From 17999be4aa408e7ff3b9d32c735649676567a3cd Mon Sep 17 00:00:00 2001 From: NeilBrown Date: Fri, 6 Jan 2006 00:20:12 -0800 Subject: [PATCH] md: improve raid1 "IO Barrier" concept raid1 needs to put up a barrier to new requests while it does resync or other background recovery. The code for this is currently open-coded, slighty obscure by its use of two waitqueues, and not documented. This patch gathers all the related code into 4 functions, and includes a comment which (hopefully) explains what is happening. Signed-off-by: Neil Brown Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- include/linux/raid/raid1.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'include/linux/raid') diff --git a/include/linux/raid/raid1.h b/include/linux/raid/raid1.h index 292b98f2b408..c55674252533 100644 --- a/include/linux/raid/raid1.h +++ b/include/linux/raid/raid1.h @@ -45,6 +45,7 @@ struct r1_private_data_s { spinlock_t resync_lock; int nr_pending; + int nr_waiting; int barrier; sector_t next_resync; int fullsync; /* set to 1 if a full sync is needed, @@ -52,8 +53,7 @@ struct r1_private_data_s { * Cleared when a sync completes. */ - wait_queue_head_t wait_idle; - wait_queue_head_t wait_resume; + wait_queue_head_t wait_barrier; struct pool_info *poolinfo; -- cgit v1.2.3 From 0a27ec96b6fb1abf867e36d7b0b681d67588767a Mon Sep 17 00:00:00 2001 From: NeilBrown Date: Fri, 6 Jan 2006 00:20:13 -0800 Subject: [PATCH] md: improve raid10 "IO Barrier" concept raid10 needs to put up a barrier to new requests while it does resync or other background recovery. The code for this is currently open-coded, slighty obscure by its use of two waitqueues, and not documented. This patch gathers all the related code into 4 functions, and includes a comment which (hopefully) explains what is happening. Signed-off-by: Neil Brown Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- drivers/md/raid10.c | 135 ++++++++++++++++++++++++++------------------ include/linux/raid/raid10.h | 4 +- 2 files changed, 81 insertions(+), 58 deletions(-) (limited to 'include/linux/raid') diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c index 713dc9c2c730..50bd7b152f28 100644 --- a/drivers/md/raid10.c +++ b/drivers/md/raid10.c @@ -47,6 +47,9 @@ static void unplug_slaves(mddev_t *mddev); +static void allow_barrier(conf_t *conf); +static void lower_barrier(conf_t *conf); + static void * r10bio_pool_alloc(gfp_t gfp_flags, void *data) { conf_t *conf = data; @@ -175,20 +178,13 @@ static void put_all_bios(conf_t *conf, r10bio_t *r10_bio) static inline void free_r10bio(r10bio_t *r10_bio) { - unsigned long flags; - conf_t *conf = mddev_to_conf(r10_bio->mddev); /* * Wake up any possible resync thread that waits for the device * to go idle. */ - spin_lock_irqsave(&conf->resync_lock, flags); - if (!--conf->nr_pending) { - wake_up(&conf->wait_idle); - wake_up(&conf->wait_resume); - } - spin_unlock_irqrestore(&conf->resync_lock, flags); + allow_barrier(conf); put_all_bios(conf, r10_bio); mempool_free(r10_bio, conf->r10bio_pool); @@ -197,22 +193,10 @@ static inline void free_r10bio(r10bio_t *r10_bio) static inline void put_buf(r10bio_t *r10_bio) { conf_t *conf = mddev_to_conf(r10_bio->mddev); - unsigned long flags; mempool_free(r10_bio, conf->r10buf_pool); - spin_lock_irqsave(&conf->resync_lock, flags); - if (!conf->barrier) - BUG(); - --conf->barrier; - wake_up(&conf->wait_resume); - wake_up(&conf->wait_idle); - - if (!--conf->nr_pending) { - wake_up(&conf->wait_idle); - wake_up(&conf->wait_resume); - } - spin_unlock_irqrestore(&conf->resync_lock, flags); + lower_barrier(conf); } static void reschedule_retry(r10bio_t *r10_bio) @@ -640,30 +624,82 @@ static int raid10_issue_flush(request_queue_t *q, struct gendisk *disk, return ret; } -/* - * Throttle resync depth, so that we can both get proper overlapping of - * requests, but are still able to handle normal requests quickly. +/* Barriers.... + * Sometimes we need to suspend IO while we do something else, + * either some resync/recovery, or reconfigure the array. + * To do this we raise a 'barrier'. + * The 'barrier' is a counter that can be raised multiple times + * to count how many activities are happening which preclude + * normal IO. + * We can only raise the barrier if there is no pending IO. + * i.e. if nr_pending == 0. + * We choose only to raise the barrier if no-one is waiting for the + * barrier to go down. This means that as soon as an IO request + * is ready, no other operations which require a barrier will start + * until the IO request has had a chance. + * + * So: regular IO calls 'wait_barrier'. When that returns there + * is no backgroup IO happening, It must arrange to call + * allow_barrier when it has finished its IO. + * backgroup IO calls must call raise_barrier. Once that returns + * there is no normal IO happeing. It must arrange to call + * lower_barrier when the particular background IO completes. */ #define RESYNC_DEPTH 32 -static void device_barrier(conf_t *conf, sector_t sect) +static void raise_barrier(conf_t *conf) { spin_lock_irq(&conf->resync_lock); - wait_event_lock_irq(conf->wait_idle, !waitqueue_active(&conf->wait_resume), - conf->resync_lock, unplug_slaves(conf->mddev)); - - if (!conf->barrier++) { - wait_event_lock_irq(conf->wait_idle, !conf->nr_pending, - conf->resync_lock, unplug_slaves(conf->mddev)); - if (conf->nr_pending) - BUG(); + + /* Wait until no block IO is waiting */ + wait_event_lock_irq(conf->wait_barrier, !conf->nr_waiting, + conf->resync_lock, + raid10_unplug(conf->mddev->queue)); + + /* block any new IO from starting */ + conf->barrier++; + + /* No wait for all pending IO to complete */ + wait_event_lock_irq(conf->wait_barrier, + !conf->nr_pending && conf->barrier < RESYNC_DEPTH, + conf->resync_lock, + raid10_unplug(conf->mddev->queue)); + + spin_unlock_irq(&conf->resync_lock); +} + +static void lower_barrier(conf_t *conf) +{ + unsigned long flags; + spin_lock_irqsave(&conf->resync_lock, flags); + conf->barrier--; + spin_unlock_irqrestore(&conf->resync_lock, flags); + wake_up(&conf->wait_barrier); +} + +static void wait_barrier(conf_t *conf) +{ + spin_lock_irq(&conf->resync_lock); + if (conf->barrier) { + conf->nr_waiting++; + wait_event_lock_irq(conf->wait_barrier, !conf->barrier, + conf->resync_lock, + raid10_unplug(conf->mddev->queue)); + conf->nr_waiting--; } - wait_event_lock_irq(conf->wait_resume, conf->barrier < RESYNC_DEPTH, - conf->resync_lock, unplug_slaves(conf->mddev)); - conf->next_resync = sect; + conf->nr_pending++; spin_unlock_irq(&conf->resync_lock); } +static void allow_barrier(conf_t *conf) +{ + unsigned long flags; + spin_lock_irqsave(&conf->resync_lock, flags); + conf->nr_pending--; + spin_unlock_irqrestore(&conf->resync_lock, flags); + wake_up(&conf->wait_barrier); +} + static int make_request(request_queue_t *q, struct bio * bio) { mddev_t *mddev = q->queuedata; @@ -719,10 +755,7 @@ static int make_request(request_queue_t *q, struct bio * bio) * thread has put up a bar for new requests. * Continue immediately if no resync is active currently. */ - spin_lock_irq(&conf->resync_lock); - wait_event_lock_irq(conf->wait_resume, !conf->barrier, conf->resync_lock, ); - conf->nr_pending++; - spin_unlock_irq(&conf->resync_lock); + wait_barrier(conf); disk_stat_inc(mddev->gendisk, ios[rw]); disk_stat_add(mddev->gendisk, sectors[rw], bio_sectors(bio)); @@ -897,13 +930,8 @@ static void print_conf(conf_t *conf) static void close_sync(conf_t *conf) { - spin_lock_irq(&conf->resync_lock); - wait_event_lock_irq(conf->wait_resume, !conf->barrier, - conf->resync_lock, unplug_slaves(conf->mddev)); - spin_unlock_irq(&conf->resync_lock); - - if (conf->barrier) BUG(); - if (waitqueue_active(&conf->wait_idle)) BUG(); + wait_barrier(conf); + allow_barrier(conf); mempool_destroy(conf->r10buf_pool); conf->r10buf_pool = NULL; @@ -1395,9 +1423,10 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i * If there is non-resync activity waiting for us then * put in a delay to throttle resync. */ - if (!go_faster && waitqueue_active(&conf->wait_resume)) + if (!go_faster && conf->nr_waiting) msleep_interruptible(1000); - device_barrier(conf, sector_nr + RESYNC_SECTORS); + raise_barrier(conf); + conf->next_resync = sector_nr; /* Again, very different code for resync and recovery. * Both must result in an r10bio with a list of bios that @@ -1427,7 +1456,6 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i r10_bio = mempool_alloc(conf->r10buf_pool, GFP_NOIO); spin_lock_irq(&conf->resync_lock); - conf->nr_pending++; if (rb2) conf->barrier++; spin_unlock_irq(&conf->resync_lock); atomic_set(&r10_bio->remaining, 0); @@ -1500,10 +1528,6 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i int count = 0; r10_bio = mempool_alloc(conf->r10buf_pool, GFP_NOIO); - spin_lock_irq(&conf->resync_lock); - conf->nr_pending++; - spin_unlock_irq(&conf->resync_lock); - r10_bio->mddev = mddev; atomic_set(&r10_bio->remaining, 0); @@ -1713,8 +1737,7 @@ static int run(mddev_t *mddev) INIT_LIST_HEAD(&conf->retry_list); spin_lock_init(&conf->resync_lock); - init_waitqueue_head(&conf->wait_idle); - init_waitqueue_head(&conf->wait_resume); + init_waitqueue_head(&conf->wait_barrier); /* need to check that every block has at least one working mirror */ if (!enough(conf)) { diff --git a/include/linux/raid/raid10.h b/include/linux/raid/raid10.h index 60708789c8f9..08317b77802b 100644 --- a/include/linux/raid/raid10.h +++ b/include/linux/raid/raid10.h @@ -39,11 +39,11 @@ struct r10_private_data_s { spinlock_t resync_lock; int nr_pending; + int nr_waiting; int barrier; sector_t next_resync; - wait_queue_head_t wait_idle; - wait_queue_head_t wait_resume; + wait_queue_head_t wait_barrier; mempool_t *r10bio_pool; mempool_t *r10buf_pool; -- cgit v1.2.3 From 6ff8d8ec06690f4011a6c3ad9e0759b9094f0601 Mon Sep 17 00:00:00 2001 From: NeilBrown Date: Fri, 6 Jan 2006 00:20:15 -0800 Subject: [PATCH] md: allow dirty raid[456] arrays to be started at boot See patch to md.txt for more details Signed-off-by: Neil Brown Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- Documentation/md.txt | 24 ++++++++++++++++++++++++ drivers/md/md.c | 4 ++++ drivers/md/raid5.c | 15 +++++++++++---- drivers/md/raid6main.c | 13 +++++++++---- include/linux/raid/md_k.h | 1 + 5 files changed, 49 insertions(+), 8 deletions(-) (limited to 'include/linux/raid') diff --git a/Documentation/md.txt b/Documentation/md.txt index 23e6cce40f9c..1dd0fb6021cf 100644 --- a/Documentation/md.txt +++ b/Documentation/md.txt @@ -51,6 +51,30 @@ superblock can be autodetected and run at boot time. The kernel parameter "raid=partitionable" (or "raid=part") means that all auto-detected arrays are assembled as partitionable. +Boot time assembly of degraded/dirty arrays +------------------------------------------- + +If a raid5 or raid6 array is both dirty and degraded, it could have +undetectable data corruption. This is because the fact that it is +'dirty' means that the parity cannot be trusted, and the fact that it +is degraded means that some datablocks are missing and cannot reliably +be reconstructed (due to no parity). + +For this reason, md will normally refuse to start such an array. This +requires the sysadmin to take action to explicitly start the array +desipite possible corruption. This is normally done with + mdadm --assemble --force .... + +This option is not really available if the array has the root +filesystem on it. In order to support this booting from such an +array, md supports a module parameter "start_dirty_degraded" which, +when set to 1, bypassed the checks and will allows dirty degraded +arrays to be started. + +So, to boot with a root filesystem of a dirty degraded raid[56], use + + md-mod.start_dirty_degraded=1 + Superblock formats ------------------ diff --git a/drivers/md/md.c b/drivers/md/md.c index 8175a2a222da..b4fb7247b3ed 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -1937,6 +1937,7 @@ static void md_safemode_timeout(unsigned long data) md_wakeup_thread(mddev->thread); } +static int start_dirty_degraded; static int do_md_run(mddev_t * mddev) { @@ -2048,6 +2049,7 @@ static int do_md_run(mddev_t * mddev) mddev->recovery = 0; mddev->resync_max_sectors = mddev->size << 1; /* may be over-ridden by personality */ mddev->barriers_work = 1; + mddev->ok_start_degraded = start_dirty_degraded; if (start_readonly) mddev->ro = 2; /* read-only, but switch on first write */ @@ -4509,6 +4511,8 @@ static int set_ro(const char *val, struct kernel_param *kp) } module_param_call(start_ro, set_ro, get_ro, NULL, 0600); +module_param(start_dirty_degraded, int, 0644); + EXPORT_SYMBOL(register_md_personality); EXPORT_SYMBOL(unregister_md_personality); diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 334ff7a07283..53a0f2ce76c8 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -1904,10 +1904,17 @@ static int run(mddev_t *mddev) if (mddev->degraded == 1 && mddev->recovery_cp != MaxSector) { - printk(KERN_ERR - "raid5: cannot start dirty degraded array for %s\n", - mdname(mddev)); - goto abort; + if (mddev->ok_start_degraded) + printk(KERN_WARNING + "raid5: starting dirty degraded array: %s" + "- data corruption possible.\n", + mdname(mddev)); + else { + printk(KERN_ERR + "raid5: cannot start dirty degraded array for %s\n", + mdname(mddev)); + goto abort; + } } { diff --git a/drivers/md/raid6main.c b/drivers/md/raid6main.c index 0000d162d198..9ac6dcd55127 100644 --- a/drivers/md/raid6main.c +++ b/drivers/md/raid6main.c @@ -1929,13 +1929,18 @@ static int run(mddev_t *mddev) goto abort; } -#if 0 /* FIX: For now */ if (mddev->degraded > 0 && mddev->recovery_cp != MaxSector) { - printk(KERN_ERR "raid6: cannot start dirty degraded array for %s\n", mdname(mddev)); - goto abort; + if (mddev->ok_start_degraded) + printk(KERN_WARNING "raid6: starting dirty degraded array:%s" + "- data corruption possible.