diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2011-05-25 09:15:35 -0700 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2011-05-25 09:15:35 -0700 |
commit | 929cfdd5d3bdc772aff32e5a3fb4e3894394aa75 (patch) | |
tree | f67202d079eaf1f8d65b2e1bfac70b768ae34bc4 /drivers | |
parent | 798ce8f1cca29dcc3f4b55947f611f4ffb32ac2b (diff) | |
parent | a1c15c59feee36267c43142a41152fbf7402afb6 (diff) |
Merge branch 'for-2.6.40/drivers' of git://git.kernel.dk/linux-2.6-block
* 'for-2.6.40/drivers' of git://git.kernel.dk/linux-2.6-block: (110 commits)
loop: handle on-demand devices correctly
loop: limit 'max_part' module param to DISK_MAX_PARTS
drbd: fix warning
drbd: fix warning
drbd: Fix spelling
drbd: fix schedule in atomic
drbd: Take a more conservative approach when deciding max_bio_size
drbd: Fixed state transitions after async outdate-peer-handler returned
drbd: Disallow the peer_disk_state to be D_OUTDATED while connected
drbd: Fix for the connection problems on high latency links
drbd: fix potential activity log refcount imbalance in error path
drbd: Only downgrade the disk state in case of disk failures
drbd: fix disconnect/reconnect loop, if ping-timeout == ping-int
drbd: fix potential distributed deadlock
lru_cache.h: fix comments referring to ts_ instead of lc_
drbd: Fix for application IO with the on-io-error=pass-on policy
xen/p2m: Add EXPORT_SYMBOL_GPL to the M2P override functions.
xen/p2m/m2p/gnttab: Support GNTMAP_host_map in the M2P override.
xen/blkback: don't fail empty barrier requests
xen/blkback: fix xenbus_transaction_start() hang caused by double xenbus_transaction_end()
...
Diffstat (limited to 'drivers')
-rw-r--r-- | drivers/block/Kconfig | 21 | ||||
-rw-r--r-- | drivers/block/Makefile | 1 | ||||
-rw-r--r-- | drivers/block/cciss.c | 571 | ||||
-rw-r--r-- | drivers/block/cciss.h | 11 | ||||
-rw-r--r-- | drivers/block/cciss_cmd.h | 11 | ||||
-rw-r--r-- | drivers/block/cciss_scsi.c | 41 | ||||
-rw-r--r-- | drivers/block/cciss_scsi.h | 4 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_actlog.c | 2 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_bitmap.c | 6 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_int.h | 19 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_main.c | 37 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_nl.c | 127 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_receiver.c | 68 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_req.c | 20 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_req.h | 5 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_worker.c | 98 | ||||
-rw-r--r-- | drivers/block/loop.c | 11 | ||||
-rw-r--r-- | drivers/block/xen-blkback/Makefile | 3 | ||||
-rw-r--r-- | drivers/block/xen-blkback/blkback.c | 824 | ||||
-rw-r--r-- | drivers/block/xen-blkback/common.h | 233 | ||||
-rw-r--r-- | drivers/block/xen-blkback/xenbus.c | 768 | ||||
-rw-r--r-- | drivers/block/xen-blkfront.c | 51 |
22 files changed, 2637 insertions, 295 deletions
