 0xFFFFFFFFULL) { - cciss_read_capacity_16(ctlr, drv_index, + cciss_read_capacity_16(h, drv_index, &total_size, &block_size); h->cciss_read = CCISS_READ_16; h->cciss_write = CCISS_WRITE_16; @@ -1939,14 +1980,14 @@ static void cciss_update_drive_info(int ctlr, int drv_index, int first_time, } } - cciss_geometry_inquiry(ctlr, drv_index, total_size, block_size, + cciss_geometry_inquiry(h, drv_index, total_size, block_size, inq_buff, drvinfo); drvinfo->block_size = block_size; drvinfo->nr_blocks = total_size + 1; - cciss_get_device_descr(ctlr, drv_index, drvinfo->vendor, + cciss_get_device_descr(h, drv_index, drvinfo->vendor, drvinfo->model, drvinfo->rev); - cciss_get_serial_no(ctlr, drv_index, drvinfo->serial_no, + cciss_get_serial_no(h, drv_index, drvinfo->serial_no, sizeof(drvinfo->serial_no)); /* Save the lunid in case we deregister the disk, below. */ memcpy(drvinfo->LunID, h->drv[drv_index]->LunID, @@ -1971,10 +2012,10 @@ static void cciss_update_drive_info(int ctlr, int drv_index, int first_time, * (unless it's the first disk (for the controller node). */ if (h->drv[drv_index]->raid_level != -1 && drv_index != 0) { - printk(KERN_WARNING "disk %d has changed.\n", drv_index); - spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags); + dev_warn(&h->pdev->dev, "disk %d has changed.\n", drv_index); + spin_lock_irqsave(&h->lock, flags); h->drv[drv_index]->busy_configuring = 1; - spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); /* deregister_disk sets h->drv[drv_index]->queue = NULL * which keeps the interrupt handler from starting @@ -2024,8 +2065,8 @@ static void cciss_update_drive_info(int ctlr, int drv_index, int first_time, if (cciss_add_disk(h, disk, drv_index) != 0) { cciss_free_gendisk(h, drv_index); cciss_free_drive_info(h, drv_index); - printk(KERN_WARNING "cciss:%d could not update " - "disk %d\n", h->ctlr, drv_index); + dev_warn(&h->pdev->dev, "could not update disk %d\n", + drv_index); --h->num_luns; } } @@ -2035,7 +2076,7 @@ freeret: kfree(drvinfo); return; mem_msg: - printk(KERN_ERR "cciss: out of memory\n"); + dev_err(&h->pdev->dev, "out of memory\n"); goto freeret; } @@ -2127,9 +2168,9 @@ static int cciss_add_gendisk(ctlr_info_t *h, unsigned char lunid[], h->gendisk[drv_index] = alloc_disk(1 << NWD_SHIFT); if (!h->gendisk[drv_index]) { - printk(KERN_ERR "cciss%d: could not " - "allocate a new disk %d\n", - h->ctlr, drv_index); + dev_err(&h->pdev->dev, + "could not allocate a new disk %d\n", + drv_index); goto err_free_drive_info; } } @@ -2180,8 +2221,7 @@ static void cciss_add_controller_node(ctlr_info_t *h) cciss_free_gendisk(h, drv_index); cciss_free_drive_info(h, drv_index); error: - printk(KERN_WARNING "cciss%d: could not " - "add disk 0.\n", h->ctlr); + dev_warn(&h->pdev->dev, "could not add disk 0.\n"); return; } @@ -2196,7 +2236,6 @@ error: static int rebuild_lun_table(ctlr_info_t *h, int first_time, int via_ioctl) { - int ctlr = h->ctlr; int num_luns; ReportLunData_struct *ld_buff = NULL; int return_code; @@ -2211,27 +2250,27 @@ static int rebuild_lun_table(ctlr_info_t *h, int first_time, return -EPERM; /* Set busy_configuring flag for this operation */ - spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags); + spin_lock_irqsave(&h->lock, flags); if (h->busy_configuring) { - spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); return -EBUSY; } h->busy_configuring = 1; - spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); ld_buff = kzalloc(sizeof(ReportLunData_struct), GFP_KERNEL); if (ld_buff == NULL) goto mem_msg; - return_code = sendcmd_withirq(CISS_REPORT_LOG, ctlr, ld_buff, + return_code = sendcmd_withirq(h, CISS_REPORT_LOG, ld_buff, sizeof(ReportLunData_struct), 0, CTLR_LUNID, TYPE_CMD); if (return_code == IO_OK) listlength = be32_to_cpu(*(__be32 *) ld_buff->LUNListLength); else { /* reading number of logical volumes failed */ - printk(KERN_WARNING "cciss: report logical volume" - " command failed\n"); + dev_warn(&h->pdev->dev, + "report logical volume command failed\n"); listlength = 0; goto freeret; } @@ -2239,7 +2278,7 @@ static int rebuild_lun_table(ctlr_info_t *h, int first_time, num_luns = listlength / 8; /* 8 bytes per entry */ if (num_luns > CISS_MAX_LUN) { num_luns = CISS_MAX_LUN; - printk(KERN_WARNING "cciss: more luns configured" + dev_warn(&h->pdev->dev, "more luns configured" " on controller than can be handled by" " this driver.\n"); } @@ -2270,9 +2309,9 @@ static int rebuild_lun_table(ctlr_info_t *h, int first_time, } if (!drv_found) { /* Deregister it from the OS, it's gone. */ - spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags); + spin_lock_irqsave(&h->lock, flags); h->drv[i]->busy_configuring = 1; - spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags); + spin_unlock_irqrestore(&h->lock, flags); return_code = deregister_disk(h, i, 1, via_ioctl); if (h->drv[i] != NULL) h->drv[i]->busy_configuring = 0; @@ -2311,8 +2350,7 @@ static int rebuild_lun_table(ctlr_info_t *h, int first_time, if (drv_index == -1) goto freeret; } - cciss_update_drive_info(ctlr, drv_index, first_time, - via_ioctl); + cciss_update_drive_info(h, drv_index, first_time, via_ioctl); } /* end for */ freeret: @@ -2324,7 +2362,7 @@ freeret: */ return -1; mem_msg: - printk(KERN_ERR "cciss: out of memory\n"); + dev_err(&h->pdev->dev, "out of memory\n"); h->busy_configuring = 0; goto freeret; } @@ -2444,11 +2482,10 @@ static int deregister_disk(ctlr_info_t *h, int drv_index, return 0; } -static int fill_cmd(CommandList_struct *c, __u8 cmd, int ctlr, void *buff, +static int fill_cmd(ctlr_info_t *h, CommandList_struct *c, __u8 cmd, void *buff, size_t size, __u8 page_code, unsigned char *scsi3addr, int cmd_type) { - ctlr_info_t *h = hba[ctlr]; u64bit buff_dma_handle; int status = IO_OK; @@ -2532,8 +2569,7 @@ static int fill_cmd(CommandList_struct *c, __u8 cmd, int ctlr, void *buff, c->Request.Timeout = 0; break; default: - printk(KERN_WARNING - "cciss%d: Unknown Command 0x%c\n", ctlr, cmd); + dev_warn(&h->pdev->dev, "Unknown Command 0x%c\n", cmd); return IO_ERROR; } } else if (cmd_type == TYPE_MSG) { @@ -2565,13 +2601,12 @@ static int fill_cmd(CommandList_struct *c, __u8 cmd, int ctlr, void *buff, c->Request.CDB[0] = cmd; break; default: - printk(KERN_WARNING - "cciss%d: unknown message type %d\n", ctlr, cmd); + dev_warn(&h->pdev->dev, + "unknown message type %d\n", cmd); return IO_ERROR; } } else { - printk(KERN_WARNING - "cciss%d: unknown command type %d\n", ctlr, cmd_type); + dev_warn(&h->pdev->dev, "unknown command type %d\n", cmd_type); return IO_ERROR; } /* Fill in the scatter gather information */ @@ -2599,15 +2634,14 @@ static int check_target_status(ctlr_info_t *h, CommandList_struct *c) default: if (check_for_unit_attention(h, c)) return IO_NEEDS_RETRY; - printk(KERN_WARNING "cciss%d: cmd 0x%02x " + dev_warn(&h->pdev->dev, "cmd 0x%02x " "check condition, sense key = 0x%02x\n", - h->ctlr, c->Request.CDB[0], - c->err_info->SenseInfo[2]); + c->Request.CDB[0], c->err_info->SenseInfo[2]); } break; default: - printk(KERN_WARNING "cciss%d: cmd 0x%02x" - "scsi status = 0x%02x\n", h->ctlr, + dev_warn(&h->pdev->dev, "cmd 0x%02x" + "scsi status = 0x%02x\n", c->Request.CDB[0], c->err_info->ScsiStatus); break; } @@ -2630,43 +2664,42 @@ static int process_sendcmd_error(ctlr_info_t *h, CommandList_struct *c) /* expected for inquiry and report lun commands */ break; case CMD_INVALID: - printk(KERN_WARNING "cciss: cmd 0x%02x is " + dev_warn(&h->pdev->dev, "cmd 0x%02x is " "reported invalid\n", c->Request.CDB[0]); return_status = IO_ERROR; break; case CMD_PROTOCOL_ERR: - printk(KERN_WARNING "cciss: cmd 0x%02x has " - "protocol error \n", c->Request.CDB[0]); + dev_warn(&h->pdev->dev, "cmd 0x%02x has " + "protocol error\n", c->Request.CDB[0]); return_status = IO_ERROR; break; case CMD_HARDWARE_ERR: - printk(KERN_WARNING "cciss: cmd 0x%02x had " + dev_warn(&h->pdev->dev, "cmd 0x%02x had " " hardware error\n", c->Request.CDB[0]); return_status = IO_ERROR; break; case CMD_CONNECTION_LOST: - printk(KERN_WARNING "cciss: cmd 0x%02x had " + dev_warn(&h->pdev->dev, "cmd 0x%02x had " "connection lost\n", c->Request.CDB[0]); return_status = IO_ERROR; break; case CMD_ABORTED: - printk(KERN_WARNING "cciss: cmd 0x%02x was " + dev_warn(&h->pdev->dev, "cmd 0x%02x was " "aborted\n", c->Request.CDB[0]); return_status = IO_ERROR; break; case CMD_ABORT_FAILED: - printk(KERN_WARNING "cciss: cmd 0x%02x reports " + dev_warn(&h->pdev->dev, "cmd 0x%02x reports " "abort failed\n", c->Request.CDB[0]); return_status = IO_ERROR; break; case CMD_UNSOLICITED_ABORT: - printk(KERN_WARNING - "cciss%d: unsolicited abort 0x%02x\n", h->ctlr, + dev_warn(&h->pdev->dev, "unsolicited abort 0x%02x\n", c->Request.CDB[0]); return_status = IO_NEEDS_RETRY; break; default: - printk(KERN_WARNING "cciss: cmd 0x%02x returned " + dev_warn(&h->pdev->dev, "cmd 0x%02x returned " "unknown status %x\n", c->Request.CDB[0], c->err_info->CommandStatus); return_status = IO_ERROR; @@ -2679,17 +2712,11 @@ static int sendcmd_withirq_core(ctlr_info_t *h, CommandList_struct *c, { DECLARE_COMPLETION_ONSTACK(wait); u64bit buff_dma_handle; - unsigned long flags; int return_status = IO_OK; resend_cmd2: c->waiting = &wait; - /* Put the request on the tail of the queue and send it */ - spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags); - addQ(&h->reqQ, c); - h->Qdepth++; - start_io(h); - spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags); + enqueue_cmd_and_start_io(h, c); wait_for_completion(&wait); @@ -2700,7 +2727,7 @@ resend_cmd2: if (return_status == IO_NEEDS_RETRY && c->retry_count < MAX_CMD_RETRIES) { - printk(KERN_WARNING "cciss%d: retrying 0x%02x\n", h->ctlr, + dev_warn(&h->pdev->dev, "retrying 0x%02x\n", c->Request.CDB[0]); c->retry_count++; /* erase the old error information */ @@ -2719,27 +2746,26 @@ command_done: return return_status; } -static int sendcmd_withirq(__u8 cmd, int ctlr, void *buff, size_t size, +static int sendcmd_withirq(ctlr_info_t *h, __u8 cmd, void *buff, size_t size, __u8 page_code, unsigned char scsi3addr[], int cmd_type) { - ctlr_info_t *h = hba[ctlr]; CommandList_struct *c; int return_status; - c = cmd_alloc(h, 0); + c = cmd_special_alloc(h); if (!c) return -ENOMEM; - return_status = fill_cmd(c, cmd, ctlr, buff, size, page_code, + return_status = fill_cmd(h, c, cmd, buff, size, page_code, scsi3addr, cmd_type); if (return_status == IO_OK) return_status = sendcmd_withirq_core(h, c, 