\n", + mdname(mddev)); + else { + printk(KERN_ERR "raid6: cannot start dirty degraded array" + " for %s\n", mdname(mddev)); + goto abort; + } } -#endif { mddev->thread = md_register_thread(raid6d, mddev, "%s_raid6"); diff --git a/include/linux/raid/md_k.h b/include/linux/raid/md_k.h index 46629a275ba9..1dd587b5975a 100644 --- a/include/linux/raid/md_k.h +++ b/include/linux/raid/md_k.h @@ -183,6 +183,7 @@ struct mddev_s sector_t resync_mismatches; /* count of sectors where * parity/replica mismatch found */ + int ok_start_degraded; /* recovery/resync flags * NEEDED: we might need to start a resync/recover * RUNNING: a thread is running, or about to be started -- cgit v1.2.3 From 6cce3b23f6f8e974c00af7a9b88f1d413ba368a8 Mon Sep 17 00:00:00 2001 From: NeilBrown Date: Fri, 6 Jan 2006 00:20:16 -0800 Subject: [PATCH] md: write intent bitmap support for raid10 Signed-off-by: Neil Brown Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- drivers/md/md.c | 10 ++- drivers/md/raid10.c | 178 ++++++++++++++++++++++++++++++++++++++------ include/linux/raid/raid10.h | 9 ++- 3 files changed, 171 insertions(+), 26 deletions(-) (limited to 'include/linux/raid') diff --git a/drivers/md/md.c b/drivers/md/md.c index ee199d462520..64e7da3701a5 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -714,9 +714,10 @@ static int super_90_validate(mddev_t *mddev, mdk_rdev_t *rdev) if (sb->state & (1<bitmap_file == NULL) { - if (mddev->level != 1 && mddev->level != 5 && mddev->level != 6) { + if (mddev->level != 1 && mddev->level != 5 && mddev->level != 6 + && mddev->level != 10) { /* FIXME use a better test */ - printk(KERN_WARNING "md: bitmaps only support for raid1\n"); + printk(KERN_WARNING "md: bitmaps not supported for this level.\n"); return -EINVAL; } mddev->bitmap_offset = mddev->default_bitmap_offset; @@ -1037,8 +1038,9 @@ static int super_1_validate(mddev_t *mddev, mdk_rdev_t *rdev) if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_BITMAP_OFFSET) && mddev->bitmap_file == NULL ) { - if (mddev->level != 1) { - printk(KERN_WARNING "md: bitmaps only supported for raid1\n"); + if (mddev->level != 1 && mddev->level != 5 && mddev->level != 6 + && mddev->level != 10) { + printk(KERN_WARNING "md: bitmaps not supported for this level.\n"); return -EINVAL; } mddev->bitmap_offset = (__s32)le32_to_cpu(sb->bitmap_offset); diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c index 50bd7b152f28..8f58a447d9f0 100644 --- a/drivers/md/raid10.c +++ b/drivers/md/raid10.c @@ -18,7 +18,9 @@ * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ +#include "dm-bio-list.h" #include +#include /* * RAID10 provides a combination of RAID0 and RAID1 functionality. @@ -306,9 +308,11 @@ static int raid10_end_write_request(struct bio *bio, unsigned int bytes_done, in /* * this branch is our 'one mirror IO has finished' event handler: */ - if (!uptodate) + if (!uptodate) { md_error(r10_bio->mddev, conf->mirrors[dev].rdev); - else + /* an I/O failed, we can't clear the bitmap */ + set_bit(R10BIO_Degraded, &r10_bio->state); + } else /* * Set R10BIO_Uptodate in our master bio, so that * we will return a good error code for to the higher @@ -328,6 +332,11 @@ static int raid10_end_write_request(struct bio *bio, unsigned int bytes_done, in * already. */ if (atomic_dec_and_test(&r10_bio->remaining)) { + /* clear the bitmap if all writes complete successfully */ + bitmap_endwrite(r10_bio->mddev->bitmap, r10_bio->sector, + r10_bio->sectors, + !test_bit(R10BIO_Degraded, &r10_bio->state), + 0); md_write_end(r10_bio->mddev); raid_end_bio_io(r10_bio); } @@ -486,8 +495,9 @@ static int read_balance(conf_t *conf, r10bio_t *r10_bio) rcu_read_lock(); /* * Check if we can balance. We can balance on the whole - * device if no resync is going on, or below the resync window. - * We take the first readable disk when above the resync window. + * device if no resync is going on (recovery is ok), or below + * the resync window. We take the first readable disk when + * above the resync window. */ if (conf->mddev->recovery_cp < MaxSector && (this_sector + sectors >= conf->next_resync)) { @@ -591,7 +601,10 @@ static void unplug_slaves(mddev_t *mddev) static void raid10_unplug(request_queue_t *q) { + mddev_t *mddev = q->queuedata; + unplug_slaves(q->queuedata); + md_wakeup_thread(mddev->thread); } static int raid10_issue_flush(request_queue_t *q, struct gendisk *disk, @@ -647,12 +660,13 @@ static int raid10_issue_flush(request_queue_t *q, struct gendisk *disk, */ #define RESYNC_DEPTH 32 -static void raise_barrier(conf_t *conf) +static void raise_barrier(conf_t *conf, int force) { + BUG_ON(force && !conf->barrier); spin_lock_irq(&conf->resync_lock); - /* Wait until no block IO is waiting */ - wait_event_lock_irq(conf->wait_barrier, !conf->nr_waiting, + /* Wait until no block IO is waiting (unless 'force') */ + wait_event_lock_irq(conf->wait_barrier, force || !conf->nr_waiting, conf->resync_lock, raid10_unplug(conf->mddev->queue)); @@ -710,6 +724,8 @@ static int make_request(request_queue_t *q, struct bio * bio) int i; int chunk_sects = conf->chunk_mask + 1; const int rw = bio_data_dir(bio); + struct bio_list bl; + unsigned long flags; if (unlikely(bio_barrier(bio))) { bio_endio(bio, bio->bi_size, -EOPNOTSUPP); @@ -767,6 +783,7 @@ static int make_request(request_queue_t *q, struct bio * bio) r10_bio->mddev = mddev; r10_bio->sector = bio->bi_sector; + r10_bio->state = 0; if (rw == READ) { /* @@ -811,13 +828,16 @@ static int make_request(request_queue_t *q, struct bio * bio) !test_bit(Faulty, &rdev->flags)) { atomic_inc(&rdev->nr_pending); r10_bio->devs[i].bio = bio; - } else + } else { r10_bio->devs[i].bio = NULL; + set_bit(R10BIO_Degraded, &r10_bio->state); + } } rcu_read_unlock(); - atomic_set(&r10_bio->remaining, 1); + atomic_set(&r10_bio->remaining, 0); + bio_list_init(&bl); for (i = 0; i < conf->copies; i++) { struct bio *mbio; int d = r10_bio->devs[i].devnum; @@ -835,13 +855,14 @@ static int make_request(request_queue_t *q, struct bio * bio) mbio->bi_private = r10_bio; atomic_inc(&r10_bio->remaining); - generic_make_request(mbio); + bio_list_add(&bl, mbio); } - if (atomic_dec_and_test(&r10_bio->remaining)) { - md_write_end(mddev); - raid_end_bio_io(r10_bio); - } + bitmap_startwrite(mddev->bitmap, bio->bi_sector, r10_bio->sectors, 0); + spin_lock_irqsave(&conf->device_lock, flags); + bio_list_merge(&conf->pending_bio_list, &bl); + blk_plug_device(mddev->queue); + spin_unlock_irqrestore(&conf->device_lock, flags); return 0; } @@ -999,7 +1020,12 @@ static int raid10_add_disk(mddev_t *mddev, mdk_rdev_t *rdev) if (!enough(conf)) return 0; - for (mirror=0; mirror < mddev->raid_disks; mirror++) + if (rdev->saved_raid_disk >= 0 && + conf->mirrors[rdev->saved_raid_disk].rdev == NULL) + mirror = rdev->saved_raid_disk; + else + mirror = 0; + for ( ; mirror < mddev->raid_disks; mirror++) if ( !(p=conf->mirrors+mirror)->rdev) { blk_queue_stack_limits(mddev->queue, @@ -1015,6 +1041,8 @@ static int raid10_add_disk(mddev_t *mddev, mdk_rdev_t *rdev) p->head_position = 0; rdev->raid_disk = mirror; found = 1; + if (rdev->saved_raid_disk != mirror) + conf->fullsync = 1; rcu_assign_pointer(p->rdev, rdev); break; } @@ -1282,6 +1310,26 @@ static void raid10d(mddev_t *mddev) for (;;) { char b[BDEVNAME_SIZE]; spin_lock_irqsave(&conf->device_lock, flags); + + if (conf->pending_bio_list.head) { + bio = bio_list_get(&conf->pending_bio_list); + blk_remove_plug(mddev->queue); + spin_unlock_irqrestore(&conf->device_lock, flags); + /* flush any pending bitmap writes to disk before proceeding w/ I/O */ + if (bitmap_unplug(mddev->bitmap) != 0) + printk("%s: bitmap file write failed!\n", mdname(mddev)); + + while (bio) { /* submit pending writes */ + struct bio *next = bio->bi_next; + bio->bi_next = NULL; + generic_make_request(bio); + bio = next; + } + unplug = 1; + + continue; + } + if (list_empty(head)) break; r10_bio = list_entry(head->prev, r10bio_t, retry_list); @@ -1388,6 +1436,8 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i sector_t max_sector, nr_sectors; int disk; int i; + int max_sync; + int sync_blocks; sector_t sectors_skipped = 0; int chunks_skipped = 0; @@ -1401,6 +1451,29 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) max_sector = mddev->resync_max_sectors; if (sector_nr >= max_sector) { + /* If we aborted, we need to abort the + * sync on the 'current' bitmap chucks (there can + * be several when recovering multiple devices). + * as we may have started syncing it but not finished. + * We can find the current address in + * mddev->curr_resync, but for recovery, + * we need to convert that to several + * virtual addresses. + */ + if (mddev->curr_resync < max_sector) { /* aborted */ + if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) + bitmap_end_sync(mddev->bitmap, mddev->curr_resync, + &sync_blocks, 1); + else for (i=0; iraid_disks; i++) { + sector_t sect = + raid10_find_virt(conf, mddev->curr_resync, i); + bitmap_end_sync(mddev->bitmap, sect, + &sync_blocks, 1); + } + } else /* completed sync */ + conf->fullsync = 0; + + bitmap_close_sync(mddev->bitmap); close_sync(conf); *skipped = 1; return sectors_skipped; @@ -1425,8 +1498,6 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i */ if (!go_faster && conf->nr_waiting) msleep_interruptible(1000); - raise_barrier(conf); - conf->next_resync = sector_nr; /* Again, very different code for resync and recovery. * Both must result in an r10bio with a list of bios that @@ -1443,6 +1514,7 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i * end_sync_write if we will want to write. */ + max_sync = RESYNC_PAGES << (PAGE_SHIFT-9); if (!test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) { /* recovery... the complicated one */ int i, j, k; @@ -1451,13 +1523,29 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i for (i=0 ; iraid_disks; i++) if (conf->mirrors[i].rdev && !test_bit(In_sync, &conf->mirrors[i].rdev->flags)) { + int still_degraded = 0; /* want to reconstruct this device */ r10bio_t *rb2 = r10_bio; + sector_t sect = raid10_find_virt(conf, sector_nr, i); + int must_sync; + /* Unless we are doing a full sync, we only need + * to recover the block if it is set in the bitmap + */ + must_sync = bitmap_start_sync(mddev->bitmap, sect, + &sync_blocks, 1); + if (sync_blocks < max_sync) + max_sync = sync_blocks; + if (!must_sync && + !conf->fullsync) { + /* yep, skip the sync_blocks here, but don't assume + * that there will never be anything to do here + */ + chunks_skipped = -1; + continue; + } r10_bio = mempool_alloc(conf->r10buf_pool, GFP_NOIO); - spin_lock_irq(&conf->resync_lock); - if (rb2) conf->barrier++; - spin_unlock_irq(&conf->resync_lock); + raise_barrier(conf, rb2 != NULL); atomic_set(&r10_bio->remaining, 0); r10_bio->master_bio = (struct bio*)rb2; @@ -1465,8 +1553,21 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i atomic_inc(&rb2->remaining); r10_bio->mddev = mddev; set_bit(R10BIO_IsRecover, &r10_bio->state); - r10_bio->sector = raid10_find_virt(conf, sector_nr, i); + r10_bio->sector = sect; + raid10_find_phys(conf, r10_bio); + /* Need to check if this section will still be + * degraded + */ + for (j=0; jcopies;j++) { + int d = r10_bio->devs[j].devnum; + if (conf->mirrors[d].rdev == NULL || + test_bit(Faulty, &conf->mirrors[d].rdev->flags)) + still_degraded = 1; + } + must_sync = bitmap_start_sync(mddev->bitmap, sect, + &sync_blocks, still_degraded); + for (j=0; jcopies;j++) { int d = r10_bio->devs[j].devnum; if (conf->mirrors[d].rdev && @@ -1526,10 +1627,22 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i } else { /* resync. Schedule a read for every block at this virt offset */ int count = 0; + + if (!bitmap_start_sync(mddev->bitmap, sector_nr, + &sync_blocks, mddev->degraded) && + !conf->fullsync && !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) { + /* We can skip this block */ + *skipped = 1; + return sync_blocks + sectors_skipped; + } + if (sync_blocks < max_sync) + max_sync = sync_blocks; r10_bio = mempool_alloc(conf->r10buf_pool, GFP_NOIO); r10_bio->mddev = mddev; atomic_set(&r10_bio->remaining, 0); + raise_barrier(conf, 0); + conf->next_resync = sector_nr; r10_bio->master_bio = NULL; r10_bio->sector = sector_nr; @@ -1582,6 +1695,8 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i } nr_sectors = 0; + if (sector_nr + max_sync < max_sector) + max_sector = sector_nr + max_sync; do { struct page *page; int len = PAGE_SIZE; @@ -1821,6 +1936,26 @@ static int stop(mddev_t *mddev) return 0; } +static void raid10_quiesce(mddev_t *mddev, int state) +{ + conf_t *conf = mddev_to_conf(mddev); + + switch(state) { + case 1: + raise_barrier(conf, 0); + break; + case 0: + lower_barrier(conf); + break; + } + if (mddev->thread) { + if (mddev->bitmap) + mddev->thread->timeout = mddev->bitmap->daemon_sleep * HZ; + else + mddev->thread->timeout = MAX_SCHEDULE_TIMEOUT; + md_wakeup_thread(mddev->thread); + } +} static mdk_personality_t raid10_personality = { @@ -1835,6 +1970,7 @@ static mdk_personality_t raid10_personality = .hot_remove_disk= raid10_remove_disk, .spare_active = raid10_spare_active, .sync_request = sync_request, + .quiesce = raid10_quiesce, }; static int __init raid_init(void) diff --git a/include/linux/raid/raid10.h b/include/linux/raid/raid10.h index 08317b77802b..b660cbf628d8 100644 --- a/include/linux/raid/raid10.h +++ b/include/linux/raid/raid10.h @@ -35,13 +35,19 @@ struct r10_private_data_s { sector_t chunk_mask; struct list_head retry_list; - /* for use when syncing mirrors: */ + /* queue pending writes and submit them on unplug */ + struct bio_list pending_bio_list; + spinlock_t resync_lock; int nr_pending; int nr_waiting; int barrier; sector_t next_resync; + int fullsync; /* set to 1 if a full sync is needed, + * (fresh device added). + * Cleared when a sync completes. + */ wait_queue_head_t wait_barrier; @@ -100,4 +106,5 @@ struct r10bio_s { #define R10BIO_Uptodate 0 #define R10BIO_IsSync 1 #define R10BIO_IsRecover 2 +#define R10BIO_Degraded 3 #endif -- cgit v1.2.3 From ca65b73bd9c301d243df93780f7b26579e6c9204 Mon Sep 17 00:00:00 2001 From: NeilBrown Date: Fri, 6 Jan 2006 00:20:17 -0800 Subject: [PATCH] md: fix raid6 resync check/repair code raid6 currently does not check the P/Q syndromes when doing a resync, it just calculates the correct value and writes it. Doing the check can reduce writes (often to 0) for a resync, and it is needed to properly implement the echo check > sync_action operation. This patch implements the appropriate checks and tidies up some related code. It also allows raid6 user-requested resync to bypass the intent bitmap. Signed-off-by: Neil Brown Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- drivers/md/raid6main.c | 182 ++++++++++++++++++++++++++------------------- include/linux/raid/raid5.h | 2 + 2 files changed, 108 insertions(+), 76 deletions(-) (limited to 'include/linux/raid') diff --git a/drivers/md/raid6main.c b/drivers/md/raid6main.c index 304455d236f9..52e8796bb8ac 100644 --- a/drivers/md/raid6main.c +++ b/drivers/md/raid6main.c @@ -805,7 +805,7 @@ static void compute_parity(struct stripe_head *sh, int method) } /* Compute one missing block */ -static void compute_block_1(struct stripe_head *sh, int dd_idx) +static void compute_block_1(struct stripe_head *sh, int dd_idx, int nozero) { raid6_conf_t *conf = sh->raid_conf; int i, count, disks = conf->raid_disks; @@ -821,7 +821,7 @@ static void compute_block_1(struct stripe_head *sh, int dd_idx) compute_parity(sh, UPDATE_PARITY); } else { ptr[0] = page_address(sh->dev[dd_idx].page); - memset(ptr[0], 0, STRIPE_SIZE); + if (!nozero) memset(ptr[0], 0, STRIPE_SIZE); count = 1; for (i = disks ; i--; ) { if (i == dd_idx || i == qd_idx) @@ -838,7 +838,8 @@ static void compute_block_1(struct stripe_head *sh, int dd_idx) } if (count != 1) xor_block(count, STRIPE_SIZE, ptr); - set_bit(R5_UPTODATE, &sh->dev[dd_idx].flags); + if (!nozero) set_bit(R5_UPTODATE, &sh->dev[dd_idx].flags); + else clear_bit(R5_UPTODATE, &sh->dev[dd_idx].flags); } } @@ -871,7 +872,7 @@ static void compute_block_2(struct stripe_head *sh, int dd_idx1, int dd_idx2) return; } else { /* We're missing D+Q; recompute D from P */ - compute_block_1(sh, (dd_idx1 == qd_idx) ? dd_idx2 : dd_idx1); + compute_block_1(sh, (dd_idx1 == qd_idx) ? dd_idx2 : dd_idx1, 0); compute_parity(sh, UPDATE_PARITY); /* Is this necessary? */ return; } @@ -982,6 +983,12 @@ static int add_stripe_bio(struct stripe_head *sh, struct bio *bi, int dd_idx, in } +static int page_is_zero(struct page *p) +{ + char *a = page_address(p); + return ((*(u32*)a) == 0 && + memcmp(a, a+4, STRIPE_SIZE-4)==0); +} /* * handle_stripe - do things to a stripe. * @@ -1000,7 +1007,7 @@ static int add_stripe_bio(struct stripe_head *sh, struct bio *bi, int dd_idx, in * */ -static void handle_stripe(struct stripe_head *sh) +static void handle_stripe(struct stripe_head *sh, struct page *tmp_page) { raid6_conf_t *conf = sh->raid_conf; int disks = conf->raid_disks; @@ -1228,7 +1235,7 @@ static void handle_stripe(struct stripe_head *sh) if (uptodate == disks-1) { PRINTK("Computing stripe %llu block %d\n", (unsigned long long)sh->sector, i); - compute_block_1(sh, i); + compute_block_1(sh, i, 0); uptodate++; } else if ( uptodate == disks-2 && failed >= 2 ) { /* Computing 2-failure is *very* expensive; only do it if failed >= 2 */ @@ -1323,7 +1330,7 @@ static void handle_stripe(struct stripe_head *sh) /* We have failed blocks and need to compute them */ switch ( failed ) { case 0: BUG(); - case 1: compute_block_1(sh, failed_num[0]); break; + case 1: compute_block_1(sh, failed_num[0], 0); break; case 2: compute_block_2(sh, failed_num[0], failed_num[1]); break; default: BUG(); /* This request should have been failed? */ } @@ -1338,12 +1345,10 @@ static void handle_stripe(struct stripe_head *sh) (unsigned long long)sh->sector, i); locked++; set_bit(R5_Wantwrite, &sh->dev[i].flags); -#if 0 /**** FIX: I don't understand the logic here... ****/ - if (!test_bit(R5_Insync, &sh->dev[i].flags) - || ((i==pd_idx || i==qd_idx) && failed == 0)) /* FIX? */ - set_bit(STRIPE_INSYNC, &sh->state); -#endif } + /* after a RECONSTRUCT_WRITE, the stripe MUST be in-sync */ + set_bit(STRIPE_INSYNC, &sh->state); + if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) { atomic_dec(&conf->preread_active_stripes); if (atomic_read(&conf->preread_active_stripes) < IO_THRESHOLD) @@ -1356,79 +1361,97 @@ static void handle_stripe(struct stripe_head *sh) * Any reads will already have been scheduled, so we just see if enough data * is available */ - if (syncing && locked == 0 && - !test_bit(STRIPE_INSYNC, &sh->state) && failed <= 2) { - set_bit(STRIPE_HANDLE, &sh->state); -#if 0 /* RAID-6: Don't support CHECK PARITY yet */ - if (failed == 0) { - char *pagea; - if (uptodate != disks) - BUG(); - compute_parity(sh, CHECK_PARITY); - uptodate--; - pagea = page_address(sh->dev[pd_idx].page); - if ((*(u32*)pagea) == 0 && - !memcmp(pagea, pagea+4, STRIPE_SIZE-4)) { - /* parity is correct (on disc, not in buffer any more) */ - set_bit(STRIPE_INSYNC, &sh->state); - } - } -#endif - if (!test_bit(STRIPE_INSYNC, &sh->state)) { - int failed_needupdate[2]; - struct r5dev *adev, *bdev; - - if ( failed < 1 ) - failed_num[0] = pd_idx; - if ( failed < 2 ) - failed_num[1] = (failed_num[0] == qd_idx) ? pd_idx : qd_idx; + if (syncing && locked == 0 && !test_bit(STRIPE_INSYNC, &sh->state)) { + int update_p = 0, update_q = 0; + struct r5dev *dev; - failed_needupdate[0] = !test_bit(R5_UPTODATE, &sh->dev[failed_num[0]].flags); - failed_needupdate[1] = !test_bit(R5_UPTODATE, &sh->dev[failed_num[1]].flags); + set_bit(STRIPE_HANDLE, &sh->state); - PRINTK("sync: failed=%d num=%d,%d fnu=%u%u\n", - failed, failed_num[0], failed_num[1], failed_needupdate[0], failed_needupdate[1]); + BUG_ON(failed>2); + BUG_ON(uptodate < disks); + /* Want to check and possibly repair P and Q. + * However there could be one 'failed' device, in which + * case we can only check one of them, possibly using the + * other to generate missing data + */ -#if 0 /* RAID-6: This code seems to require that CHECK_PARITY destroys the uptodateness of the parity */ - /* should be able to compute the missing block(s) and write to spare */ - if ( failed_needupdate[0] ^ failed_needupdate[1] ) { - if (uptodate+1 != disks) - BUG(); - compute_block_1(sh, failed_needupdate[0] ? failed_num[0] : failed_num[1]); - uptodate++; - } else if ( failed_needupdate[0] & failed_needupdate[1] ) { - if (uptodate+2 != disks) - BUG(); - compute_block_2(sh, failed_num[0], failed_num[1]); - uptodate += 2; + /* If !tmp_page, we cannot do the calculations, + * but as we have set STRIPE_HANDLE, we will soon be called + * by stripe_handle with a tmp_page - just wait until then. + */ + if (tmp_page) { + if (failed == q_failed) { + /* The only possible failed device holds 'Q', so it makes + * sense to check P (If anything else were failed, we would + * have used P to recreate it). + */ + compute_block_1(sh, pd_idx, 1); + if (!page_is_zero(sh->dev[pd_idx].page)) { + compute_block_1(sh,pd_idx,0); + update_p = 1; + } + } + if (!q_failed && failed < 2) { + /* q is not failed, and we didn't use it to generate + * anything, so it makes sense to check it + */ + memcpy(page_address(tmp_page), + page_address(sh->dev[qd_idx].page), + STRIPE_SIZE); + compute_parity(sh, UPDATE_PARITY); + if (memcmp(page_address(tmp_page), + page_address(sh->dev[qd_idx].page), + STRIPE_SIZE)!= 0) { + clear_bit(STRIPE_INSYNC, &sh->state); + update_q = 1; + } + } + if (update_p || update_q) { + conf->mddev->resync_mismatches += STRIPE_SECTORS; + if (test_bit(MD_RECOVERY_CHECK, &conf->mddev->recovery)) + /* don't try to repair!! */ + update_p = update_q = 0; } -#else - compute_block_2(sh, failed_num[0], failed_num[1]); - uptodate += failed_needupdate[0] + failed_needupdate[1]; -#endif - if (uptodate != disks) - BUG(); + /* now write out any block on a failed drive, + * or P or Q if they need it + */ - PRINTK("Marking for sync stripe %llu blocks %d,%d\n", - (unsigned long long)sh->sector, failed_num[0], failed_num[1]); + if (failed == 2) { + dev = &sh->dev[failed_num[1]]; + locked++; + set_bit(R5_LOCKED, &dev->flags); + set_bit(R5_Wantwrite, &dev->flags); + set_bit(R5_Syncio, &dev->flags); + } + if (failed >= 1) { + dev = &sh->dev[failed_num[0]]; + locked++; + set_bit(R5_LOCKED, &dev->flags); + set_bit(R5_Wantwrite, &dev->flags); + set_bit(R5_Syncio, &dev->flags); + } - /**** FIX: Should we really do both of these unconditionally? ****/ - adev = &sh->dev[failed_num[0]]; - locked += !test_bit(R5_LOCKED, &adev->flags); - set_bit(R5_LOCKED, &adev->flags); - set_bit(R5_Wantwrite, &adev->flags); - bdev = &sh->dev[failed_num[1]]; - locked += !test_bit(R5_LOCKED, &bdev->flags); - set_bit(R5_LOCKED, &bdev->flags); + if (update_p) { + dev = &sh->dev[pd_idx]; + locked ++; + set_bit(R5_LOCKED, &dev->flags); + set_bit(R5_Wantwrite, &dev->flags); + set_bit(R5_Syncio, &dev->flags); + } + if (update_q) { + dev = &sh->dev[qd_idx]; + locked++; + set_bit(R5_LOCKED, &dev->flags); + set_bit(R5_Wantwrite, &dev->flags); + set_bit(R5_Syncio, &dev->flags); + } clear_bit(STRIPE_DEGRADED, &sh->state); - set_bit(R5_Wantwrite, &bdev->flags); set_bit(STRIPE_INSYNC, &sh->state); - set_bit(R5_Syncio, &adev->flags); - set_bit(R5_Syncio, &bdev->flags); } } + if (syncing && locked == 0 && test_bit(STRIPE_INSYNC, &sh->state)) { md_done_sync(conf->mddev, STRIPE_SECTORS,1); clear_bit(STRIPE_SYNCING, &sh->state); @@ -1664,7 +1687,7 @@ static int make_request (request_queue_t *q, struct bio * bi) } finish_wait(&conf->wait_for_overlap, &w); raid6_plug_device(conf); - handle_stripe(sh); + handle_stripe(sh, NULL); release_stripe(sh); } else { /* cannot get stripe for read-ahead, just give-up */ @@ -1728,6 +1751,7 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i return rv; } if (!bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) && + !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) && !conf->fullsync && sync_blocks >= STRIPE_SECTORS) { /* we can skip this block, and probably more */ sync_blocks /= STRIPE_SECTORS; @@ -1765,7 +1789,7 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i clear_bit(STRIPE_INSYNC, &sh->state); spin_unlock(&sh->lock); - handle_stripe(sh); + handle_stripe(sh, NULL); release_stripe(sh); return STRIPE_SECTORS; @@ -1821,7 +1845,7 @@ static void raid6d (mddev_t *mddev) spin_unlock_irq(&conf->device_lock); handled++; - handle_stripe(sh); + handle_stripe(sh, conf->spare_page); release_stripe(sh); spin_lock_irq(&conf->device_lock); @@ -1860,6 +1884,10 @@ static int run(mddev_t *mddev) goto abort; memset(conf->stripe_hashtbl, 0, HASH_PAGES * PAGE_SIZE); + conf->spare_page = alloc_page(GFP_KERNEL); + if (!conf->spare_page) + goto abort; + spin_lock_init(&conf->device_lock); init_waitqueue_head(&conf->wait_for_stripe); init_waitqueue_head(&conf->wait_for_overlap); @@ -1996,6 +2024,8 @@ static int run(mddev_t *mddev) abort: if (conf) { print_raid6_conf(conf); + if (conf->spare_page) + page_cache_release(conf->spare_page); if (conf->stripe_hashtbl) free_pages((unsigned long) conf->stripe_hashtbl, HASH_PAGES_ORDER); diff --git a/include/linux/raid/raid5.h b/include/linux/raid/raid5.h index f025ba6fb14c..e9c1c0d4f90b 100644 --- a/include/linux/raid/raid5.h +++ b/include/linux/raid/raid5.h @@ -228,6 +228,8 @@ struct raid5_private_data { * Cleared when a sync completes. */ + struct page *spare_page; /* Used when checking P/Q in raid6 */ + /* * Free stripes pool */ -- cgit v1.2.3 From ddaf22abaa831763e75775e6d4c7693504237997 Mon Sep 17 00:00:00 2001 From: NeilBrown Date: Fri, 6 Jan 2006 00:20:19 -0800 Subject: [PATCH] md: attempt to auto-correct read errors in raid1 On a read-error we suspend the array, then synchronously read the block from other arrays until we find one where we can read it. Then we try writing the good data back everywhere and make sure it works. If any write or subsequent read fails, only then do we fail the device out of the array. To be able to suspend the array, we need to also keep track of how many requests are queued for handling by raid1d. Signed-off-by: Neil Brown Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- drivers/md/md.c | 1 + drivers/md/raid1.c | 115 +++++++++++++++++++++++++++++++++++++++++---- include/linux/raid/raid1.h | 3 ++ 3 files changed, 109 insertions(+), 10 deletions(-) (limited to 'include/linux/raid') diff --git a/drivers/md/md.c b/drivers/md/md.c index 64e7da3701a5..1364a1c97e6f 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -461,6 +461,7 @@ int sync_page_io(struct block_device *bdev, sector_t sector, int size, bio_put(bio); return ret; } +EXPORT_SYMBOL(sync_page_io); static int read_disk_sb(mdk_rdev_t * rdev, int size) { diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c index c618015f07f6..b3856db8d6c2 100644 --- a/drivers/md/raid1.c +++ b/drivers/md/raid1.c @@ -191,6 +191,7 @@ static void reschedule_retry(r1bio_t *r1_bio) spin_lock_irqsave(&conf->device_lock, flags); list_add(&r1_bio->retry_list, &conf->retry_list); + conf->nr_queued ++; spin_unlock_irqrestore(&conf->device_lock, flags); wake_up(&conf->wait_barrier); @@ -245,9 +246,9 @@ static int raid1_end_read_request(struct bio *bio, unsigned int bytes_done, int /* * this branch is our 'one mirror IO has finished' event handler: */ - if (!uptodate) - md_error(r1_bio->mddev, conf->mirrors[mirror].rdev); - else + update_head_pos(mirror, r1_bio); + + if (uptodate || conf->working_disks <= 1) { /* * Set R1BIO_Uptodate in our master bio, so that * we will return a good error code for to the higher @@ -259,14 +260,8 @@ static int raid1_end_read_request(struct bio *bio, unsigned int bytes_done, int */ set_bit(R1BIO_Uptodate, &r1_bio->state); - update_head_pos(mirror, r1_bio); - - /* - * we have only one bio on the read side - */ - if (uptodate) raid_end_bio_io(r1_bio); - else { + } else { /* * oops, read error: */ @@ -653,6 +648,32 @@ static void allow_barrier(conf_t *conf) wake_up(&conf->wait_barrier); } +static void freeze_array(conf_t *conf) +{ + /* stop syncio and normal IO and wait for everything to + * go