diff --git a/drivers/block/Kconfig b/drivers/block/Kconfig index 83c32cb72582..717d6e4e18d3 100644 --- a/drivers/block/Kconfig +++ b/drivers/block/Kconfig @@ -470,6 +470,27 @@ config XEN_BLKDEV_FRONTEND block device driver. It communicates with a back-end driver in another domain which drives the actual block device. +config XEN_BLKDEV_BACKEND + tristate "Block-device backend driver" + depends on XEN_BACKEND + help + The block-device backend driver allows the kernel to export its + block devices to other guests via a high-performance shared-memory + interface. + + The corresponding Linux frontend driver is enabled by the + CONFIG_XEN_BLKDEV_FRONTEND configuration option. + + The backend driver attaches itself to a any block device specified + in the XenBus configuration. There are no limits to what the block + device as long as it has a major and minor. + + If you are compiling a kernel to run in a Xen block backend driver + domain (often this is domain 0) you should say Y here. To + compile this driver as a module, chose M here: the module + will be called xen-blkback. + + config VIRTIO_BLK tristate "Virtio block driver (EXPERIMENTAL)" depends on EXPERIMENTAL && VIRTIO diff --git a/drivers/block/Makefile b/drivers/block/Makefile index 40528ba56d1b..76646e9a1c91 100644 --- a/drivers/block/Makefile +++ b/drivers/block/Makefile @@ -36,6 +36,7 @@ obj-$(CONFIG_BLK_DEV_UB) += ub.o obj-$(CONFIG_BLK_DEV_HD) += hd.o obj-$(CONFIG_XEN_BLKDEV_FRONTEND) += xen-blkfront.o +obj-$(CONFIG_XEN_BLKDEV_BACKEND) += xen-blkback/ obj-$(CONFIG_BLK_DEV_DRBD) += drbd/ obj-$(CONFIG_BLK_DEV_RBD) += rbd.o diff --git a/drivers/block/cciss.c b/drivers/block/cciss.c index 9bf13988f1a2..8f4ef656a1af 100644 --- a/drivers/block/cciss.c +++ b/drivers/block/cciss.c @@ -64,6 +64,10 @@ MODULE_DESCRIPTION("Driver for HP Smart Array Controllers"); MODULE_SUPPORTED_DEVICE("HP Smart Array Controllers"); MODULE_VERSION("3.6.26"); MODULE_LICENSE("GPL"); +static int cciss_tape_cmds = 6; +module_param(cciss_tape_cmds, int, 0644); +MODULE_PARM_DESC(cciss_tape_cmds, + "number of commands to allocate for tape devices (default: 6)"); static DEFINE_MUTEX(cciss_mutex); static struct proc_dir_entry *proc_cciss; @@ -194,6 +198,8 @@ static int __devinit cciss_find_cfg_addrs(struct pci_dev *pdev, static int __devinit cciss_pci_find_memory_BAR(struct pci_dev *pdev, unsigned long *memory_bar); static inline u32 cciss_tag_discard_error_bits(ctlr_info_t *h, u32 tag); +static __devinit int write_driver_ver_to_cfgtable( + CfgTable_struct __iomem *cfgtable); /* performant mode helper functions */ static void calc_bucket_map(int *bucket, int num_buckets, int nsgs, @@ -556,7 +562,7 @@ static void __devinit cciss_procinit(ctlr_info_t *h) #define to_hba(n) container_of(n, struct ctlr_info, dev) #define to_drv(n) container_of(n, drive_info_struct, dev) -/* List of controllers which cannot be reset on kexec with reset_devices */ +/* List of controllers which cannot be hard reset on kexec with reset_devices */ static u32 unresettable_controller[] = { 0x324a103C, /* Smart Array P712m */ 0x324b103C, /* SmartArray P711m */ @@ -574,23 +580,45 @@ static u32 unresettable_controller[] = { 0x409D0E11, /* Smart Array 6400 EM */ }; -static int ctlr_is_resettable(struct ctlr_info *h) +/* List of controllers which cannot even be soft reset */ +static u32 soft_unresettable_controller[] = { + 0x409C0E11, /* Smart Array 6400 */ + 0x409D0E11, /* Smart Array 6400 EM */ +}; + +static int ctlr_is_hard_resettable(u32 board_id) { int i; for (i = 0; i < ARRAY_SIZE(unresettable_controller); i++) - if (unresettable_controller[i] == h->board_id) + if (unresettable_controller[i] == board_id) return 0; return 1; } +static int ctlr_is_soft_resettable(u32 board_id) +{ + int i; + + for (i = 0; i < ARRAY_SIZE(soft_unresettable_controller); i++) + if (soft_unresettable_controller[i] == board_id) + return 