1); - cmd_free(h, c, 0); + cmd_special_free(h, c); return return_status; } -static void cciss_geometry_inquiry(int ctlr, int logvol, +static void cciss_geometry_inquiry(ctlr_info_t *h, int logvol, sector_t total_size, unsigned int block_size, InquiryData_struct *inq_buff, @@ -2750,13 +2776,13 @@ static void cciss_geometry_inquiry(int ctlr, int logvol, unsigned char scsi3addr[8]; memset(inq_buff, 0, sizeof(InquiryData_struct)); - log_unit_to_scsi3addr(hba[ctlr], scsi3addr, logvol); - return_code = sendcmd_withirq(CISS_INQUIRY, ctlr, inq_buff, + log_unit_to_scsi3addr(h, scsi3addr, logvol); + return_code = sendcmd_withirq(h, CISS_INQUIRY, inq_buff, sizeof(*inq_buff), 0xC1, scsi3addr, TYPE_CMD); if (return_code == IO_OK) { if (inq_buff->data_byte[8] == 0xFF) { - printk(KERN_WARNING - "cciss: reading geometry failed, volume " + dev_warn(&h->pdev->dev, + "reading geometry failed, volume " "does not support reading geometry\n"); drv->heads = 255; drv->sectors = 32; /* Sectors per track */ @@ -2780,12 +2806,12 @@ static void cciss_geometry_inquiry(int ctlr, int logvol, drv->cylinders = real_size; } } else { /* Get geometry failed */ - printk(KERN_WARNING "cciss: reading geometry failed\n"); + dev_warn(&h->pdev->dev, "reading geometry failed\n"); } } static void -cciss_read_capacity(int ctlr, int logvol, sector_t *total_size, +cciss_read_capacity(ctlr_info_t *h, int logvol, sector_t *total_size, unsigned int *block_size) { ReadCapdata_struct *buf; @@ -2794,25 +2820,25 @@ cciss_read_capacity(int ctlr, int logvol, sector_t *total_size, buf = kzalloc(sizeof(ReadCapdata_struct), GFP_KERNEL); if (!buf) { - printk(KERN_WARNING "cciss: out of memory\n"); + dev_warn(&h->pdev->dev, "out of memory\n"); return; } - log_unit_to_scsi3addr(hba[ctlr], scsi3addr, logvol); - return_code = sendcmd_withirq(CCISS_READ_CAPACITY, ctlr, buf, + log_unit_to_scsi3addr(h, scsi3addr, logvol); + return_code = sendcmd_withirq(h, CCISS_READ_CAPACITY, buf, sizeof(ReadCapdata_struct), 0, scsi3addr, TYPE_CMD); if (return_code == IO_OK) { *total_size = be32_to_cpu(*(__be32 *) buf->total_size); *block_size = be32_to_cpu(*(__be32 *) buf->block_size); } else { /* read capacity command failed */ - printk(KERN_WARNING "cciss: read capacity failed\n"); + dev_warn(&h->pdev->dev, "read capacity failed\n"); *total_size = 0; *block_size = BLOCK_SIZE; } kfree(buf); } -static void cciss_read_capacity_16(int ctlr, int logvol, +static void cciss_read_capacity_16(ctlr_info_t *h, int logvol, sector_t *total_size, unsigned int *block_size) { ReadCapdata_struct_16 *buf; @@ -2821,23 +2847,23 @@ static void cciss_read_capacity_16(int ctlr, int logvol, buf = kzalloc(sizeof(ReadCapdata_struct_16), GFP_KERNEL); if (!buf) { - printk(KERN_WARNING "cciss: out of memory\n"); + dev_warn(&h->pdev->dev, "out of memory\n"); return; } - log_unit_to_scsi3addr(hba[ctlr], scsi3addr, logvol); - return_code = sendcmd_withirq(CCISS_READ_CAPACITY_16, - ctlr, buf, sizeof(ReadCapdata_struct_16), + log_unit_to_scsi3addr(h, scsi3addr, logvol); + return_code = sendcmd_withirq(h, CCISS_READ_CAPACITY_16, + buf, sizeof(ReadCapdata_struct_16), 0, scsi3addr, TYPE_CMD); if (return_code == IO_OK) { *total_size = be64_to_cpu(*(__be64 *) buf->total_size); *block_size = be32_to_cpu(*(__be32 *) buf->block_size); } else { /* read capacity command failed */ - printk(KERN_WARNING "cciss: read capacity failed\n"); + dev_warn(&h->pdev->dev, "read capacity failed\n"); *total_size = 0; *block_size = BLOCK_SIZE; } - printk(KERN_INFO " blocks= %llu block_size= %d\n", + dev_info(&h->pdev->dev, " blocks= %llu block_size= %d\n", (unsigned long long)*total_size+1, *block_size); kfree(buf); } @@ -2865,17 +2891,17 @@ static int cciss_revalidate(struct gendisk *disk) inq_buff = kmalloc(sizeof(InquiryData_struct), GFP_KERNEL); if (inq_buff == NULL) { - printk(KERN_WARNING "cciss: out of memory\n"); + dev_warn(&h->pdev->dev, "out of memory\n"); return 1; } if (h->cciss_read == CCISS_READ_10) { - cciss_read_capacity(h->ctlr, logvol, + cciss_read_capacity(h, logvol, &total_size, &block_size); } else { - cciss_read_capacity_16(h->ctlr, logvol, + cciss_read_capacity_16(h, logvol, &total_size, &block_size); } - cciss_geometry_inquiry(h->ctlr, logvol, total_size, block_size, + cciss_geometry_inquiry(h, logvol, total_size, block_size, inq_buff, drv); blk_queue_logical_block_size(drv->queue, drv->block_size); @@ -2909,7 +2935,7 @@ static void start_io(ctlr_info_t *h) c = hlist_entry(h->reqQ.first, CommandList_struct, list); /* can't do anything if fifo is full */ if ((h->access.fifo_full(h))) { - printk(KERN_WARNING "cciss: fifo full\n"); + dev_warn(&h->pdev->dev, "fifo full\n"); break; } @@ -2925,7 +2951,7 @@ static void start_io(ctlr_info_t *h) } } -/* Assumes that CCISS_LOCK(h->ctlr) is held. */ +/* Assumes that h->lock is held. */ /* Zeros out the error record and then resends the command back */ /* to the controller */ static inline void resend_cciss_cmd(ctlr_info_t *h, CommandList_struct *c) @@ -2966,7 +2992,7 @@ static inline int evaluate_target_status(ctlr_info_t *h, driver_byte = DRIVER_OK; msg_byte = cmd->err_info->CommandStatus; /* correct? seems too device specific */ - if (blk_pc_request(cmd->rq)) + if (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) host_byte = DID_PASSTHROUGH; else host_byte = DID_OK; @@ -2975,8 +3001,8 @@ static inline int evaluate_target_status(ctlr_info_t *h, host_byte, driver_byte); if (cmd->err_info->ScsiStatus != SAM_STAT_CHECK_CONDITION) { - if (!blk_pc_request(cmd->rq)) - printk(KERN_WARNING "cciss: cmd %p " + if (cmd->rq->cmd_type != REQ_TYPE_BLOCK_PC) + dev_warn(&h->pdev->dev, "cmd %p " "has SCSI Status 0x%x\n", cmd, cmd->err_info->ScsiStatus); return error_value; @@ -2985,17 +3011,19 @@ static inline int evaluate_target_status(ctlr_info_t *h, /* check the sense key */ sense_key = 0xf & cmd->err_info->SenseInfo[2]; /* no status or recovered error */ - if (((sense_key == 0x0) || (sense_key == 0x1)) && !blk_pc_request(cmd->rq)) + if (((sense_key == 0x0) || (sense_key == 0x1)) && + (cmd->rq->cmd_type != REQ_TYPE_BLOCK_PC)) error_value = 0; if (check_for_unit_attention(h, cmd)) { - *retry_cmd = !blk_pc_request(cmd->rq); + *retry_cmd = !(cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC); return 0; } - if (!blk_pc_request(cmd->rq)) { /* Not SG_IO or similar? */ + /* Not SG_IO or similar? */ + if (cmd->rq->cmd_type != REQ_TYPE_BLOCK_PC) { if (error_value != 0) - printk(KERN_WARNING "cciss: cmd %p has CHECK CONDITION" + dev_warn(&h->pdev->dev, "cmd %p has CHECK CONDITION" " sense key = 0x%x\n", cmd, sense_key); return error_value; } @@ -3035,90 +3063,97 @@ static inline void complete_command(ctlr_info_t *h, CommandList_struct *cmd, rq->errors = evaluate_target_status(h, cmd, &retry_cmd); break; case CMD_DATA_UNDERRUN: - if (blk_fs_request(cmd->rq)) { - printk(KERN_WARNING "cciss: cmd %p has" + if (cmd->rq->cmd_type == REQ_TYPE_FS) { + dev_warn(&h->pdev->dev, "cmd %p has" " completed with data underrun " "reported\n", cmd); cmd->rq->resid_len = cmd->err_info->ResidualCnt; } break; case CMD_DATA_OVERRUN: - if (blk_fs_request(cmd->rq)) - printk(KERN_WARNING "cciss: cmd %p has" + if (cmd->rq->cmd_type == REQ_TYPE_FS) + dev_warn(&h->pdev->dev, "cciss: cmd %p has" " completed with data overrun " "reported\n", cmd); break; case CMD_INVALID: - printk(KERN_WARNING "cciss: cmd %p is " + dev_warn(&h->pdev->dev, "cciss: cmd %p is " "reported invalid\n", cmd); rq->errors = make_status_bytes(SAM_STAT_GOOD, cmd->err_info->CommandStatus, DRIVER_OK, - blk_pc_request(cmd->rq) ? DID_PASSTHROUGH : DID_ERROR); + (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ? + DID_PASSTHROUGH : DID_ERROR); break; case CMD_PROTOCOL_ERR: - printk(KERN_WARNING "cciss: cmd %p has " - "protocol error \n", cmd); + dev_warn(&h->pdev->dev, "cciss: cmd %p has " + "protocol error\n", cmd); rq->errors = make_status_bytes(SAM_STAT_GOOD, cmd->err_info->CommandStatus, DRIVER_OK, - blk_pc_request(cmd->rq) ? DID_PASSTHROUGH : DID_ERROR); + (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ? + DID_PASSTHROUGH : DID_ERROR); break; case CMD_HARDWARE_ERR: - printk(KERN_WARNING "cciss: cmd %p had " + dev_warn(&h->pdev->dev, "cciss: cmd %p had " " hardware error\n", cmd); rq->errors = make_status_bytes(SAM_STAT_GOOD, cmd->err_info->CommandStatus, DRIVER_OK, - blk_pc_request(cmd->rq) ? DID_PASSTHROUGH : DID_ERROR); + (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ? + DID_PASSTHROUGH : DID_ERROR); break; case CMD_CONNECTION_LOST: - printk(KERN_WARNING "cciss: cmd %p had " + dev_warn(&h->pdev->dev, "cciss: cmd %p had " "connection lost\n", cmd); rq->errors = make_status_bytes(SAM_STAT_GOOD, cmd->err_info->CommandStatus, DRIVER_OK, - blk_pc_request(cmd->rq) ? DID_PASSTHROUGH : DID_ERROR); + (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ? + DID_PASSTHROUGH : DID_ERROR); break; case CMD_ABORTED: - printk(KERN_WARNING "cciss: cmd %p was " + dev_warn(&h->pdev->dev, "cciss: cmd %p was " "aborted\n", cmd); rq->errors = make_status_bytes(SAM_STAT_GOOD, cmd->err_info->CommandStatus, DRIVER_OK, - blk_pc_request(cmd->rq) ? DID_PASSTHROUGH : DID_ABORT); + (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ? + DID_PASSTHROUGH : DID_ABORT); break; case CMD_ABORT_FAILED: - printk(KERN_WARNING "cciss: cmd %p reports " + dev_warn(&h->pdev->dev, "cciss: cmd %p reports " "abort failed\n", cmd); rq->errors = make_status_bytes(SAM_STAT_GOOD, cmd->err_info->CommandStatus, DRIVER_OK, - blk_pc_request(cmd->rq) ? DID_PASSTHROUGH : DID_ERROR); + (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ? + DID_PASSTHROUGH : DID_ERROR); break; case CMD_UNSOLICITED_ABORT: - printk(KERN_WARNING "cciss%d: unsolicited " + dev_warn(&h->pdev->dev, "cciss%d: unsolicited " "abort %p\n", h->ctlr, cmd); if (cmd->retry_count < MAX_CMD_RETRIES) { retry_cmd = 1; - printk(KERN_WARNING - "cciss%d: retrying %p\n", h->ctlr, cmd); + dev_warn(&h->pdev->dev, "retrying %p\n", cmd); cmd->retry_count++; } else - printk(KERN_WARNING - "cciss%d: %p retried too " - "many times\n", h->ctlr, cmd); + dev_warn(&h->pdev->dev, + "%p retried too many times\n", cmd); rq->errors = make_status_bytes(SAM_STAT_GOOD, cmd->err_info->CommandStatus, DRIVER_OK, - blk_pc_request(cmd->rq) ? DID_PASSTHROUGH : DID_ABORT); + (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ? + DID_PASSTHROUGH : DID_ABORT); break; case CMD_TIMEOUT: - printk(KERN_WARNING "cciss: cmd %p