diff options
Diffstat (limited to 'drivers/block')
-rw-r--r-- | drivers/block/amiflop.c | 2 | ||||
-rw-r--r-- | drivers/block/aoe/aoeblk.c | 3 | ||||
-rw-r--r-- | drivers/block/ataflop.c | 2 | ||||
-rw-r--r-- | drivers/block/cciss.c | 136 | ||||
-rw-r--r-- | drivers/block/cciss.h | 4 | ||||
-rw-r--r-- | drivers/block/cciss_scsi.c | 8 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_actlog.c | 42 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_int.h | 52 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_main.c | 148 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_nl.c | 25 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_proc.c | 1 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_receiver.c | 232 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_req.c | 38 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_req.h | 3 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_worker.c | 34 | ||||
-rw-r--r-- | drivers/block/floppy.c | 4 | ||||
-rw-r--r-- | drivers/block/loop.c | 6 | ||||
-rw-r--r-- | drivers/block/rbd.c | 748 | ||||
-rw-r--r-- | drivers/block/xen-blkfront.c | 57 |
19 files changed, 802 insertions, 743 deletions
diff --git a/drivers/block/amiflop.c b/drivers/block/amiflop.c index a1725e6..7888501 100644 --- a/drivers/block/amiflop.c +++ b/drivers/block/amiflop.c @@ -1341,7 +1341,7 @@ static struct request *set_next_request(void) { struct request_queue *q; int cnt = FD_MAX_UNITS; - struct request *rq; + struct request *rq = NULL; /* Find next queue we can dispatch from */ fdc_queue = fdc_queue + 1; diff --git a/drivers/block/aoe/aoeblk.c b/drivers/block/aoe/aoeblk.c index 541e188..528f631 100644 --- a/drivers/block/aoe/aoeblk.c +++ b/drivers/block/aoe/aoeblk.c @@ -180,9 +180,6 @@ aoeblk_make_request(struct request_queue *q, struct bio *bio) BUG(); bio_endio(bio, -ENXIO); return 0; - } else if (bio->bi_rw & REQ_HARDBARRIER) { - bio_endio(bio, -EOPNOTSUPP); - return 0; } else if (bio->bi_io_vec == NULL) { printk(KERN_ERR "aoe: bi_io_vec is NULL\n"); BUG(); diff --git a/drivers/block/ataflop.c b/drivers/block/ataflop.c index 4e4cc6c..605a67e 100644 --- a/drivers/block/ataflop.c +++ b/drivers/block/ataflop.c @@ -1399,7 +1399,7 @@ static struct request *set_next_request(void) { struct request_queue *q; int old_pos = fdc_queue; - struct request *rq; + struct request *rq = NULL; do { q = unit[fdc_queue].disk->queue; diff --git a/drivers/block/cciss.c b/drivers/block/cciss.c index 2cc4dda4..8e0f925 100644 --- a/drivers/block/cciss.c +++ b/drivers/block/cciss.c @@ -66,6 +66,7 @@ MODULE_VERSION("3.6.26"); MODULE_LICENSE("GPL"); static DEFINE_MUTEX(cciss_mutex); +static struct proc_dir_entry *proc_cciss; #include "cciss_cmd.h" #include "cciss.h" @@ -113,6 +114,8 @@ static struct board_type products[] = { {0x409D0E11, "Smart Array 6400 EM", &SA5_access}, {0x40910E11, "Smart Array 6i", &SA5_access}, {0x3225103C, "Smart Array P600", &SA5_access}, + {0x3223103C, "Smart Array P800", &SA5_access}, + {0x3234103C, "Smart Array P400", &SA5_access}, {0x3235103C, "Smart Array P400i", &SA5_access}, {0x3211103C, "Smart Array E200i", &SA5_access}, {0x3212103C, "Smart Array E200", &SA5_access}, @@ -361,8 +364,6 @@ static const char *raid_label[] = { "0", "4", "1(1+0)", "5", "5+1", "ADG", #define ENG_GIG_FACTOR (ENG_GIG/512) #define ENGAGE_SCSI "engage scsi" -static struct proc_dir_entry *proc_cciss; - static void cciss_seq_show_header(struct seq_file *seq) { ctlr_info_t *h = seq->private; @@ -2833,6 +2834,8 @@ static int cciss_revalidate(struct gendisk *disk) InquiryData_struct *inq_buff = NULL; for (logvol = 0; logvol < CISS_MAX_LUN; logvol++) { + if (!h->drv[logvol]) + continue; if (memcmp(h->drv[logvol]->LunID, drv->LunID, sizeof(drv->LunID)) == 0) { FOUND = 1; @@ -3753,7 +3756,7 @@ static void __devinit cciss_wait_for_mode_change_ack(ctlr_info_t *h) for (i = 0; i < MAX_CONFIG_WAIT; i++) { if (!(readl(h->vaddr + SA5_DOORBELL) & CFGTBL_ChangeReq)) break; - msleep(10); + usleep_range(10000, 20000); } } @@ -3937,10 +3940,9 @@ static int __devinit cciss_lookup_board_id(struct pci_dev *pdev, u32 *board_id) *board_id = ((subsystem_device_id << 16) & 0xffff0000) | subsystem_vendor_id; - for (i = 0; i < ARRAY_SIZE(products); i++) { + for (i = 0; i < ARRAY_SIZE(products); i++) if (*board_id == products[i].board_id) return i; - } dev_warn(&pdev->dev, "unrecognized board ID: 0x%08x, ignoring.\n", *board_id); return -ENODEV; @@ -3971,18 +3973,31 @@ static int __devinit cciss_pci_find_memory_BAR(struct pci_dev *pdev, return -ENODEV; } -static int __devinit cciss_wait_for_board_ready(ctlr_info_t *h) +static int __devinit cciss_wait_for_board_state(struct pci_dev *pdev, + void __iomem *vaddr, int wait_for_ready) +#define BOARD_READY 1 +#define BOARD_NOT_READY 0 { - int i; + int i, iterations; u32 scratchpad; - for (i = 0; i < CCISS_BOARD_READY_ITERATIONS; i++) { - scratchpad = readl(h->vaddr + SA5_SCRATCHPAD_OFFSET); - if (scratchpad == CCISS_FIRMWARE_READY) - return 0; + if (wait_for_ready) + iterations = CCISS_BOARD_READY_ITERATIONS; + else + iterations = CCISS_BOARD_NOT_READY_ITERATIONS; + + for (i = 0; i < iterations; i++) { + scratchpad = readl(vaddr + SA5_SCRATCHPAD_OFFSET); + if (wait_for_ready) { + if (scratchpad == CCISS_FIRMWARE_READY) + return 0; + } else { + 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"); + dev_warn(&pdev->dev, "board not ready, timed out.\n"); return -ENODEV; } @@ -4031,6 +4046,11 @@ static int __devinit cciss_find_cfgtables(ctlr_info_t *h) static void __devinit cciss_get_max_perf_mode_cmds(struct ctlr_info *h) { h->max_commands = readl(&(h->cfgtable->MaxPerformantModeCommands)); + + /* Limit commands in memory limited kdump scenario. */ + if (reset_devices && h->max_commands > 32) + h->max_commands = 32; + if (h->max_commands < 16) { dev_warn(&h->pdev->dev, "Controller reports " "max supported commands of %d, an obvious lie. " @@ -4148,7 +4168,7 @@ static int __devinit cciss_pci_init(ctlr_info_t *h) err = -ENOMEM; goto err_out_free_res; } - err = cciss_wait_for_board_ready(h); + err = cciss_wait_for_board_state(h->pdev, h->vaddr, BOARD_READY); if (err) goto err_out_free_res; err = cciss_find_cfgtables(h); @@ -4313,36 +4333,6 @@ static __devinit int cciss_message(struct pci_dev *pdev, unsigned char opcode, u #define cciss_soft_reset_controller(p) cciss_message(p, 1, 0) #define cciss_noop(p) cciss_message(p, 3, 0) -static __devinit int cciss_reset_msi(struct pci_dev *pdev) -{ -/* the #defines are stolen from drivers/pci/msi.h. */ -#define msi_control_reg(base) (base + PCI_MSI_FLAGS) -#define PCI_MSIX_FLAGS_ENABLE (1 << 15) - - int pos; - u16 control = 0; - - pos = pci_find_capability(pdev, PCI_CAP_ID_MSI); - if (pos) { - pci_read_config_word(pdev, msi_control_reg(pos), &control); - if (control & PCI_MSI_FLAGS_ENABLE) { - dev_info(&pdev->dev, "resetting MSI\n"); - pci_write_config_word(pdev, msi_control_reg(pos), control & ~PCI_MSI_FLAGS_ENABLE); - } - } - - pos = pci_find_capability(pdev, PCI_CAP_ID_MSIX); - if (pos) { - pci_read_config_word(pdev, msi_control_reg(pos), &control); - if (control & PCI_MSIX_FLAGS_ENABLE) { - dev_info(&pdev->dev, "resetting MSI-X\n"); - pci_write_config_word(pdev, msi_control_reg(pos), control & ~PCI_MSIX_FLAGS_ENABLE); - } - } - - return 0; -} - static int cciss_controller_hard_reset(struct pci_dev *pdev, void * __iomem vaddr, bool use_doorbell) { @@ -4397,17 +4387,17 @@ static int cciss_controller_hard_reset(struct pci_dev *pdev, * 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; + int rc; CfgTable_struct __iomem *cfgtable; bool use_doorbell; u32 board_id; + u16 command_register; /* For controllers as old a the p600, this is very nearly * the same thing as @@ -4417,14 +4407,6 @@ static __devinit int cciss_kdump_hard_reset_controller(struct pci_dev *pdev) * 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. @@ -4443,8 +4425,13 @@ static __devinit int cciss_kdump_hard_reset_controller(struct pci_dev *pdev) return -ENODEV; } - for (i = 0; i < 32; i++) - pci_read_config_word(pdev, 2*i, &saved_config_space[i]); + /* 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() + * won't turn the board on before the rest of config space is ready. + */ + pci_disable_device(pdev); + pci_save_state(pdev); /* find the first memory BAR, so we can find the cfg table */ rc = cciss_pci_find_memory_BAR(pdev, &paddr); @@ -4479,26 +4466,32 @@ static __devinit int cciss_kdump_hard_reset_controller(struct pci_dev *pdev) 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. - */ - for (i = 0; i < 32; i++) { - if (i == 2 || i == 3) - continue; - pci_write_config_word(pdev, 2*i, saved_config_space[i]); + pci_restore_state(pdev); + rc = pci_enable_device(pdev); + if (rc) { + dev_warn(&pdev->dev, "failed to enable device.