quite. + * We increment barrier and nr_waiting, and then + * wait until barrier+nr_pending match nr_queued+2 + */ + spin_lock_irq(&conf->resync_lock); + conf->barrier++; + conf->nr_waiting++; + wait_event_lock_irq(conf->wait_barrier, + conf->barrier+conf->nr_pending == conf->nr_queued+2, + conf->resync_lock, + raid1_unplug(conf->mddev->queue)); + spin_unlock_irq(&conf->resync_lock); +} +static void unfreeze_array(conf_t *conf) +{ + /* reverse the effect of the freeze */ + spin_lock_irq(&conf->resync_lock); + conf->barrier--; + conf->nr_waiting--; + wake_up(&conf->wait_barrier); + spin_unlock_irq(&conf->resync_lock); +} + /* duplicate the data pages for behind I/O */ static struct page **alloc_behind_pages(struct bio *bio) @@ -1196,6 +1217,7 @@ static void raid1d(mddev_t *mddev) break; r1_bio = list_entry(head->prev, r1bio_t, retry_list); list_del(head->prev); + conf->nr_queued--; spin_unlock_irqrestore(&conf->device_lock, flags); mddev = r1_bio->mddev; @@ -1235,6 +1257,74 @@ static void raid1d(mddev_t *mddev) } } else { int disk; + + /* we got a read error. Maybe the drive is bad. Maybe just + * the block and we can fix it. + * We freeze all other IO, and try reading the block from + * other devices. When we find one, we re-write + * and check it that fixes the read error. + * This is all done synchronously while the array is + * frozen + */ + sector_t sect = r1_bio->sector; + int sectors = r1_bio->sectors; + freeze_array(conf); + while(sectors) { + int s = sectors; + int d = r1_bio->read_disk; + int success = 0; + + if (s > (PAGE_SIZE>>9)) + s = PAGE_SIZE >> 9; + + do { + rdev = conf->mirrors[d].rdev; + if (rdev && + test_bit(In_sync, &rdev->flags) && + sync_page_io(rdev->bdev, + sect + rdev->data_offset, + s<<9, + conf->tmppage, READ)) + success = 1; + else { + d++; + if (d == conf->raid_disks) + d = 0; + } + } while (!success && d != r1_bio->read_disk); + + if (success) { + /* write it back and re-read */ + while (d != r1_bio->read_disk) { + if (d==0) + d = conf->raid_disks; + d--; + rdev = conf->mirrors[d].rdev; + if (rdev && + test_bit(In_sync, &rdev->flags)) { + if (sync_page_io(rdev->bdev, + sect + rdev->data_offset, + s<<9, conf->tmppage, WRITE) == 0 || + sync_page_io(rdev->bdev, + sect + rdev->data_offset, + s<<9, conf->tmppage, READ) == 0) { + /* Well, this device is dead */ + md_error(mddev, rdev); + } + } + } + } else { + /* Cannot read from anywhere -- bye bye array */ + md_error(mddev, conf->mirrors[r1_bio->read_disk].rdev); + break; + } + sectors -= s; + sect += s; + } + + + unfreeze_array(conf); + bio = r1_bio->bios[r1_bio->read_disk]; if ((disk=read_balance(conf, r1_bio)) == -1) { printk(KERN_ALERT "raid1: %s: unrecoverable I/O" @@ -1529,6 +1619,10 @@ static int run(mddev_t *mddev) memset(conf->mirrors, 0, sizeof(struct mirror_info)*mddev->raid_disks); + conf->tmppage = alloc_page(GFP_KERNEL); + if (!conf->tmppage) + goto out_no_mem; + conf->poolinfo = kmalloc(sizeof(*conf->poolinfo), GFP_KERNEL); if (!conf->poolinfo) goto out_no_mem; @@ -1635,6 +1729,7 @@ out_free_conf: if (conf->r1bio_pool) mempool_destroy(conf->r1bio_pool); kfree(conf->mirrors); + __free_page(conf->tmppage); kfree(conf->poolinfo); kfree(conf); mddev->private = NULL; diff --git a/include/linux/raid/raid1.h b/include/linux/raid/raid1.h index c55674252533..cbe4238d3f9f 100644 --- a/include/linux/raid/raid1.h +++ b/include/linux/raid/raid1.h @@ -46,6 +46,7 @@ struct r1_private_data_s { spinlock_t resync_lock; int nr_pending; int nr_waiting; + int nr_queued; int barrier; sector_t next_resync; int fullsync; /* set to 1 if a full sync is needed, @@ -57,6 +58,8 @@ struct r1_private_data_s { struct pool_info *poolinfo; + struct page *tmppage; + mempool_t *r1bio_pool; mempool_t *r1buf_pool; }; -- cgit v1.2.3 From cf30a473a02901fe4db37abc0b0fa26dd5ba3f72 Mon Sep 17 00:00:00 2001 From: NeilBrown Date: Fri, 6 Jan 2006 00:20:23 -0800 Subject: [PATCH] md: handle errors when read-only Signed-off-by: Neil Brown Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- drivers/md/raid1.c | 18 +++++++++++------- include/linux/raid/raid1.h | 7 +++++++ 2 files changed, 18 insertions(+), 7 deletions(-) (limited to 'include/linux/raid') diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c index 14a8fe0349c7..a8bc93d6ff63 100644 --- a/drivers/md/raid1.c +++ b/drivers/md/raid1.c @@ -154,7 +154,7 @@ static void put_all_bios(conf_t *conf, r1bio_t *r1_bio) for (i = 0; i < conf->raid_disks; i++) { struct bio **bio = r1_bio->bios + i; - if (*bio) + if (*bio && *bio != IO_BLOCKED) bio_put(*bio); *bio = NULL; } @@ -419,11 +419,13 @@ static int read_balance(conf_t *conf, r1bio_t *r1_bio) new_disk = 0; for (rdev = rcu_dereference(conf->mirrors[new_disk].rdev); + r1_bio->bios[new_disk] == IO_BLOCKED || !rdev || !test_bit(In_sync, &rdev->flags) || test_bit(WriteMostly, &rdev->flags); rdev = rcu_dereference(conf->mirrors[++new_disk].rdev)) { - if (rdev && test_bit(In_sync, &rdev->flags)) + if (rdev && test_bit(In_sync, &rdev->flags) && + r1_bio->bios[new_disk] != IO_BLOCKED) wonly_disk = new_disk; if (new_disk == conf->raid_disks - 1) { @@ -437,11 +439,13 @@ static int read_balance(conf_t *conf, r1bio_t *r1_bio) /* make sure the disk is operational */ for (rdev = rcu_dereference(conf->mirrors[new_disk].rdev); + r1_bio->bios[new_disk] == IO_BLOCKED || !rdev || !test_bit(In_sync, &rdev->flags) || test_bit(WriteMostly, &rdev->flags); rdev = rcu_dereference(conf->mirrors[new_disk].rdev)) { - if (rdev && test_bit(In_sync, &rdev->flags)) + if (rdev && test_bit(In_sync, &rdev->flags) && + r1_bio->bios[new_disk] != IO_BLOCKED) wonly_disk = new_disk; if (new_disk <= 0) @@ -478,7 +482,7 @@ static int read_balance(conf_t *conf, r1bio_t *r1_bio) rdev = rcu_dereference(conf->mirrors[disk].rdev); - if (!rdev || + if (!rdev || r1_bio->bios[disk] == IO_BLOCKED || !test_bit(In_sync, &rdev->flags) || test_bit(WriteMostly, &rdev->flags)) continue; @@ -1335,7 +1339,7 @@ static void raid1d(mddev_t *mddev) sector_t sect = r1_bio->sector; int sectors = r1_bio->sectors; freeze_array(conf); - while(sectors) { + if (mddev->ro == 0) while(sectors) { int s = sectors; int d = r1_bio->read_disk; int success = 0; @@ -1388,7 +1392,6 @@ static void raid1d(mddev_t *mddev) sect += s; } - unfreeze_array(conf); bio = r1_bio->bios[r1_bio->read_disk]; @@ -1399,7 +1402,8 @@ static void raid1d(mddev_t *mddev) (unsigned long long)r1_bio->sector); raid_end_bio_io(r1_bio); } else { - r1_bio->bios[r1_bio->read_disk] = NULL; + r1_bio->bios[r1_bio->read_disk] = + mddev->ro ? IO_BLOCKED : NULL; r1_bio->read_disk = disk; bio_put(bio); bio = bio_clone(r1_bio->master_bio, GFP_NOIO); diff --git a/include/linux/raid/raid1.h b/include/linux/raid/raid1.h index cbe4238d3f9f..9d5494aaac0f 100644 --- a/include/linux/raid/raid1.h +++ b/include/linux/raid/raid1.h @@ -109,6 +109,13 @@ struct r1bio_s { /* DO NOT PUT ANY NEW FIELDS HERE - bios array is contiguously alloced*/ }; +/* when we get a read error on a read-only array, we redirect to another + * device without failing the first device, or trying to over-write to + * correct the read error. To keep track of bad blocks on a per-bio + * level, we store IO_BLOCKED in the appropriate 'bios' pointer + */ +#define IO_BLOCKED ((struct bio*)1) + /* bits for r1bio.state */ #define R1BIO_Uptodate 0 #define R1BIO_IsSync 1 -- cgit v1.2.3 From 9910f16af35419a5382fa7850eecc220103036fa Mon Sep 17 00:00:00 2001 From: NeilBrown Date: Fri, 6 Jan 2006 00:20:24 -0800 Subject: [PATCH] md: fix up some rdev rcu locking in raid5/6 There is this "FIXME" comment with a typo in it!! that been annoying me for days, so I just had to remove it. conf->disks[i].rdev should only be accessed if - we know we hold a reference or - the mddev->reconfig_sem is down or - we have a rcu_readlock handle_stripe was referencing rdev in three places without any of these. For the first two, get an rcu_readlock. For the last, the same access (md_sync_acct call) is made a little later after the rdev has been claimed under and rcu_readlock, if R5_Syncio is set. So just use that access... However R5_Syncio isn't really needed as the 'syncing' variable contains the same information. So use that instead. Issues, comment, and fix are identical in raid5 and raid6. Signed-off-by: Neil Brown Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- drivers/md/raid5.c | 16 ++++++++-------- drivers/md/raid6main.c | 19 ++++++++----------- include/linux/raid/raid5.h | 1 - 3 files changed, 16 insertions(+), 20 deletions(-) (limited to 'include/linux/raid') diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 0d016a844ec6..0222ba1a6d35 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -961,11 +961,11 @@ static void handle_stripe(struct stripe_head *sh) syncing = test_bit(STRIPE_SYNCING, &sh->state); /* Now to look around and see what can be done */ + rcu_read_lock(); for (i=disks; i--; ) { mdk_rdev_t *rdev; dev = &sh->dev[i]; clear_bit(R5_Insync, &dev->flags); - clear_bit(R5_Syncio, &dev->flags); PRINTK("check %d: state 0x%lx read %p write %p written %p\n", i, dev->flags, dev->toread, dev->towrite, dev->written); @@ -1004,7 +1004,7 @@ static void handle_stripe(struct stripe_head *sh) non_overwrite++; } if (dev->written) written++; - rdev = conf->disks[i].rdev; /* FIXME, should I be looking rdev */ + rdev = rcu_dereference(conf->disks[i].rdev); if (!rdev || !test_bit(In_sync, &rdev->flags)) { /* The ReadError flag will just be confusing now */ clear_bit(R5_ReadError, &dev->flags); @@ -1017,6 +1017,7 @@ static void handle_stripe(struct stripe_head *sh) } else set_bit(R5_Insync, &dev->flags); } + rcu_read_unlock(); PRINTK("locked=%d uptodate=%d to_read=%d" " to_write=%d failed=%d failed_num=%d\n", locked, uptodate, to_read, to_write, failed, failed_num); @@ -1028,10 +1029,13 @@ static void handle_stripe(struct stripe_head *sh) int bitmap_end = 0; if (test_bit(R5_ReadError, &sh->dev[i].flags)) { - mdk_rdev_t *rdev = conf->disks[i].rdev; + mdk_rdev_t *rdev; + rcu_read_lock(); + rdev = rcu_dereference(conf->disks[i].rdev); if (rdev && test_bit(In_sync, &rdev->flags)) /* multiple read failures in one stripe */ md_error(conf->mddev, rdev); + rcu_read_unlock(); } spin_lock_irq(&conf->device_lock); @@ -1180,9 +1184,6 @@ static void handle_stripe(struct stripe_head *sh) locked++; PRINTK("Reading block %d (sync=%d)\n", i, syncing); - if (syncing) - md_sync_acct(conf->disks[i].rdev->bdev, - STRIPE_SECTORS); } } } @@ -1326,7 +1327,6 @@ static void handle_stripe(struct stripe_head *sh) clear_bit(STRIPE_DEGRADED, &sh->state); locked++; set_bit(STRIPE_INSYNC, &sh->state); - set_bit(R5_Syncio, &dev->flags); } } if (syncing && locked == 0 && test_bit(STRIPE_INSYNC, &sh->state)) { @@ -1392,7 +1392,7 @@ static void handle_stripe(struct stripe_head *sh) rcu_read_unlock(); if (rdev) { - if (test_bit(R5_Syncio, &sh->dev[i].flags)) + if (syncing) md_sync_acct(rdev->bdev, STRIPE_SECTORS); bi->bi_bdev = rdev->bdev; diff --git a/drivers/md/raid6main.c b/drivers/md/raid6main.c index 7a51553d8be5..b5b7a8d0b165 100644 --- a/drivers/md/raid6main.c +++ b/drivers/md/raid6main.c @@ -1060,11 +1060,11 @@ static void handle_stripe(struct stripe_head *sh, struct page *tmp_page) syncing = test_bit(STRIPE_SYNCING, &sh->state); /* Now to look around and see what can be done */ + rcu_read_lock(); for (i=disks; i--; ) { mdk_rdev_t *rdev; dev = &sh->dev[i]; clear_bit(R5_Insync, &dev->flags); - clear_bit(R5_Syncio, &dev->flags); PRINTK("check %d: state 0x%lx read %p write %p written %p\n", i, dev->flags, dev->toread, dev->towrite, dev->written); @@ -1103,7 +1103,7 @@ static void handle_stripe(struct stripe_head *sh, struct page *tmp_page) non_overwrite++; } if (dev->written) written++; - rdev = conf->disks[i].rdev; /* FIXME, should I be looking rdev */ + rdev = rcu_dereference(conf->disks[i].rdev); if (!rdev || !test_bit(In_sync, &rdev->flags)) { /* The ReadError flag will just be confusing now */ clear_bit(R5_ReadError, &dev->flags); @@ -1117,6 +1117,7 @@ static void handle_stripe(struct stripe_head *sh, struct page *tmp_page) } else set_bit(R5_Insync, &dev->flags); } + rcu_read_unlock(); PRINTK("locked=%d uptodate=%d to_read=%d" " to_write=%d failed=%d failed_num=%d,%d\n", locked, uptodate, to_read, to_write, failed, @@ -1129,10 +1130,13 @@ static void handle_stripe(struct stripe_head *sh, struct page *tmp_page) int bitmap_end = 0; if (test_bit(R5_ReadError, &sh->dev[i].flags)) { - mdk_rdev_t *rdev = conf->disks[i].rdev; + mdk_rdev_t *rdev; + rcu_read_lock(); + rdev = rcu_dereference(conf->disks[i].rdev); if (rdev && test_bit(In_sync, &rdev->flags)) /* multiple read failures in one stripe */ md_error(conf->mddev, rdev); + rcu_read_unlock(); } spin_lock_irq(&conf->device_lock); @@ -1307,9 +1311,6 @@ static void handle_stripe(struct stripe_head *sh, struct page *tmp_page) locked++; PRINTK("Reading block %d (sync=%d)\n", i, syncing); - if (syncing) - md_sync_acct(conf->disks[i].rdev->bdev, - STRIPE_SECTORS); } } } @@ -1463,14 +1464,12 @@ static void handle_stripe(struct stripe_head *sh, struct page *tmp_page) locked++; set_bit(R5_LOCKED, &dev->flags); set_bit(R5_Wantwrite, &dev->flags); - set_bit(R5_Syncio, &dev->flags); } if (failed >= 1) { dev = &sh->dev[failed_num[0]]; locked++; set_bit(R5_LOCKED, &dev->flags); set_bit(R5_Wantwrite, &dev->flags); - set_bit(R5_Syncio, &dev->flags); } if (update_p) { @@ -1478,14 +1477,12 @@ static void handle_stripe(struct stripe_head *sh, struct page *tmp_page) locked ++; set_bit(R5_LOCKED, &dev->flags); set_bit(R5_Wantwrite, &dev->flags); - set_bit(R5_Syncio, &dev->flags); } if (update_q) { dev = &sh->dev[qd_idx]; locked++; set_bit(R5_LOCKED, &dev->flags); set_bit(R5_Wantwrite, &dev->flags); - set_bit(R5_Syncio, &dev->flags); } clear_bit(STRIPE_DEGRADED, &sh->state); @@ -1557,7 +1554,7 @@ static void handle_stripe(struct stripe_head *sh, struct page *tmp_page) rcu_read_unlock(); if (rdev) { - if (test_bit(R5_Syncio, &sh->dev[i].flags)) + if (syncing) md_sync_acct(rdev->bdev, STRIPE_SECTORS); bi->bi_bdev = rdev->bdev; diff --git a/include/linux/raid/raid5.h b/include/linux/raid/raid5.h index e9c1c0d4f90b..28fcd7533ac4 100644 --- a/include/linux/raid/raid5.h +++ b/include/linux/raid/raid5.h @@ -152,7 +152,6 @@ struct stripe_head { #define R5_Insync 3 /* rdev && rdev->in_sync at start */ #define R5_Wantread 4 /* want to schedule a read */ #define R5_Wantwrite 5 -#define R5_Syncio 6 /* this io need to be accounted as resync io */ #define R5_Overlap 7 /* There is a pending overlapping request on this block */ #define R5_ReadError 8 /* seen a read error here recently */ #define R5_ReWrite 9 /* have tried to over-write the readerror */ -- cgit v1.2.3 From 4443ae10ca15d07922ceda622f03db8865fa3d13 Mon Sep 17 00:00:00 2001 From: NeilBrown Date: Fri, 6 Jan 2006 00:20:28 -0800 Subject: [PATCH] md: auto-correct correctable read errors in raid10 Largely just a cross-port from raid1. Signed-off-by: Neil Brown Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- drivers/md/raid10.c | 127 ++++++++++++++++++++++++++++++++++++++------ include/linux/raid/raid10.h | 2 + 2 files changed, 114 insertions(+), 15 deletions(-) (limited to 'include/linux/raid') diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c index 1fa70c34b7d2..64bb4ddc6798 100644 --- a/drivers/md/raid10.c +++ b/drivers/md/raid10.c @@ -209,6 +209,7 @@ static void reschedule_retry(r10bio_t *r10_bio) spin_lock_irqsave(&conf->device_lock, flags); list_add(&r10_bio->retry_list, &conf->retry_list); + conf->nr_queued ++; spin_unlock_irqrestore(&conf->device_lock, flags); md_wakeup_thread(mddev->thread); @@ -254,9 +255,9 @@ static int raid10_end_read_request(struct bio *bio, unsigned int bytes_done, int /* * this branch is our 'one mirror IO has finished' event handler: */ - if (!uptodate) - md_error(r10_bio->mddev, conf->mirrors[dev].rdev); - else + update_head_pos(slot, r10_bio); + + if (uptodate) { /* * Set R10BIO_Uptodate in our master bio, so that * we will return a good error code to the higher @@ -267,15 +268,8 @@ static int raid10_end_read_request(struct bio *bio, unsigned int bytes_done, int * wait for the 'master' bio. */ set_bit(R10BIO_Uptodate, &r10_bio->state); - - update_head_pos(slot, r10_bio); - - /* - * we have only one bio on the read side - */ - if (uptodate) raid_end_bio_io(r10_bio); - else { + } else { /* * oops, read error: */ @@ -714,6 +708,33 @@ static void allow_barrier(conf_t *conf) wake_up(&conf->wait_barrier); } +static void freeze_array(conf_t *conf) +{ + /* stop syncio and normal IO and wait for everything to + * go quite. + * We increment barrier and nr_waiting, and then + * wait until barrier+nr_pending match nr_queued+2 + */ + spin_lock_irq(&conf->resync_lock); + conf->barrier++; + conf->nr_waiting++; + wait_event_lock_irq(conf->wait_barrier, + conf->barrier+conf->nr_pending == conf->nr_queued+2, + conf->resync_lock, + raid10_unplug(conf->mddev->queue)); + spin_unlock_irq(&conf->resync_lock); +} + +static void unfreeze_array(conf_t *conf) +{ + /* reverse the effect of the freeze */ + spin_lock_irq(&conf->resync_lock); + conf->barrier--; + conf->nr_waiting--; + wake_up(&conf->wait_barrier); + spin_unlock_irq(&conf->resync_lock); +} + static int make_request(request_queue_t *q, struct bio * bio) { mddev_t *mddev = q->queuedata; @@ -1338,6 +1359,7 @@ static void raid10d(mddev_t *mddev) break; r10_bio = list_entry(head->prev, r10bio_t, retry_list); list_del(head->prev); + conf->nr_queued--; spin_unlock_irqrestore(&conf->device_lock, flags); mddev = r10_bio->mddev; @@ -1350,6 +1372,78 @@ static void raid10d(mddev_t *mddev) unplug = 1; } else { int mirror; + /* we got a read error. Maybe the drive is bad. Maybe just + * the block and we can fix it. + * We freeze all other IO, and try reading the block from + * other devices. When we find one, we re-write + * and check it that fixes the read error. + * This is all done synchronously while the array is + * frozen. + */ + int sect = 0; /* Offset from r10_bio->sector */ + int sectors = r10_bio->sectors; + freeze_array(conf); + if (mddev->ro == 0) while(sectors) { + int s = sectors; + int sl = r10_bio->read_slot; + int success = 0; + + if (s > (PAGE_SIZE>>9)) + s = PAGE_SIZE >> 9; + + do { + int d = r10_bio->devs[sl].devnum; + rdev = conf->mirrors[d].rdev; + if (rdev && + test_bit(In_sync, &rdev->flags) && + sync_page_io(rdev->bdev, + r10_bio->devs[sl].addr + + sect + rdev->data_offset, + s<<9, + conf->tmppage, READ)) + success = 1; + else { + sl++; + if (sl == conf->copies) + sl = 0; + } + } while (!success && sl != r10_bio->read_slot); + + if (success) { + /* write it back and re-read */ + while (sl != r10_bio->read_slot) { + int d; + if (sl==0) + sl = conf->copies; + sl--; + d = r10_bio->devs[sl].devnum; + rdev = conf->mirrors[d].rdev; + if (rdev && + test_bit(In_sync, &rdev->flags)) { + if (sync_page_io(rdev->bdev, + r10_bio->devs[sl].addr + + sect + rdev->data_offset, + s<<9, conf->tmppage, WRITE) == 0 || + sync_page_io(rdev->bdev, + r10_bio->devs[sl].addr + + sect + rdev->data_offset, + s<<9, conf->tmppage, READ) == 0) { + /* Well, this device is dead */ + md_error(mddev, rdev); + } + } + } + } else { + /* Cannot read from anywhere -- bye bye array */ + md_error(mddev, conf->mirrors[r10_bio->devs[r10_bio->read_slot].devnum].rdev); + break; + } + sectors -= s; + sect += s; + } + + unfreeze_array(conf); + bio = r10_bio->devs[r10_bio->read_slot].bio; r10_bio->devs[r10_bio->read_slot].bio = NULL; bio_put(bio); @@ -1793,22 +1887,24 @@ static int run(mddev_t *mddev) * bookkeeping area. [whatever we allocate in run(), * should be freed in stop()] */ - conf = kmalloc(sizeof(conf_t), GFP_KERNEL); + conf = kzalloc(sizeof(conf_t), GFP_KERNEL); mddev->private = conf; if (!conf) { printk(KERN_ERR "raid10: couldn't allocate memory for %s\n", mdname(mddev)); goto out; } - memset(conf, 0, sizeof(*conf)); - conf->mirrors = kmalloc(sizeof(struct mirror_info)*mddev->raid_disks, + conf->mirrors = kzalloc(sizeof(struct mirror_info)*mddev->raid_disks, GFP_KERNEL); if (!conf->mirrors) { printk(KERN_ERR "raid10: couldn't allocate memory for %s\n", mdname(mddev)); goto out_free_conf; } - memset(conf->mirrors, 0, sizeof(struct mirror_info)*mddev->raid_disks); + + conf->tmppage = alloc_page(GFP_KERNEL); + if (!conf->tmppage) + goto out_free_conf; conf->near_copies = nc; conf->far_copies = fc; @@ -1918,6 +2014,7 @@ static int run(mddev_t *mddev) out_free_conf: if (conf->r10bio_pool) mempool_destroy(conf->r10bio_pool); + put_page(conf->tmppage); kfree(conf->mirrors); kfree(conf); mddev->private = NULL; diff --git a/include/linux/raid/raid10.h b/include/linux/raid/raid10.h index b660cbf628d8..dfa528385e3f 100644 --- a/include/linux/raid/raid10.h +++ b/include/linux/raid/raid10.h @@ -42,6 +42,7 @@ struct r10_private_data_s { spinlock_t resync_lock; int nr_pending; int nr_waiting; + int nr_queued; int barrier; sector_t next_resync; int fullsync; /* set to 1 if a full sync is needed, @@ -53,6 +54,7 @@ struct r10_private_data_s { mempool_t *r10bio_pool; mempool_t *r10buf_pool; + struct page *tmppage; }; typedef struct r10_private_data_s conf_t; -- cgit v1.2.3 From 0eb3ff12aa8a12538ef681dc83f4361636a0699f Mon Sep 17 00:00:00 2001 From: NeilBrown Date: Fri, 6 Jan 2006 00:20:29 -0800 Subject: [PATCH] md: raid10 read-error handling - resync and read-only Add in correct read-error handling for resync and read-only situations. When read-only, we don't over-write, so we need to mark the failed drive in the r10_bio so we don't re-try it. During resync, we always read all blocks, so if there is a read error, we simply over-write it with the good block that we found (assuming we found one). Note that the recovery case still isn't handled in an interesting way. There is nothing useful to do for the 2-copies case. If there are 3 or more copies, then we could try reading from one of the non-missing copies, but this is a bit complicated and very rarely would be used, so I'm leaving it for now. Signed-off-by: Neil Brown Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- drivers/md/raid10.c | 56 ++++++++++++++++++++++++++++----------------- include/linux/raid/raid10.h | 7 ++++++ 2 files changed, 42 insertions(+), 21 deletions(-) (limited to 'include/linux/raid') diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c index 64bb4ddc6798..3f8df2ecbae3 100644 --- a/drivers/md/raid10.c +++ b/drivers/md/raid10.c @@ -172,7 +172,7 @@ static void put_all_bios(conf_t *conf, r10bio_t *r10_bio) for (i = 0; i < conf->copies; i++) { struct bio **bio = & r10_bio->devs[i].bio; - if (*bio) + if (*bio && *bio != IO_BLOCKED) bio_put(*bio); *bio = NULL; } @@ -500,6 +500,7 @@ static int read_balance(conf_t *conf, r10bio_t *r10_bio) disk = r10_bio->devs[slot].devnum; while ((rdev = rcu_dereference(conf->mirrors[disk].rdev)) == NULL || + r10_bio->devs[slot].bio == IO_BLOCKED || !test_bit(In_sync, &rdev->flags)) { slot++; if (slot == conf->copies) { @@ -517,6 +518,7 @@ static int read_balance(conf_t *conf, r10bio_t *r10_bio) slot = 0; disk = r10_bio->devs[slot].devnum; while ((rdev=rcu_dereference(conf->mirrors[disk].rdev)) == NULL || + r10_bio->devs[slot].bio == IO_BLOCKED || !test_bit(In_sync, &rdev->flags)) { slot ++; if (slot == conf->copies) { @@ -537,6 +539,7 @@ static int read_balance(conf_t *conf, r10bio_t *r10_bio) if ((rdev=rcu_dereference(conf->mirrors[ndisk].rdev)) == NULL || + r10_bio->devs[nslot].bio == IO_BLOCKED || !test_bit(In_sync, &rdev->flags)) continue; @@ -1104,7 +1107,6 @@ abort: static int end_sync_read(struct bio *bio, unsigned int bytes_done, int error) { - int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private); conf_t *conf = mddev_to_conf(r10_bio->mddev); int i,d; @@ -1119,7 +1121,10 @@ static int end_sync_read(struct bio *bio, unsigned int bytes_done, int error) BUG(); update_head_pos(i, r10_bio); d = r10_bio->devs[i].devnum; - if (!uptodate) + + if (test_bit(BIO_UPTODATE, &bio->bi_flags)) + set_bit(R10BIO_Uptodate, &r10_bio->state); + else if (!test_bit(MD_RECOVERY_SYNC, &conf->mddev->recovery)) md_error(r10_bio->mddev, conf->mirrors[d].rdev); @@ -1209,25 +1214,30 @@ static void sync_request_write(mddev_t *mddev, r10bio_t *r10_bio) fbio = r10_bio->devs[i].bio; /* now find blocks with errors */ - for (i=first+1 ; i < conf->copies ; i++) { - int vcnt, j, d; + for (i=0 ; i < conf->copies ; i++) { + int j, d; + int vcnt = r10_bio->sectors >> (PAGE_SHIFT-9); - if (!test_bit(BIO_UPTODATE, &r10_bio->devs[i].bio->bi_flags)) - continue; - /* We know that the bi_io_vec layout is the same for - * both 'first' and 'i', so we just compare them. - * All vec entries are PAGE_SIZE; - */ tbio = r10_bio->devs[i].bio; - vcnt = r10_bio->sectors >> (PAGE_SHIFT-9); - for (j = 0; j < vcnt; j++) - if (memcmp(page_address(fbio->bi_io_vec[j].bv_page), - page_address(tbio->bi_io_vec[j].bv_page), - PAGE_SIZE)) - break; - if (j == vcnt) + + if (tbio->bi_end_io != end_sync_read) + continue; + if (i == first) continue; - mddev->resync_mismatches += r10_bio->sectors; + if (test_bit(BIO_UPTODATE, &r10_bio->devs[i].bio->bi_flags)) { + /* We know that the bi_io_vec layout is the same for + * both 'first' and 'i', so we just compare them. + * All vec entries are PAGE_SIZE; + */ + for (j = 0; j < vcnt; j++) + if (memcmp(page_address(fbio->bi_io_vec[j].bv_page), + page_address(tbio->bi_io_vec[j].bv_page), + PAGE_SIZE)) + break; + if (j == vcnt) + continue; + mddev->resync_mismatches += r10_bio->sectors; + } if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)) /* Don't fix anything. */ continue; @@ -1308,7 +1318,10 @@ static void recovery_request_write(mddev_t *mddev, r10bio_t *r10_bio) atomic_inc(&conf->mirrors[d].rdev->nr_pending); md_sync_acct(conf->mirrors[d].rdev->bdev, wbio->bi_size >> 9); - generic_make_request(wbio); + if (test_bit(R10BIO_Uptodate, &r10_bio->state)) + generic_make_request(wbio); + else + bio_endio(wbio, wbio->bi_size, -EIO); } @@ -1445,7 +1458,8 @@ static void raid10d(mddev_t *mddev) unfreeze_array(conf); bio = r10_bio->devs[r10_bio->read_slot].bio; - r10_bio->devs[r10_bio->read_slot].bio = NULL; + r10_bio->devs[r10_bio->read_slot].bio = + mddev->ro ? IO_BLOCKED : NULL; bio_put(bio); mirror = read_balance(conf, r10_bio); if (mirror == -1) { diff --git a/include/linux/raid/raid10.h b/include/linux/raid/raid10.h index dfa528385e3f..b1103298a8c2 100644 --- a/include/linux/raid/raid10.h +++ b/include/linux/raid/raid10.h @@ -104,6 +104,13 @@ struct r10bio_s { } devs[0]; }; +/* when we get a read error on a read-only array, we redirect to another + * device without failing the first device, or trying to over-write to + * correct the read error. To keep track of bad blocks on a per-bio + * level, we store IO_BLOCKED in the appropriate 'bios' pointer + */ +#define IO_BLOCKED ((struct bio*)1) + /* bits for r10bio.state */ #define R10BIO_Uptodate 0 #define R10BIO_IsSync 1 -- cgit v1.2.3 From fccddba060f2b4916a30aa27acc3d03b01bb981e Mon Sep 17 00:00:00 2001 From: NeilBrown Date: Fri, 6 Jan 2006 00:20:33 -0800 Subject: [PATCH] md: tidy up raid5/6 hash table code - replace open-coded hash chain with hlist macros - Fix hash-table size at one page - it is already quite generous, so there will never be a need to use multiple pages, so no need for __get_free_pages No functional change. Signed-off-by: Neil Brown Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- drivers/md/raid5.c | 40 ++++++++++++++-------------------------- drivers/md/raid6main.c | 46 +++++++++++++++++----------------------------- include/linux/raid/raid5.h | 4 ++-- 3 files changed, 33 insertions(+), 57 deletions(-) (limited to 'include/linux/raid') diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 9fc50487e2ed..6e4db95cebb1 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -35,12 +35,10 @@ #define STRIPE_SHIFT (PAGE_SHIFT - 9) #define STRIPE_SECTORS (STRIPE_SIZE>>9) #define IO_THRESHOLD 1 -#define HASH_PAGES 1 -#define HASH_PAGES_ORDER 0 -#define NR_HASH (HASH_PAGES * PAGE_SIZE / sizeof(struct stripe_head *)) +#define NR_HASH (PAGE_SIZE / sizeof(struct hlist_head)) #define HASH_MASK (NR_HASH - 1) -#define stripe_hash(conf, sect) ((conf)->stripe_hashtbl[((sect) >> STRIPE_SHIFT) & HASH_MASK]) +#define stripe_hash(conf, sect) (&((conf)->stripe_hashtbl[((sect) >> STRIPE_SHIFT) & HASH_MASK])) /* bio's attached to a stripe+device for I/O are linked together in bi_sector * order without overlap. There may be several bio's per stripe+device, and @@ -113,29 +111,21 @@ static void release_stripe(struct stripe_head *sh) spin_unlock_irqrestore(&conf->device_lock, flags); } -static void remove_hash(struct stripe_head *sh) +static inline void remove_hash(struct stripe_head *sh) { PRINTK("remove_hash(), stripe %llu\n", (unsigned long long)sh->sector); - if (sh->hash_pprev) { - if (sh->hash_next) - sh->hash_next->hash_pprev = sh->hash_pprev; - *sh->hash_pprev = sh->hash_next; - sh->hash_pprev = NULL; - } + hlist_del_init(&sh->hash); } -static __inline__ void insert_hash(raid5_conf_t *conf, struct stripe_head *sh) +static inline void insert_hash(raid5_conf_t *conf, struct stripe_head *sh) { - struct stripe_head **shp = &stripe_hash(conf, sh->sector); + struct hlist_head *hp = stripe_hash(conf, sh->sector); PRINTK("insert_hash(), stripe %llu\n", (unsigned long long)sh->sector); CHECK_DEVLOCK(); - if ((sh->hash_next = *shp) != NULL) - (*shp)->hash_pprev = &sh->hash_next; - *shp = sh; - sh->hash_pprev = shp; + hlist_add_head(&sh->hash, hp); } @@ -228,10 +218,11 @@ static inline void init_stripe(struct stripe_head *sh, sector_t sector, int pd_i static struct stripe_head *__find_stripe(raid5_conf_t *conf, sector_t sector) { struct stripe_head *sh; + struct hlist_node *hn; CHECK_DEVLOCK(); PRINTK("__find_stripe, sector %llu\n", (unsigned long long)sector); - for (sh = stripe_hash(conf, sector); sh; sh = sh->hash_next) + hlist_for_each_entry(sh, hn, stripe_hash(conf, sector), hash) if (sh->sector == sector) return sh; PRINTK("__stripe %llu not in cache\n", (unsigned long long)sector); @@ -1835,9 +1826,8 @@ static int run(mddev_t *mddev) conf->mddev = mddev; - if ((conf->stripe_hashtbl = (struct stripe_head **) __get_free_pages(GFP_ATOMIC, HASH_PAGES_ORDER)) == NULL) + if ((conf->stripe_hashtbl = kzalloc(PAGE_SIZE, GFP_KERNEL)) == NULL) goto abort; - memset(conf->stripe_hashtbl, 0, HASH_PAGES * PAGE_SIZE); spin_lock_init(&conf->device_lock); init_waitqueue_head(&conf->wait_for_stripe); @@ -1972,9 +1962,7 @@ static int run(mddev_t *mddev) abort: if (conf) { print_raid5_conf(conf); - if (conf->stripe_hashtbl) - free_pages((unsigned long) conf->stripe_hashtbl, - HASH_PAGES_ORDER); + kfree(conf->stripe_hashtbl); kfree(conf); } mddev->private = NULL; @@ -1991,7 +1979,7 @@ static int stop(mddev_t *mddev) md_unregister_thread(mddev->thread); mddev->thread = NULL; shrink_stripes(conf); - free_pages((unsigned long) conf->stripe_hashtbl, HASH_PAGES_ORDER); + kfree(conf->stripe_hashtbl); blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/ sysfs_remove_group(&mddev->kobj, &raid5_attrs_group); kfree(conf); @@ -2019,12 +2007,12 @@ static void print_sh (struct stripe_head *sh) static void printall (raid5_conf_t *conf) { struct stripe_head *sh; + struct hlist_node *hn; int i; spin_lock_irq(&conf->device_lock); for (i = 0; i < NR_HASH; i++) { - sh = conf->stripe_hashtbl[i]; - for (; sh; sh = sh->hash_next) { + hlist_for_each_entry(sh, hn, &conf->stripe_hashtbl[i], hash) { if (sh->raid_conf != conf) continue; print_sh(sh); diff --git a/drivers/md/raid6main.c b/drivers/md/raid6main.c index 4062fc16ac2b..79b5244f44f4 100644 --- a/drivers/md/raid6main.c +++ b/drivers/md/raid6main.c @@ -40,12 +40,10 @@ #define STRIPE_SHIFT (PAGE_SHIFT - 9) #define STRIPE_SECTORS (STRIPE_SIZE>>9) #define IO_THRESHOLD 1 -#define HASH_PAGES 1 -#define HASH_PAGES_ORDER 0 -#define NR_HASH (HASH_PAGES * PAGE_SIZE / sizeof(struct stripe_head *)) +#define NR_HASH (PAGE_SIZE / sizeof(struct hlist_head)) #define HASH_MASK (NR_HASH - 1) -#define stripe_hash(conf, sect) ((conf)->stripe_hashtbl[((sect) >> STRIPE_SHIFT) & HASH_MASK]) +#define stripe_hash(conf, sect) (&((conf)->stripe_hashtbl[((sect) >> STRIPE_SHIFT) & HASH_MASK])) /* bio's attached to a stripe+device for I/O are linked together in bi_sector * order without overlap. There may be several bio's per stripe+device, and @@ -132,29 +130,21 @@ static void release_stripe(struct stripe_head *sh) spin_unlock_irqrestore(&conf->device_lock, flags); } -static void remove_hash(struct stripe_head *sh) +static inline void remove_hash(struct stripe_head *sh) { PRINTK("remove_hash(), stripe %llu\n", (unsigned long long)sh->sector); - if (sh->hash_pprev) { - if (sh->hash_next) - sh->hash_next->hash_pprev = sh->hash_pprev; - *sh->hash_pprev = sh->hash_next; - sh->hash_pprev = NULL; - } + hlist_del_init(&sh->hash); } -static __inline__ void insert_hash(raid6_conf_t *conf, struct stripe_head *sh) +static inline void insert_hash(raid6_conf_t *conf, struct stripe_head *sh) { - struct stripe_head **shp = &stripe_hash(conf, sh->sector); + struct hlist_head *hp = stripe_hash(conf, sh->sector); PRINTK("insert_hash(), stripe %llu\n", (unsigned long long)sh->sector); CHECK_DEVLOCK(); - if ((sh->hash_next = *shp) != NULL) - (*shp)->hash_pprev = &sh->hash_next; - *shp = sh; - sh->hash_pprev = shp; + hlist_add_head(&sh->hash, hp); } @@ -247,10 +237,11 @@ static inline void init_stripe(struct stripe_head *sh, sector_t sector, int pd_i static struct stripe_head *__find_stripe(raid6_conf_t *conf, sector_t sector) { struct stripe_head *sh; + struct hlist_node *hn; CHECK_DEVLOCK(); PRINTK("__find_stripe, sector %llu\n", (unsigned long long)sector); - for (sh = stripe_hash(conf, sector); sh; sh = sh->hash_next) + hlist_for_each_entry (sh, hn, stripe_hash(conf, sector), hash) if (sh->sector == sector) return sh; PRINTK("__stripe %llu not in cache\n", (unsigned long long)sector); @@ -1931,17 +1922,15 @@ static int run(mddev_t *mddev) return -EIO; } - mddev->private = kmalloc (sizeof (raid6_conf_t) - + mddev->raid_disks * sizeof(struct disk_info), - GFP_KERNEL); + mddev->private = kzalloc(sizeof (raid6_conf_t) + + mddev->raid_disks * sizeof(struct disk_info), + GFP_KERNEL); if ((conf = mddev->private) == NULL) goto abort; - memset (conf, 0, sizeof (*conf) + mddev->raid_disks * sizeof(struct disk_info) ); conf->mddev = mddev; - if ((conf->stripe_hashtbl = (struct stripe_head **) __get_free_pages(GFP_ATOMIC, HASH_PAGES_ORDER)) == NULL) + if ((conf->stripe_hashtbl = kzalloc(PAGE_SIZE, GFP_KERNEL)) == NULL) goto abort; - memset(conf->stripe_hashtbl, 0, HASH_PAGES * PAGE_SIZE); conf->spare_page = alloc_page(GFP_KERNEL); if (!conf->spare_page) @@ -2085,9 +2074,7 @@ abort: print_raid6_conf(conf); if (conf->spare_page) put_page(conf->spare_page); - if (conf->stripe_hashtbl) - free_pages((unsigned long) conf->stripe_hashtbl, - HASH_PAGES_ORDER); + kfree(conf->stripe_hashtbl); kfree(conf); } mddev->private = NULL; @@ -2104,7 +2091,7 @@ static int stop (mddev_t *mddev) md_unregister_thread(mddev->thread); mddev->thread = NULL; shrink_stripes(conf); - free_pages((unsigned long) conf->stripe_hashtbl, HASH_PAGES_ORDER); + kfree(conf->stripe_hashtbl); blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/ kfree(conf); mddev->private = NULL; @@ -2131,12 +2118,13 @@ static void print_sh (struct seq_file *seq, struct stripe_head *sh) static void printall (struct seq_file *seq, raid6_conf_t *conf) { struct stripe_head *sh; + struct hlist_node *hn; int i; spin_lock_irq(&conf->device_lock); for (i = 0; i < NR_HASH; i++) { sh = conf->stripe_hashtbl[i]; - for (; sh; sh = sh->hash_next) { + hlist_for_each_entry(sh, hn, &conf->stripe_hashtbl[i], hash) { if (sh->raid_conf != conf) continue; print_sh(seq, sh); diff --git a/include/linux/raid/raid5.h b/include/linux/raid/raid5.h index 28fcd7533ac4..394da8207b34 100644 --- a/include/linux/raid/raid5.h +++ b/include/linux/raid/raid5.h @@ -126,7 +126,7 @@ */ struct stripe_head { - struct stripe_head *hash_next, **hash_pprev; /* hash pointers */ + struct hlist_node hash; struct list_head lru; /* inactive_list or handle_list */ struct raid5_private_data *raid_conf; sector_t sector; /* sector of this row */ @@ -204,7 +204,7 @@ struct disk_info { }; struct raid5_private_data { - struct stripe_head **stripe_hashtbl; + struct hlist_head *stripe_hashtbl; mddev_t *mddev; struct disk_info *spare; int chunk_size, level, algorithm; -- cgit v1.2.3 From 2604b703b6b3db80e3c75ce472a54dfd0b7bf9f4 Mon Sep 17 00:00:00 2001 From: NeilBrown Date: Fri, 6 Jan 2006 00:20:36 -0800 Subject: [PATCH] md: remove personality numbering from md md supports multiple different RAID level, each being implemented by a 'personality' (which is often in a separate module). These personalities have fairly artificial 'numbers'. The numbers are use to: 1- provide an index into an array where the various personalities are recorded 2- identify the module (via an alias) which implements are particular personality. Neither of these uses really justify the existence of personality numbers. The array can be replaced by a linked list which is searched (array lookup only happens very rarely). Module identification can be done using an alias based on level rather than 'personality' number. The current 'raid5' modules support two level (4 and 5) but only one personality. This slight awkwardness (which was handled in the mapping from level to personality) can be better handled by allowing raid5 to register 2 personalities. With this change in place, the core md module does not need to have an exhaustive list of all possible personalities, so other personalities can be added independently. This patch also moves the check for chunksize being non-zero into the ->run routines for the personalities that need it, rather than having it in core-md. This has a side effect of allowing 'faulty' and 'linear' not to have a chunk-size set. Signed-off-by: Neil Brown Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- drivers/md/faulty.c | 8 +++-- drivers/md/linear.c | 10 +++--- drivers/md/md.c | 79 +++++++++++++++++------------------------------ drivers/md/multipath.c | 11 +++---- drivers/md/raid0.c | 14 ++++++--- drivers/md/raid1.c | 9 +++--- drivers/md/raid10.c | 16 +++++----- drivers/md/raid5.c | 34 +++++++++++++++++--- drivers/md/raid6main.c | 10 +++--- include/linux/raid/md.h | 4 +-- include/linux/raid/md_k.h | 63 ++++++------------------------------- init/do_mounts_md.c | 22 ++++++------- 12 files changed, 125 insertions(+), 155 deletions(-) (limited to 'include/linux/raid') diff --git a/drivers/md/faulty.c b/drivers/md/faulty.c index 0248f8e7eac0..f12e83086897 100644 --- a/drivers/md/faulty.c +++ b/drivers/md/faulty.c @@ -316,9 +316,10 @@ static int stop(mddev_t *mddev) return 0; } -static mdk_personality_t faulty_personality = +static struct mdk_personality faulty_personality = { .name = "faulty", + .level = LEVEL_FAULTY, .owner = THIS_MODULE, .make_request = make_request, .run = run, @@ -329,15 +330,16 @@ static mdk_personality_t faulty_personality = static int __init raid_init(void) { - return register_md_personality(FAULTY, &faulty_personality); + return register_md_personality(&faulty_personality); } static void raid_exit(void) { - unregister_md_personality(FAULTY); + unregister_md_personality(&faulty_personality); } module_init(raid_init); module_exit(raid_exit); MODULE_LICENSE("GPL"); MODULE_ALIAS("md-personality-10"); /* faulty */ +MODULE_ALIAS("md-level--5"); diff --git a/drivers/md/linear.c b/drivers/md/linear.c index f46c98d05b44..79dee8159217 100644 --- a/drivers/md/linear.c +++ b/drivers/md/linear.c @@ -351,9 +351,10 @@ static void linear_status (struct seq_file *seq, mddev_t *mddev) } -static mdk_personality_t linear_personality= +static struct mdk_personality linear_personality = { .name = "linear", + .level = LEVEL_LINEAR, .owner = THIS_MODULE, .make_request = linear_make_request, .run = linear_run, @@ -363,16 +364,17 @@ static mdk_personality_t linear_personality= static int __init linear_init (void) { - return register_md_personality (LINEAR, &linear_personality); + return register_md_personality (&linear_personality); } static void linear_exit (void) { - unregister_md_personality (LINEAR); + unregister_md_personality (&linear_personality); } module_init(linear_init); module_exit(linear_exit); MODULE_LICENSE("GPL"); -MODULE_ALIAS("md-personality-1"); /* LINEAR */ +MODULE_ALIAS("md-personality-1"); /* LINEAR - degrecated*/ +MODULE_ALIAS("md-level--1"); diff --git a/drivers/md/md.c b/drivers/md/md.c index a6a066fc92e3..07f180f95b47 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -68,7 +68,7 @@ static void autostart_arrays (int part); #endif -static mdk_personality_t *pers[MAX_PERSONALITY]; +static LIST_HEAD(pers_list); static DEFINE_SPINLOCK(pers_lock); /* @@ -303,6 +303,15 @@ static mdk_rdev_t * find_rdev(mddev_t * mddev, dev_t dev) return NULL; } +static struct mdk_personality *find_pers(int level) +{ + struct mdk_personality *pers; + list_for_each_entry(pers, &pers_list, list) + if (pers->level == level) + return pers; + return NULL; +} + static inline sector_t calc_dev_sboffset(struct block_device *bdev) { sector_t size = bdev->bd_inode->i_size >> BLOCK_SIZE_BITS; @@ -1744,7 +1753,7 @@ static void analyze_sbs(mddev_t * mddev) static ssize_t level_show(mddev_t *mddev, char *page) { - mdk_personality_t *p = mddev->pers; + struct mdk_personality *p = mddev->pers; if (p == NULL && mddev->raid_disks == 0) return 0; if (mddev->level >= 0) @@ -1960,11 +1969,12 @@ static int start_dirty_degraded; static int do_md_run(mddev_t * mddev) { - int pnum, err; + int err; int chunk_size; struct list_head *tmp; mdk_rdev_t *rdev; struct gendisk *disk; + struct mdk_personality *pers; char b[BDEVNAME_SIZE]; if (list_empty(&mddev->disks)) @@ -1981,20 +1991,8 @@ static int do_md_run(mddev_t * mddev) analyze_sbs(mddev); chunk_size = mddev->chunk_size; - pnum = level_to_pers(mddev->level); - if ((pnum != MULTIPATH) && (pnum != RAID1)) { - if (!chunk_size) { - /* - * 'default chunksize' in the old md code used to - * be PAGE_SIZE, baaad. - * we abort here to be on the safe side. We don't - * want to continue the bad practice. - */ - printk(KERN_ERR - "no chunksize specified, see 'man raidtab'\n"); - return -EINVAL; - } + if (chunk_size) { if (chunk_size > MAX_CHUNK_SIZE) { printk(KERN_ERR "too big chunk_size: %d > %d\n", chunk_size, MAX_CHUNK_SIZE); @@ -2030,10 +2028,7 @@ static int do_md_run(mddev_t * mddev) } #ifdef CONFIG_KMOD - if (!pers[pnum]) - { - request_module("md-personality-%d", pnum); - } + request_module("md-level-%d", mddev->level); #endif /* @@ -2055,14 +2050,14 @@ static int do_md_run(mddev_t * mddev) return -ENOMEM; spin_lock(&pers_lock); - if (!pers[pnum] || !try_module_get(pers[pnum]->owner)) { + pers = find_pers(mddev->level); + if (!pers || !try_module_get(pers->owner)) { spin_unlock(&pers_lock); - printk(KERN_WARNING "md: personality %d is not loaded!