timedout\n", cmd); + dev_warn(&h->pdev->dev, "cmd %p timedout\n", cmd); rq->errors = make_status_bytes(SAM_STAT_GOOD, cmd->err_info->CommandStatus, DRIVER_OK, - blk_pc_request(cmd->rq) ? DID_PASSTHROUGH : DID_ERROR); + (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ? + DID_PASSTHROUGH : DID_ERROR); break; default: - printk(KERN_WARNING "cciss: cmd %p returned " + dev_warn(&h->pdev->dev, "cmd %p returned " "unknown status %x\n", cmd, cmd->err_info->CommandStatus); rq->errors = make_status_bytes(SAM_STAT_GOOD, cmd->err_info->CommandStatus, DRIVER_OK, - blk_pc_request(cmd->rq) ? DID_PASSTHROUGH : DID_ERROR); + (cmd->rq->cmd_type == REQ_TYPE_BLOCK_PC) ? + DID_PASSTHROUGH : DID_ERROR); } after_error_processing: @@ -3132,6 +3167,34 @@ after_error_processing: blk_complete_request(cmd->rq); } +static inline u32 cciss_tag_contains_index(u32 tag) +{ +#define DIRECT_LOOKUP_BIT 0x10 + return tag & DIRECT_LOOKUP_BIT; +} + +static inline u32 cciss_tag_to_index(u32 tag) +{ +#define DIRECT_LOOKUP_SHIFT 5 + return tag >> DIRECT_LOOKUP_SHIFT; +} + +static inline u32 cciss_tag_discard_error_bits(u32 tag) +{ +#define CCISS_ERROR_BITS 0x03 + return tag & ~CCISS_ERROR_BITS; +} + +static inline void cciss_mark_tag_indexed(u32 *tag) +{ + *tag |= DIRECT_LOOKUP_BIT; +} + +static inline void cciss_set_tag_index(u32 *tag, u32 index) +{ + *tag |= (index << DIRECT_LOOKUP_SHIFT); +} + /* * Get a request and submit it to the controller. */ @@ -3163,7 +3226,8 @@ static void do_cciss_request(struct request_queue *q) BUG_ON(creq->nr_phys_segments > h->maxsgentries); - if ((c = cmd_alloc(h, 1)) == NULL) + c = cmd_alloc(h); + if (!c) goto full; blk_start_request(creq); @@ -3180,8 +3244,8 @@ static void do_cciss_request(struct request_queue *q) /* got command from pool, so use the command block index instead */ /* for direct lookups. */ /* The first 2 bits are reserved for controller error reporting. */ - c->Header.Tag.lower = (c->cmdindex << 3); - c->Header.Tag.lower |= 0x04; /* flag for direct lookup. */ + cciss_set_tag_index(&c->Header.Tag.lower, c->cmdindex); + cciss_mark_tag_indexed(&c->Header.Tag.lower); memcpy(&c->Header.LUN, drv->LunID, sizeof(drv->LunID)); c->Request.CDBLen = 10; /* 12 byte commands not in FW yet; */ c->Request.Type.Type = TYPE_CMD; /* It is a command. */ @@ -3192,11 +3256,8 @@ static void do_cciss_request(struct request_queue *q) c->Request.CDB[0] = (rq_data_dir(creq) == READ) ? h->cciss_read : h->cciss_write; start_blk = blk_rq_pos(creq); -#ifdef CCISS_DEBUG - printk(KERN_DEBUG "ciss: sector =%d nr_sectors=%d\n", + dev_dbg(&h->pdev->dev, "sector =%d nr_sectors=%d\n", (int)blk_rq_pos(creq), (int)blk_rq_sectors(creq)); -#endif /* CCISS_DEBUG */ - sg_init_table(tmp_sg, h->maxsgentries); seg = blk_rq_map_sg(q, creq, tmp_sg); @@ -3236,17 +3297,18 @@ static void do_cciss_request(struct request_queue *q) if (seg > h->maxSG) h->maxSG = seg; -#ifdef CCISS_DEBUG - printk(KERN_DEBUG "cciss: Submitting %ld sectors in %d segments " + dev_dbg(&h->pdev->dev, "Submitting %u sectors in %d segments " "chained[%d]\n", blk_rq_sectors(creq), seg, chained); -#endif /* CCISS_DEBUG */ - c->Header.SGList = c->Header.SGTotal = seg + chained; - if (seg > h->max_cmd_sgentries) + c->Header.SGTotal = seg + chained; + if (seg <= h->max_cmd_sgentries) + c->Header.SGList = c->Header.SGTotal; + else c->Header.SGList = h->max_cmd_sgentries; + set_performant_mode(h, c); - if (likely(blk_fs_request(creq))) { + if (likely(creq->cmd_type == REQ_TYPE_FS)) { if(h->cciss_read == CCISS_READ_10) { c->Request.CDB[1] = 0; c->Request.CDB[2] = (start_blk >> 24) & 0xff; /* MSB */ @@ -3276,11 +3338,12 @@ static void do_cciss_request(struct request_queue *q) c->Request.CDB[13]= blk_rq_sectors(creq) & 0xff; c->Request.CDB[14] = c->Request.CDB[15] = 0; } - } else if (blk_pc_request(creq)) { + } else if (creq->cmd_type == REQ_TYPE_BLOCK_PC) { c->Request.CDBLen = creq->cmd_len; memcpy(c->Request.CDB, creq->cmd, BLK_MAX_CDB); } else { - printk(KERN_WARNING "cciss%d: bad request type %d\n", h->ctlr, creq->cmd_type); + dev_warn(&h->pdev->dev, "bad request type %d\n", + creq->cmd_type); BUG(); } @@ -3313,72 +3376,131 @@ static inline int interrupt_pending(ctlr_info_t *h) static inline long interrupt_not_for_us(ctlr_info_t *h) { - return (((h->access.intr_pending(h) == 0) || - (h->interrupts_enabled == 0))); + return ((h->access.intr_pending(h) == 0) || + (h->interrupts_enabled == 0)); } -static irqreturn_t do_cciss_intr(int irq, void *dev_id) +static inline int bad_tag(ctlr_info_t *h, u32 tag_index, + u32 raw_tag) { - ctlr_info_t *h = dev_id; + if (unlikely(tag_index >= h->nr_cmds)) { + dev_warn(&h->pdev->dev, "bad tag 0x%08x ignored.\n", raw_tag); + return 1; + } + return 0; +} + +static inline void finish_cmd(ctlr_info_t *h, CommandList_struct *c, + u32 raw_tag) +{ + removeQ(c); + if (likely(c->cmd_type == CMD_RWREQ)) + complete_command(h, c, 0); + else if (c->cmd_type == CMD_IOCTL_PEND) + complete(c->waiting); +#ifdef CONFIG_CISS_SCSI_TAPE + else if (c->cmd_type == CMD_SCSI) + complete_scsi_command(c, 0, raw_tag); +#endif +} + +static inline u32 next_command(ctlr_info_t *h) +{ + u32 a; + + if (unlikely(h->transMethod != CFGTBL_Trans_Performant)) + return h->access.command_completed(h); + + if ((*(h->reply_pool_head) & 1) == (h->reply_pool_wraparound)) { + a = *(h->reply_pool_head); /* Next cmd in ring buffer */ + (h->reply_pool_head)++; + h->commands_outstanding--; + } else { + a = FIFO_EMPTY; + } + /* Check for wraparound */ + if (h->reply_pool_head == (h->reply_pool + h->max_commands)) { + h->reply_pool_head = h->reply_pool; + h->reply_pool_wraparound ^= 1; + } + return a; +} + +/* process completion of an indexed ("direct lookup") command */ +static inline u32 process_indexed_cmd(ctlr_info_t *h, u32 raw_tag) +{ + u32 tag_index; CommandList_struct *c; + + tag_index = cciss_tag_to_index(raw_tag); + if (bad_tag(h, tag_index, raw_tag)) + return next_command(h); + c = h->cmd_pool + tag_index; + finish_cmd(h, c, raw_tag); + return next_command(h); +} + +/* process completion of a non-indexed command */ +static inline u32 process_nonindexed_cmd(ctlr_info_t *h, u32 raw_tag) +{ + u32 tag; + CommandList_struct *c = NULL; + struct hlist_node *tmp; + __u32 busaddr_masked, tag_masked; + + tag = cciss_tag_discard_error_bits(raw_tag); + hlist_for_each_entry(c, tmp, &h->cmpQ, list) { + busaddr_masked = cciss_tag_discard_error_bits(c->busaddr); + tag_masked = cciss_tag_discard_error_bits(tag); + if (busaddr_masked == tag_masked) { + finish_cmd(h, c, raw_tag); + return next_command(h); + } + } + bad_tag(h, h->nr_cmds + 1, raw_tag); + return next_command(h); +} + +static irqreturn_t do_cciss_intx(int irq, void *dev_id) +{ + ctlr_info_t *h = dev_id; unsigned long flags; - __u32 a, a1, a2; + u32 raw_tag; if (interrupt_not_for_us(h)) return IRQ_NONE; - /* - * If there are completed commands in the completion queue, - * we had better do something about it. - */ - spin_lock_irqsave(CCISS_LOCK(h->ctlr), flags); + spin_lock_irqsave(&h->lock, flags); while (interrupt_pending(h)) { - while ((a = get_next_completion(h)) != FIFO_EMPTY) { - a1 = a; - if ((a & 0x04)) { - a2 = (a >> 3); - if (a2 >= h->nr_cmds) { - printk(KERN_WARNING - "cciss: controller cciss%d failed, stopping.\n", - h->ctlr); - spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags); - fail_all_cmds(h->ctlr); - return IRQ_HANDLED; - } - - c = h->cmd_pool + a2; - a = c->busaddr; - - } else { - struct hlist_node *tmp; - - a &= ~3; - c = NULL; - hlist_for_each_entry(c, tmp, &h->cmpQ, list) { - if (c->busaddr == a) - break; - } - } - /* - * If we've found the command, take it off the - * completion Q and free it - */ - if (c && c->busaddr == a) { - removeQ(c); - if (c->cmd_type == CMD_RWREQ) { - complete_command(h, c, 0); - } else if (c->cmd_type == CMD_IOCTL_PEND) { - complete(c->waiting); - } -# ifdef CONFIG_CISS_SCSI_TAPE - else if (c->cmd_type == CMD_SCSI) - complete_scsi_command(c, 0, a1); -# endif - continue; - } + raw_tag = get_next_completion(h); + while (raw_tag != FIFO_EMPTY) { + if (cciss_tag_contains_index(raw_tag)) + raw_tag = process_indexed_cmd(h, raw_tag); + else + raw_tag = process_nonindexed_cmd(h, raw_tag); } } + spin_unlock_irqrestore(&h->lock, flags); + return IRQ_HANDLED; +} - spin_unlock_irqrestore(CCISS_LOCK(h->ctlr), flags); +/* Add a second interrupt handler for MSI/MSI-X mode. In this mode we never + * check the interrupt pending register because it is not set. + */ +static irqreturn_t do_cciss_msix_intr(int irq, void *dev_id) +{ + ctlr_info_t *h = dev_id; + unsigned long flags; + u32 raw_tag; + + spin_lock_irqsave(&h->lock, flags); + raw_tag = get_next_completion(h); + while (raw_tag != FIFO_EMPTY) { + if (cciss_tag_contains_index(raw_tag)) + raw_tag = process_indexed_cmd(h, raw_tag); + else + raw_tag = process_nonindexed_cmd(h, raw_tag); + } + spin_unlock_irqrestore(&h->lock, flags); return IRQ_HANDLED; } @@ -3510,18 +3632,17 @@ static int check_for_unit_attention(ctlr_info_t *h, CommandList_struct *c) switch (c->err_info->SenseInfo[12]) { case STATE_CHANGED: - printk(KERN_WARNING "cciss%d: a state change " - "detected, command retried\n", h->ctlr); + dev_warn(&h->pdev->dev, "a state change " + "detected, command retried\n"); return 1; break; case LUN_FAILED: - printk(KERN_WARNING "cciss%d: LUN failure " - "detected, action required\n", h->ctlr); + dev_warn(&h->pdev->dev, "LUN failure " + "detected, action required\n"); return 1; break; case REPORT_LUNS_CHANGED: - printk(KERN_WARNING "cciss%d: report LUN data " - "changed\n", h->ctlr); + dev_warn(&h->pdev->dev, "report LUN data changed\n"); /* * Here, we could call add_to_scan_list and wake up the scan thread, * except that it's quite likely that we will get more than one @@ -3541,19 +3662,18 @@ static int check_for_unit_attention(ctlr_info_t *h, CommandList_struct *c) return 1; break; case POWER_OR_RESET: - printk(KERN_WARNING "cciss%d: a power on " - "or device reset detected\n", h->ctlr); + dev_warn(&h->pdev->dev, + "a power on or device reset detected\n"); return 1; break; case UNIT_ATTENTION_CLEARED: - printk(KERN_WARNING "cciss%d: unit attention " - "cleared by another initiator\n", h->ctlr); + dev_warn(&h->pdev->dev, + "unit attention cleared by another initiator\n"); return 1; break; default: - printk(KERN_WARNING "cciss%d: unknown " - "unit attention detected\n", h->ctlr); - return 1; + dev_warn(&h->pdev->dev, "unknown unit attention