\n"); + goto unmap_cfgtable; } - wmb(); - pci_write_config_word(pdev, 4, saved_config_space[2]); + pci_write_config_word(pdev, 4, command_register); /* Some devices (notably the HP Smart Array 5i Controller) need a little pause here */ 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"); + 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"); + rc = cciss_wait_for_board_state(pdev, vaddr, BOARD_READY); + if (rc) { + dev_warn(&pdev->dev, + "failed waiting for board to become ready\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 @@ -4539,8 +4532,6 @@ static __devinit int cciss_init_reset_devices(struct pci_dev *pdev) 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++) { @@ -4936,7 +4927,8 @@ static void __exit cciss_cleanup(void) } } kthread_stop(cciss_scan_thread); - remove_proc_entry("driver/cciss", NULL); + if (proc_cciss) + remove_proc_entry("driver/cciss", NULL); bus_unregister(&cciss_bus_type); } diff --git a/drivers/block/cciss.h b/drivers/block/cciss.h index ae340ff..4b8933d 100644 --- a/drivers/block/cciss.h +++ b/drivers/block/cciss.h @@ -200,10 +200,14 @@ struct ctlr_info * the above. */ #define CCISS_BOARD_READY_WAIT_SECS (120) +#define CCISS_BOARD_NOT_READY_WAIT_SECS (10) #define CCISS_BOARD_READY_POLL_INTERVAL_MSECS (100) #define CCISS_BOARD_READY_ITERATIONS \ ((CCISS_BOARD_READY_WAIT_SECS * 1000) / \ CCISS_BOARD_READY_POLL_INTERVAL_MSECS) +#define CCISS_BOARD_NOT_READY_ITERATIONS \ + ((CCISS_BOARD_NOT_READY_WAIT_SECS * 1000) / \ + CCISS_BOARD_READY_POLL_INTERVAL_MSECS) #define CCISS_POST_RESET_PAUSE_MSECS (3000) #define CCISS_POST_RESET_NOOP_INTERVAL_MSECS (1000) #define CCISS_POST_RESET_NOOP_RETRIES (12) diff --git a/drivers/block/cciss_scsi.c b/drivers/block/cciss_scsi.c index 575495f..727d022 100644 --- a/drivers/block/cciss_scsi.c +++ b/drivers/block/cciss_scsi.c @@ -62,8 +62,8 @@ static int cciss_scsi_proc_info( int length, /* length of data in buffer */ int func); /* 0 == read, 1 == write */ -static int cciss_scsi_queue_command (struct scsi_cmnd *cmd, - void (* done)(struct scsi_cmnd *)); +static int cciss_scsi_queue_command (struct Scsi_Host *h, + struct scsi_cmnd *cmd); static int cciss_eh_device_reset_handler(struct scsi_cmnd *); static int cciss_eh_abort_handler(struct scsi_cmnd *); @@ -1406,7 +1406,7 @@ static void cciss_scatter_gather(ctlr_info_t *h, CommandList_struct *c, static int -cciss_scsi_queue_command (struct scsi_cmnd *cmd, void (* done)(struct scsi_cmnd *)) +cciss_scsi_queue_command_lck(struct scsi_cmnd *cmd, void (*done)(struct scsi_cmnd *)) { ctlr_info_t *h; int rc; @@ -1504,6 +1504,8 @@ cciss_scsi_queue_command (struct scsi_cmnd *cmd, void (* done)(struct scsi_cmnd return 0; } +static DEF_SCSI_QCMD(cciss_scsi_queue_command) + static void cciss_unregister_scsi(ctlr_info_t *h) { struct cciss_scsi_adapter_data_t *sa; diff --git a/drivers/block/drbd/drbd_actlog.c b/drivers/block/drbd/drbd_actlog.c index ac04ef9..ba95cba 100644 --- a/drivers/block/drbd/drbd_actlog.c +++ b/drivers/block/drbd/drbd_actlog.c @@ -78,11 +78,10 @@ static int _drbd_md_sync_page_io(struct drbd_conf *mdev, init_completion(&md_io.event); md_io.error = 0; - if ((rw & WRITE) && !test_bit(MD_NO_BARRIER, &mdev->flags)) - rw |= REQ_HARDBARRIER; + if ((rw & WRITE) && !test_bit(MD_NO_FUA, &mdev->flags)) + rw |= REQ_FUA; rw |= REQ_UNPLUG | REQ_SYNC; - retry: bio = bio_alloc(GFP_NOIO, 1); bio->bi_bdev = bdev->md_bdev; bio->bi_sector = sector; @@ -100,17 +99,6 @@ static int _drbd_md_sync_page_io(struct drbd_conf *mdev, wait_for_completion(&md_io.event); ok = bio_flagged(bio, BIO_UPTODATE) && md_io.error == 0; - /* check for unsupported barrier op. - * would rather check on EOPNOTSUPP, but that is not reliable. - * don't try again for ANY return value != 0 */ - if (unlikely((bio->bi_rw & REQ_HARDBARRIER) && !ok)) { - /* Try again with no barrier */ - dev_warn(DEV, "Barriers not supported on meta data device - disabling\n"); - set_bit(MD_NO_BARRIER, &mdev->flags); - rw &= ~REQ_HARDBARRIER; - bio_put(bio); - goto retry; - } out: bio_put(bio); return ok; @@ -284,18 +272,32 @@ w_al_write_transaction(struct drbd_conf *mdev, struct drbd_work *w, int unused) u32 xor_sum = 0; if (!get_ldev(mdev)) { - dev_err(DEV, "get_ldev() failed in w_al_write_transaction\n"); + dev_err(DEV, + "disk is %s, cannot start al transaction (-%d +%d)\n", + drbd_disk_str(mdev->state.disk), evicted, new_enr); complete(&((struct update_al_work *)w)->event); return 1; } /* do we have to do a bitmap write, first? * TODO reduce maximum latency: * submit both bios, then wait for both, - * instead of doing two synchronous sector writes. */ + * instead of doing two synchronous sector writes. + * For now, we must not write the transaction, + * if we cannot write out the bitmap of the evicted extent. */ if (mdev->state.conn < C_CONNECTED && evicted != LC_FREE) drbd_bm_write_sect(mdev, evicted/AL_EXT_PER_BM_SECT); - mutex_lock(&mdev->md_io_mutex); /* protects md_io_page, al_tr_cycle, ... */ + /* The bitmap write may have failed, causing a state change. */ + if (mdev->state.disk < D_INCONSISTENT) { + dev_err(DEV, + "disk is %s, cannot write al transaction (-%d +%d)\n", + drbd_disk_str(mdev->state.disk), evicted, new_enr); + complete(&((struct update_al_work *)w)->event); + put_ldev(mdev); + return 1; + } + + mutex_lock(&mdev->md_io_mutex); /* protects md_io_buffer, al_tr_cycle, ... */ buffer = (struct al_transaction *)page_address(mdev->md_io_page); buffer->magic = __constant_cpu_to_be32(DRBD_MAGIC); @@ -739,7 +741,7 @@ void drbd_al_apply_to_bm(struct drbd_conf *mdev) unsigned int enr; unsigned long add = 0; char ppb[10]; - int i; + int i, tmp; wait_event(mdev->al_wait, lc_try_lock(mdev->act_log)); @@ -747,7 +749,9 @@ void drbd_al_apply_to_bm(struct drbd_conf *mdev) enr = lc_element_by_index(mdev->act_log, i)->lc_number; if (enr == LC_FREE) continue; - add += drbd_bm_ALe_set_all(mdev, enr); + tmp = drbd_bm_ALe_set_all(mdev, enr); + dynamic_dev_dbg(DEV, "AL: set %d bits in extent %u\n", tmp, enr); + add += tmp; } lc_unlock(mdev->act_log); diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index 9bdcf43..1ea1a34 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -114,11 +114,11 @@ struct drbd_conf; #define D_ASSERT(exp) if (!(exp)) \ dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__) -#define ERR_IF(exp) if (({ \ - int _b = (exp) != 0; \ - if (_b) dev_err(DEV, "%s: (%s) in %s:%d\n", \ - __func__, #exp, __FILE__, __LINE__); \ - _b; \ +#define ERR_IF(exp) if (({ \ + int _b = (exp) != 0; \ + if (_b) dev_err(DEV, "ASSERT FAILED: %s: (%s) in %s:%d\n", \ + __func__, #exp, __FILE__, __LINE__); \ + _b; \ })) /* Defines to control fault insertion */ @@ -749,17 +749,12 @@ struct drbd_epoch { /* drbd_epoch flag bits */ enum { - DE_BARRIER_IN_NEXT_EPOCH_ISSUED, - DE_BARRIER_IN_NEXT_EPOCH_DONE, - DE_CONTAINS_A_BARRIER, DE_HAVE_BARRIER_NUMBER, - DE_IS_FINISHING, }; enum epoch_event { EV_PUT, EV_GOT_BARRIER_NR, - EV_BARRIER_DONE, EV_BECAME_LAST, EV_CLEANUP = 32, /* used as flag */ }; @@ -801,11 +796,6 @@ enum { __EE_CALL_AL_COMPLETE_IO, __EE_MAY_SET_IN_SYNC, - /* This epoch entry closes an epoch using a barrier. - * On sucessful completion, the epoch is released, - * and the P_BARRIER_ACK send. */ - __EE_IS_BARRIER, - /* In case a barrier failed, * we need to resubmit without the barrier flag. */ __EE_RESUBMITTED, @@ -820,7 +810,6 @@ enum { }; #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO) #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC) -#define EE_IS_BARRIER (1<<__EE_IS_BARRIER) #define EE_RESUBMITTED (1<<__EE_RESUBMITTED) #define EE_WAS_ERROR (1<<__EE_WAS_ERROR) #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST) @@ -843,16 +832,15 @@ enum { * Gets cleared when the state.conn * goes into C_CONNECTED state. */ WRITE_BM_AFTER_RESYNC, /* A kmalloc() during resync failed */ - NO_BARRIER_SUPP, /* underlying block device doesn't implement barriers */ CONSIDER_RESYNC, - MD_NO_BARRIER, /* meta data device does not support barriers, - so don't even try */ + MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */ SUSPEND_IO, /* suspend application io */ BITMAP_IO, /* suspend application io; once no more io in flight, start bitmap io */ BITMAP_IO_QUEUED, /* Started bitmap IO */ - GO_DISKLESS, /* Disk failed, local_cnt reached zero, we are going diskless */ + GO_DISKLESS, /* Disk is being detached, on io-error or admin request. */ + WAS_IO_ERROR, /* Local disk failed returned IO error */ RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */ NET_CONGESTED, /* The data socket is congested */ @@ -947,7 +935,6 @@ enum write_ordering_e { WO_none, WO_drain_io, WO_bdev_flush, - WO_bio_barrier }; struct fifo_buffer { @@ -1281,6 +1268,7 @@ extern int drbd_bmio_set_n_write(struct drbd_conf *mdev); extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev); extern int drbd_bitmap_io(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), char *why); extern void drbd_go_diskless(struct drbd_conf *mdev); +extern void drbd_ldev_destroy(struct drbd_conf *mdev); /* Meta data layout @@ -1798,17 +1786,17 @@ static inline void __drbd_chk_io_error_(struct drbd_conf *mdev, int forcedetach, case EP_PASS_ON: if (!forcedetach) { if (__ratelimit(&drbd_ratelimit_state)) - dev_err(DEV, "Local IO failed in %s." - "Passing error on...\n", where); + dev_err(DEV, "Local IO failed in %s.\n", where); break; } /* NOTE fall through to detach case if forcedetach set */ case EP_DETACH: case EP_CALL_HELPER: + set_bit(WAS_IO_ERROR, &mdev->flags); if (mdev->state.disk > D_FAILED) { _drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL); - dev_err(DEV, "Local IO failed in %s." - "Detaching...\n", where); + dev_err(DEV, + "Local IO failed in %s. Detaching...