0; + return 1; +} + +static int ctlr_is_resettable(u32 board_id) +{ + return ctlr_is_hard_resettable(board_id) || + ctlr_is_soft_resettable(board_id); +} + static ssize_t host_show_resettable(struct device *dev, struct device_attribute *attr, char *buf) { struct ctlr_info *h = to_hba(dev); - return snprintf(buf, 20, "%d\n", ctlr_is_resettable(h)); + return snprintf(buf, 20, "%d\n", ctlr_is_resettable(h->board_id)); } static DEVICE_ATTR(resettable, S_IRUGO, host_show_resettable, NULL); @@ -2567,7 +2595,7 @@ static int fill_cmd(ctlr_info_t *h, CommandList_struct *c, __u8 cmd, void *buff, } } else if (cmd_type == TYPE_MSG) { switch (cmd) { - case 0: /* ABORT message */ + case CCISS_ABORT_MSG: c->Request.CDBLen = 12; c->Request.Type.Attribute = ATTR_SIMPLE; c->Request.Type.Direction = XFER_WRITE; @@ -2577,16 +2605,16 @@ static int fill_cmd(ctlr_info_t *h, CommandList_struct *c, __u8 cmd, void *buff, /* buff contains the tag of the command to abort */ memcpy(&c->Request.CDB[4], buff, 8); break; - case 1: /* RESET message */ + case CCISS_RESET_MSG: c->Request.CDBLen = 16; c->Request.Type.Attribute = ATTR_SIMPLE; c->Request.Type.Direction = XFER_NONE; c->Request.Timeout = 0; memset(&c->Request.CDB[0], 0, sizeof(c->Request.CDB)); c->Request.CDB[0] = cmd; /* reset */ - c->Request.CDB[1] = 0x03; /* reset a target */ + c->Request.CDB[1] = CCISS_RESET_TYPE_TARGET; break; - case 3: /* No-Op message */ + case CCISS_NOOP_MSG: c->Request.CDBLen = 1; c->Request.Type.Attribute = ATTR_SIMPLE; c->Request.Type.Direction = XFER_WRITE; @@ -2615,6 +2643,31 @@ static int fill_cmd(ctlr_info_t *h, CommandList_struct *c, __u8 cmd, void *buff, return status; } +static int __devinit cciss_send_reset(ctlr_info_t *h, unsigned char *scsi3addr, + u8 reset_type) +{ + CommandList_struct *c; + int return_status; + + c = cmd_alloc(h); + if (!c) + return -ENOMEM; + return_status = fill_cmd(h, c, CCISS_RESET_MSG, NULL, 0, 0, + CTLR_LUNID, TYPE_MSG); + c->Request.CDB[1] = reset_type; /* fill_cmd defaults to target reset */ + if (return_status != IO_OK) { + cmd_special_free(h, c); + return return_status; + } + c->waiting = NULL; + enqueue_cmd_and_start_io(h, c); + /* Don't wait for completion, the reset won't complete. Don't free + * the command either. This is the last command we will send before + * re-initializing everything, so it doesn't matter and won't leak. + */ + return 0; +} + static int check_target_status(ctlr_info_t *h, CommandList_struct *c) { switch (c->err_info->ScsiStatus) { @@ -3461,6 +3514,63 @@ static inline u32 process_nonindexed_cmd(ctlr_info_t *h, u32 raw_tag) return next_command(h); } +/* Some controllers, like p400, will give us one interrupt + * after a soft reset, even if we turned interrupts off. + * Only need to check for this in the cciss_xxx_discard_completions + * functions. + */ +static int ignore_bogus_interrupt(ctlr_info_t *h) +{ + if (likely(!reset_devices)) + return 0; + + if (likely(h->interrupts_enabled)) + return 0; + + dev_info(&h->pdev->dev, "Received interrupt while interrupts disabled " + "(known firmware bug.) Ignoring.