detected\n"); + return 1; } } @@ -3562,39 +3682,41 @@ static int check_for_unit_attention(ctlr_info_t *h, CommandList_struct *c) * the io functions. * This is for debug only. */ -#ifdef CCISS_DEBUG -static void print_cfg_table(CfgTable_struct *tb) +static void print_cfg_table(ctlr_info_t *h) { int i; char temp_name[17]; + CfgTable_struct *tb = h->cfgtable; - printk("Controller Configuration information\n"); - printk("------------------------------------\n"); + dev_dbg(&h->pdev->dev, "Controller Configuration information\n"); + dev_dbg(&h->pdev->dev, "------------------------------------\n"); for (i = 0; i < 4; i++) temp_name[i] = readb(&(tb->Signature[i])); temp_name[4] = '\0'; - printk(" Signature = %s\n", temp_name); - printk(" Spec Number = %d\n", readl(&(tb->SpecValence))); - printk(" Transport methods supported = 0x%x\n", + dev_dbg(&h->pdev->dev, " Signature = %s\n", temp_name); + dev_dbg(&h->pdev->dev, " Spec Number = %d\n", + readl(&(tb->SpecValence))); + dev_dbg(&h->pdev->dev, " Transport methods supported = 0x%x\n", readl(&(tb->TransportSupport))); - printk(" Transport methods active = 0x%x\n", + dev_dbg(&h->pdev->dev, " Transport methods active = 0x%x\n", readl(&(tb->TransportActive))); - printk(" Requested transport Method = 0x%x\n", + dev_dbg(&h->pdev->dev, " Requested transport Method = 0x%x\n", readl(&(tb->HostWrite.TransportRequest))); - printk(" Coalesce Interrupt Delay = 0x%x\n", + dev_dbg(&h->pdev->dev, " Coalesce Interrupt Delay = 0x%x\n", readl(&(tb->HostWrite.CoalIntDelay))); - printk(" Coalesce Interrupt Count = 0x%x\n", + dev_dbg(&h->pdev->dev, " Coalesce Interrupt Count = 0x%x\n", readl(&(tb->HostWrite.CoalIntCount))); - printk(" Max outstanding commands = 0x%d\n", + dev_dbg(&h->pdev->dev, " Max outstanding commands = 0x%d\n", readl(&(tb->CmdsOutMax))); - printk(" Bus Types = 0x%x\n", readl(&(tb->BusTypes))); + dev_dbg(&h->pdev->dev, " Bus Types = 0x%x\n", + readl(&(tb->BusTypes))); for (i = 0; i < 16; i++) temp_name[i] = readb(&(tb->ServerName[i])); temp_name[16] = '\0'; - printk(" Server Name = %s\n", temp_name); - printk(" Heartbeat Counter = 0x%x\n\n\n", readl(&(tb->HeartBeat))); + dev_dbg(&h->pdev->dev, " Server Name = %s\n", temp_name); + dev_dbg(&h->pdev->dev, " Heartbeat Counter = 0x%x\n\n\n", + readl(&(tb->HeartBeat))); } -#endif /* CCISS_DEBUG */ static int find_PCI_BAR_index(struct pci_dev *pdev, unsigned long pci_bar_addr) { @@ -3618,7 +3740,7 @@ static int find_PCI_BAR_index(struct pci_dev *pdev, unsigned long pci_bar_addr) offset += 8; break; default: /* reserved in PCI 2.2 */ - printk(KERN_WARNING + dev_warn(&pdev->dev, "Base address is invalid\n"); return -1; break; @@ -3630,12 +3752,182 @@ static int find_PCI_BAR_index(struct pci_dev *pdev, unsigned long pci_bar_addr) return -1; } +/* Fill in bucket_map[], given nsgs (the max number of + * scatter gather elements supported) and bucket[], + * which is an array of 8 integers. The bucket[] array + * contains 8 different DMA transfer sizes (in 16 + * byte increments) which the controller uses to fetch + * commands. This function fills in bucket_map[], which + * maps a given number of scatter gather elements to one of + * the 8 DMA transfer sizes. The point of it is to allow the + * controller to only do as much DMA as needed to fetch the + * command, with the DMA transfer size encoded in the lower + * bits of the command address. + */ +static void calc_bucket_map(int bucket[], int num_buckets, + int nsgs, int *bucket_map) +{ + int i, j, b, size; + + /* even a command with 0 SGs requires 4 blocks */ +#define MINIMUM_TRANSFER_BLOCKS 4 +#define NUM_BUCKETS 8 + /* Note, bucket_map must have nsgs+1 entries. */ + for (i = 0; i <= nsgs; i++) { + /* Compute size of a command with i SG entries */ + size = i + MINIMUM_TRANSFER_BLOCKS; + b = num_buckets; /* Assume the biggest bucket */ + /* Find the bucket that is just big enough */ + for (j = 0; j < 8; j++) { + if (bucket[j] >= size) { + b = j; + break; + } + } + /* for a command with i SG entries, use bucket b. */ + bucket_map[i] = b; + } +} + +static void __devinit cciss_wait_for_mode_change_ack(ctlr_info_t *h) +{ + int i; + + /* under certain very rare conditions, this can take awhile. + * (e.g.: hot replace a failed 144GB drive in a RAID 5 set right + * as we enter this code.) */ + for (i = 0; i < MAX_CONFIG_WAIT; i++) { + if (!(readl(h->vaddr + SA5_DOORBELL) & CFGTBL_ChangeReq)) + break; + msleep(10); + } +} + +static __devinit void cciss_enter_performant_mode(ctlr_info_t *h) +{ + /* This is a bit complicated. There are 8 registers on + * the controller which we write to to tell it 8 different + * sizes of commands which there may be. It's a way of + * reducing the DMA done to fetch each command. Encoded into + * each command's tag are 3 bits which communicate to the controller + * which of the eight sizes that command fits within. The size of + * each command depends on how many scatter gather entries there are. + * Each SG entry requires 16 bytes. The eight registers are programmed + * with the number of 16-byte blocks a command of that size requires. + * The smallest command possible requires 5 such 16 byte blocks. + * the largest command possible requires MAXSGENTRIES + 4 16-byte + * blocks. Note, this only extends to the SG entries contained + * within the command block, and does not extend to chained blocks + * of SG elements. bft[] contains the eight values we write to + * the registers. They are not evenly distributed, but have more + * sizes for small commands, and fewer sizes for larger commands. + */ + __u32 trans_offset; + int bft[8] = { 5, 6, 8, 10, 12, 20, 28, MAXSGENTRIES + 4}; + /* + * 5 = 1 s/g entry or 4k + * 6 = 2 s/g entry or 8k + * 8 = 4 s/g entry or 16k + * 10 = 6 s/g entry or 24k + */ + unsigned long register_value; + BUILD_BUG_ON(28 > MAXSGENTRIES + 4); + + h->reply_pool_wraparound = 1; /* spec: init to 1 */ + + /* Controller spec: zero out this buffer. */ + memset(h->reply_pool, 0, h->max_commands * sizeof(__u64)); + h->reply_pool_head = h->reply_pool; + + trans_offset = readl(&(h->cfgtable->TransMethodOffset)); + calc_bucket_map(bft, ARRAY_SIZE(bft), h->maxsgentries, + h->blockFetchTable); + writel(bft[0], &h->transtable->BlockFetch0); + writel(bft[1], &h->transtable->BlockFetch1); + writel(bft[2], &h->transtable->BlockFetch2); + writel(bft[3], &h->transtable->BlockFetch3); + writel(bft[4], &h->transtable->BlockFetch4); + writel(bft[5], &h->transtable->BlockFetch5); + writel(bft[6], &h->transtable->BlockFetch6); + writel(bft[7], &h->transtable->BlockFetch7); + + /* size of controller ring buffer */ + writel(h->max_commands, &h->transtable->RepQSize); + writel(1, &h->transtable->RepQCount); + writel(0, &h->transtable->RepQCtrAddrLow32); + writel(0, &h->transtable->RepQCtrAddrHigh32); + writel(h->reply_pool_dhandle, &h->transtable->RepQAddr0Low32); + writel(0, &h->transtable->RepQAddr0High32); + writel(CFGTBL_Trans_Performant, + &(h->cfgtable->HostWrite.TransportRequest)); + + writel(CFGTBL_ChangeReq, h->vaddr + SA5_DOORBELL); + cciss_wait_for_mode_change_ack(h); + register_value = readl(&(h->cfgtable->TransportActive)); + if (!(register_value & CFGTBL_Trans_Performant)) + dev_warn(&h->pdev->dev, "cciss: unable to get board into" + " performant mode\n"); +} + +static void __devinit cciss_put_controller_into_performant_mode(ctlr_info_t *h) +{ + __u32 trans_support; + + dev_dbg(&h->pdev->dev, "Trying to put board into Performant mode\n"); + /* Attempt to put controller into performant mode if supported */ + /* Does board support performant mode? */ + trans_support = readl(&(h->cfgtable->TransportSupport)); + if (!(trans_support & PERFORMANT_MODE)) + return; + + dev_dbg(&h->pdev->dev, "Placing controller into performant mode\n"); + /* Performant mode demands commands on a 32 byte boundary + * pci_alloc_consistent aligns on page boundarys already. + * Just need to check if divisible by 32 + */ + if ((sizeof(CommandList_struct) % 32) != 0) { + dev_warn(&h->pdev->dev, "%s %d %s\n", + "cciss info: command size[", + (int)sizeof(CommandList_struct), + "] not divisible by 32, no performant mode..\n"); + return; + } + + /* Performant mode ring buffer and supporting data structures */ + h->reply_pool = (__u64 *)pci_alloc_consistent( + h->pdev, h->max_commands * sizeof(__u64), + &(h->reply_pool_dhandle)); + + /* Need a block fetch table for performant mode */ + h->blockFetchTable = kmalloc(((h->maxsgentries+1) * + sizeof(__u32)), GFP_KERNEL); + + if ((h->reply_pool == NULL) || (h->blockFetchTable == NULL)) + goto clean_up; + + cciss_enter_performant_mode(h); + + /* Change the access methods to the performant access methods */ + h->access = SA5_performant_access; + h->transMethod = CFGTBL_Trans_Performant; + + return; +clean_up: + kfree(h->blockFetchTable); + if (h->reply_pool) + pci_free_consistent(h->pdev, + h->max_commands * sizeof(__u64), + h->reply_pool, + h->reply_pool_dhandle); + return; + +} /* cciss_put_controller_into_performant_mode */ + /* If MSI/MSI-X is supported by the kernel we will try to enable it on * controllers that are capable. If not, we use IO-APIC mode. */ -static void __devinit cciss_interrupt_mode(ctlr_info_t *c, - struct pci_dev *pdev, __u32 board_id) +static void __devinit cciss_interrupt_mode(ctlr_info_t *h) { #ifdef CONFIG_PCI_MSI int err; @@ -3644,268 +3936,283 @@ static void __devinit cciss_interrupt_mode(ctlr_info_t *c, }; /* Some boards advertise MSI but don't really support it */ - if ((board_id == 0x40700E11) || - (board_id == 0x40800E11) || - (board_id == 0x40820E11) || (board_id == 0x40830E11)) + if ((h->board_id == 0x40700E11) || (h->board_id == 0x40800E11) || + (h->board_id == 0x40820E11) || (h->board_id == 0x40830E11)) goto default_int_mode; - if (pci_find_capability(pdev, PCI_CAP_ID_MSIX)) { - err = pci_enable_msix(pdev, cciss_msix_entries, 4); + if (pci_find_capability(h->pdev, PCI_CAP_ID_MSIX)) { + err = pci_enable_msix(h->pdev, cciss_msix_entries, 4); if (!err) { - c->intr[0] = cciss_msix_entries[0].vector; - c->intr[1] = cciss_msix_entries[1].vector; - c->intr[2] = cciss_msix_entries[2].vector; - c->intr[3] = cciss_msix_entries[3].vector; - c->msix_vector = 1; + h->intr[0] = cciss_msix_entries[0].vector; + h->intr[1] = cciss_msix_entries[1].vector; + h->intr[2] = cciss_msix_entries[2].vector; + h->intr[3] = cciss_msix_entries[3].vector; + h->msix_vector = 1; return; } if (err > 0) { - printk(KERN_WARNING "cciss: only %d MSI-X