\n", where); } break; } @@ -1874,7 +1862,7 @@ static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev) static inline sector_t drbd_get_capacity(struct block_device *bdev) { /* return bdev ? get_capacity(bdev->bd_disk) : 0; */ - return bdev ? bdev->bd_inode->i_size >> 9 : 0; + return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0; } /** @@ -2127,7 +2115,11 @@ static inline void put_ldev(struct drbd_conf *mdev) __release(local); D_ASSERT(i >= 0); if (i == 0) { + if (mdev->state.disk == D_DISKLESS) + /* even internal references gone, safe to destroy */ + drbd_ldev_destroy(mdev); if (mdev->state.disk == D_FAILED) + /* all application IO references gone. */ drbd_go_diskless(mdev); wake_up(&mdev->misc_wait); } @@ -2138,6 +2130,10 @@ static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_stat { int io_allowed; + /* never get a reference while D_DISKLESS */ + if (mdev->state.disk == D_DISKLESS) + return 0; + atomic_inc(&mdev->local_cnt); io_allowed = (mdev->state.disk >= mins); if (!io_allowed) @@ -2406,12 +2402,12 @@ static inline void drbd_md_flush(struct drbd_conf *mdev) { int r; - if (test_bit(MD_NO_BARRIER, &mdev->flags)) + if (test_bit(MD_NO_FUA, &mdev->flags)) return; r = blkdev_issue_flush(mdev->ldev->md_bdev, GFP_KERNEL, NULL); if (r) { - set_bit(MD_NO_BARRIER, &mdev->flags); + set_bit(MD_NO_FUA, &mdev->flags); dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r); } } diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index 25c7a73..6be5401 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -835,6 +835,15 @@ static union drbd_state sanitize_state(struct drbd_conf *mdev, union drbd_state ns.conn != C_UNCONNECTED && ns.conn != C_DISCONNECTING && ns.conn <= C_TEAR_DOWN) ns.conn = os.conn; + /* we cannot fail (again) if we already detached */ + if (ns.disk == D_FAILED && os.disk == D_DISKLESS) + ns.disk = D_DISKLESS; + + /* if we are only D_ATTACHING yet, + * we can (and should) go directly to D_DISKLESS. */ + if (ns.disk == D_FAILED && os.disk == D_ATTACHING) + ns.disk = D_DISKLESS; + /* After C_DISCONNECTING only C_STANDALONE may follow */ if (os.conn == C_DISCONNECTING && ns.conn != C_STANDALONE) ns.conn = os.conn; @@ -1056,7 +1065,15 @@ int __drbd_set_state(struct drbd_conf *mdev, !test_and_set_bit(CONFIG_PENDING, &mdev->flags)) set_bit(DEVICE_DYING, &mdev->flags); - mdev->state.i = ns.i; + /* if we are going -> D_FAILED or D_DISKLESS, grab one extra reference + * on the ldev here, to be sure the transition -> D_DISKLESS resp. + * drbd_ldev_destroy() won't happen before our corresponding + * after_state_ch works run, where we put_ldev again. */ + if ((os.disk != D_FAILED && ns.disk == D_FAILED) || + (os.disk != D_DISKLESS && ns.disk == D_DISKLESS)) + atomic_inc(&mdev->local_cnt); + + mdev->state = ns; wake_up(&mdev->misc_wait); wake_up(&mdev->state_wait); @@ -1268,7 +1285,6 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, if (test_bit(NEW_CUR_UUID, &mdev->flags)) { drbd_uuid_new_current(mdev); clear_bit(NEW_CUR_UUID, &mdev->flags); - drbd_md_sync(mdev); } spin_lock_irq(&mdev->req_lock); _drbd_set_state(_NS(mdev, susp_fen, 0), CS_VERBOSE, NULL); @@ -1365,63 +1381,64 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, os.disk > D_INCONSISTENT && ns.disk == D_INCONSISTENT) drbd_queue_bitmap_io(mdev, &drbd_bmio_set_n_write, NULL, "set_n_write from invalidate"); - /* first half of local IO error */ - if (os.disk > D_FAILED && ns.disk == D_FAILED) { - enum drbd_io_error_p eh = EP_PASS_ON; + /* first half of local IO error, failure to attach, + * or administrative detach */ + if (os.disk != D_FAILED && ns.disk == D_FAILED) { + enum drbd_io_error_p eh; + int was_io_error; + /* corresponding get_ldev was in __drbd_set_state, to serialize + * our cleanup here with the transition to D_DISKLESS, + * so it is safe to dreference ldev here. */ + eh = mdev->ldev->dc.on_io_error; + was_io_error = test_and_clear_bit(WAS_IO_ERROR, &mdev->flags); + + /* current state still has to be D_FAILED, + * there is only one way out: to D_DISKLESS, + * and that may only happen after our put_ldev below. */ + if (mdev->state.disk != D_FAILED) + dev_err(DEV, + "ASSERT FAILED: disk is %s during detach\n", + drbd_disk_str(mdev->state.disk)); if (drbd_send_state(mdev)) - dev_warn(DEV, "Notified peer that my disk is broken.\n"); + dev_warn(DEV, "Notified peer that I am detaching my disk\n"); else - dev_err(DEV, "Sending state for drbd_io_error() failed\n"); + dev_err(DEV, "Sending state for detaching disk failed\n"); drbd_rs_cancel_all(mdev); - if (get_ldev_if_state(mdev, D_FAILED)) { - eh = mdev->ldev->dc.on_io_error; - put_ldev(mdev); - } - if (eh == EP_CALL_HELPER) + /* In case we want to get something to stable storage still, + * this may be the last chance. + * Following put_ldev may transition to D_DISKLESS. */ + drbd_md_sync(mdev); + put_ldev(mdev); + + if (was_io_error && eh == EP_CALL_HELPER) drbd_khelper(mdev, "local-io-error"); } + /* second half of local IO error, failure to attach, + * or administrative detach, + * after local_cnt references have reached zero again */ + if (os.disk != D_DISKLESS && ns.disk == D_DISKLESS) { + /* We must still be diskless, + * re-attach has to be serialized with this! */ + if (mdev->state.disk != D_DISKLESS) + dev_err(DEV, + "ASSERT FAILED: disk is %s while going diskless\n", + drbd_disk_str(mdev->state.disk)); - /* second half of local IO error handling, - * after local_cnt references have reached zero: */ - if (os.disk == D_FAILED && ns.disk == D_DISKLESS) { - mdev->rs_total = 0; - mdev->rs_failed = 0; - atomic_set(&mdev->rs_pending_cnt, 0); - } - - if (os.disk > D_DISKLESS && ns.disk == D_DISKLESS) { - /* We must still be diskless, - * re-attach has to be serialized with this! */ - if (mdev->state.disk != D_DISKLESS) - dev_err(DEV, - "ASSERT FAILED: disk is %s while going diskless\n", - drbd_disk_str(mdev->state.disk)); + mdev->rs_total = 0; + mdev->rs_failed = 0; + atomic_set(&mdev->rs_pending_cnt, 0); - /* we cannot assert local_cnt == 0 here, as get_ldev_if_state - * will inc/dec it frequently. Since we became D_DISKLESS, no - * one has touched the protected members anymore, though, so we - * are safe to free them here. */ if (drbd_send_state(mdev)) - dev_warn(DEV, "Notified peer that I detached my disk.\n"); + dev_warn(DEV, "Notified peer that I'm now diskless.\n"); else - dev_err(DEV, "Sending state for detach failed\n"); - - lc_destroy(mdev->resync); - mdev->resync = NULL; - lc_destroy(mdev->act_log); - mdev->act_log = NULL; - __no_warn(local, - drbd_free_bc(mdev->ldev); - mdev->ldev = NULL;); - - if (mdev->md_io_tmpp) { - __free_page(mdev->md_io_tmpp); - mdev->md_io_tmpp = NULL; - } + dev_err(DEV, "Sending state for being diskless failed\n"); + /* corresponding get_ldev in __drbd_set_state + * this may finaly trigger drbd_ldev_destroy. */ + put_ldev(mdev); } /* Disks got bigger while they were detached */ @@ -2772,11 +2789,6 @@ void drbd_init_set_defaults(struct drbd_conf *mdev) drbd_set_defaults(mdev); - /* for now, we do NOT yet support it, - * even though we start some framework - * to eventually support barriers */ - set_bit(NO_BARRIER_SUPP, &mdev->flags); - atomic_set(&mdev->ap_bio_cnt, 0); atomic_set(&mdev->ap_pending_cnt, 0); atomic_set(&mdev->rs_pending_cnt, 0); @@ -2842,7 +2854,7 @@ void drbd_init_set_defaults(struct drbd_conf *mdev) drbd_thread_init(mdev, &mdev->asender, drbd_asender); mdev->agreed_pro_version = PRO_VERSION_MAX; - mdev->write_ordering = WO_bio_barrier; + mdev->write_ordering = WO_bdev_flush; mdev->resync_wenr = LC_FREE; } @@ -2899,7 +2911,6 @@ void drbd_mdev_cleanup(struct drbd_conf *mdev) D_ASSERT(list_empty(&mdev->resync_work.list)); D_ASSERT(list_empty(&mdev->unplug_work.list)); D_ASSERT(list_empty(&mdev->go_diskless.list)); - } @@ -3660,6 +3671,8 @@ void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local) get_random_bytes(&val, sizeof(u64)); _drbd_uuid_set(mdev, UI_CURRENT, val); + /* get it to stable storage _now_ */ + drbd_md_sync(mdev); } void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local) @@ -3756,19 +3769,31 @@ static int w_bitmap_io(struct drbd_conf *mdev, struct drbd_work *w, int unused) return 1; } +void drbd_ldev_destroy(struct drbd_conf *mdev) +{ + lc_destroy(mdev->resync); + mdev->resync = NULL; + lc_destroy(mdev->act_log); + mdev->act_log = NULL; + __no_warn(local, + drbd_free_bc(mdev->ldev); + mdev->ldev = NULL;); + + if (mdev->md_io_tmpp) { + __free_page(mdev->md_io_tmpp); + mdev->md_io_tmpp = NULL; + } + clear_bit(GO_DISKLESS, &mdev->flags); +} + static int w_go_diskless(struct drbd_conf *mdev, struct drbd_work *w, int unused) { D_ASSERT(mdev->state.disk == D_FAILED); /* we cannot assert local_cnt == 0 here, as get_ldev_if_state will * inc/dec it frequently. Once we are D_DISKLESS, no one will touch - * the protected members anymore, though, so in the after_state_ch work - * it will be safe to free them. */ + * the protected members anymore, though, so once put_ldev reaches zero + * again, it will be safe to free them. */ drbd_force_state(mdev, NS(disk, D_DISKLESS)); - /* We need to wait for return of references checked out while we still - * have been D_FAILED, though (drbd_md_sync, bitmap io). */ - wait_event(mdev->misc_wait, !atomic_read(&mdev->local_cnt)); - - clear_bit(GO_DISKLESS, &mdev->flags); return 1; } @@ -3777,9 +3802,6 @@ void drbd_go_diskless(struct drbd_conf *mdev) D_ASSERT(mdev->state.disk == D_FAILED); if (!test_and_set_bit(GO_DISKLESS, &mdev->flags)) drbd_queue_work(&mdev->data.work, &mdev->go_diskless); - /* don't drbd_queue_work_front, - * we need to serialize with the after_state_ch work - * of the -> D_FAILED transition. */ } /** diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c index 87925e9..29e5c70 100644 --- a/drivers/block/drbd/drbd_nl.c +++ b/drivers/block/drbd/drbd_nl.c @@ -870,6 +870,11 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp retcode = ERR_DISK_CONFIGURED; goto fail; } + /* It may just now have detached because of IO error. Make sure + * drbd_ldev_destroy is done already, we may end up here very fast, + * e.g. if someone calls attach from the on-io-error handler, + * to realize a "hot spare" feature (not that I'd recommend that) */ + wait_event(mdev->misc_wait, !atomic_read(&mdev->local_cnt)); /* allocation not in the IO path, cqueue thread context */ nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL); @@ -1098,9 +1103,9 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp /* Reset the "barriers don't work" bits here, then force meta data to * be written, to ensure we determine if barriers are supported. */ if (nbc->dc.no_md_flush) - set_bit(MD_NO_BARRIER, &mdev->flags); + set_bit(MD_NO_FUA, &mdev->flags); else - clear_bit(MD_NO_BARRIER, &mdev->flags); + clear_bit(MD_NO_FUA, &mdev->flags); /* Point of no return reached. * Devices and memory are no longer released by error cleanup below. @@ -1112,8 +1117,8 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp nbc = NULL; resync_lru = NULL; - mdev->write_ordering = WO_bio_barrier; - drbd_bump_write_ordering(mdev, WO_bio_barrier); + mdev->write_ordering = WO_bdev_flush; + drbd_bump_write_ordering(mdev, WO_bdev_flush); if (drbd_md_test_flag(mdev->ldev, MDF_CRASHED_PRIMARY)) set_bit(CRASHED_PRIMARY, &mdev->flags); @@ -1262,7 +1267,7 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp force_diskless_dec: put_ldev(mdev); force_diskless: - drbd_force_state(mdev, NS(disk, D_DISKLESS)); + drbd_force_state(mdev, NS(disk, D_FAILED)); drbd_md_sync(mdev); release_bdev2_fail: if (nbc) @@ -1285,10 +1290,19 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp return 0; } +/* Detaching the disk is a process in multiple stages. First we need to lock + * out application IO, in-flight IO, IO stuck in drbd_al_begin_io. + * Then we transition to D_DISKLESS, and wait for put_ldev() to return all + * internal references as well. + * Only then we have finally detached. */ static int drbd_nl_detach(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp, struct drbd_nl_cfg_reply *reply) { + drbd_suspend_io(mdev); /* so no-one is stuck in drbd_al_begin_io */ reply->ret_code = drbd_request_state(mdev, NS(disk, D_DISKLESS)); + if (mdev->state.disk == D_DISKLESS) + wait_event(mdev->misc_wait, !atomic_read(&mdev->local_cnt)); + drbd_resume_io(mdev); return 0; } @@ -1953,7 +1967,6 @@ static int drbd_nl_resume_io(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp if (test_bit(NEW_CUR_UUID, &mdev->flags)) { drbd_uuid_new_current(mdev); clear_bit(NEW_CUR_UUID, &mdev->flags); - drbd_md_sync(mdev); } drbd_suspend_io(mdev); reply->ret_code = drbd_request_state(mdev, NS3(susp, 0, susp_nod, 0, susp_fen, 0)); diff --git a/drivers/block/drbd/drbd_proc.c b/drivers/block/drbd/drbd_proc.c index ad325c5..7e6ac30 100644 --- a/drivers/block/drbd/drbd_proc.c +++ b/drivers/block/drbd/drbd_proc.c @@ -158,7 +158,6 @@ static int drbd_seq_show(struct seq_file *seq, void *v) [WO_none] = 'n', [WO_drain_io] = 'd', [WO_bdev_flush] = 'f', - [WO_bio_barrier] = 'b', }; seq_printf(seq, "version: " REL_VERSION " (api:%d/proto:%d-%d)\n%s\n", diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index efd6169..24487d4 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -36,7 +36,6 @@ #include <linux/memcontrol.h> #include <linux/mm_inline.h> #include <linux/slab.h> -#include <linux/smp_lock.h> #include <linux/pkt_sched.h> #define __KERNEL_SYSCALLS__ #include <linux/unistd.h> @@ -49,11 +48,6 @@ #include "drbd_vli.h" -struct flush_work { - struct drbd_work w; - struct drbd_epoch *epoch; -}; - enum finish_epoch { FE_STILL_LIVE, FE_DESTROYED, @@ -66,16 +60,6 @@ static int drbd_do_auth(struct drbd_conf *mdev); static enum finish_epoch drbd_may_finish_epoch(struct drbd_conf *, struct drbd_epoch *, enum epoch_event); static int e_end_block(struct drbd_conf *, struct drbd_work *, int); -static struct drbd_epoch *previous_epoch(struct drbd_conf *mdev, struct drbd_epoch *epoch) -{ - struct drbd_epoch *prev; - spin_lock(&mdev->epoch_lock); - prev = list_entry(epoch->list.prev, struct drbd_epoch, list); - if (prev == epoch || prev == mdev->current_epoch) - prev = NULL; - spin_unlock(&mdev->epoch_lock); - return prev; -} #define GFP_TRY (__GFP_HIGHMEM | __GFP_NOWARN) @@ -981,7 +965,7 @@ static int drbd_recv_header(struct drbd_conf *mdev, enum drbd_packets *cmd, unsi return TRUE; } -static enum finish_epoch drbd_flush_after_epoch(struct drbd_conf *mdev, struct drbd_epoch *epoch) +static void drbd_flush(struct drbd_conf *mdev) { int rv; @@ -997,24 +981,6 @@ static enum finish_epoch drbd_flush_after_epoch(struct drbd_conf *mdev, struct d } put_ldev(mdev); } - - return drbd_may_finish_epoch(mdev, epoch, EV_BARRIER_DONE); -} - -static int w_flush(struct drbd_conf *mdev, struct drbd_work *w, int cancel) -{ - struct flush_work *fw = (struct flush_work *)w; - struct drbd_epoch *epoch = fw->epoch; - - kfree(w); - - if (!test_and_set_bit(DE_BARRIER_IN_NEXT_EPOCH_ISSUED, &epoch->flags)) - drbd_flush_after_epoch(mdev, epoch); - - drbd_may_finish_epoch(mdev, epoch, EV_PUT | - (mdev->state.conn < C_CONNECTED ? EV_CLEANUP : 0)); - - return 1; } /** @@ -1027,15 +993,13 @@ static enum finish_epoch drbd_may_finish_epoch(struct drbd_conf *mdev, struct drbd_epoch *epoch, enum epoch_event ev) { - int finish, epoch_size; + int epoch_size; struct drbd_epoch *next_epoch; - int schedule_flush = 0; enum finish_epoch rv = FE_STILL_LIVE; spin_lock(&mdev->epoch_lock); do { next_epoch = NULL; - finish = 0; epoch_size = atomic_read(&epoch->epoch_size); @@ -1045,16 +1009,6 @@ static enum finish_epoch drbd_may_finish_epoch(struct drbd_conf *mdev, break; case EV_GOT_BARRIER_NR: set_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags); - - /* Special case: If we just switched from WO_bio_barrier to - WO_bdev_flush we should not finish the current epoch */ - if (test_bit(DE_CONTAINS_A_BARRIER, &epoch->flags) && epoch_size == 1 && - mdev->write_ordering != WO_bio_barrier && - epoch == mdev->current_epoch) - clear_bit(DE_CONTAINS_A_BARRIER, &epoch->flags); - break; - case EV_BARRIER_DONE: - set_bit(DE_BARRIER_IN_NEXT_EPOCH_DONE, &epoch->flags); break; case EV_BECAME_LAST: /* nothing to do*/ @@ -1063,23 +1017,7 @@ static enum finish_epoch drbd_may_finish_epoch(struct drbd_conf *mdev, if (epoch_size != 0 && atomic_read(&epoch->active) == 0 && - test_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags) && - epoch->list.prev == &mdev->current_epoch->list && - !test_bit(DE_IS_FINISHING, &epoch->flags)) { - /* Nearly all conditions are met to finish that epoch... */ - if (test_bit(DE_BARRIER_IN_NEXT_EPOCH_DONE, &epoch->flags) || - mdev->write_ordering == WO_none || - (epoch_size == 1 && test_bit(DE_CONTAINS_A_BARRIER, &epoch->flags)) || - ev & EV_CLEANUP) { - finish = 1; - set_bit(DE_IS_FINISHING, &epoch->flags); - } else if (!test_bit(DE_BARRIER_IN_NEXT_EPOCH_ISSUED, &epoch->flags) && - mdev->write_ordering == WO_bio_barrier) { - atomic_inc(&epoch->active); - schedule_flush = 1; - } - } - if (finish) { + test_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags)) { if (!(ev & EV_CLEANUP)) { spin_unlock(&mdev->epoch_lock); drbd_send_b_ack(mdev, epoch->barrier_nr, epoch_size); @@ -1102,6 +1040,7 @@ static enum finish_epoch drbd_may_finish_epoch(struct drbd_conf *mdev, /* atomic_set(&epoch->active, 0); is already zero */ if (rv == FE_STILL_LIVE) rv = FE_RECYCLED; + wake_up(&mdev->ee_wait); } } @@ -1113,22 +1052,6 @@ static enum finish_epoch drbd_may_finish_epoch(struct drbd_conf *mdev, spin_unlock(&mdev->epoch_lock); - if (schedule_flush) { - struct flush_work *fw; - fw = kmalloc(sizeof(*fw), GFP_ATOMIC); - if (fw) { - fw->w.cb = w_flush; - fw->epoch = epoch; - drbd_queue_work(&mdev->data.work, &fw->w); - } else { - dev_warn(DEV, "Could not kmalloc a flush_work obj\n"); - set_bit(DE_BARRIER_IN_NEXT_EPOCH_ISSUED, &epoch->flags); - /* That is not a recursion, only one level */ - drbd_may_finish_epoch(mdev, epoch, EV_BARRIER_DONE); - drbd_may_finish_epoch(mdev, epoch, EV_PUT); - } - } - return rv; } @@ -1144,19 +1067,16 @@ void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo) [WO_none] = "none", [WO_drain_io] = "drain", [WO_bdev_flush] = "flush", - [WO_bio_barrier] = "barrier", }; pwo = mdev->write_ordering; wo = min(pwo, wo); - if (wo == WO_bio_barrier && mdev->ldev->dc.no_disk_barrier) - wo = WO_bdev_flush; if (wo == WO_bdev_flush && mdev->ldev->dc.no_disk_flush) wo = WO_drain_io; if (wo == WO_drain_io && mdev->ldev->dc.no_disk_drain) wo = WO_none; mdev->write_ordering = wo; - if (pwo != mdev->write_ordering || wo == WO_bio_barrier) + if (pwo != mdev->write_ordering || wo == WO_bdev_flush) dev_info(DEV, "Method to ensure write ordering: %s\n", write_ordering_str[mdev->write_ordering]); } @@ -1192,7 +1112,7 @@ next_bio: bio->bi_sector = sector; bio->bi_bdev = mdev->ldev->backing_bdev; /* we special case some flags in the multi-bio case, see below - * (REQ_UNPLUG, REQ_HARDBARRIER) */ + * (REQ_UNPLUG) */ bio->bi_rw = rw; bio->bi_private = e; bio->bi_end_io = drbd_endio_sec; @@ -1226,11 +1146,6 @@ next_bio: bio->bi_rw &= ~REQ_UNPLUG; drbd_generic_make_request(mdev, fault_type, bio); - - /* strip off REQ_HARDBARRIER, - * unless it is the first or last bio */ - if (bios && bios->bi_next) - bios->bi_rw &= ~REQ_HARDBARRIER; } while (bios); maybe_kick_lo(mdev); return 0; @@ -1244,45 +1159,9 @@ fail: return -ENOMEM; } -/** - * w_e_reissue() - Worker callback; Resubmit a bio, without REQ_HARDBARRIER set - * @mdev: DRBD device. - * @w: work object. - * @cancel: The connection will be closed anyways (unused in this callback) - */ -int w_e_reissue(struct drbd_conf *mdev, struct drbd_work *w, int cancel) __releases(local) -{ - struct drbd_epoch_entry *e = (struct drbd_epoch_entry *)w; - /* We leave DE_CONTAINS_A_BARRIER and EE_IS_BARRIER in place, - (and DE_BARRIER_IN_NEXT_EPOCH_ISSUED in the previous Epoch) - so that we can finish that epoch in drbd_may_finish_epoch(). - That is necessary if we already have a long chain of Epochs, before - we realize that REQ_HARDBARRIER is actually not supported */ - - /* As long as the -ENOTSUPP on the barrier is reported immediately - that will never trigger. If it is reported late, we will just - print that warning and continue correctly for all future requests - with WO_bdev_flush */ - if (previous_epoch(mdev, e->epoch)) - dev_warn(DEV, "Write ordering was not enforced (one time event)\n"); - - /* we still have a local reference, - * get_ldev was done in receive_Data. */ - - e->w.cb = e_end_block; - if (drbd_submit_ee(mdev, e, WRITE, DRBD_FAULT_DT_WR) != 0) { - /* drbd_submit_ee fails for one reason only: - * if was not able to allocate sufficient bios. - * requeue, try again later. */ - e->w.cb = w_e_reissue; - drbd_queue_work(&mdev->data.work, &e->w); - } - return 1; -} - static int receive_Barrier(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) { - int rv, issue_flush; + int rv; struct p_barrier *p = &mdev->data.rbuf.barrier; struct drbd_epoch *epoch; @@ -1300,44 +1179,40 @@ static int receive_Barrier(struct drbd_conf *mdev, enum drbd_packets cmd, unsign * Therefore we must send the barrier_ack after the barrier request was * completed. */ switch (mdev->write_ordering) { - case WO_bio_barrier: case WO_none: if (rv == FE_RECYCLED) return TRUE; - break; + + /* receiver context, in the writeout path of the other node. + * avoid potential distributed deadlock */ + epoch = kmalloc(sizeof(struct drbd_epoch), GFP_NOIO); + if (epoch) + break; + else + dev_warn(DEV, "Allocation of an epoch failed, slowing down\n"); + /* Fall through */ case WO_bdev_flush: case WO_drain_io: - if (rv == FE_STILL_LIVE) { - set_bit(DE_BARRIER_IN_NEXT_EPOCH_ISSUED, &mdev->current_epoch->flags); - drbd_wait_ee_list_empty(mdev, &mdev->active_ee); - rv = drbd_flush_after_epoch(mdev, mdev->current_epoch); - } - if (rv == FE_RECYCLED) - return TRUE; - - /* The asender will send all the ACKs and barrier ACKs out, since - all EEs moved from the active_ee to the done_ee. We need to - provide a new epoch object for the EEs that come in soon */ - break; - } - - /* receiver context, in the writeout path of the other node. - * avoid potential distributed deadlock */ - epoch = kmalloc(sizeof(struct drbd_epoch), GFP_NOIO); - if (!epoch) { - dev_warn(DEV, "Allocation of an epoch failed, slowing down\n"); - issue_flush = !test_and_set_bit(DE_BARRIER_IN_NEXT_EPOCH_ISSUED, &mdev->current_epoch->flags); drbd_wait_ee_list_empty(mdev, &mdev->active_ee); - if (issue_flush) { - rv = drbd_flush_after_epoch(mdev, mdev->current_epoch); - if (rv == FE_RECYCLED) - return TRUE; + drbd_flush(mdev); + + if (atomic_read(&mdev->current_epoch->epoch_size)) { + epoch = kmalloc(sizeof(struct drbd_epoch), GFP_NOIO); + if (epoch) + break; } - drbd_wait_ee_list_empty(mdev, &mdev->done_ee); + epoch = mdev->current_epoch; + wait_event(mdev->ee_wait, atomic_read(&epoch->epoch_size) == 0); + + D_ASSERT(atomic_read(&epoch->active) == 0); + D_ASSERT(epoch->flags == 0); return TRUE; + default: + dev_err(DEV, "Strangeness in mdev->write_ordering %d\n", mdev->write_ordering); + return FALSE; } epoch->flags = 0; @@ -1652,15 +1527,8 @@ static int e_end_block(struct drbd_conf *mdev, struct drbd_work *w, int cancel) { struct drbd_epoch_entry *e = (struct drbd_epoch_entry *)w; sector_t sector = e->sector; - struct drbd_epoch *epoch; int ok = 1, pcmd; - if (e->flags & EE_IS_BARRIER) { - epoch = previous_epoch(mdev, e->epoch); - if (epoch) - drbd_may_finish_epoch(mdev, epoch, EV_BARRIER_DONE + (cancel ? EV_CLEANUP : 0)); - } - if (mdev->net_conf->wire_protocol == DRBD_PROT_C) { if (likely((e->flags & EE_WAS_ERROR) == 0)) { pcmd = (mdev->state.conn >= C_SYNC_SOURCE && @@ -1817,27 +1685,6 @@ static int receive_Data(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned e->epoch = mdev->current_epoch; atomic_inc(&e->epoch->epoch_size); atomic_inc(&e->epoch->active); - - if (mdev->write_ordering == WO_bio_barrier && atomic_read(&e->epoch->epoch_size) == 1) { - struct drbd_epoch *epoch; - /* Issue a barrier if we start a new epoch, and the previous epoch - was not a epoch containing a single request which already was - a Barrier. */ - epoch = list_entry(e->epoch->list.prev, struct drbd_epoch, list); - if (epoch == e->epoch) { - set_bit(DE_CONTAINS_A_BARRIER, &e->epoch->flags); - rw |= REQ_HARDBARRIER; - e->flags |= EE_IS_BARRIER; - } else { - if (atomic_read(&epoch->epoch_size) > 1 || - !test_bit(DE_CONTAINS_A_BARRIER, &epoch->flags)) { - set_bit(DE_BARRIER_IN_NEXT_EPOCH_ISSUED, &epoch->flags); - set_bit(DE_CONTAINS_A_BARRIER, &e->epoch->flags); - rw |= REQ_HARDBARRIER; - e->flags |= EE_IS_BARRIER; - } - } - } spin_unlock(&mdev->epoch_lock); dp_flags = be32_to_cpu(p->dp_flags); @@ -1995,10 +1842,11 @@ static int receive_Data(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned break; } - if (mdev->state.pdsk == D_DISKLESS) { + if (mdev->state.pdsk < D_INCONSISTENT) { /* In case we have the only disk of the cluster, */ drbd_set_out_of_sync(mdev, e->sector, e->size); e->flags |= EE_CALL_AL_COMPLETE_IO; + e->flags &= ~EE_MAY_SET_IN_SYNC; drbd_al_begin_io(mdev, e->sector); } @@ -3362,7 +3210,7 @@ static int receive_state(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned if (ns.conn == C_MASK) { ns.conn = C_CONNECTED; if (mdev->state.disk == D_NEGOTIATING) { - drbd_force_state(mdev, NS(disk, D_DISKLESS)); + drbd_force_state(mdev, NS(disk, D_FAILED)); } else if (peer_state.disk == D_NEGOTIATING) { dev_err(DEV, "Disk attach process on the peer node was aborted.\n"); peer_state.disk = D_DISKLESS; @@ -3779,17 +3627,19 @@ static void drbdd(struct drbd_conf *mdev) } shs = drbd_cmd_handler[cmd].pkt_size - sizeof(union p_header); - rv = drbd_recv(mdev, &header->h80.payload, shs); - if (unlikely(rv != shs)) { - dev_err(DEV, "short read while reading sub header: rv=%d\n", rv); - goto err_out; - } - if (packet_size - shs > 0 && !drbd_cmd_handler[cmd].expect_payload) { dev_err(DEV, "No payload expected %s l:%d\n", cmdname(cmd), packet_size); goto err_out; } + if (shs) { + rv = drbd_recv(mdev, &header->h80.payload, shs); + if (unlikely(rv != shs)) { + dev_err(DEV, "short read while reading sub header: rv=%d\n", rv); + goto err_out; + } + } + rv = drbd_cmd_handler[cmd].function(mdev, cmd, packet_size - shs); if (unlikely(!rv)) { diff --git a/drivers/block/drbd/drbd_req.c b/drivers/block/drbd/drbd_req.c index 9e91a25..11a75d3 100644 --- a/drivers/block/drbd/drbd_req.c +++ b/drivers/block/drbd/drbd_req.c @@ -258,7 +258,7 @@ void _req_may_be_done(struct drbd_request *req, struct bio_and_error *m) if (!hlist_unhashed(&req->colision)) hlist_del(&req->colision); else - D_ASSERT((s & RQ_NET_MASK) == 0); + D_ASSERT((s & (RQ_NET_MASK & ~RQ_NET_DONE)) == 0); /* for writes we need to do some extra housekeeping */ if (rw == WRITE) @@ -813,7 +813,8 @@ static int drbd_make_request_common(struct drbd_conf *mdev, struct bio *bio) mdev->state.conn >= C_CONNECTED)); if (!(local || remote) && !is_susp(mdev->state)) { - dev_err(DEV, "IO ERROR: neither local nor remote disk\n"); + if (__ratelimit(&drbd_ratelimit_state)) + dev_err(DEV, "IO ERROR: neither local nor remote disk\n"); goto fail_free_complete; } @@ -942,12 +943,21 @@ allocate_barrier: if (local) { req->private_bio->bi_bdev = mdev->ldev->backing_bdev; - if (FAULT_ACTIVE(mdev, rw == WRITE ? DRBD_FAULT_DT_WR - : rw == READ ? DRBD_FAULT_DT_RD - : DRBD_FAULT_DT_RA)) + /* State may have changed since we grabbed our reference on the + * mdev->ldev member. Double check, and short-circuit to endio. + * In case the last activity log transaction failed to get on + * stable storage, and this is a WRITE, we may not even submit + * this bio. */ + if (get_ldev(mdev)) { + if (FAULT_ACTIVE(mdev, rw == WRITE ? DRBD_FAULT_DT_WR + : rw == READ ? DRBD_FAULT_DT_RD + : DRBD_FAULT_DT_RA)) + bio_endio(req->private_bio, -EIO); + else + generic_make_request(req->private_bio); + put_ldev(mdev); + } else bio_endio(req->private_bio, -EIO); - else - generic_make_request(req->private_bio); } /* we need to plug ALWAYS since we possibly need to kick lo_dev. @@ -1022,20 +1032,6 @@ int drbd_make_request_26(struct request_queue *q, struct bio *bio) return 0; } - /* Reject barrier requests if we know the underlying device does - * not support them. - * XXX: Need to get this info from peer as well some how so we - * XXX: reject if EITHER side/data/metadata area does not support them. - * - * because of those XXX, this is not yet enabled, - * i.e. in drbd_init_set_defaults we set the NO_BARRIER_SUPP bit. - */ - if (unlikely(bio->bi_rw & REQ_HARDBARRIER) && test_bit(NO_BARRIER_SUPP, &mdev->flags)) { - /* dev_warn(DEV, "Rejecting barrier request as underlying device does not support\n"); */ - bio_endio(bio, -EOPNOTSUPP); - return 0; - } - /* * what we "blindly" assume: */ diff --git a/drivers/block/drbd/drbd_req.h b/drivers/block/drbd/drbd_req.h index 181ea03..ab2bd09 100644 --- a/drivers/block/drbd/drbd_req.h +++ b/drivers/block/drbd/drbd_req.h @@ -339,7 +339,8 @@ static inline int _req_mod(struct drbd_request *req, enum drbd_req_event what) } /* completion of master bio is outside of spinlock. - * If you need it irqsave, do it your self! */ + * If you need it irqsave, do it your self! + * Which means: don't use from bio endio callback. */ static inline int req_mod(struct drbd_request *req, enum drbd_req_event what) { diff --git a/drivers/block/drbd/drbd_worker.c b/drivers/block/drbd/drbd_worker.c index 108d580..34f224b 100644 --- a/drivers/block/drbd/drbd_worker.c +++ b/drivers/block/drbd/drbd_worker.c @@ -26,7 +26,6 @@ #include <linux/module.h> #include <linux/drbd.h> #include <linux/sched.h> -#include <linux/smp_lock.h> #include <linux/wait.h> #include <linux/mm.h> #include <linux/memcontrol.h> @@ -102,12 +101,6 @@ void drbd_endio_read_sec_final(struct drbd_epoch_entry *e) __releases(local) put_ldev(mdev); } -static int is_failed_barrier(int ee_flags) -{ - return (ee_flags & (EE_IS_BARRIER|EE_WAS_ERROR|EE_RESUBMITTED)) - == (EE_IS_BARRIER|EE_WAS_ERROR); -} - /* writes on behalf of the partner, or resync writes, * "submitted" by the receiver, final stage. */ static void drbd_endio_write_sec_final(struct drbd_epoch_entry *e) __releases(local) @@ -119,21 +112,6 @@ static void drbd_endio_write_sec_final(struct drbd_epoch_entry *e) __releases(lo int is_syncer_req; int do_al_complete_io; - /* if this is a failed barrier request, disable use of barriers, - * and schedule for resubmission */ - if (is_failed_barrier(e->flags)) { - drbd_bump_write_ordering(mdev, WO_bdev_flush); - spin_lock_irqsave(&mdev->req_lock, flags); - list_del(&e->w.list); - e->flags = (e->flags & ~EE_WAS_ERROR) | EE_RESUBMITTED; - e->w.cb = w_e_reissue; - /* put_ldev actually happens below, once we come here again. */ - __release(local); - spin_unlock_irqrestore(&mdev->req_lock, flags); - drbd_queue_work(&mdev->data.work, &e->w); - return; - } - D_ASSERT(e->block_id != ID_VACANT); /* after we moved e to