\n"); + + return 1; +} + +static irqreturn_t cciss_intx_discard_completions(int irq, void *dev_id) +{ + ctlr_info_t *h = dev_id; + unsigned long flags; + u32 raw_tag; + + if (ignore_bogus_interrupt(h)) + return IRQ_NONE; + + if (interrupt_not_for_us(h)) + return IRQ_NONE; + spin_lock_irqsave(&h->lock, flags); + while (interrupt_pending(h)) { + raw_tag = get_next_completion(h); + while (raw_tag != FIFO_EMPTY) + raw_tag = next_command(h); + } + spin_unlock_irqrestore(&h->lock, flags); + return IRQ_HANDLED; +} + +static irqreturn_t cciss_msix_discard_completions(int irq, void *dev_id) +{ + ctlr_info_t *h = dev_id; + unsigned long flags; + u32 raw_tag; + + if (ignore_bogus_interrupt(h)) + return IRQ_NONE; + + spin_lock_irqsave(&h->lock, flags); + raw_tag = get_next_completion(h); + while (raw_tag != FIFO_EMPTY) + raw_tag = next_command(h); + spin_unlock_irqrestore(&h->lock, flags); + return IRQ_HANDLED; +} + static irqreturn_t do_cciss_intx(int irq, void *dev_id) { ctlr_info_t *h = dev_id; @@ -4078,6 +4188,9 @@ static int __devinit cciss_find_cfgtables(ctlr_info_t *h) cfg_base_addr_index) + cfg_offset, sizeof(h->cfgtable)); if (!h->cfgtable) return -ENOMEM; + rc = write_driver_ver_to_cfgtable(h->cfgtable); + if (rc) + return rc; /* Find performant mode table. */ trans_offset = readl(&h->cfgtable->TransMethodOffset); h->transtable = remap_pci_mem(pci_resource_start(h->pdev, @@ -4112,7 +4225,7 @@ static void __devinit cciss_get_max_perf_mode_cmds(struct ctlr_info *h) 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->nr_cmds = h->max_commands - 4 - cciss_tape_cmds; h->maxsgentries = readl(&(h->cfgtable->MaxSGElements)); /* * Limit in-command s/g elements to 32 save dma'able memory. @@ -4348,7 +4461,7 @@ static __devinit int cciss_message(struct pci_dev *pdev, unsigned char opcode, u tag = readl(vaddr + SA5_REPLY_PORT_OFFSET); if ((tag & ~3) == paddr32) break; - schedule_timeout_uninterruptible(HZ); + msleep(CCISS_POST_RESET_NOOP_TIMEOUT_MSECS); } iounmap(vaddr); @@ -4375,11 +4488,10 @@ static __devinit int cciss_message(struct pci_dev *pdev, unsigned char opcode, u return 0; } -#define cciss_soft_reset_controller(p) cciss_message(p, 1, 0) #define cciss_noop(p) cciss_message(p, 3, 0) static int cciss_controller_hard_reset(struct pci_dev *pdev, - void * __iomem vaddr, bool use_doorbell) + void * __iomem vaddr, u32 use_doorbell) { u16 pmcsr; int pos; @@ -4390,8 +4502,7 @@ static int cciss_controller_hard_reset(struct pci_dev *pdev, * 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); + writel(use_doorbell, vaddr + SA5_DOORBELL); } else { /* Try to do it the PCI power state way */ /* Quoting from the Open CISS Specification: "The Power @@ -4422,12 +4533,64 @@ static int cciss_controller_hard_reset(struct pci_dev *pdev, pmcsr &= ~PCI_PM_CTRL_STATE_MASK; pmcsr |= PCI_D0; pci_write_config_word(pdev, pos + PCI_PM_CTRL, pmcsr); - - msleep(500); } return 0; } +static __devinit void init_driver_version(char *driver_version, int len) +{ + memset(driver_version, 0, len); + strncpy(driver_version, "cciss " DRIVER_NAME, len - 1); +} + +static __devinit int write_driver_ver_to_cfgtable( + CfgTable_struct __iomem *cfgtable) +{ + char *driver_version; + int i, size = sizeof(cfgtable->driver_version); + + driver_version = kmalloc(size, GFP_KERNEL); + if (!driver_version) + return -ENOMEM; + + init_driver_version(driver_version, size); + for (i = 0; i < size; i++) + writeb(driver_version[i], &cfgtable->driver_version[i]); + kfree(driver_version); + return 0; +} + +static __devinit void read_driver_ver_from_cfgtable( + CfgTable_struct __iomem *cfgtable, unsigned char *driver_ver) +{ + int i; + + for (i = 0; i < sizeof(cfgtable->driver_version); i++) + driver_ver[i] = readb(&cfgtable->driver_version[i]); +} + +static __devinit int controller_reset_failed( + CfgTable_struct __iomem *cfgtable) +{ + + char *driver_ver, *old_driver_ver; + int rc, size = sizeof(cfgtable->driver_version); + + old_driver_ver = kmalloc(2 * size, GFP_KERNEL); + if (!old_driver_ver) + return -ENOMEM; + driver_ver = old_driver_ver + size; + + /* After a reset, the 32 bytes of "driver version" in the cfgtable + * should have been changed, otherwise we know the reset failed. + */ + init_driver_version(old_driver_ver, size); + read_driver_ver_from_cfgtable(cfgtable, driver_ver); + rc = !memcmp(driver_ver, old_driver_ver, size); + kfree(old_driver_ver); + return rc; +} + /* 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) @@ -4437,10 +4600,10 @@ static __devinit int cciss_kdump_hard_reset_controller(struct pci_dev *pdev) u64 cfg_base_addr_index; void __iomem *vaddr; unsigned long paddr; - u32 misc_fw_support, active_transport; + u32 misc_fw_support; int rc; CfgTable_struct __iomem *cfgtable; - bool use_doorbell; + u32 use_doorbell; u32 board_id; u16 command_register; @@ -4464,12 +4627,16 @@ static __devinit int cciss_kdump_hard_reset_controller(struct pci_dev *pdev) * likely not be happy. Just forbid resetting this conjoined mess. */ cciss_lookup_board_id(pdev, &board_id); - if (board_id == 0x409C0E11 || board_id == 0x409D0E11) { + if (!ctlr_is_resettable(board_id)) { dev_warn(&pdev->dev, "Cannot reset Smart Array 640x " "due to shared cache module."); return -ENODEV; } + /* if controller is soft- but not hard resettable... */ + if (!ctlr_is_hard_resettable(board_id)) + return -ENOTSUPP; /* try soft reset later. */ + /* Save the PCI command register */ pci_read_config_word(pdev, 4, &command_register); /* Turn the board off. This is so that later pci_restore_state() @@ -4497,16 +4664,28 @@ static __devinit int cciss_kdump_hard_reset_controller(struct pci_dev *pdev) rc = -ENOMEM; goto unmap_vaddr; } + rc = write_driver_ver_to_cfgtable(cfgtable); + if (rc) + goto unmap_vaddr; - /* 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; - - /* The doorbell reset seems to cause lockups on some Smart - * Arrays (e.g. P410, P410i, maybe others). Until this is - * fixed or at least isolated, avoid the doorbell reset. + /* If reset via doorbell register is supported, use that. + * There are two such methods. Favor the newest method. */ - use_doorbell = 0; + misc_fw_support = readl(&cfgtable->misc_fw_support); + use_doorbell = misc_fw_support & MISC_FW_DOORBELL_RESET2; + if (use_doorbell) { + use_doorbell = DOORBELL_CTLR_RESET2; + } else { + use_doorbell = misc_fw_support & MISC_FW_DOORBELL_RESET; + if (use_doorbell) { + dev_warn(&pdev->dev, "Controller claims that " + "'Bit 2 doorbell reset' is " + "supported, but not 'bit 5 doorbell reset'. " + "Firmware update is recommended.\n"); + rc = -ENOTSUPP; /* use the soft reset */ + goto unmap_cfgtable; + } + } rc = cciss_controller_hard_reset(pdev, vaddr, use_doorbell); if (rc) @@ -4524,30 +4703,31 @@ static __devinit int cciss_kdump_hard_reset_controller(struct pci_dev *pdev) msleep(CCISS_POST_RESET_PAUSE_MSECS); /* Wait for board to become not ready, then ready. */ - dev_info(&pdev->dev, "Waiting for board to become ready.\n"); + dev_info(&pdev->dev, "Waiting for board to reset.\n"); rc = cciss_wait_for_board_state(pdev, vaddr, BOARD_NOT_READY); - if (rc) /* Don't bail, might be E500, etc. which can't be reset */ - dev_warn(&pdev->dev, - "failed waiting for board to become not ready\n"); + if (rc) { + dev_warn(&pdev->dev, "Failed waiting for board to hard reset." + " Will try soft reset.