vectors " - "available\n", err); + dev_warn(&h->pdev->dev, + "only %d MSI-X vectors available\n", err); goto default_int_mode; } else { - printk(KERN_WARNING "cciss: MSI-X init failed %d\n", - err); + dev_warn(&h->pdev->dev, + "MSI-X init failed %d\n", err); goto default_int_mode; } } - if (pci_find_capability(pdev, PCI_CAP_ID_MSI)) { - if (!pci_enable_msi(pdev)) { - c->msi_vector = 1; - } else { - printk(KERN_WARNING "cciss: MSI init failed\n"); - } + if (pci_find_capability(h->pdev, PCI_CAP_ID_MSI)) { + if (!pci_enable_msi(h->pdev)) + h->msi_vector = 1; + else + dev_warn(&h->pdev->dev, "MSI init failed\n"); } default_int_mode: #endif /* CONFIG_PCI_MSI */ /* if we get here we're going to use the default interrupt mode */ - c->intr[SIMPLE_MODE_INT] = pdev->irq; + h->intr[PERF_MODE_INT] = h->pdev->irq; return; } -static int __devinit cciss_pci_init(ctlr_info_t *c, struct pci_dev *pdev) +static int __devinit cciss_lookup_board_id(struct pci_dev *pdev, u32 *board_id) { - ushort subsystem_vendor_id, subsystem_device_id, command; - __u32 board_id, scratchpad = 0; - __u64 cfg_offset; - __u32 cfg_base_addr; - __u64 cfg_base_addr_index; - int i, prod_index, err; + int i; + u32 subsystem_vendor_id, subsystem_device_id; subsystem_vendor_id = pdev->subsystem_vendor; subsystem_device_id = pdev->subsystem_device; - board_id = (((__u32) (subsystem_device_id << 16) & 0xffff0000) | - subsystem_vendor_id); + *board_id = ((subsystem_device_id << 16) & 0xffff0000) | + subsystem_vendor_id; for (i = 0; i < ARRAY_SIZE(products); i++) { /* Stand aside for hpsa driver on request */ if (cciss_allow_hpsa && products[i].board_id == HPSA_BOUNDARY) return -ENODEV; - if (board_id == products[i].board_id) - break; - } - prod_index = i; - if (prod_index == ARRAY_SIZE(products)) { - dev_warn(&pdev->dev, - "unrecognized board ID: 0x%08lx, ignoring.\n", - (unsigned long) board_id); - return -ENODEV; + if (*board_id == products[i].board_id) + return i; } + dev_warn(&pdev->dev, "unrecognized board ID: 0x%08x, ignoring.\n", + *board_id); + return -ENODEV; +} - /* check to see if controller has been disabled */ - /* BEFORE trying to enable it */ - (void)pci_read_config_word(pdev, PCI_COMMAND, &command); - if (!(command & 0x02)) { - printk(KERN_WARNING - "cciss: controller appears to be disabled\n"); - return -ENODEV; - } +static inline bool cciss_board_disabled(ctlr_info_t *h) +{ + u16 command; - err = pci_enable_device(pdev); - if (err) { - printk(KERN_ERR "cciss: Unable to Enable PCI device\n"); - return err; - } + (void) pci_read_config_word(h->pdev, PCI_COMMAND, &command); + return ((command & PCI_COMMAND_MEMORY) == 0); +} - err = pci_request_regions(pdev, "cciss"); - if (err) { - printk(KERN_ERR "cciss: Cannot obtain PCI resources, " - "aborting\n"); - return err; - } +static int __devinit cciss_pci_find_memory_BAR(struct pci_dev *pdev, + unsigned long *memory_bar) +{ + int i; -#ifdef CCISS_DEBUG - printk("command = %x\n", command); - printk("irq = %x\n", pdev->irq); - printk("board_id = %x\n", board_id); -#endif /* CCISS_DEBUG */ + for (i = 0; i < DEVICE_COUNT_RESOURCE; i++) + if (pci_resource_flags(pdev, i) & IORESOURCE_MEM) { + /* addressing mode bits already removed */ + *memory_bar = pci_resource_start(pdev, i); + dev_dbg(&pdev->dev, "memory BAR = %lx\n", + *memory_bar); + return 0; + } + dev_warn(&pdev->dev, "no memory BAR found\n"); + return -ENODEV; +} -/* If the kernel supports MSI/MSI-X we will try to enable that functionality, - * else we use the IO-APIC interrupt assigned to us by system ROM. - */ - cciss_interrupt_mode(c, pdev, board_id); +static int __devinit cciss_wait_for_board_ready(ctlr_info_t *h) +{ + int i; + u32 scratchpad; - /* find the memory BAR */ - for (i = 0; i < DEVICE_COUNT_RESOURCE; i++) { - if (pci_resource_flags(pdev, i) & IORESOURCE_MEM) - break; - } - if (i == DEVICE_COUNT_RESOURCE) { - printk(KERN_WARNING "cciss: No memory BAR found\n"); - err = -ENODEV; - goto err_out_free_res; + for (i = 0; i < CCISS_BOARD_READY_ITERATIONS; i++) { + scratchpad = readl(h->vaddr + SA5_SCRATCHPAD_OFFSET); + if (scratchpad == CCISS_FIRMWARE_READY) + return 0; + msleep(CCISS_BOARD_READY_POLL_INTERVAL_MSECS); } + dev_warn(&h->pdev->dev, "board not ready, timed out.\n"); + return -ENODEV; +} - c->paddr = pci_resource_start(pdev, i); /* addressing mode bits - * already removed - */ +static int __devinit cciss_find_cfg_addrs(struct pci_dev *pdev, + void __iomem *vaddr, u32 *cfg_base_addr, u64 *cfg_base_addr_index, + u64 *cfg_offset) +{ + *cfg_base_addr = readl(vaddr + SA5_CTCFG_OFFSET); + *cfg_offset = readl(vaddr + SA5_CTMEM_OFFSET); + *cfg_base_addr &= (u32) 0x0000ffff; + *cfg_base_addr_index = find_PCI_BAR_index(pdev, *cfg_base_addr); + if (*cfg_base_addr_index == -1) { + dev_warn(&pdev->dev, "cannot find cfg_base_addr_index, " + "*cfg_base_addr = 0x%08x\n", *cfg_base_addr); + return -ENODEV; + } + return 0; +} -#ifdef CCISS_DEBUG - printk("address 0 = %lx\n", c->paddr); -#endif /* CCISS_DEBUG */ - c->vaddr = remap_pci_mem(c->paddr, 0x250); +static int __devinit cciss_find_cfgtables(ctlr_info_t *h) +{ + u64 cfg_offset; + u32 cfg_base_addr; + u64 cfg_base_addr_index; + u32 trans_offset; + int rc; - /* Wait for the board to become ready. (PCI hotplug needs this.) - * We poll for up to 120 secs, once per 100ms. */ - for (i = 0; i < 1200; i++) { - scratchpad = readl(c->vaddr + SA5_SCRATCHPAD_OFFSET); - if (scratchpad == CCISS_FIRMWARE_READY) - break; - set_current_state(TASK_INTERRUPTIBLE); - schedule_timeout(msecs_to_jiffies(100)); /* wait 100ms */ - } - if (scratchpad != CCISS_FIRMWARE_READY) { - printk(KERN_WARNING "cciss: Board not ready. Timed out.\n"); - err = -ENODEV; - goto err_out_free_res; - } + rc = cciss_find_cfg_addrs(h->pdev, h->vaddr, &cfg_base_addr, + &cfg_base_addr_index, &cfg_offset); + if (rc) + return rc; + h->cfgtable = remap_pci_mem(pci_resource_start(h->pdev, + cfg_base_addr_index) + cfg_offset, sizeof(h->cfgtable)); + if (!h->cfgtable) + return -ENOMEM; + /* Find performant mode table. */ + trans_offset = readl(&h->cfgtable->TransMethodOffset); + h->transtable = remap_pci_mem(pci_resource_start(h->pdev, + cfg_base_addr_index)+cfg_offset+trans_offset, + sizeof(*h->transtable)); + if (!h->transtable) + return -ENOMEM; + return 0; +} - /* get the address index number */ - cfg_base_addr = readl(c->vaddr + SA5_CTCFG_OFFSET); - cfg_base_addr &= (__u32) 0x0000ffff; -#ifdef CCISS_DEBUG - printk("cfg base address = %x\n", cfg_base_addr); -#endif /* CCISS_DEBUG */ - cfg_base_addr_index = find_PCI_BAR_index(pdev, cfg_base_addr); -#ifdef CCISS_DEBUG - printk("cfg base address index = %llx\n", - (unsigned long long)cfg_base_addr_index); -#endif /* CCISS_DEBUG */ - if (cfg_base_addr_index == -1) { - printk(KERN_WARNING "cciss: Cannot find cfg_base_addr_index\n"); - err = -ENODEV; - goto err_out_free_res; +static void __devinit cciss_get_max_perf_mode_cmds(struct ctlr_info *h) +{ + h->max_commands = readl(&(h->cfgtable->MaxPerformantModeCommands)); + if (h->max_commands < 16) { + dev_warn(&h->pdev->dev, "Controller reports " + "max supported commands of %d, an obvious lie. " + "Using 16. Ensure that firmware is up to date.\n", + h->max_commands); + h->max_commands = 16; } +} - cfg_offset = readl(c->vaddr + SA5_CTMEM_OFFSET); -#ifdef CCISS_DEBUG - printk("cfg offset = %llx\n", (unsigned long long)cfg_offset); -#endif /* CCISS_DEBUG */ - c->cfgtable = remap_pci_mem(pci_resource_start(pdev, - cfg_base_addr_index) + - cfg_offset, sizeof(CfgTable_struct)); - c->board_id = board_id; - -#ifdef CCISS_DEBUG - print_cfg_table(c->cfgtable); -#endif /* CCISS_DEBUG */ - - /* Some controllers support Zero Memory Raid (ZMR). - * When configured in ZMR mode the number of supported - * commands drops to 64. So instead of just setting an - * arbitrary value we make the driver a little smarter. - * We read the config table to tell us how many commands - * are supported on the controller then subtract 4 to - * leave a little room for ioctl calls. - */ - c->max_commands = readl(&(c->cfgtable->CmdsOutMax)); - c->maxsgentries = readl(&(c->cfgtable->MaxSGElements)); - +/* Interrogate the hardware for some limits: + * max commands, max SG elements without chaining, and with chaining, + * SG chain block size, etc. + */ +static void __devinit cciss_find_board_params(ctlr_info_t *h) +{ + cciss_get_max_perf_mode_cmds(h); + h->nr_cmds = h->max_commands - 4; /* Allow room for some ioctls */ + h->maxsgentries = readl(&(h->cfgtable->MaxSGElements)); /* - * Limit native command to 32 s/g elements to save dma'able memory. + * Limit in-command s/g elements to 32 save dma'able memory. * Howvever spec says if 0, use 31 */ - - c->max_cmd_sgentries = 31; - if (c->maxsgentries > 512) { - c->max_cmd_sgentries = 32; - c->chainsize = c->maxsgentries - c->max_cmd_sgentries + 1; - c->maxsgentries -= 1; /* account for chain pointer */ + h->max_cmd_sgentries = 31; + if (h->maxsgentries > 512) { + h->max_cmd_sgentries = 32; + h->chainsize = h->maxsgentries - h->max_cmd_sgentries + 1; + h->maxsgentries--; /* save one for chain pointer */ } else { - c->maxsgentries = 31; /* Default to traditional value */ - c->chainsize = 0; /* traditional */ + h->maxsgentries = 31; /* default to traditional values */ + h->chainsize = 0; } +} - c->product_name = products[prod_index].product_name; - c->access = *(products[prod_index].access); - c->nr_cmds = c->max_commands - 4; - if ((readb(&c->cfgtable->Signature[0]) != 'C') || - (readb(&c->cfgtable->Signature[1]) != 'I') || - (readb(&c->cfgtable->Signature[2]) != 'S') || - (readb(&c->cfgtable->Signature[3]) != 'S')) { - printk("Does not appear to be a valid CISS config table\n"); - err = -ENODEV; - goto err_out_free_res; +static inline bool CISS_signature_present(ctlr_info_t *h) +{ + if ((readb(&h->cfgtable->Signature[0]) != 'C') || + (readb(&h->cfgtable->Signature[1]) != 'I') || + (readb(&h->cfgtable->Signature[2]) != 'S') || + (readb(&h->cfgtable->Signature[3]) != 'S')) { + dev_warn(&h->pdev->dev, "not a valid CISS config table\n"); + return false; } + return true; +} + +/* Need to enable prefetch in the SCSI core for 6400 in x86 */ +static inline void cciss_enable_scsi_prefetch(ctlr_info_t *h) +{ #ifdef CONFIG_X86 - { - /* Need to enable prefetch in the SCSI core for 6400 in x86 */ - __u32 prefetch; - prefetch = readl(&(c->cfgtable->SCSI_Prefetch)); - prefetch |= 0x100; - writel(prefetch, &(c->cfgtable->SCSI_Prefetch)); - } + u32 