done_ee, @@ -215,8 +193,10 @@ void drbd_endio_sec(struct bio *bio, int error) */ void drbd_endio_pri(struct bio *bio, int error) { + unsigned long flags; struct drbd_request *req = bio->bi_private; struct drbd_conf *mdev = req->mdev; + struct bio_and_error m; enum drbd_req_event what; int uptodate = bio_flagged(bio, BIO_UPTODATE); @@ -242,7 +222,13 @@ void drbd_endio_pri(struct bio *bio, int error) bio_put(req->private_bio); req->private_bio = ERR_PTR(error); - req_mod(req, what); + /* not req_mod(), we need irqsave here! */ + spin_lock_irqsave(&mdev->req_lock, flags); + __req_mod(req, what, &m); + spin_unlock_irqrestore(&mdev->req_lock, flags); + + if (m.bio) + complete_master_bio(mdev, &m); } int w_read_retry_remote(struct drbd_conf *mdev, struct drbd_work *w, int cancel) @@ -925,7 +911,7 @@ out: drbd_md_sync(mdev); if (test_and_clear_bit(WRITE_BM_AFTER_RESYNC, &mdev->flags)) { - dev_warn(DEV, "Writing the whole bitmap, due to failed kmalloc\n"); + dev_info(DEV, "Writing the whole bitmap\n"); drbd_queue_bitmap_io(mdev, &drbd_bm_write, NULL, "write from resync_finished"); } diff --git a/drivers/block/floppy.c b/drivers/block/floppy.c index 767107c..3951020 100644 --- a/drivers/block/floppy.c +++ b/drivers/block/floppy.c @@ -4363,9 +4363,9 @@ out_unreg_blkdev: out_put_disk: while (dr--) { del_timer(&motor_off_timer[dr]); - put_disk(disks[dr]); if (disks[dr]->queue) blk_cleanup_queue(disks[dr]->queue); + put_disk(disks[dr]); } return err; } @@ -4573,8 +4573,8 @@ static void __exit floppy_module_exit(void) device_remove_file(&floppy_device[drive].dev, &dev_attr_cmos); platform_device_unregister(&floppy_device[drive]); } - put_disk(disks[drive]); blk_cleanup_queue(disks[drive]->queue); + put_disk(disks[drive]); } del_timer_sync(&fd_timeout); diff --git a/drivers/block/loop.c b/drivers/block/loop.c index 1e5284e..7ea0bea 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -481,12 +481,6 @@ static int do_bio_filebacked(struct loop_device *lo, struct bio *bio) if (bio_rw(bio) == WRITE) { struct file *file = lo->lo_backing_file; - /* REQ_HARDBARRIER is deprecated */ - if (bio->bi_rw & REQ_HARDBARRIER) { - ret = -EOPNOTSUPP; - goto out; - } - if (bio->bi_rw & REQ_FLUSH) { ret = vfs_fsync(file, 0); if (unlikely(ret && ret != -EINVAL)) { diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c index 6ec9d53..008d4a0 100644 --- a/drivers/block/rbd.c +++ b/drivers/block/rbd.c @@ -21,80 +21,9 @@ - Instructions for use - -------------------- + For usage instructions, please refer to: - 1) Map a Linux block device to an existing rbd image. - - Usage: <mon ip addr> <options> <pool name> <rbd image name> [snap name] - - $ echo "192.168.0.1 name=admin rbd foo" > /sys/class/rbd/add - - The snapshot name can be "-" or omitted to map the image read/write. - - 2) List all active blkdev<->object mappings. - - In this example, we have performed step #1 twice, creating two blkdevs, - mapped to two separate rados objects in the rados rbd pool - - $ cat /sys/class/rbd/list - #id major client_name pool name snap KB - 0 254 client4143 rbd foo - 1024000 - - The columns, in order, are: - - blkdev unique id - - blkdev assigned major - - rados client id - - rados pool name - - rados block device name - - mapped snapshot ("-" if none) - - device size in KB - - - 3) Create a snapshot. - - Usage: <blkdev id> <snapname> - - $ echo "0 mysnap" > /sys/class/rbd/snap_create - - - 4) Listing a snapshot. - - $ cat /sys/class/rbd/snaps_list - #id snap KB - 0 - 1024000 (*) - 0 foo 1024000 - - The columns, in order, are: - - blkdev unique id - - snapshot name, '-' means none (active read/write version) - - size of device at time of snapshot - - the (*) indicates this is the active version - - 5) Rollback to snapshot. - - Usage: <blkdev id> <snapname> - - $ echo "0 mysnap" > /sys/class/rbd/snap_rollback - - - 6) Mapping an image using snapshot. - - A snapshot mapping is read-only. This is being done by passing - snap=<snapname> to the options when adding a device. - - $ echo "192.168.0.1 name=admin,snap=mysnap rbd foo" > /sys/class/rbd/add - - - 7) Remove an active blkdev<->rbd image mapping. - - In this example, we remove the mapping with blkdev unique id 1. - - $ echo 1 > /sys/class/rbd/remove - - - NOTE: The actual creation and deletion of rados objects is outside the scope - of this driver. + Documentation/ABI/testing/sysfs-bus-rbd */ @@ -163,6 +92,14 @@ struct rbd_request { u64 len; }; +struct rbd_snap { + struct device dev; + const char *name; + size_t size; + struct list_head node; + u64 id; +}; + /* * a single device */ @@ -193,21 +130,60 @@ struct rbd_device { int read_only; struct list_head node; + + /* list of snapshots */ + struct list_head snaps; + + /* sysfs related */ + struct device dev; +}; + +static struct bus_type rbd_bus_type = { + .name = "rbd", }; static spinlock_t node_lock; /* protects client get/put */ -static struct class *class_rbd; /* /sys/class/rbd */ static DEFINE_MUTEX(ctl_mutex); /* Serialize open/close/setup/teardown */ static LIST_HEAD(rbd_dev_list); /* devices */ static LIST_HEAD(rbd_client_list); /* clients */ +static int __rbd_init_snaps_header(struct rbd_device *rbd_dev); +static void rbd_dev_release(struct device *dev); +static ssize_t rbd_snap_rollback(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t size); +static ssize_t rbd_snap_add(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t count); +static void __rbd_remove_snap_dev(struct rbd_device *rbd_dev, + struct rbd_snap *snap);; + + +static struct rbd_device *dev_to_rbd(struct device *dev) +{ + return container_of(dev, struct rbd_device, dev); +} + +static struct device *rbd_get_dev(struct rbd_device *rbd_dev) +{ + return get_device(&rbd_dev->dev); +} + +static void rbd_put_dev(struct rbd_device *rbd_dev) +{ + put_device(&rbd_dev->dev); +} static int rbd_open(struct block_device *bdev, fmode_t mode) { struct gendisk *disk = bdev->bd_disk; struct rbd_device *rbd_dev = disk->private_data; + rbd_get_dev(rbd_dev); + set_device_ro(bdev, rbd_dev->read_only); if ((mode & FMODE_WRITE) && rbd_dev->read_only) @@ -216,9 +192,19 @@ static int rbd_open(struct block_device *bdev, fmode_t mode) return 0; } +static int rbd_release(struct gendisk *disk, fmode_t mode) +{ + struct rbd_device *rbd_dev = disk->private_data; + + rbd_put_dev(rbd_dev); + + return 0; +} + static const struct block_device_operations rbd_bd_ops = { .owner = THIS_MODULE, .open = rbd_open, + .release = rbd_release, }; /* @@ -361,7 +347,6 @@ static int rbd_header_from_disk(struct rbd_image_header *header, int ret = -ENOMEM; init_rwsem(&header->snap_rwsem); - header->snap_names_len = le64_to_cpu(ondisk->snap_names_len); header->snapc = kmalloc(sizeof(struct ceph_snap_context) + snap_count * @@ -1256,10 +1241,20 @@ bad: return -ERANGE; } +static void __rbd_remove_all_snaps(struct rbd_device *rbd_dev) +{ + struct rbd_snap *snap; + + while (!list_empty(&rbd_dev->snaps)) { + snap = list_first_entry(&rbd_dev->snaps, struct rbd_snap, node); + __rbd_remove_snap_dev(rbd_dev, snap); + } +} + /* * only read the first part of the ondisk header, without the snaps info */ -static int rbd_update_snaps(struct rbd_device *rbd_dev) +static int __rbd_update_snaps(struct rbd_device *rbd_dev) { int ret; struct rbd_image_header h; @@ -1280,12 +1275,15 @@ static int rbd_update_snaps(struct rbd_device *rbd_dev) rbd_dev->header.total_snaps = h.total_snaps; rbd_dev->header.snapc = h.snapc; rbd_dev->header.snap_names = h.snap_names; + rbd_dev->header.snap_names_len = h.snap_names_len; rbd_dev->header.snap_sizes = h.snap_sizes; rbd_dev->header.snapc->seq = snap_seq; + ret = __rbd_init_snaps_header(rbd_dev); + up_write(&rbd_dev->header.snap_rwsem); - return 0; + return ret; } static int rbd_init_disk(struct rbd_device *rbd_dev) @@ -1300,6 +1298,11 @@ static int rbd_init_disk(struct rbd_device *rbd_dev) if (rc) return rc; + /* no need to lock here, as rbd_dev is not registered yet */ + rc = __rbd_init_snaps_header(rbd_dev); + if (rc) + return rc; + rc = rbd_header_set_snap(rbd_dev, rbd_dev->snap_name, &total_size); if (rc) return rc; @@ -1343,54 +1346,360 @@ out: return rc; } -/******************************************************************** - * /sys/class/rbd/ - * add map rados objects to blkdev - * remove unmap rados objects - * list show mappings - *******************************************************************/ +/* + sysfs +*/ + +static ssize_t rbd_size_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct rbd_device *rbd_dev = dev_to_rbd(dev); + + return sprintf(buf, "%llu\n", (unsigned long long)rbd_dev->header.image_size); +} + +static ssize_t rbd_major_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct rbd_device *rbd_dev = dev_to_rbd(dev); -static void class_rbd_release(struct class *cls) + return sprintf(buf, "%d\n", rbd_dev->major); +} + +static ssize_t rbd_client_id_show(struct device *dev, + struct device_attribute *attr, char *buf) { - kfree(cls); + struct rbd_device *rbd_dev = dev_to_rbd(dev); + + return sprintf(buf, "client%lld\n", ceph_client_id(rbd_dev->client)); } -static ssize_t class_rbd_list(struct class *c, - struct class_attribute *attr, - char *data) +static ssize_t rbd_pool_show(struct device *dev, + struct device_attribute *attr, char *buf) { - int n = 0; - struct list_head *tmp; - int max = PAGE_SIZE; + struct rbd_device *rbd_dev = dev_to_rbd(dev); + + return sprintf(buf, "%s\n", rbd_dev->pool_name); +} + +static ssize_t rbd_name_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct rbd_device *rbd_dev = dev_to_rbd(dev); + + return sprintf(buf, "%s\n", rbd_dev->obj); +} + +static ssize_t rbd_snap_show(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct rbd_device *rbd_dev = dev_to_rbd(dev); + + return sprintf(buf, "%s\n", rbd_dev->snap_name); +} + +static ssize_t rbd_image_refresh(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t size) +{ + struct rbd_device *rbd_dev = dev_to_rbd(dev); + int rc; + int ret = size; mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); - n += snprintf(data, max, - "#id\tmajor\tclient_name\tpool\tname\tsnap\tKB\n"); + rc = __rbd_update_snaps(rbd_dev); + if (rc < 0) + ret = rc; - list_for_each(tmp, &rbd_dev_list) { - struct rbd_device *rbd_dev; + mutex_unlock(&ctl_mutex); + return ret; +} - rbd_dev = list_entry(tmp, struct rbd_device, node); - n += snprintf(data+n, max-n, - "%d\t%d\tclient%lld\t%s\t%s\t%s\t%lld\n", - rbd_dev->id, - rbd_dev->major, - ceph_client_id(rbd_dev->client), - rbd_dev->pool_name, - rbd_dev->obj, rbd_dev->snap_name, - rbd_dev->header.image_size >> 10); - if (n == max) +static DEVICE_ATTR(size, S_IRUGO, rbd_size_show, NULL); +static DEVICE_ATTR(major, S_IRUGO, rbd_major_show, NULL); +static DEVICE_ATTR(client_id, S_IRUGO, rbd_client_id_show, NULL); +static DEVICE_ATTR(pool, S_IRUGO, rbd_pool_show, NULL); +static DEVICE_ATTR(name, S_IRUGO, rbd_name_show, NULL); +static DEVICE_ATTR(refresh, S_IWUSR, NULL, rbd_image_refresh); +static DEVICE_ATTR(current_snap, S_IRUGO, rbd_snap_show, NULL); +static DEVICE_ATTR(create_snap, S_IWUSR, NULL, rbd_snap_add); +static DEVICE_ATTR(rollback_snap, S_IWUSR, NULL, rbd_snap_rollback); + +static struct attribute *rbd_attrs[] = { + &dev_attr_size.attr, + &dev_attr_major.attr, + &dev_attr_client_id.attr, + &dev_attr_pool.attr, + &dev_attr_name.attr, + &dev_attr_current_snap.attr, + &dev_attr_refresh.attr, + &dev_attr_create_snap.attr, + &dev_attr_rollback_snap.attr, + NULL +}; + +static struct attribute_group rbd_attr_group = { + .attrs = rbd_attrs, +}; + +static const struct attribute_group *rbd_attr_groups[] = { + &rbd_attr_group, + NULL +}; + +static void rbd_sysfs_dev_release(struct device *dev) +{ +} + +static struct device_type rbd_device_type = { + .name = "rbd", + .groups = rbd_attr_groups, + .release = rbd_sysfs_dev_release, +}; + + +/* + sysfs - snapshots +*/ + +static ssize_t rbd_snap_size_show(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); + + return sprintf(buf, "%lld\n", (long long)snap->size); +} + +static ssize_t rbd_snap_id_show(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); + + return sprintf(buf, "%lld\n", (long long)snap->id); +} + +static DEVICE_ATTR(snap_size, S_IRUGO, rbd_snap_size_show, NULL); +static DEVICE_ATTR(snap_id, S_IRUGO, rbd_snap_id_show, NULL); + +static struct attribute *rbd_snap_attrs[] = { + &dev_attr_snap_size.attr, + &dev_attr_snap_id.attr, + NULL, +}; + +static struct attribute_group rbd_snap_attr_group = { + .attrs = rbd_snap_attrs, +}; + +static void rbd_snap_dev_release(struct device *dev) +{ + struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); + kfree(snap->name); + kfree(snap); +} + +static const struct attribute_group *rbd_snap_attr_groups[] = { + &rbd_snap_attr_group, + NULL +}; + +static struct device_type rbd_snap_device_type = { + .groups = rbd_snap_attr_groups, + .release = rbd_snap_dev_release, +}; + +static void __rbd_remove_snap_dev(struct rbd_device *rbd_dev, + struct rbd_snap *snap) +{ + list_del(&snap->node); + device_unregister(&snap->dev); +} + +static int rbd_register_snap_dev(struct rbd_device *rbd_dev, + struct rbd_snap *snap, + struct device *parent) +{ + struct device *dev = &snap->dev; + int ret; + + dev->type = &rbd_snap_device_type; + dev->parent = parent; + dev->release = rbd_snap_dev_release; + dev_set_name(dev, "snap_%s", snap->name); + ret = device_register(dev); + + return ret; +} + +static int __rbd_add_snap_dev(struct rbd_device *rbd_dev, + int i, const char *name, + struct rbd_snap **snapp) +{ + int ret; + struct rbd_snap *snap = kzalloc(sizeof(*snap), GFP_KERNEL); + if (!snap) + return -ENOMEM; + snap->name = kstrdup(name, GFP_KERNEL); + snap->size = rbd_dev->header.snap_sizes[i]; + snap->id = rbd_dev->header.snapc->snaps[i]; + if (device_is_registered(&rbd_dev->dev)) { + ret = rbd_register_snap_dev(rbd_dev, snap, + &rbd_dev->dev); + if (ret < 0) + goto err; + } + *snapp = snap; + return 0; +err: + kfree(snap->name); + kfree(snap); + return ret; +} + +/* + * search for the previous snap in a null delimited string list + */ +const char *rbd_prev_snap_name(const char *name, const char *start) +{ + if (name < start + 2) + return NULL; + + name -= 2; + while (*name) { + if (name == start) + return start; + name--; + } + return name + 1; +} + +/* + * compare the old list of snapshots that we have to what's in the header + * and update it accordingly. Note that the header holds the snapshots + * in a reverse order (from newest to oldest) and we need to go from + * older to new so that we don't get a duplicate snap name when + * doing the process (e.g., removed snapshot and recreated a new + * one with the same name. + */ +static int __rbd_init_snaps_header(struct rbd_device *rbd_dev) +{ + const char *name, *first_name; + int i = rbd_dev->header.total_snaps; + struct rbd_snap *snap, *old_snap = NULL; + int ret; + struct list_head *p, *n; + + first_name = rbd_dev->header.snap_names; + name = first_name + rbd_dev->header.snap_names_len; + + list_for_each_prev_safe(p, n, &rbd_dev->snaps) { + u64 cur_id; + + old_snap = list_entry(p, struct rbd_snap, node); + + if (i) + cur_id = rbd_dev->header.snapc->snaps[i - 1]; + + if (!i || old_snap->id < cur_id) { + /* old_snap->id was skipped, thus was removed */ + __rbd_remove_snap_dev(rbd_dev, old_snap); + continue; + } + if (old_snap->id == cur_id) { + /* we have this snapshot already */ + i--; + name = rbd_prev_snap_name(name, first_name); + continue; + } + for (; i > 0; + i--, name = rbd_prev_snap_name(name, first_name)) { + if (!name) { + WARN_ON(1); + return -EINVAL; + } + cur_id = rbd_dev->header.snapc->snaps[i]; + /* snapshot removal? handle it above */ + if (cur_id >= old_snap->id) + break; + /* a new snapshot */ + ret = __rbd_add_snap_dev(rbd_dev, i - 1, name, &snap); + if (ret < 0) + return ret; + + /* note that we add it backward so using n and not p */ + list_add(&snap->node, n); + p = &snap->node; + } + } + /* we're done going over the old snap list, just add what's left */ + for (; i > 0; i--) { + name = rbd_prev_snap_name(name, first_name); + if (!name) { + WARN_ON(1); + return -EINVAL; + } + ret = __rbd_add_snap_dev(rbd_dev, i - 1, name, &snap); + if (ret < 0) + return ret; + list_add(&snap->node, &rbd_dev->snaps); + } + + return 0; +} + + +static void rbd_root_dev_release(struct device *dev) +{ +} + +static struct device rbd_root_dev = { + .init_name = "rbd", + .release = rbd_root_dev_release, +}; + +static int rbd_bus_add_dev(struct rbd_device *rbd_dev) +{ + int ret = -ENOMEM; + struct device *dev; + struct rbd_snap *snap; + + mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); + dev = &rbd_dev->dev; + + dev->bus = &rbd_bus_type; + dev->type = &rbd_device_type; + dev->parent = &rbd_root_dev; + dev->release = rbd_dev_release; + dev_set_name(dev, "%d", rbd_dev->id); + ret = device_register(dev); + if (ret < 0) + goto done_free; + + list_for_each_entry(snap, &rbd_dev->snaps, node) { + ret = rbd_register_snap_dev(rbd_dev, snap, + &rbd_dev->dev); + if (ret < 0) break; } mutex_unlock(&ctl_mutex); - return n; + return 0; +done_free: + mutex_unlock(&ctl_mutex); + return ret; } -static ssize_t class_rbd_add(struct class *c, - struct class_attribute *attr, - const char *buf, size_t count) +static void rbd_bus_del_dev(struct rbd_device *rbd_dev) +{ + device_unregister(&rbd_dev->dev); +} + +static ssize_t rbd_add(struct bus_type *bus, const char *buf, size_t count) { struct ceph_osd_client *osdc; struct rbd_device *rbd_dev; @@ -1419,6 +1728,7 @@ static ssize_t class_rbd_add(struct class *c, /* static rbd_device initialization */ spin_lock_init(&rbd_dev->lock); INIT_LIST_HEAD(&rbd_dev->node); + INIT_LIST_HEAD(&rbd_dev->snaps); /* generate unique id: find highest unique id, add one */ mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); @@ -1478,6 +1788,9 @@ static ssize_t class_rbd_add(struct class *c, } rbd_dev->major = irc; + rc = rbd_bus_add_dev(rbd_dev); + if (rc) + goto err_out_disk; /* set up and announce blkdev mapping */ rc = rbd_init_disk(rbd_dev); if (rc) @@ -1487,6 +1800,8 @@ static ssize_t class_rbd_add(struct class *c, err_out_blkdev: unregister_blkdev(rbd_dev->major, rbd_dev->name); +err_out_disk: + rbd_free_disk(rbd_dev); err_out_client: rbd_put_client(rbd_dev); mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); @@ -1518,35 +1833,10 @@ static struct rbd_device *__rbd_get_dev(unsigned long id) return NULL; } -static ssize_t class_rbd_remove(struct class *c, - struct class_attribute *attr, - const char *buf, - size_t count) +static void rbd_dev_release(struct device *dev) { - struct rbd_device *rbd_dev = NULL; - int target_id, rc; - unsigned long ul; - - rc = strict_strtoul(buf, 10, &ul); - if (rc) - return rc; - - /* convert to int; abort if we lost anything in the conversion */ - target_id = (int) ul; - if (target_id != ul) - return -EINVAL; - - /* remove object from list immediately */ - mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); - - rbd_dev = __rbd_get_dev(target_id); - if (rbd_dev) - list_del_init(&rbd_dev->node); - - mutex_unlock(&ctl_mutex); - - if (!rbd_dev) - return -ENOENT; + struct rbd_device *rbd_dev = + container_of(dev, struct rbd_device, dev); rbd_put_client(rbd_dev); @@ -1557,67 +1847,11 @@ static ssize_t class_rbd_remove(struct class *c, /* release module ref */ module_put(THIS_MODULE); - - return count; } -static ssize_t class_rbd_snaps_list(struct class *c, - struct class_attribute *attr, - char *data) -{ - struct rbd_device *rbd_dev = NULL; - struct list_head *tmp; - struct rbd_image_header *header; - int i, n = 0, max = PAGE_SIZE; - int ret; - - mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); - - n += snprintf(data, max, "#id\tsnap\tKB\n"); - - list_for_each(tmp, &rbd_dev_list) { - char *names, *p; - struct