\n"); + rc = -ENOTSUPP; /* Not expected, but try soft reset later */ + goto unmap_cfgtable; + } rc = cciss_wait_for_board_state(pdev, vaddr, BOARD_READY); if (rc) { dev_warn(&pdev->dev, - "failed waiting for board to become ready\n"); + "failed waiting for board to become ready " + "after hard reset\n"); goto unmap_cfgtable; } - dev_info(&pdev->dev, "board ready.\n"); - /* 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; + rc = controller_reset_failed(vaddr); + if (rc < 0) + goto unmap_cfgtable; + if (rc) { + dev_warn(&pdev->dev, "Unable to successfully hard reset " + "controller. Will try soft reset.\n"); + rc = -ENOTSUPP; /* Not expected, but try soft reset later */ + } else { + dev_info(&pdev->dev, "Board ready after hard reset.\n"); } unmap_cfgtable: @@ -4574,11 +4754,12 @@ static __devinit int cciss_init_reset_devices(struct pci_dev *pdev) * due to concerns about shared bbwc between 6402/6404 pair. */ if (rc == -ENOTSUPP) - return 0; /* just try to do the kdump anyhow. */ + return rc; /* just try to do the kdump anyhow. */ if (rc) return -ENODEV; /* Now try to get the controller to respond to a no-op */ + dev_warn(&pdev->dev, "Waiting for controller to respond to no-op\n"); for (i = 0; i < CCISS_POST_RESET_NOOP_RETRIES; i++) { if (cciss_noop(pdev) == 0) break; @@ -4591,6 +4772,148 @@ static __devinit int cciss_init_reset_devices(struct pci_dev *pdev) return 0; } +static __devinit int cciss_allocate_cmd_pool(ctlr_info_t *h) +{ + h->cmd_pool_bits = kmalloc( + DIV_ROUND_UP(h->nr_cmds, BITS_PER_LONG) * + sizeof(unsigned long), GFP_KERNEL); + h->cmd_pool = pci_alloc_consistent(h->pdev, + h->nr_cmds * sizeof(CommandList_struct), + &(h->cmd_pool_dhandle)); + h->errinfo_pool = 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"); + return -ENOMEM; + } + return 0; +} + +static __devinit int cciss_allocate_scatterlists(ctlr_info_t *h) +{ + int i; + + /* zero it, so that on free we need not know how many were alloc'ed */ + h->scatter_list = kzalloc(h->max_commands * + sizeof(struct scatterlist *), GFP_KERNEL); + if (!h->scatter_list) + return -ENOMEM; + + for (i = 0; i < h->nr_cmds; i++) { + h->scatter_list[i] = kmalloc(sizeof(struct scatterlist) * + h->maxsgentries, GFP_KERNEL); + if (h->scatter_list[i] == NULL) { + dev_err(&h->pdev->dev, "could not allocate " + "s/g lists\n"); + return -ENOMEM; + } + } + return 0; +} + +static void cciss_free_scatterlists(ctlr_info_t *h) +{ + int i; + + if (h->scatter_list) { + for (i = 0; i < h->nr_cmds; i++) + kfree(h->scatter_list[i]); + kfree(h->scatter_list); + } +} + +static void cciss_free_cmd_pool(ctlr_info_t *h) +{ + kfree(h->cmd_pool_bits); + 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); +} + +static int cciss_request_irq(ctlr_info_t *h, + irqreturn_t (*msixhandler)(int, void *), + irqreturn_t (*intxhandler)(int, void *)) +{ + if (h->msix_vector || h->msi_vector) { + if (!request_irq(h->intr[PERF_MODE_INT], msixhandler, + IRQF_DISABLED, h->devname, h)) + return 0; + dev_err(&h->pdev->dev, "Unable to get msi irq %d" + " for %s\n", h->intr[PERF_MODE_INT], + h->devname); + return -1; + } + + if (!request_irq(h->intr[PERF_MODE_INT], intxhandler, + IRQF_DISABLED, h->devname, h)) + return 0; + dev_err(&h->pdev->dev, "Unable to get irq %d for %s\n", + h->intr[PERF_MODE_INT], h->devname); + return -1; +} + +static int __devinit cciss_kdump_soft_reset(ctlr_info_t *h) +{ + if (cciss_send_reset(h, CTLR_LUNID, CCISS_RESET_TYPE_CONTROLLER)) { + dev_warn(&h->pdev->dev, "Resetting array controller failed.