prefetch; + + prefetch = readl(&(h->cfgtable->SCSI_Prefetch)); + prefetch |= 0x100; + writel(prefetch, &(h->cfgtable->SCSI_Prefetch)); #endif +} - /* Disabling DMA prefetch and refetch for the P600. - * An ASIC bug may result in accesses to invalid memory addresses. - * We've disabled prefetch for some time now. Testing with XEN - * kernels revealed a bug in the refetch if dom0 resides on a P600. - */ - if(board_id == 0x3225103C) { - __u32 dma_prefetch; - __u32 dma_refetch; - dma_prefetch = readl(c->vaddr + I2O_DMA1_CFG); - dma_prefetch |= 0x8000; - writel(dma_prefetch, c->vaddr + I2O_DMA1_CFG); - pci_read_config_dword(pdev, PCI_COMMAND_PARITY, &dma_refetch); - dma_refetch |= 0x1; - pci_write_config_dword(pdev, PCI_COMMAND_PARITY, dma_refetch); +/* Disable DMA prefetch for the P600. Otherwise an ASIC bug may result + * in a prefetch beyond physical memory. + */ +static inline void cciss_p600_dma_prefetch_quirk(ctlr_info_t *h) +{ + u32 dma_prefetch; + __u32 dma_refetch; + + if (h->board_id != 0x3225103C) + return; + dma_prefetch = readl(h->vaddr + I2O_DMA1_CFG); + dma_prefetch |= 0x8000; + writel(dma_prefetch, h->vaddr + I2O_DMA1_CFG); + pci_read_config_dword(h->pdev, PCI_COMMAND_PARITY, &dma_refetch); + dma_refetch |= 0x1; + pci_write_config_dword(h->pdev, PCI_COMMAND_PARITY, dma_refetch); +} + +static int __devinit cciss_pci_init(ctlr_info_t *h) +{ + int prod_index, err; + + prod_index = cciss_lookup_board_id(h->pdev, &h->board_id); + if (prod_index < 0) + return -ENODEV; + h->product_name = products[prod_index].product_name; + h->access = *(products[prod_index].access); + + if (cciss_board_disabled(h)) { + dev_warn(&h->pdev->dev, "controller appears to be disabled\n"); + return -ENODEV; + } + err = pci_enable_device(h->pdev); + if (err) { + dev_warn(&h->pdev->dev, "Unable to Enable PCI device\n"); + return err; } -#ifdef CCISS_DEBUG - printk("Trying to put board into Simple mode\n"); -#endif /* CCISS_DEBUG */ - c->max_commands = readl(&(c->cfgtable->CmdsOutMax)); - /* Update the field, and then ring the doorbell */ - writel(CFGTBL_Trans_Simple, &(c->cfgtable->HostWrite.TransportRequest)); - writel(CFGTBL_ChangeReq, c->vaddr + SA5_DOORBELL); + err = pci_request_regions(h->pdev, "cciss"); + if (err) { + dev_warn(&h->pdev->dev, + "Cannot obtain PCI resources, aborting\n"); + return err; + } - /* under certain very rare conditions, this can take awhile. - * (e.g.: hot replace a failed 144GB drive in a RAID 5 set right - * as we enter this code.) */ - for (i = 0; i < MAX_CONFIG_WAIT; i++) { - if (!(readl(c->vaddr + SA5_DOORBELL) & CFGTBL_ChangeReq)) - break; - /* delay and try again */ - set_current_state(TASK_INTERRUPTIBLE); - schedule_timeout(msecs_to_jiffies(1)); + dev_dbg(&h->pdev->dev, "irq = %x\n", h->pdev->irq); + dev_dbg(&h->pdev->dev, "board_id = %x\n", h->board_id); + +/* If the kernel supports MSI/MSI-X we will try to enable that functionality, + * else we use the IO-APIC interrupt assigned to us by system ROM. + */ + cciss_interrupt_mode(h); + err = cciss_pci_find_memory_BAR(h->pdev, &h->paddr); + if (err) + goto err_out_free_res; + h->vaddr = remap_pci_mem(h->paddr, 0x250); + if (!h->vaddr) { + err = -ENOMEM; + goto err_out_free_res; } + err = cciss_wait_for_board_ready(h); + if (err) + goto err_out_free_res; + err = cciss_find_cfgtables(h); + if (err) + goto err_out_free_res; + print_cfg_table(h); + cciss_find_board_params(h); -#ifdef CCISS_DEBUG - printk(KERN_DEBUG "I counter got to %d %x\n", i, - readl(c->vaddr + SA5_DOORBELL)); -#endif /* CCISS_DEBUG */ -#ifdef CCISS_DEBUG - print_cfg_table(c->cfgtable); -#endif /* CCISS_DEBUG */ - - if (!(readl(&(c->cfgtable->TransportActive)) & CFGTBL_Trans_Simple)) { - printk(KERN_WARNING "cciss: unable to get board into" - " simple mode\n"); + if (!CISS_signature_present(h)) { err = -ENODEV; goto err_out_free_res; } + cciss_enable_scsi_prefetch(h); + cciss_p600_dma_prefetch_quirk(h); + cciss_put_controller_into_performant_mode(h); return 0; err_out_free_res: @@ -3913,42 +4220,47 @@ err_out_free_res: * Deliberately omit pci_disable_device(): it does something nasty to * Smart Array controllers that pci_enable_device does not undo */ - pci_release_regions(pdev); + if (h->transtable) + iounmap(h->transtable); + if (h->cfgtable) + iounmap(h->cfgtable); + if (h->vaddr) + iounmap(h->vaddr); + pci_release_regions(h->pdev); return err; } /* Function to find the first free pointer into our hba[] array * Returns -1 if no free entries are left. */ -static int alloc_cciss_hba(void) +static int alloc_cciss_hba(struct pci_dev *pdev) { int i; for (i = 0; i < MAX_CTLR; i++) { if (!hba[i]) { - ctlr_info_t *p; + ctlr_info_t *h; - p = kzalloc(sizeof(ctlr_info_t), GFP_KERNEL); - if (!p) + h = kzalloc(sizeof(ctlr_info_t), GFP_KERNEL); + if (!h) goto Enomem; - hba[i] = p; + hba[i] = h; return i; } } - printk(KERN_WARNING "cciss: This driver supports a maximum" + dev_warn(&pdev->dev, "This driver supports a maximum" " of %d controllers.\n", MAX_CTLR); return -1; Enomem: - printk(KERN_ERR "cciss: out of memory.\n"); + dev_warn(&pdev->dev, "out of memory.\n"); return -1; } -static void free_hba(int n) +static void free_hba(ctlr_info_t *h) { - ctlr_info_t *h = hba[n]; int i; - hba[n] = NULL; + hba[h->ctlr] = NULL; for (i = 0; i < h->highest_lun + 1; i++) if (h->gendisk[i] != NULL) put_disk(h->gendisk[i]); @@ -4028,7 +4340,8 @@ static __devinit int cciss_message(struct pci_dev *pdev, unsigned char opcode, u /* we leak the DMA buffer here ... no choice since the controller could still complete the command. */ if (i == 10) { - printk(KERN_ERR "cciss: controller message %02x:%02x timed out\n", + dev_err(&pdev->dev, + "controller message %02x:%02x timed out\n", opcode, type); return -ETIMEDOUT; } @@ -4036,12 +4349,12 @@ static __devinit int cciss_message(struct pci_dev *pdev, unsigned char opcode, u pci_free_consistent(pdev, cmd_sz, cmd, paddr64); if (tag & 2) { - printk(KERN_ERR "cciss: controller message %02x:%02x failed\n", + dev_err(&pdev->dev, "controller message %02x:%02x failed\n", opcode, type); return -EIO; } - printk(KERN_INFO "cciss: controller message %02x:%02x succeeded\n", + dev_info(&pdev->dev, "controller message %02x:%02x succeeded\n", opcode, type); return 0; } @@ -4062,7 +4375,7 @@ static __devinit int cciss_reset_msi(struct pci_dev *pdev) if (pos) { pci_read_config_word(pdev, msi_control_reg(pos), &control); if (control & PCI_MSI_FLAGS_ENABLE) { - printk(KERN_INFO "cciss: resetting MSI\n"); + dev_info(&pdev->dev, "resetting MSI\n"); pci_write_config_word(pdev, msi_control_reg(pos), control & ~PCI_MSI_FLAGS_ENABLE); } } @@ -4071,7 +4384,7 @@ static __devinit int cciss_reset_msi(struct pci_dev *pdev) if (pos) { pci_read_config_word(pdev, msi_control_reg(pos), &control); if (control & PCI_MSIX_FLAGS_ENABLE) { - printk(KERN_INFO "cciss: resetting MSI-X\n"); + dev_info(&pdev->dev, "resetting MSI-X\n"); pci_write_config_word(pdev, msi_control_reg(pos), control & ~PCI_MSIX_FLAGS_ENABLE); } } @@ -4079,68 +4392,144 @@ static __devinit int cciss_reset_msi(struct pci_dev *pdev) return 0; } -/* This does a hard reset of the controller using PCI power management - * states. */ -static __devinit int cciss_hard_reset_controller(struct pci_dev *pdev) +static int cciss_controller_hard_reset(struct pci_dev *pdev, + void * __iomem vaddr, bool use_doorbell) { - u16 pmcsr, saved_config_space[32]; - int i, pos; + u16 pmcsr; + int pos; - printk(KERN_INFO "cciss: using PCI PM to reset controller\n"); + if (use_doorbell) { + /* For everything after the P600, the PCI power state method + * of resetting the controller doesn't work, so we have this + * other way using the doorbell register. + */ + dev_info(&pdev->dev, "using doorbell to reset controller\n"); + writel(DOORBELL_CTLR_RESET, vaddr + SA5_DOORBELL); + msleep(1000); + } else { /* Try to do it the PCI power state way */ + + /* Quoting from the Open CISS Specification: "The Power + * Management Control/Status Register (CSR) controls the power + * state of the device. The normal operating state is D0, + * CSR=00h. The software off state is D3, CSR=03h. To reset + * the controller, place the interface device in D3 then to D0, + * this causes a secondary PCI reset which will reset the + * controller." */ + + pos = pci_find_capability(pdev, PCI_CAP_ID_PM); + if (pos == 0) { + dev_err(&pdev->dev, + "cciss_controller_hard_reset: " + "PCI PM not supported\n"); + return -ENODEV; + } + dev_info(&pdev->dev, "using PCI PM to reset controller\n"); + /* enter the D3hot power management state */ + pci_read_config_word(pdev, pos + PCI_PM_CTRL, &pmcsr); + pmcsr &= ~PCI_PM_CTRL_STATE_MASK; + pmcsr |= PCI_D3hot; + pci_write_config_word(pdev, pos + PCI_PM_CTRL, pmcsr); - /* This is very nearly the same thing as + msleep(500); - pci_save_state(pci_dev); - pci_set_power_state(pci_dev, PCI_D3hot); - pci_set_power_state(pci_dev, PCI_D0); - pci_restore_state(pci_dev); + /* enter the D0 power management state */ + pmcsr &= ~PCI_PM_CTRL_STATE_MASK; + pmcsr |= PCI_D0; + pci_write_config_word(pdev, pos + PCI_PM_CTRL, pmcsr); - but we can't use these nice canned kernel routines on - kexec, because they also check the MSI/MSI-X state in PCI - configuration space and do the wrong thing when it is - set/cleared. Also, the pci_save/restore_state functions - violate the ordering requirements for restoring the - configuration space from the CCISS document (see the - comment below). So we roll our own .... */ + msleep(500); + } + return 0; +} - for (i = 0; i < 32; i++) - pci_read_config_word(pdev, 2*i, &saved_config_space[i]); +/* This does a hard reset of the controller using PCI power management + * states or using the doorbell register. */ +static __devinit int cciss_kdump_hard_reset_controller(struct pci_dev *pdev) +{ + u16 saved_config_space[32]; + u64 cfg_offset; + u32 cfg_base_addr; + u64 cfg_base_addr_index; + void __iomem *vaddr; + unsigned long paddr; + u32 misc_fw_support, active_transport; + int rc, i; + CfgTable_struct __iomem *cfgtable; + bool use_doorbell; + u32 board_id; + + /* For controllers as old a the p600, this is very nearly + * the same thing as + * + * pci_save_state(pci_dev); + * pci_set_power_state(pci_dev, PCI_D3hot); + * pci_set_power_state(pci_dev, PCI_D0); + * pci_restore_state(pci_dev); + * + * but we can't use these nice canned kernel routines on + * kexec, because they also check the MSI/MSI-X state in PCI + * configuration space