ceph_snap_context *snapc; - - rbd_dev = list_entry(tmp, struct rbd_device, node); - header = &rbd_dev->header; - - down_read(&header->snap_rwsem); - - names = header->snap_names; - snapc = header->snapc; - - n += snprintf(data + n, max - n, "%d\t%s\t%lld%s\n", - rbd_dev->id, RBD_SNAP_HEAD_NAME, - header->image_size >> 10, - (!rbd_dev->cur_snap ? " (*)" : "")); - if (n == max) - break; - - p = names; - for (i = 0; i < header->total_snaps; i++, p += strlen(p) + 1) { - n += snprintf(data + n, max - n, "%d\t%s\t%lld%s\n", - rbd_dev->id, p, header->snap_sizes[i] >> 10, - (rbd_dev->cur_snap && - (snap_index(header, i) == rbd_dev->cur_snap) ? - " (*)" : "")); - if (n == max) - break; - } - - up_read(&header->snap_rwsem); - } - - - ret = n; - mutex_unlock(&ctl_mutex); - return ret; -} - -static ssize_t class_rbd_snaps_refresh(struct class *c, - struct class_attribute *attr, - const char *buf, - size_t count) +static ssize_t rbd_remove(struct bus_type *bus, + const char *buf, + size_t count) { struct rbd_device *rbd_dev = NULL; int target_id, rc; @@ -1641,95 +1875,70 @@ static ssize_t class_rbd_snaps_refresh(struct class *c, goto done; } - rc = rbd_update_snaps(rbd_dev); - if (rc < 0) - ret = rc; + list_del_init(&rbd_dev->node); + + __rbd_remove_all_snaps(rbd_dev); + rbd_bus_del_dev(rbd_dev); done: mutex_unlock(&ctl_mutex); return ret; } -static ssize_t class_rbd_snap_create(struct class *c, - struct class_attribute *attr, - const char *buf, - size_t count) +static ssize_t rbd_snap_add(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t count) { - struct rbd_device *rbd_dev = NULL; - int target_id, ret; - char *name; - - name = kmalloc(RBD_MAX_SNAP_NAME_LEN + 1, GFP_KERNEL); + struct rbd_device *rbd_dev = dev_to_rbd(dev); + int ret; + char *name = kmalloc(count + 1, GFP_KERNEL); if (!name) return -ENOMEM; - /* parse snaps add command */ - if (sscanf(buf, "%d " - "%" __stringify(RBD_MAX_SNAP_NAME_LEN) "s", - &target_id, - name) != 2) { - ret = -EINVAL; - goto done; - } + snprintf(name, count, "%s", buf); mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); - rbd_dev = __rbd_get_dev(target_id); - if (!rbd_dev) { - ret = -ENOENT; - goto done_unlock; - } - ret = rbd_header_add_snap(rbd_dev, name, GFP_KERNEL); if (ret < 0) goto done_unlock; - ret = rbd_update_snaps(rbd_dev); + ret = __rbd_update_snaps(rbd_dev); if (ret < 0) goto done_unlock; ret = count; done_unlock: mutex_unlock(&ctl_mutex); -done: kfree(name); return ret; } -static ssize_t class_rbd_rollback(struct class *c, - struct class_attribute *attr, - const char *buf, - size_t count) +static ssize_t rbd_snap_rollback(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t count) { - struct rbd_device *rbd_dev = NULL; - int target_id, ret; + struct rbd_device *rbd_dev = dev_to_rbd(dev); + int ret; u64 snapid; - char snap_name[RBD_MAX_SNAP_NAME_LEN]; u64 cur_ofs; - char *seg_name; + char *seg_name = NULL; + char *snap_name = kmalloc(count + 1, GFP_KERNEL); + ret = -ENOMEM; + if (!snap_name) + return ret; /* parse snaps add command */ - if (sscanf(buf, "%d " - "%" __stringify(RBD_MAX_SNAP_NAME_LEN) "s", - &target_id, - snap_name) != 2) { - return -EINVAL; - } - - ret = -ENOMEM; + snprintf(snap_name, count, "%s", buf); seg_name = kmalloc(RBD_MAX_SEG_NAME_LEN + 1, GFP_NOIO); if (!seg_name) - return ret; + goto done; mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); - rbd_dev = __rbd_get_dev(target_id); - if (!rbd_dev) { - ret = -ENOENT; - goto done_unlock; - } - ret = snap_by_name(&rbd_dev->header, snap_name, &snapid, NULL); if (ret < 0) goto done_unlock; @@ -1750,7 +1959,7 @@ static ssize_t class_rbd_rollback(struct class *c, seg_name, ret); } - ret = rbd_update_snaps(rbd_dev); + ret = __rbd_update_snaps(rbd_dev); if (ret < 0) goto done_unlock; @@ -1758,57 +1967,42 @@ static ssize_t class_rbd_rollback(struct class *c, done_unlock: mutex_unlock(&ctl_mutex); +done: kfree(seg_name); + kfree(snap_name); return ret; } -static struct class_attribute class_rbd_attrs[] = { - __ATTR(add, 0200, NULL, class_rbd_add), - __ATTR(remove, 0200, NULL, class_rbd_remove), - __ATTR(list, 0444, class_rbd_list, NULL), - __ATTR(snaps_refresh, 0200, NULL, class_rbd_snaps_refresh), - __ATTR(snap_create, 0200, NULL, class_rbd_snap_create), - __ATTR(snaps_list, 0444, class_rbd_snaps_list, NULL), - __ATTR(snap_rollback, 0200, NULL, class_rbd_rollback), +static struct bus_attribute rbd_bus_attrs[] = { + __ATTR(add, S_IWUSR, NULL, rbd_add), + __ATTR(remove, S_IWUSR, NULL, rbd_remove), __ATTR_NULL }; /* * create control files in sysfs - * /sys/class/rbd/... + * /sys/bus/rbd/... */ static int rbd_sysfs_init(void) { - int ret = -ENOMEM; + int ret; - class_rbd = kzalloc(sizeof(*class_rbd), GFP_KERNEL); - if (!class_rbd) - goto out; + rbd_bus_type.bus_attrs = rbd_bus_attrs; - class_rbd->name = DRV_NAME; - class_rbd->owner = THIS_MODULE; - class_rbd->class_release = class_rbd_release; - class_rbd->class_attrs = class_rbd_attrs; + ret = bus_register(&rbd_bus_type); + if (ret < 0) + return ret; - ret = class_register(class_rbd); - if (ret) - goto out_class; - return 0; + ret = device_register(&rbd_root_dev); -out_class: - kfree(class_rbd); - class_rbd = NULL; - pr_err(DRV_NAME ": failed to create class rbd\n"); -out: return ret; } static void rbd_sysfs_cleanup(void) { - if (class_rbd) - class_destroy(class_rbd); - class_rbd = NULL; + device_unregister(&rbd_root_dev); + bus_unregister(&rbd_bus_type); } int __init rbd_init(void) diff --git a/drivers/block/xen-blkfront.c b/drivers/block/xen-blkfront.c index 06e2812..657873e 100644 --- a/drivers/block/xen-blkfront.c +++ b/drivers/block/xen-blkfront.c @@ -65,14 +65,14 @@ enum blkif_state { struct blk_shadow { struct blkif_request req; - unsigned long request; + struct request *request; unsigned long frame[BLKIF_MAX_SEGMENTS_PER_REQUEST]; }; static DEFINE_MUTEX(blkfront_mutex); static const struct block_device_operations xlvbd_block_fops; -#define BLK_RING_SIZE __RING_SIZE((struct blkif_sring *)0, PAGE_SIZE) +#define BLK_RING_SIZE __CONST_RING_SIZE(blkif, PAGE_SIZE) /* * We have one of these per vbd, whether ide, scsi or 'other'. They @@ -136,7 +136,7 @@ static void add_id_to_freelist(struct blkfront_info *info, unsigned long id) { info->shadow[id].req.id = info->shadow_free; - info->shadow[id].request = 0; + info->shadow[id].request = NULL; info->shadow_free = id; } @@ -245,14 +245,11 @@ static int blkif_ioctl(struct block_device *bdev, fmode_t mode, } /* - * blkif_queue_request + * Generate a Xen blkfront IO request from a blk layer request. Reads + * and writes are handled as expected. Since we lack a loose flush + * request, we map flushes into a full ordered barrier. * - * request block io - * - * id: for guest use only. - * operation: BLKIF_OP_{READ,WRITE,PROBE} - * buffer: buffer to read/write into. this should be a - * virtual address in the guest os. + * @req: a request struct */ static int blkif_queue_request(struct request *req) { @@ -281,7 +278,7 @@ static int blkif_queue_request(struct request *req) /* Fill out a communications ring structure. */ ring_req = RING_GET_REQUEST(&info->ring, info->ring.req_prod_pvt); id = get_id_from_freelist(info); - info->shadow[id].request = (unsigned long)req; + info->shadow[id].request = req; ring_req->id = id; ring_req->sector_number = (blkif_sector_t)blk_rq_pos(req); @@ -289,8 +286,18 @@ static int blkif_queue_request(struct request *req) ring_req->operation = rq_data_dir(req) ? BLKIF_OP_WRITE : BLKIF_OP_READ; - if (req->cmd_flags & REQ_HARDBARRIER) + + if (req->cmd_flags & (REQ_FLUSH | REQ_FUA)) { + /* + * Ideally we could just do an unordered + * flush-to-disk, but all we have is a full write + * barrier at the moment. However, a barrier write is + * a superset of FUA, so we can implement it the same + * way. (It's also a FLUSH+FUA, since it is + * guaranteed ordered WRT previous writes.) + */ ring_req->operation = BLKIF_OP_WRITE_BARRIER; + } ring_req->nr_segments = blk_rq_map_sg(req->q, req, info->sg); BUG_ON(ring_req->nr_segments > BLKIF_MAX_SEGMENTS_PER_REQUEST); @@ -636,7 +643,7 @@ static irqreturn_t blkif_interrupt(int irq, void *dev_id) bret = RING_GET_RESPONSE(&info->ring, i); id = bret->id; - req = (struct request *)info->shadow[id].request; + req = info->shadow[id].request; blkif_completion(&info->shadow[id]); @@ -649,6 +656,16 @@ static irqreturn_t blkif_interrupt(int irq, void *dev_id) printk(KERN_WARNING "blkfront: %s: write barrier op failed\n", info->gd->disk_name); error = -EOPNOTSUPP; + } + if (unlikely(bret->status == BLKIF_RSP_ERROR && + info->shadow[id].req.nr_segments == 0)) { + printk(KERN_WARNING "blkfront: %s: empty write barrier op failed\n", + info->gd->disk_name); + error = -EOPNOTSUPP; + } + if (unlikely(error)) { + if (error == -EOPNOTSUPP) + error = 0; info->feature_flush = 0; xlvbd_flush(info); } @@ -901,7 +918,7 @@ static int blkif_recover(struct blkfront_info *info) /* Stage 3: Find pending requests and requeue them. */ for (i = 0; i < BLK_RING_SIZE; i++) { /* Not in use? */ - if (copy[i].request == 0) + if (!copy[i].request) continue; /* Grab a request slot and copy shadow state into it. */ @@ -918,9 +935,7 @@ static int blkif_recover(struct blkfront_info *info) req->seg[j].gref, info->xbdev->otherend_id, pfn_to_mfn(info->shadow[req->id].frame[j]), - rq_data_dir( - (struct request *) - info->shadow[req->id].request)); + rq_data_dir(info->shadow[req->id].request)); info->shadow[req->id].req = *req; info->ring.req_prod_pvt++; @@ -1069,14 +1084,8 @@ static void blkfront_connect(struct blkfront_info *info) */ info->feature_flush = 0; - /* - * The driver doesn't properly handled empty flushes, so - * lets disable barrier support for now. - */ -#if 0 if (!err && barrier) - info->feature_flush = REQ_FLUSH; -#endif + info->feature_flush = REQ_FLUSH | REQ_FUA; err = xlvbd_alloc_gendisk(sectors, info, binfo, sector_size); if (err) { |