\n"); + return -EIO; + } + + dev_info(&h->pdev->dev, "Waiting for board to soft reset.\n"); + if (cciss_wait_for_board_state(h->pdev, h->vaddr, BOARD_NOT_READY)) { + dev_warn(&h->pdev->dev, "Soft reset had no effect.\n"); + return -1; + } + + dev_info(&h->pdev->dev, "Board reset, awaiting READY status.\n"); + if (cciss_wait_for_board_state(h->pdev, h->vaddr, BOARD_READY)) { + dev_warn(&h->pdev->dev, "Board failed to become ready " + "after soft reset.\n"); + return -1; + } + + return 0; +} + +static void cciss_undo_allocations_after_kdump_soft_reset(ctlr_info_t *h) +{ + int ctlr = h->ctlr; + + free_irq(h->intr[PERF_MODE_INT], h); +#ifdef CONFIG_PCI_MSI + if (h->msix_vector) + pci_disable_msix(h->pdev); + else if (h->msi_vector) + pci_disable_msi(h->pdev); +#endif /* CONFIG_PCI_MSI */ + cciss_free_sg_chain_blocks(h->cmd_sg_list, h->nr_cmds); + cciss_free_scatterlists(h); + cciss_free_cmd_pool(h); + kfree(h->blockFetchTable); + if (h->reply_pool) + pci_free_consistent(h->pdev, h->max_commands * sizeof(__u64), + h->reply_pool, h->reply_pool_dhandle); + if (h->transtable) + iounmap(h->transtable); + if (h->cfgtable) + iounmap(h->cfgtable); + if (h->vaddr) + iounmap(h->vaddr); + unregister_blkdev(h->major, h->devname); + cciss_destroy_hba_sysfs_entry(h); + pci_release_regions(h->pdev); + kfree(h); + hba[ctlr] = NULL; +} + /* * This is it. Find all the controllers and register them. I really hate * stealing all these major device numbers. @@ -4601,15 +4924,28 @@ static int __devinit cciss_init_one(struct pci_dev *pdev, { int i; int j = 0; - int k = 0; int rc; + int try_soft_reset = 0; int dac, return_code; InquiryData_struct *inq_buff; ctlr_info_t *h; + unsigned long flags; rc = cciss_init_reset_devices(pdev); - if (rc) - return rc; + if (rc) { + if (rc != -ENOTSUPP) + return rc; + /* If the reset fails in a particular way (it has no way to do + * a proper hard reset, so returns -ENOTSUPP) we can try to do + * a soft reset once we get the controller configured up to the + * point that it can accept a command. + */ + try_soft_reset = 1; + rc = 0; + } + +reinit_after_soft_reset: + i = alloc_cciss_hba(pdev); if (i < 0) return -1; @@ -4627,6 +4963,11 @@ static int __devinit cciss_init_one(struct pci_dev *pdev, sprintf(h->devname, "cciss%d", i); h->ctlr = i; + if (cciss_tape_cmds < 2) + cciss_tape_cmds = 2; + if (cciss_tape_cmds > 16) + cciss_tape_cmds = 16; + init_completion(&h->scan_wait); if (cciss_create_hba_sysfs_entry(h)) @@ -4662,62 +5003,20 @@ static int __devinit cciss_init_one(struct pci_dev *pdev, /* make sure the board interrupts are off */ 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; - } - } + rc = cciss_request_irq(h, do_cciss_msix_intr, do_cciss_intx); + if (rc) + goto clean2; 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"); - h->cmd_pool_bits = - kmalloc(DIV_ROUND_UP(h->nr_cmds, BITS_PER_LONG) - * sizeof(unsigned long), GFP_KERNEL); - 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"); + if (cciss_allocate_cmd_pool(h)) goto clean4; - } - /* Need space for temp scatter list */ - h->scatter_list = kmalloc(h->max_commands * - sizeof(struct scatterlist *), - GFP_KERNEL); - if (!h->scatter_list) + if (cciss_allocate_scatterlists(h)) goto clean4; - for (k = 0; k < h->nr_cmds; k++) { - h->scatter_list[k] = kmalloc(sizeof(struct scatterlist) * - h->maxsgentries, - GFP_KERNEL); - if (h->scatter_list[k] == NULL) { - dev_err(&h->pdev->dev, - "could not allocate s/g lists\n"); - goto clean4; - } - } h->cmd_sg_list = cciss_allocate_sg_chain_blocks(h, h->chainsize, h->nr_cmds); if (!h->cmd_sg_list && h->chainsize > 0) @@ -4741,6 +5040,62 @@ static int __devinit cciss_init_one(struct pci_dev *pdev, h->gendisk[j] = NULL; } + /* At this point, the controller is ready to take commands. + * Now, if reset_devices and the hard reset didn't work, try + * the soft reset and see if that works. + */ + if (try_soft_reset) { + + /* This is kind of gross. We may or may not get a completion + * from the soft reset command, and if we do, then the value + * from the fifo may or may not be valid. So, we wait 10 secs + * after the reset throwing away any completions we get during + * that time. Unregister the interrupt handler and register + * fake ones to scoop up any residual completions. + */ + spin_lock_irqsave(&h->lock, flags); + h->access.set_intr_mask(h, CCISS_INTR_OFF); + spin_unlock_irqrestore(&h->lock, flags); + free_irq(h->intr[PERF_MODE_INT], h); + rc = cciss_request_irq(h, cciss_msix_discard_completions, + cciss_intx_discard_completions); + if (rc) { + dev_warn(&h->pdev->dev, "Failed to request_irq after " + "soft reset.\n"); + goto clean4; + } + + rc = cciss_kdump_soft_reset(h); + if (rc) { + dev_warn(&h->pdev->dev, "Soft reset failed.\n"); + goto clean4; + } + + dev_info(&h->pdev->dev, "Board READY.\n"); + dev_info(&h->pdev->dev, + "Waiting for stale completions to drain.\n"); + h->access.set_intr_mask(h, CCISS_INTR_ON); + msleep(10000); + h->access.set_intr_mask(h, CCISS_INTR_OFF); + + rc = controller_reset_failed(h->cfgtable); + if (rc) + dev_info(&h->pdev->dev, + "Soft reset appears to have failed.\n"); + + /* since the controller's reset, we have to go back and re-init + * everything. Easiest to just forget what we've done and do it + * all over again. + */ + cciss_undo_allocations_after_kdump_soft_reset(h); + try_soft_reset = 0; + if (rc) + /* don't go to clean4, we already unallocated */ + return -ENODEV; + + goto reinit_after_soft_reset; + } + cciss_scsi_setup(h); /* Turn the interrupts on so we can service requests */ @@ -4775,21 +5130,9 @@ static int __devinit cciss_init_one(struct pci_dev *pdev, return 1; clean4: - kfree(h->cmd_pool_bits); - /* Free up sg elements */ - for (k-- ; k >= 0; k--) - kfree(h->scatter_list[k]); - kfree(h->scatter_list); + cciss_free_cmd_pool(h); + cciss_free_scatterlists(h); 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(h->major, h->devname); @@ -4887,16 +5230,16 @@ static void __devexit cciss_remove_one(struct pci_dev *pdev) iounmap(h->cfgtable); iounmap(h->vaddr); - 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); + cciss_free_cmd_pool(h); /* Free up sg elements */ 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); + kfree(h->blockFetchTable); + if (h->reply_pool) + pci_free_consistent(h->pdev, h->max_commands * sizeof(__u64), + h->reply_pool, h->reply_pool_dhandle); /* * Deliberately omit pci_disable_device(): it does something nasty to * Smart Array controllers that pci_enable_device does not undo diff --git a/drivers/block/cciss.h b/drivers/block/cciss.h index 554bbd907d14..16b4d58d84dd 100644 --- a/drivers/block/cciss.h + |