\n", - pnum); + printk(KERN_WARNING "md: personality for level %d is not loaded!\n", + mddev->level); return -EINVAL; } - - mddev->pers = pers[pnum]; + mddev->pers = pers; spin_unlock(&pers_lock); mddev->recovery = 0; @@ -3701,15 +3696,14 @@ static int md_seq_show(struct seq_file *seq, void *v) struct list_head *tmp2; mdk_rdev_t *rdev; struct mdstat_info *mi = seq->private; - int i; struct bitmap *bitmap; if (v == (void*)1) { + struct mdk_personality *pers; seq_printf(seq, "Personalities : "); spin_lock(&pers_lock); - for (i = 0; i < MAX_PERSONALITY; i++) - if (pers[i]) - seq_printf(seq, "[%s] ", pers[i]->name); + list_for_each_entry(pers, &pers_list, list) + seq_printf(seq, "[%s] ", pers->name); spin_unlock(&pers_lock); seq_printf(seq, "\n"); @@ -3870,35 +3864,20 @@ static struct file_operations md_seq_fops = { .poll = mdstat_poll, }; -int register_md_personality(int pnum, mdk_personality_t *p) +int register_md_personality(struct mdk_personality *p) { - if (pnum >= MAX_PERSONALITY) { - printk(KERN_ERR - "md: tried to install personality %s as nr %d, but max is %lu\n", - p->name, pnum, MAX_PERSONALITY-1); - return -EINVAL; - } - spin_lock(&pers_lock); - if (pers[pnum]) { - spin_unlock(&pers_lock); - return -EBUSY; - } - - pers[pnum] = p; - printk(KERN_INFO "md: %s personality registered as nr %d\n", p->name, pnum); + list_add_tail(&p->list, &pers_list); + printk(KERN_INFO "md: %s personality registered for level %d\n", p->name, p->level); spin_unlock(&pers_lock); return 0; } -int unregister_md_personality(int pnum) +int unregister_md_personality(struct mdk_personality *p) { - if (pnum >= MAX_PERSONALITY) - return -EINVAL; - - printk(KERN_INFO "md: %s personality unregistered\n", pers[pnum]->name); + printk(KERN_INFO "md: %s personality unregistered\n", p->name); spin_lock(&pers_lock); - pers[pnum] = NULL; + list_del_init(&p->list); spin_unlock(&pers_lock); return 0; } diff --git a/drivers/md/multipath.c b/drivers/md/multipath.c index 97a56aaaef6d..d4d838e3f9f8 100644 --- a/drivers/md/multipath.c +++ b/drivers/md/multipath.c @@ -35,9 +35,6 @@ #define NR_RESERVED_BUFS 32 -static mdk_personality_t multipath_personality; - - static void *mp_pool_alloc(gfp_t gfp_flags, void *data) { struct multipath_bh *mpb; @@ -553,9 +550,10 @@ static int multipath_stop (mddev_t *mddev) return 0; } -static mdk_personality_t multipath_personality= +static struct mdk_personality multipath_personality = { .name = "multipath", + .level = LEVEL_MULTIPATH, .owner = THIS_MODULE, .make_request = multipath_make_request, .run = multipath_run, @@ -568,15 +566,16 @@ static mdk_personality_t multipath_personality= static int __init multipath_init (void) { - return register_md_personality (MULTIPATH, &multipath_personality); + return register_md_personality (&multipath_personality); } static void __exit multipath_exit (void) { - unregister_md_personality (MULTIPATH); + unregister_md_personality (&multipath_personality); } module_init(multipath_init); module_exit(multipath_exit); MODULE_LICENSE("GPL"); MODULE_ALIAS("md-personality-7"); /* MULTIPATH */ +MODULE_ALIAS("md-level--4"); diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c index b4eaa67fabde..7fb69e29391b 100644 --- a/drivers/md/raid0.c +++ b/drivers/md/raid0.c @@ -275,7 +275,11 @@ static int raid0_run (mddev_t *mddev) mdk_rdev_t *rdev; struct list_head *tmp; - printk("%s: setting max_sectors to %d, segment boundary to %d\n", + if (mddev->chunk_size == 0) { + printk(KERN_ERR "md/raid0: non-zero chunk size required.\n"); + return -EINVAL; + } + printk(KERN_INFO "%s: setting max_sectors to %d, segment boundary to %d\n", mdname(mddev), mddev->chunk_size >> 9, (mddev->chunk_size>>1)-1); @@ -507,9 +511,10 @@ static void raid0_status (struct seq_file *seq, mddev_t *mddev) return; } -static mdk_personality_t raid0_personality= +static struct mdk_personality raid0_personality= { .name = "raid0", + .level = 0, .owner = THIS_MODULE, .make_request = raid0_make_request, .run = raid0_run, @@ -519,15 +524,16 @@ static mdk_personality_t raid0_personality= static int __init raid0_init (void) { - return register_md_personality (RAID0, &raid0_personality); + return register_md_personality (&raid0_personality); } static void raid0_exit (void) { - unregister_md_personality (RAID0); + unregister_md_personality (&raid0_personality); } module_init(raid0_init); module_exit(raid0_exit); MODULE_LICENSE("GPL"); MODULE_ALIAS("md-personality-2"); /* RAID0 */ +MODULE_ALIAS("md-level-0"); diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c index c42ef1c99fa0..6e0f59ed3d80 100644 --- a/drivers/md/raid1.c +++ b/drivers/md/raid1.c @@ -47,7 +47,6 @@ */ #define NR_RAID1_BIOS 256 -static mdk_personality_t raid1_personality; static void unplug_slaves(mddev_t *mddev); @@ -2036,9 +2035,10 @@ static void raid1_quiesce(mddev_t *mddev, int state) } -static mdk_personality_t raid1_personality = +static struct mdk_personality raid1_personality = { .name = "raid1", + .level = 1, .owner = THIS_MODULE, .make_request = make_request, .run = run, @@ -2056,15 +2056,16 @@ static mdk_personality_t raid1_personality = static int __init raid_init(void) { - return register_md_personality(RAID1, &raid1_personality); + return register_md_personality(&raid1_personality); } static void raid_exit(void) { - unregister_md_personality(RAID1); + unregister_md_personality(&raid1_personality); } module_init(raid_init); module_exit(raid_exit); MODULE_LICENSE("GPL"); MODULE_ALIAS("md-personality-3"); /* RAID1 */ +MODULE_ALIAS("md-level-1"); diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c index 253322ae9195..f23d52c5df94 100644 --- a/drivers/md/raid10.c +++ b/drivers/md/raid10.c @@ -1883,11 +1883,11 @@ static int run(mddev_t *mddev) int nc, fc; sector_t stride, size; - if (mddev->level != 10) { - printk(KERN_ERR "raid10: %s: raid level not set correctly... (%d)\n", - mdname(mddev), mddev->level); - goto out; + if (mddev->chunk_size == 0) { + printk(KERN_ERR "md/raid10: non-zero chunk size required.\n"); + return -EINVAL; } + nc = mddev->layout & 255; fc = (mddev->layout >> 8) & 255; if ((nc*fc) <2 || (nc*fc) > mddev->raid_disks || @@ -2072,9 +2072,10 @@ static void raid10_quiesce(mddev_t *mddev, int state) } } -static mdk_personality_t raid10_personality = +static struct mdk_personality raid10_personality = { .name = "raid10", + .level = 10, .owner = THIS_MODULE, .make_request = make_request, .run = run, @@ -2090,15 +2091,16 @@ static mdk_personality_t raid10_personality = static int __init raid_init(void) { - return register_md_personality(RAID10, &raid10_personality); + return register_md_personality(&raid10_personality); } static void raid_exit(void) { - unregister_md_personality(RAID10); + unregister_md_personality(&raid10_personality); } module_init(raid_init); module_exit(raid_exit); MODULE_LICENSE("GPL"); MODULE_ALIAS("md-personality-9"); /* RAID10 */ +MODULE_ALIAS("md-level-10"); diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 6e4db95cebb1..b0cfd3ca9ca0 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -2187,9 +2187,10 @@ static void raid5_quiesce(mddev_t *mddev, int state) } } -static mdk_personality_t raid5_personality= +static struct mdk_personality raid5_personality = { .name = "raid5", + .level = 5, .owner = THIS_MODULE, .make_request = make_request, .run = run, @@ -2204,17 +2205,40 @@ static mdk_personality_t raid5_personality= .quiesce = raid5_quiesce, }; -static int __init raid5_init (void) +static struct mdk_personality raid4_personality = { - return register_md_personality (RAID5, &raid5_personality); + .name = "raid4", + .level = 4, + .owner = THIS_MODULE, + .make_request = make_request, + .run = run, + .stop = stop, + .status = status, + .error_handler = error, + .hot_add_disk = raid5_add_disk, + .hot_remove_disk= raid5_remove_disk, + .spare_active = raid5_spare_active, + .sync_request = sync_request, + .resize = raid5_resize, + .quiesce = raid5_quiesce, +}; + +static int __init raid5_init(void) +{ + register_md_personality(&raid5_personality); + register_md_personality(&raid4_personality); + return 0; } -static void raid5_exit (void) +static void raid5_exit(void) { - unregister_md_personality (RAID5); + unregister_md_personality(&raid5_personality); + unregister_md_personality(&raid4_personality); } module_init(raid5_init); module_exit(raid5_exit); MODULE_LICENSE("GPL"); MODULE_ALIAS("md-personality-4"); /* RAID5 */ +MODULE_ALIAS("md-level-5"); +MODULE_ALIAS("md-level-4"); diff --git a/drivers/md/raid6main.c b/drivers/md/raid6main.c index 79b5244f44f4..950e5fa6e1f2 100644 --- a/drivers/md/raid6main.c +++ b/drivers/md/raid6main.c @@ -2304,9 +2304,10 @@ static void raid6_quiesce(mddev_t *mddev, int state) } } -static mdk_personality_t raid6_personality= +static struct mdk_personality raid6_personality = { .name = "raid6", + .level = 6, .owner = THIS_MODULE, .make_request = make_request, .run = run, @@ -2321,7 +2322,7 @@ static mdk_personality_t raid6_personality= .quiesce = raid6_quiesce, }; -static int __init raid6_init (void) +static int __init raid6_init(void) { int e; @@ -2329,15 +2330,16 @@ static int __init raid6_init (void) if ( e ) return e; - return register_md_personality (RAID6, &raid6_personality); + return register_md_personality(&raid6_personality); } static void raid6_exit (void) { - unregister_md_personality (RAID6); + unregister_md_personality(&raid6_personality); } module_init(raid6_init); module_exit(raid6_exit); MODULE_LICENSE("GPL"); MODULE_ALIAS("md-personality-8"); /* RAID6 */ +MODULE_ALIAS("md-level-6"); diff --git a/include/linux/raid/md.h b/include/linux/raid/md.h index 13e7c4b62367..b6e0bcad84e1 100644 --- a/include/linux/raid/md.h +++ b/include/linux/raid/md.h @@ -71,8 +71,8 @@ */ #define MD_PATCHLEVEL_VERSION 3 -extern int register_md_personality (int p_num, mdk_personality_t *p); -extern int unregister_md_personality (int p_num); +extern int register_md_personality (struct mdk_personality *p); +extern int unregister_md_personality (struct mdk_personality *p); extern mdk_thread_t * md_register_thread (void (*run) (mddev_t *mddev), mddev_t *mddev, const char *name); extern void md_unregister_thread (mdk_thread_t *thread); diff --git a/include/linux/raid/md_k.h b/include/linux/raid/md_k.h index 1dd587b5975a..e559fb701aa1 100644 --- a/include/linux/raid/md_k.h +++ b/include/linux/raid/md_k.h @@ -18,62 +18,19 @@ /* and dm-bio-list.h is not under include/linux because.... ??? */ #include "../../../drivers/md/dm-bio-list.h" -#define MD_RESERVED 0UL -#define LINEAR 1UL -#define RAID0 2UL -#define RAID1 3UL -#define RAID5 4UL -#define TRANSLUCENT 5UL -#define HSM 6UL -#define MULTIPATH 7UL -#define RAID6 8UL -#define RAID10 9UL -#define FAULTY 10UL -#define MAX_PERSONALITY 11UL - #define LEVEL_MULTIPATH (-4) #define LEVEL_LINEAR (-1) #define LEVEL_FAULTY (-5) +/* we need a value for 'no level specified' and 0 + * means 'raid0', so we need something else. This is + * for internal use only + */ +#define LEVEL_NONE (-1000000) + #define MaxSector (~(sector_t)0) #define MD_THREAD_NAME_MAX 14 -static inline int pers_to_level (int pers) -{ - switch (pers) { - case FAULTY: return LEVEL_FAULTY; - case MULTIPATH: return LEVEL_MULTIPATH; - case HSM: return -3; - case TRANSLUCENT: return -2; - case LINEAR: return LEVEL_LINEAR; - case RAID0: return 0; - case RAID1: return 1; - case RAID5: return 5; - case RAID6: return 6; - case RAID10: return 10; - } - BUG(); - return MD_RESERVED; -} - -static inline int level_to_pers (int level) -{ - switch (level) { - case LEVEL_FAULTY: return FAULTY; - case LEVEL_MULTIPATH: return MULTIPATH; - case -3: return HSM; - case -2: return TRANSLUCENT; - case LEVEL_LINEAR: return LINEAR; - case 0: return RAID0; - case 1: return RAID1; - case 4: - case 5: return RAID5; - case 6: return RAID6; - case 10: return RAID10; - } - return MD_RESERVED; -} - typedef struct mddev_s mddev_t; typedef struct mdk_rdev_s mdk_rdev_t; @@ -140,12 +97,10 @@ struct mdk_rdev_s */ }; -typedef struct mdk_personality_s mdk_personality_t; - struct mddev_s { void *private; - mdk_personality_t *pers; + struct mdk_personality *pers; dev_t unit; int md_minor; struct list_head disks; @@ -266,9 +221,11 @@ static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sect atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io); } -struct mdk_personality_s +struct mdk_personality { char *name; + int level; + struct list_head list; struct module *owner; int (*make_request)(request_queue_t *q, struct bio *bio); int (*run)(mddev_t *mddev); diff --git a/init/do_mounts_md.c b/init/do_mounts_md.c index 3fbc3555ce96..f6f36806f84a 100644 --- a/init/do_mounts_md.c +++ b/init/do_mounts_md.c @@ -17,7 +17,7 @@ static int __initdata raid_noautodetect, raid_autopart; static struct { int minor; int partitioned; - int pers; + int level; int chunk; char *device_names; } md_setup_args[MAX_MD_DEVS] __initdata; @@ -47,7 +47,7 @@ extern int mdp_major; */ static int __init md_setup(char *str) { - int minor, level, factor, fault, pers, partitioned = 0; + int minor, level, factor, fault, partitioned = 0; char *pername = ""; char *str1; int ent; @@ -78,7 +78,7 @@ static int __init md_setup(char *str) } if (ent >= md_setup_ents) md_setup_ents++; - switch (get_option(&str, &level)) { /* RAID Personality */ + switch (get_option(&str, &level)) { /* RAID level */ case 2: /* could be 0 or -1.. */ if (level == 0 || level == LEVEL_LINEAR) { if (get_option(&str, &factor) != 2 || /* Chunk Size */ @@ -86,16 +86,12 @@ static int __init md_setup(char *str) printk(KERN_WARNING "md: Too few arguments supplied to md=.