and do the wrong thing when it is + * set/cleared. Also, the pci_save/restore_state functions + * violate the ordering requirements for restoring the + * configuration space from the CCISS document (see the + * comment below). So we roll our own .... + * + * For controllers newer than the P600, the pci power state + * method of resetting doesn't work so we have another way + * using the doorbell register. + */ - pos = pci_find_capability(pdev, PCI_CAP_ID_PM); - if (pos == 0) { - printk(KERN_ERR "cciss_reset_controller: PCI PM not supported\n"); + /* Exclude 640x boards. These are two pci devices in one slot + * which share a battery backed cache module. One controls the + * cache, the other accesses the cache through the one that controls + * it. If we reset the one controlling the cache, the other will + * likely not be happy. Just forbid resetting this conjoined mess. + */ + cciss_lookup_board_id(pdev, &board_id); + if (board_id == 0x409C0E11 || board_id == 0x409D0E11) { + dev_warn(&pdev->dev, "Cannot reset Smart Array 640x " + "due to shared cache module."); return -ENODEV; } - /* Quoting from the Open CISS Specification: "The Power - * Management Control/Status Register (CSR) controls the power - * state of the device. The normal operating state is D0, - * CSR=00h. The software off state is D3, CSR=03h. To reset - * the controller, place the interface device in D3 then to - * D0, this causes a secondary PCI reset which will reset the - * controller." */ + for (i = 0; i < 32; i++) + pci_read_config_word(pdev, 2*i, &saved_config_space[i]); - /* enter the D3hot power management state */ - pci_read_config_word(pdev, pos + PCI_PM_CTRL, &pmcsr); - pmcsr &= ~PCI_PM_CTRL_STATE_MASK; - pmcsr |= PCI_D3hot; - pci_write_config_word(pdev, pos + PCI_PM_CTRL, pmcsr); + /* find the first memory BAR, so we can find the cfg table */ + rc = cciss_pci_find_memory_BAR(pdev, &paddr); + if (rc) + return rc; + vaddr = remap_pci_mem(paddr, 0x250); + if (!vaddr) + return -ENOMEM; - schedule_timeout_uninterruptible(HZ >> 1); + /* find cfgtable in order to check if reset via doorbell is supported */ + rc = cciss_find_cfg_addrs(pdev, vaddr, &cfg_base_addr, + &cfg_base_addr_index, &cfg_offset); + if (rc) + goto unmap_vaddr; + cfgtable = remap_pci_mem(pci_resource_start(pdev, + cfg_base_addr_index) + cfg_offset, sizeof(*cfgtable)); + if (!cfgtable) { + rc = -ENOMEM; + goto unmap_vaddr; + } - /* enter the D0 power management state */ - pmcsr &= ~PCI_PM_CTRL_STATE_MASK; - pmcsr |= PCI_D0; - pci_write_config_word(pdev, pos + PCI_PM_CTRL, pmcsr); + /* If reset via doorbell register is supported, use that. */ + misc_fw_support = readl(&cfgtable->misc_fw_support); + use_doorbell = misc_fw_support & MISC_FW_DOORBELL_RESET; - schedule_timeout_uninterruptible(HZ >> 1); + rc = cciss_controller_hard_reset(pdev, vaddr, use_doorbell); + if (rc) + goto unmap_cfgtable; /* Restore the PCI configuration space. The Open CISS * Specification says, "Restore the PCI Configuration * Registers, offsets 00h through 60h. It is important to * restore the command register, 16-bits at offset 04h, * last. Do not restore the configuration status register, - * 16-bits at offset 06h." Note that the offset is 2*i. */ + * 16-bits at offset 06h." Note that the offset is 2*i. + */ for (i = 0; i < 32; i++) { if (i == 2 || i == 3) continue; @@ -4149,6 +4538,63 @@ static __devinit int cciss_hard_reset_controller(struct pci_dev *pdev) wmb(); pci_write_config_word(pdev, 4, saved_config_space[2]); + /* Some devices (notably the HP Smart Array 5i Controller) + need a little pause here */ + msleep(CCISS_POST_RESET_PAUSE_MSECS); + + /* Controller should be in simple mode at this point. If it's not, + * It means we're on one of those controllers which doesn't support + * the doorbell reset method and on which the PCI power management reset + * method doesn't work (P800, for example.) + * In those cases, don't try to proceed, as it generally doesn't work. + */ + active_transport = readl(&cfgtable->TransportActive); + if (active_transport & PERFORMANT_MODE) { + dev_warn(&pdev->dev, "Unable to successfully reset controller," + " Ignoring controller.\n"); + rc = -ENODEV; + } + +unmap_cfgtable: + iounmap(cfgtable); + +unmap_vaddr: + iounmap(vaddr); + return rc; +} + +static __devinit int cciss_init_reset_devices(struct pci_dev *pdev) +{ + int rc, i; + + if (!reset_devices) + return 0; + + /* Reset the controller with a PCI power-cycle or via doorbell */ + rc = cciss_kdump_hard_reset_controller(pdev); + + /* -ENOTSUPP here means we cannot reset the controller + * but it's already (and still) up and running in + * "performant mode". Or, it might be 640x, which can't reset + * due to concerns about shared bbwc between 6402/6404 pair. + */ + if (rc == -ENOTSUPP) + return 0; /* just try to do the kdump anyhow. */ + if (rc) + return -ENODEV; + if (cciss_reset_msi(pdev)) + return -ENODEV; + + /* Now try to get the controller to respond to a no-op */ + for (i = 0; i < CCISS_POST_RESET_NOOP_RETRIES; i++) { + if (cciss_noop(pdev) == 0) + break; + else + dev_warn(&pdev->dev, "no-op failed%s\n", + (i < CCISS_POST_RESET_NOOP_RETRIES - 1 ? + "; re-trying" : "")); + msleep(CCISS_POST_RESET_NOOP_INTERVAL_MSECS); + } return 0; } @@ -4166,46 +4612,31 @@ static int __devinit cciss_init_one(struct pci_dev *pdev, int rc; int dac, return_code; InquiryData_struct *inq_buff; + ctlr_info_t *h; - if (reset_devices) { - /* Reset the controller with a PCI power-cycle */ - if (cciss_hard_reset_controller(pdev) || cciss_reset_msi(pdev)) - return -ENODEV; - - /* Now try to get the controller to respond to a no-op. Some - devices (notably the HP Smart Array 5i Controller) need - up to 30 seconds to respond. */ - for (i=0; i<30; i++) { - if (cciss_noop(pdev) == 0) - break; - - schedule_timeout_uninterruptible(HZ); - } - if (i == 30) { - printk(KERN_ERR "cciss: controller seems dead\n"); - return -EBUSY; - } - } - - i = alloc_cciss_hba(); + rc = cciss_init_reset_devices(pdev); + if (rc) + return rc; + i = alloc_cciss_hba(pdev); if (i < 0) return -1; - hba[i]->busy_initializing = 1; - INIT_HLIST_HEAD(&hba[i]->cmpQ); - INIT_HLIST_HEAD(&hba[i]->reqQ); - mutex_init(&hba[i]->busy_shutting_down); + h = hba[i]; + h->pdev = pdev; + h->busy_initializing = 1; + INIT_HLIST_HEAD(&h->cmpQ); + INIT_HLIST_HEAD(&h->reqQ); + mutex_init(&h->busy_shutting_down); - if (cciss_pci_init(hba[i], pdev) != 0) + if (cciss_pci_init(h) != 0) goto clean_no_release_regions; - sprintf(hba[i]->devname, "cciss%d", i); - hba[i]->ctlr = i; - hba[i]->pdev = pdev; + sprintf(h->devname, "cciss%d", i); + h->ctlr = i; - init_completion(&hba[i]->scan_wait); + init_completion(&h->scan_wait); - if (cciss_create_hba_sysfs_entry(hba[i])) + if (cciss_create_hba_sysfs_entry(h)) goto clean0; /* configure PCI DMA stuff */ @@ -4214,7 +4645,7 @@ static int __devinit cciss_init_one(struct pci_dev *pdev, else if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(32))) dac = 0; else { - printk(KERN_ERR "cciss: no suitable DMA available\n"); + dev_err(&h->pdev->dev, "no suitable DMA available\n"); goto clean1; } @@ -4224,151 +4655,161 @@ static int __devinit cciss_init_one(struct pci_dev *pdev, * 8 controller support. */ if (i < MAX_CTLR_ORIG) - hba[i]->major = COMPAQ_CISS_MAJOR + i; - rc = register_blkdev(hba[i]->major, hba[i]->devname); + h->major = COMPAQ_CISS_MAJOR + i; + rc = register_blkdev(h->major, h->devname); if (rc == -EBUSY || rc == -EINVAL) { - printk(KERN_ERR - "cciss: Unable to get major number %d for %s " - "on hba %d\n", hba[i]->major, hba[i]->devname, i); + dev_err(&h->pdev->dev, + "Unable to get major number %d for %s " + "on hba %d\n", h->major, h->devname, i); goto clean1; } else { if (i >= MAX_CTLR_ORIG) - hba[i]->major = rc; + h->major = rc; } /* make sure the board interrupts are off */ - hba[i]->access.set_intr_mask(hba[i], CCISS_INTR_OFF); - if (request_irq(hba[i]->intr[SIMPLE_MODE_INT], do_cciss_intr, - IRQF_DISABLED | IRQF_SHARED, hba[i]->devname, hba[i])) { - printk(KERN_ERR "cciss: Unable to get irq %d for %s\n", - hba[i]->intr[SIMPLE_MODE_INT], hba[i]->devname); - goto clean2; + h->access.set_intr_mask(h, CCISS_INTR_OFF); + if (h->msi_vector || h->msix_vector) { + if (request_irq(h->intr[PERF_MODE_INT], + do_cciss_msix_intr, + IRQF_DISABLED, h->devname, h)) { + dev_err(&h->pdev->dev, "Unable to get irq %d for %s\n", + h->intr[PERF_MODE_INT], h->devname); + goto clean2; + } + } else { + if (request_irq(h->intr[PERF_MODE_INT], do_cciss_intx, + IRQF_DISABLED, h->devname, h)) { + dev_err(&h->pdev->dev, "Unable to get irq %d for %s\n", + h->intr[PERF_MODE_INT], h->devname); + goto clean2; + } } - printk(KERN_INFO "%s: <0x%x> at PCI %s IRQ %d%s using DAC\n", - hba[i]->devname, pdev->device, pci_name(pdev), - hba[i]->intr[SIMPLE_MODE_INT], dac ? "" : " not"); + dev_info(&h->pdev->dev, "%s: <0x%x> at PCI %s IRQ %d%s using DAC\n", + h->devname, pdev->device, pci_name(pdev), + h->intr[PERF_MODE_INT], dac ? "" : " not"); - hba[i]->cmd_pool_bits = - kmalloc(DIV_ROUND_UP(hba[i]->nr_cmds, BITS_PER_LONG) + h->cmd_pool_bits = + kmalloc(DIV_ROUND_UP(h->nr_cmds, BITS_PER_LONG) * sizeof(unsigned long), GFP_KERNEL); - hba[i]->cmd_pool = (CommandList_struct *) - pci_alloc_consistent(hba[i]->pdev, - hba[i]->nr_cmds * sizeof(CommandList_struct), - &(hba[i]->cmd_pool_dhandle)); - hba[i]->errinfo_pool = (ErrorInfo_struct *) - pci_alloc_consistent(hba[i]->pdev, - hba[i]->nr_cmds * sizeof(ErrorInfo_struct), - &(hba[i]->errinfo_pool_dhandle)); - if ((hba[i]->cmd_pool_bits == NULL) - || (hba[i]->cmd_pool == NULL) - || (hba[i]->errinfo_pool == NULL)) { - printk(KERN_ERR "cciss: out of memory"); + h->cmd_pool = (CommandList_struct *) + pci_alloc_consistent(h->pdev, + h->nr_cmds * sizeof(CommandList_struct), + &(h->cmd_pool_dhandle)); + h->errinfo_pool = (ErrorInfo_struct *) + pci_alloc_consistent(h->pdev, + h->nr_cmds * sizeof(ErrorInfo_struct), + &(h->errinfo_pool_dhandle)); + if ((h->cmd_pool_bits == NULL) + || (h->cmd_pool == NULL) + || (h->errinfo_pool == NULL)) { + dev_err(&h->pdev->dev, "out of memory"); goto clean4; } /* Need space for temp scatter list */ - hba[i]->scatter_list = kmalloc(hba[i]->max_commands * + h->scatter_list = kmalloc(h->max_commands * sizeof(struct scatterlist *), GFP_KERNEL); - for (k = 0; k < hba[i]->nr_cmds; k++) { - hba[i]->scatter_list[k] = kmalloc(sizeof(struct scatterlist) * - hba[i]->maxsgentries, + for (k = 0; k < h->nr_cmds; k++) { + h->scatter_list[k] = kmalloc(sizeof(struct scatterlist) * + h->maxsgentries, GFP_KERNEL); - if (hba[i]->scatter_list[k] == NULL) { - printk(KERN_ERR "cciss%d: could not allocate " - "s/g lists\n", i); + if (h->scatter_list[k] == NULL) { + dev_err(&h->pdev->dev, + "could not allocate s/g lists\n"); goto clean4; } } - hba[i]->cmd_sg_list = cciss_allocate_sg_chain_blocks(hba[i], - hba[i]->chainsize, hba[i]->nr_cmds); - if (!hba[i]->cmd_sg_list && hba[i]->chainsize > 0) + h->cmd_sg_list = cciss_allocate_sg_chain_blocks(h, + h->chainsize, h->nr_cmds); + if (!h->cmd_sg_list && h->chainsize > 0) goto clean4; - spin_lock_init(&hba[i]->lock); + spin_lock_init(&h->lock); /* Initialize the pdev driver private data. - have it point to hba[i]. */ - pci_set_drvdata(pdev, hba[i]); + have it point to h. */ + pci_set_drvdata(pdev, h); /* command and error info recs zeroed out before they are used */ - memset(hba[i]->cmd_pool_bits, 0, - DIV_ROUND_UP(hba[i]->nr_cmds, BITS_PER_LONG) + memset(h->cmd_pool_bits, 0, + DIV_ROUND_UP(h->nr_cmds, BITS_PER_LONG) * sizeof(unsigned long)); - hba[i]->num_luns = 0; - hba[i]->highest_lun = -1; + h->num_luns = 0; + h->highest_lun = -1; for (j = 0; j < CISS_MAX_LUN; j++) { - hba[i]->drv[j] = NULL; - hba[i]->gendisk[j] = NULL; + h->drv[j] = NULL; + h->gendisk[j] = NULL; } - cciss_scsi_setup(i); + cciss_scsi_setup(h); /* Turn the interrupts on so we can service requests */ - hba[i]->access.set_intr_mask(hba[i], CCISS_INTR_ON); + h->access.set_intr_mask(h, CCISS_INTR_ON); /* Get the firmware version */ inq_buff = kzalloc(sizeof(InquiryData_struct), GFP_KERNEL); if (inq_buff == NULL) { - printk(KERN_ERR "cciss: out of memory\n"); + dev_err(&h->pdev->dev, "out of memory\n"); goto clean4; } - return_code = sendcmd_withirq(CISS_INQUIRY, i, inq_buff, + return_code = sendcmd_withirq(h, CISS_INQUIRY, inq_buff, sizeof(InquiryData_struct), 0, CTLR_LUNID, TYPE_CMD); if (return_code == IO_OK) { - hba[i]->firm_ver[0] = inq_buff->data_byte[32]; - hba[i]->firm_ver[1] = inq_buff->data_byte[33]; - hba[i]->firm_ver[2] = inq_buff->data_byte[34]; - hba[i]->firm_ver[3] = inq_buff->data_byte[35]; + h->firm_ver[0] = inq_buff->data_byte[32]; + h->firm_ver[1] = inq_buff->data_byte[33]; + h->firm_ver[2] = inq_buff->data_byte[34]; + h->firm_ver[3] = inq_buff->data_byte[35]; } else { /* send command failed */ - printk(KERN_WARNING "cciss: unable to determine firmware" + dev_warn(&h->pdev->dev, "unable to determine firmware" " version of controller\n"); } kfree(inq_buff); - cciss_procinit(i); + cciss_procinit(h); - hba[i]->cciss_max_sectors = 8192; + h->cciss_max_sectors = 8192; - rebuild_lun_table(hba[i], 1, 0); - hba[i]->busy_initializing = 0; + rebuild_lun_table(h, 1, 0); + h->busy_initializing = 0; return 1; clean4: - kfree(hba[i]->cmd_pool_bits); + kfree(h->cmd_pool_bits); /* Free up sg elements */ - for (k = 0; k < hba[i]->nr_cmds; k++) - kfree(hba[i]->scatter_list[k]); - kfree(hba[i]->scatter_list); - cciss_free_sg_chain_blocks(hba[i]->cmd_sg_list, hba[i]->nr_cmds); - if (hba[i]->cmd_pool) - pci_free_consistent(hba[i]->pdev, - hba[i]->nr_cmds * sizeof(CommandList_struct), - hba[i]->cmd_pool, hba[i]->cmd_pool_dhandle); - if (hba[i]->errinfo_pool) - pci_free_consistent(hba[i]->pdev, - hba[i]->nr_cmds * sizeof(ErrorInfo_struct), - hba[i]->errinfo_pool, - hba[i]->errinfo_pool_dhandle); - free_irq(hba[i]->intr[SIMPLE_MODE_INT], hba[i]); + for (k = 0; k < h->nr_cmds; k++) + kfree(h->scatter_list[k]); + kfree(h->scatter_list); + cciss_free_sg_chain_blocks(h->cmd_sg_list, h->nr_cmds); + if (h->cmd_pool) + pci_free_consistent(h->pdev, + h->nr_cmds * sizeof(CommandList_struct), + h->cmd_pool, h->cmd_pool_dhandle); + if (h->errinfo_pool) + pci_free_consistent(h->pdev, + h->nr_cmds * sizeof(ErrorInfo_struct), + h->errinfo_pool, + h->errinfo_pool_dhandle); + free_irq(h->intr[PERF_MODE_INT], h); clean2: - unregister_blkdev(hba[i]->major, hba[i]->devname); + unregister_blkdev(h->major, h->devname); clean1: - cciss_destroy_hba_sysfs_entry(hba[i]); + cciss_destroy_hba_sysfs_entry(h); clean0: pci_release_regions(pdev); clean_no_release_regions: - hba[i]->busy_initializing = 0; + h->busy_initializing = 0; /* * Deliberately omit pci_disable_device(): it does something nasty to * Smart Array controllers that pci_enable_device does not undo */ pci_set_drvdata(pdev, NULL); - free_hba(i); + free_hba(h); return -1; } @@ -4381,55 +4822,51 @@ static void cciss_shutdown(struct pci_dev *pdev) h = pci_get_drvdata(pdev); flush_buf = kzalloc(4, GFP_KERNEL); if (!flush_buf) { - printk(KERN_WARNING - "cciss:%d cache not flushed, out of memory.\n", - h->ctlr); + dev_warn(&h->pdev->dev, "cache not flushed, out of memory.\n"); return; } /* write all data in the battery backed cache to disk */ memset(flush_buf, 0, 4); - return_code = sendcmd_withirq(CCISS_CACHE_FLUSH, h->ctlr, flush_buf, + return_code = sendcmd_withirq(h, CCISS_CACHE_FLUSH, flush_buf, 4, 0, CTLR_LUNID, TYPE_CMD); kfree(flush_buf); if (return_code != IO_OK) - printk(KERN_WARNING "cciss%d: Error flushing cache\n", - h->ctlr); + dev_warn(&h->pdev->dev, "Error flushing cache\n"); h->access.set_intr_mask(h, CCISS_INTR_OFF); - free_irq(h->intr[2], h); + free_irq(h->intr[PERF_MODE_INT], h); } static void __devexit cciss_remove_one(struct pci_dev *pdev) { - ctlr_info_t *tmp_ptr; + ctlr_info_t *h; int i, j; if (pci_get_drvdata(pdev) == NULL) { - printk(KERN_ERR "cciss: Unable to remove device \n"); + dev_err(&pdev->dev, "Unable to remove device\n"); return; } - tmp_ptr = pci_get_drvdata(pdev); - i = tmp_ptr->ctlr; + h = pci_get_drvdata(pdev); + i = h->ctlr; if (hba[i] == NULL) { - printk(KERN_ERR "cciss: device appears to " - "already be removed \n"); + dev_err(&pdev->dev, "device appears to already be removed\n"); return; } - mutex_lock(&hba[i]->busy_shutting_down); + mutex_lock(&h->busy_shutting_down); - remove_from_scan_list(hba[i]); - remove_proc_entry(hba[i]->devname, proc_cciss); - unregister_blkdev(hba[i]->major, hba[i]->devname); + remove_from_scan_list(h); + remove_proc_entry(h->devname, proc_cciss); + unregister_blkdev(h->major, h->devname); /* remove it from the disk list */ for (j = 0; j < CISS_MAX_LUN; j++) { - struct gendisk *disk = hba[i]->gendisk[j]; + struct gendisk *disk = h->gendisk[j]; if (disk) { struct request_queue *q = disk->queue; if (disk->flags & GENHD_FL_UP) { - cciss_destroy_ld_sysfs_entry(hba[i], j, 1); + cciss_destroy_ld_sysfs_entry(h, j, 1); del_gendisk(disk); } if (q) @@ -4438,39 +4875,41 @@ static void __devexit cciss_remove_one(struct pci_dev *pdev) } #ifdef CONFIG_CISS_SCSI_TAPE - cciss_unregister_scsi(i); /* unhook from SCSI subsystem */ + cciss_unregister_scsi(h); /* unhook from SCSI subsystem */ #endif cciss_shutdown(pdev); #ifdef CONFIG_PCI_MSI - if (hba[i]->msix_vector) - pci_disable_msix(hba[i]->pdev); - else if (hba[i]->msi_vector) - pci_disable_msi(hba[i]->pdev); + if (h->msix_vector) + pci_disable_msix(h->pdev); + else if (h->msi_vector) + pci_disable_msi(h->pdev); #endif /* CONFIG_PCI_MSI */ - iounmap(hba[i]->vaddr); + iounmap(h->transtable); + iounmap(h->cfgtable); + iounmap(h->vaddr); - pci_free_consistent(hba[i]->pdev, hba[i]->nr_cmds * sizeof(CommandList_struct), - hba[i]->cmd_pool, hba[i]->cmd_pool_dhandle); - pci_free_consistent(hba[i]->pdev, hba[i]->nr_cmds * sizeof(ErrorInfo_struct), - hba[i]->errinfo_pool, hba[i]->errinfo_pool_dhandle); - kfree(hba[i]->cmd_pool_bits); + pci_free_consistent(h->pdev, h->nr_cmds * sizeof(CommandList_struct), + h->cmd_pool, h->cmd_pool_dhandle); + pci_free_consistent(h->pdev, h->nr_cmds * sizeof(ErrorInfo_struct), + h->errinfo_pool, h->errinfo_pool_dhandle); + kfree(h->cmd_pool_bits); /* Free up sg elements */ - for (j = 0; j < hba[i]->nr_cmds; j++) - kfree(hba[i]->scatter_list[j]); - kfree(hba[i]->scatter_list); - cciss_free_sg_chain_blocks(hba[i]->cmd_sg_list, hba[i]->nr_cmds); + for (j = 0; j < h->nr_cmds; j++) + kfree(h->scatter_list[j]); + kfree(h->scatter_list); + cciss_free_sg_chain_blocks(h->cmd_sg_list, h->nr_cmds); /* * Deliberately omit pci_disable_device(): it does something nasty to * Smart Array controllers that pci_enable_device does not undo */ pci_release_regions(pdev); pci_set_drvdata(pdev, NULL); - cciss_destroy_hba_sysfs_entry(hba[i]); - mutex_unlock(&hba[i]->busy_shutting_down); - free_hba(i); + cciss_destroy_hba_sysfs_entry(h); + mutex_unlock(&h->busy_shutting_down); + free_hba(h); } static struct pci_driver cciss_pci_driver = { @@ -4495,7 +4934,6 @@ static int __init cciss_init(void) * array of them, the size must be a multiple of 8 bytes. */ BUILD_BUG_ON(sizeof(CommandList_struct) % COMMANDLIST_ALIGNMENT); - printk(KERN_INFO DRIVER_NAME "\n"); err = bus_register(&cciss_bus_type); @@ -4532,8 +4970,8 @@ static void __exit cciss_cleanup(void) /* double check that all controller entrys have been removed */ for (i = 0; i < MAX_CTLR; i++) { if (hba[i] != NULL) { - printk(KERN_WARNING "cciss: had to remove" - " controller %d\n", i); + dev_warn(&hba[i]->pdev->dev, + "had to remove controller\n"); cciss_remove_one(hba[i]->pdev); } } @@ -4542,46 +4980,5 @@ static void __exit cciss_cleanup(void) bus_unregister(&cciss_bus_type); } -static void fail_all_cmds(unsigned long ctlr) -{ - /* If we get here, the board is apparently dead. */ - ctlr_info_t *h = hba[ctlr]; - CommandList_struct *c; - unsigned long flags; - - printk(KERN_WARNING "cciss%d: controller not responding.\n", h->ctlr); - h->alive = 0; /* the controller apparently died... */ - - spin_lock_irqsave(CCISS_LOCK(ctlr), flags); - - pci_disable_device(h->pdev); /* Make sure it is really dead. */ - - /* move everything off the request queue onto the completed queue */ - while (!hlist_empty(&h->reqQ)) { - c = hlist_entry(h->reqQ.first, CommandList_struct, list); - removeQ(c); - h->Qdepth--; - addQ(&h->cmpQ, c); - } - - /* Now, fail everything on the completed queue with a HW error */ - while (!hlist_empty(&h->cmpQ)) { - c = hlist_entry(h->cmpQ.first, CommandList_struct, list); - removeQ(c); - if (c->cmd_type != CMD_MSG_STALE) - c->err_info->CommandStatus = CMD_HARDWARE_ERR; - if (c->cmd_type == CMD_RWREQ) { - complete_command(h, c, 0); - } else if (c->cmd_type == CMD_IOCTL_PEND) - complete(c->waiting); -#ifdef CONFIG_CISS_SCSI_TAPE - else if (c->cmd_type == CMD_SCSI) - complete_scsi_command(c, 0, 0); -#endif - } - spin_unlock_irqrestore(CCISS_LOCK(ctlr), flags); - return; -} - module_init(cciss_init); module_exit(cciss_cleanup); |