\n"); return 0; } - md_setup_args[ent].pers = level; + md_setup_args[ent].level = level; md_setup_args[ent].chunk = 1 << (factor+12); - if (level == LEVEL_LINEAR) { - pers = LINEAR; + if (level == LEVEL_LINEAR) pername = "linear"; - } else { - pers = RAID0; + else pername = "raid0"; - } - md_setup_args[ent].pers = pers; break; } /* FALL THROUGH */ @@ -103,7 +99,7 @@ static int __init md_setup(char *str) str = str1; /* FALL THROUGH */ case 0: - md_setup_args[ent].pers = 0; + md_setup_args[ent].level = LEVEL_NONE; pername="super-block"; } @@ -190,10 +186,10 @@ static void __init md_setup_drive(void) continue; } - if (md_setup_args[ent].pers) { + if (md_setup_args[ent].level != LEVEL_NONE) { /* non-persistent */ mdu_array_info_t ainfo; - ainfo.level = pers_to_level(md_setup_args[ent].pers); + ainfo.level = md_setup_args[ent].level; ainfo.size = 0; ainfo.nr_disks =0; ainfo.raid_disks =0; -- cgit v1.2.3 From 1345b1d8adbdeceb1c871d9a4af5e2a700b341c6 Mon Sep 17 00:00:00 2001 From: NeilBrown Date: Fri, 6 Jan 2006 00:20:40 -0800 Subject: [PATCH] md: define and use safe_put_page for md md sometimes call put_page on NULL pointers (treating it like kfree). This is not safe, so define and use a 'safe_put_page' which checks for NULL. Signed-off-by: Neil Brown Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- drivers/md/bitmap.c | 3 +-- drivers/md/raid1.c | 8 ++++---- drivers/md/raid10.c | 8 ++++---- drivers/md/raid6main.c | 3 +-- include/linux/raid/md_k.h | 5 +++++ 5 files changed, 15 insertions(+), 12 deletions(-) (limited to 'include/linux/raid') diff --git a/drivers/md/bitmap.c b/drivers/md/bitmap.c index abe415f0c039..ee4a3424a8a3 100644 --- a/drivers/md/bitmap.c +++ b/drivers/md/bitmap.c @@ -626,8 +626,7 @@ static void bitmap_file_unmap(struct bitmap *bitmap) kfree(map); kfree(attr); - if (sb_page) - put_page(sb_page); + safe_put_page(sb_page); } static void bitmap_stop_daemon(struct bitmap *bitmap); diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c index 39c10a65683d..feea4eeca1d9 100644 --- a/drivers/md/raid1.c +++ b/drivers/md/raid1.c @@ -136,7 +136,7 @@ static void * r1buf_pool_alloc(gfp_t gfp_flags, void *data) out_free_pages: for (i=0; i < RESYNC_PAGES ; i++) for (j=0 ; j < pi->raid_disks; j++) - put_page(r1_bio->bios[j]->bi_io_vec[i].bv_page); + safe_put_page(r1_bio->bios[j]->bi_io_vec[i].bv_page); j = -1; out_free_bio: while ( ++j < pi->raid_disks ) @@ -156,7 +156,7 @@ static void r1buf_pool_free(void *__r1_bio, void *data) if (j == 0 || r1bio->bios[j]->bi_io_vec[i].bv_page != r1bio->bios[0]->bi_io_vec[i].bv_page) - put_page(r1bio->bios[j]->bi_io_vec[i].bv_page); + safe_put_page(r1bio->bios[j]->bi_io_vec[i].bv_page); } for (i=0 ; i < pi->raid_disks; i++) bio_put(r1bio->bios[i]); @@ -381,7 +381,7 @@ static int raid1_end_write_request(struct bio *bio, unsigned int bytes_done, int /* free extra copy of the data pages */ int i = bio->bi_vcnt; while (i--) - put_page(bio->bi_io_vec[i].bv_page); + safe_put_page(bio->bi_io_vec[i].bv_page); } /* clear the bitmap if all writes complete successfully */ bitmap_endwrite(r1_bio->mddev->bitmap, r1_bio->sector, @@ -1907,7 +1907,7 @@ out_free_conf: if (conf->r1bio_pool) mempool_destroy(conf->r1bio_pool); kfree(conf->mirrors); - put_page(conf->tmppage); + safe_put_page(conf->tmppage); kfree(conf->poolinfo); kfree(conf); mddev->private = NULL; diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c index 9647ebb0983a..fb952000fae2 100644 --- a/drivers/md/raid10.c +++ b/drivers/md/raid10.c @@ -132,10 +132,10 @@ static void * r10buf_pool_alloc(gfp_t gfp_flags, void *data) out_free_pages: for ( ; i > 0 ; i--) - put_page(bio->bi_io_vec[i-1].bv_page); + safe_put_page(bio->bi_io_vec[i-1].bv_page); while (j--) for (i = 0; i < RESYNC_PAGES ; i++) - put_page(r10_bio->devs[j].bio->bi_io_vec[i].bv_page); + safe_put_page(r10_bio->devs[j].bio->bi_io_vec[i].bv_page); j = -1; out_free_bio: while ( ++j < nalloc ) @@ -155,7 +155,7 @@ static void r10buf_pool_free(void *__r10_bio, void *data) struct bio *bio = r10bio->devs[j].bio; if (bio) { for (i = 0; i < RESYNC_PAGES; i++) { - put_page(bio->bi_io_vec[i].bv_page); + safe_put_page(bio->bi_io_vec[i].bv_page); bio->bi_io_vec[i].bv_page = NULL; } bio_put(bio); @@ -2042,7 +2042,7 @@ static int run(mddev_t *mddev) out_free_conf: if (conf->r10bio_pool) mempool_destroy(conf->r10bio_pool); - put_page(conf->tmppage); + safe_put_page(conf->tmppage); kfree(conf->mirrors); kfree(conf); mddev->private = NULL; diff --git a/drivers/md/raid6main.c b/drivers/md/raid6main.c index 950e5fa6e1f2..06b32bd671a3 100644 --- a/drivers/md/raid6main.c +++ b/drivers/md/raid6main.c @@ -2072,8 +2072,7 @@ static int run(mddev_t *mddev) abort: if (conf) { print_raid6_conf(conf); - if (conf->spare_page) - put_page(conf->spare_page); + safe_put_page(conf->spare_page); kfree(conf->stripe_hashtbl); kfree(conf); } diff --git a/include/linux/raid/md_k.h b/include/linux/raid/md_k.h index e559fb701aa1..12b3203e3419 100644 --- a/include/linux/raid/md_k.h +++ b/include/linux/raid/md_k.h @@ -324,5 +324,10 @@ do { \ __wait_event_lock_irq(wq, condition, lock, cmd); \ } while (0) +static inline void safe_put_page(struct page *p) +{ + if (p) put_page(p); +} + #endif -- cgit v1.2.3 From 2989ddbd6e1d9638a188311b896362c4bf7b7c25 Mon Sep 17 00:00:00 2001 From: NeilBrown Date: Fri, 6 Jan 2006 00:20:43 -0800 Subject: [PATCH] md: make a couple of names in md.c static .. because they aren't used outside md.c Signed-off-by: Neil Brown Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- drivers/md/md.c | 4 ++-- include/linux/raid/md_k.h | 2 -- 2 files changed, 2 insertions(+), 4 deletions(-) (limited to 'include/linux/raid') diff --git a/drivers/md/md.c b/drivers/md/md.c index c1613854f38d..3cf089349e77 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -144,7 +144,7 @@ static int start_readonly; * start array, stop array, error, add device, remove device, * start build, activate spare */ -DECLARE_WAIT_QUEUE_HEAD(md_event_waiters); +static DECLARE_WAIT_QUEUE_HEAD(md_event_waiters); static atomic_t md_event_count; void md_new_event(mddev_t *mddev) { @@ -279,7 +279,7 @@ static inline void mddev_unlock(mddev_t * mddev) md_wakeup_thread(mddev->thread); } -mdk_rdev_t * find_rdev_nr(mddev_t *mddev, int nr) +static mdk_rdev_t * find_rdev_nr(mddev_t *mddev, int nr) { mdk_rdev_t * rdev; struct list_head *tmp; diff --git a/include/linux/raid/md_k.h b/include/linux/raid/md_k.h index 12b3203e3419..0fb5af6d622d 100644 --- a/include/linux/raid/md_k.h +++ b/include/linux/raid/md_k.h @@ -263,8 +263,6 @@ static inline char * mdname (mddev_t * mddev) return mddev->gendisk ? mddev->gendisk->disk_name : "mdX"; } -extern mdk_rdev_t * find_rdev_nr(mddev_t *mddev, int nr); - /* * iterates through some rdev ringlist. It's safe to remove the * current 'rdev'. Dont touch 'tmp' though. -- cgit v1.2.3 From d9d166c2a9d5d01af34396793950aa695883eed4 Mon Sep 17 00:00:00 2001 From: NeilBrown Date: Fri, 6 Jan 2006 00:20:51 -0800 Subject: [PATCH] md: allow array level to be set textually via sysfs Signed-off-by: Neil Brown Acked-by: Greg KH Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- Documentation/md.txt | 8 +++++++ drivers/md/faulty.c | 1 + drivers/md/linear.c | 3 ++- drivers/md/md.c | 61 +++++++++++++++++++++++++++++++++++++---------- drivers/md/multipath.c | 1 + drivers/md/raid0.c | 1 + drivers/md/raid1.c | 1 + drivers/md/raid10.c | 1 + drivers/md/raid5.c | 2 ++ drivers/md/raid6main.c | 1 + include/linux/raid/md_k.h | 1 + 11 files changed, 67 insertions(+), 14 deletions(-) (limited to 'include/linux/raid') diff --git a/Documentation/md.txt b/Documentation/md.txt index c5512afd5917..fd43fd2cad2f 100644 --- a/Documentation/md.txt +++ b/Documentation/md.txt @@ -189,6 +189,14 @@ All md devices contain: 1.2 (newer format in varying locations) or "none" indicating that the kernel isn't managing metadata at all. + level + The raid 'level' for this array. The name will often (but not + always) be the same as the name of the module that implements the + level. To be auto-loaded the module must have an alias + md-$LEVEL e.g. md-raid5 + This can be written only while the array is being assembled, not + after it is started. + As component devices are added to an md array, they appear in the 'md' directory as new directories named dev-XXX diff --git a/drivers/md/faulty.c b/drivers/md/faulty.c index f12e83086897..a7a5ab554338 100644 --- a/drivers/md/faulty.c +++ b/drivers/md/faulty.c @@ -342,4 +342,5 @@ module_init(raid_init); module_exit(raid_exit); MODULE_LICENSE("GPL"); MODULE_ALIAS("md-personality-10"); /* faulty */ +MODULE_ALIAS("md-faulty"); MODULE_ALIAS("md-level--5"); diff --git a/drivers/md/linear.c b/drivers/md/linear.c index 79dee8159217..777585458c85 100644 --- a/drivers/md/linear.c +++ b/drivers/md/linear.c @@ -376,5 +376,6 @@ static void linear_exit (void) module_init(linear_init); module_exit(linear_exit); MODULE_LICENSE("GPL"); -MODULE_ALIAS("md-personality-1"); /* LINEAR - degrecated*/ +MODULE_ALIAS("md-personality-1"); /* LINEAR - deprecated*/ +MODULE_ALIAS("md-linear"); MODULE_ALIAS("md-level--1"); diff --git a/drivers/md/md.c b/drivers/md/md.c index ecc0166ba779..594d8c312e6a 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -303,12 +303,15 @@ static mdk_rdev_t * find_rdev(mddev_t * mddev, dev_t dev) return NULL; } -static struct mdk_personality *find_pers(int level) +static struct mdk_personality *find_pers(int level, char *clevel) { struct mdk_personality *pers; - list_for_each_entry(pers, &pers_list, list) - if (pers->level == level) + list_for_each_entry(pers, &pers_list, list) { + if (level != LEVEL_NONE && pers->level == level) return pers; + if (strcmp(pers->name, clevel)==0) + return pers; + } return NULL; } @@ -715,6 +718,7 @@ static int super_90_validate(mddev_t *mddev, mdk_rdev_t *rdev) mddev->ctime = sb->ctime; mddev->utime = sb->utime; mddev->level = sb->level; + mddev->clevel[0] = 0; mddev->layout = sb->layout; mddev->raid_disks = sb->raid_disks; mddev->size = sb->size; @@ -1051,6 +1055,7 @@ static int super_1_validate(mddev_t *mddev, mdk_rdev_t *rdev) mddev->ctime = le64_to_cpu(sb->ctime) & ((1ULL << 32)-1); mddev->utime = le64_to_cpu(sb->utime) & ((1ULL << 32)-1); mddev->level = le32_to_cpu(sb->level); + mddev->clevel[0] = 0; mddev->layout = le32_to_cpu(sb->layout); mddev->raid_disks = le32_to_cpu(sb->raid_disks); mddev->size = le64_to_cpu(sb->size)/2; @@ -1774,15 +1779,36 @@ static ssize_t level_show(mddev_t *mddev, char *page) { struct mdk_personality *p = mddev->pers; - if (p == NULL && mddev->raid_disks == 0) - return 0; - if (mddev->level >= 0) - return sprintf(page, "raid%d\n", mddev->level); - else + if (p) return sprintf(page, "%s\n", p->name); + else if (mddev->clevel[0]) + return sprintf(page, "%s\n", mddev->clevel); + else if (mddev->level != LEVEL_NONE) + return sprintf(page, "%d\n", mddev->level); + else + return 0; +} + +static ssize_t +level_store(mddev_t *mddev, const char *buf, size_t len) +{ + int rv = len; + if (mddev->pers) + return -EBUSY; + if (len == 0) + return 0; + if (len >= sizeof(mddev->clevel)) + return -ENOSPC; + strncpy(mddev->clevel, buf, len); + if (mddev->clevel[len-1] == '\n') + len--; + mddev->clevel[len] = 0; + mddev->level = LEVEL_NONE; + return rv; } -static struct md_sysfs_entry md_level = __ATTR_RO(level); +static struct md_sysfs_entry md_level = +__ATTR(level, 0644, level_show, level_store); static ssize_t raid_disks_show(mddev_t *mddev, char *page) @@ -2158,7 +2184,10 @@ static int do_md_run(mddev_t * mddev) } #ifdef CONFIG_KMOD - request_module("md-level-%d", mddev->level); + if (mddev->level != LEVEL_NONE) + request_module("md-level-%d", mddev->level); + else if (mddev->clevel[0]) + request_module("md-%s", mddev->clevel); #endif /* @@ -2180,15 +2209,21 @@ static int do_md_run(mddev_t * mddev) return -ENOMEM; spin_lock(&pers_lock); - pers = find_pers(mddev->level); + pers = find_pers(mddev->level, mddev->clevel); if (!pers || !try_module_get(pers->owner)) { spin_unlock(&pers_lock); - printk(KERN_WARNING "md: personality for level %d is not loaded!\n", - mddev->level); + if (mddev->level != LEVEL_NONE) + printk(KERN_WARNING "md: personality for level %d is not loaded!\n", + mddev->level); + else + printk(KERN_WARNING "md: personality for level %s is not loaded!\n", + mddev->clevel); return -EINVAL; } mddev->pers = pers; spin_unlock(&pers_lock); + mddev->level = pers->level; + strlcpy(mddev->clevel, pers->name, sizeof(mddev->clevel)); mddev->recovery = 0; mddev->resync_max_sectors = mddev->size << 1; /* may be over-ridden by personality */ diff --git a/drivers/md/multipath.c b/drivers/md/multipath.c index d4d838e3f9f8..e6aa309a66d7 100644 --- a/drivers/md/multipath.c +++ b/drivers/md/multipath.c @@ -578,4 +578,5 @@ module_init(multipath_init); module_exit(multipath_exit); MODULE_LICENSE("GPL"); MODULE_ALIAS("md-personality-7"); /* MULTIPATH */ +MODULE_ALIAS("md-multipath"); MODULE_ALIAS("md-level--4"); diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c index 7fb69e29391b..abbca150202b 100644 --- a/drivers/md/raid0.c +++ b/drivers/md/raid0.c @@ -536,4 +536,5 @@ module_init(raid0_init); module_exit(raid0_exit); MODULE_LICENSE("GPL"); MODULE_ALIAS("md-personality-2"); /* RAID0 */ +MODULE_ALIAS("md-raid0"); MODULE_ALIAS("md-level-0"); diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c index 7d4465f93064..181c9616d5f1 100644 --- a/drivers/md/raid1.c +++ b/drivers/md/raid1.c @@ -2092,4 +2092,5 @@ module_init(raid_init); module_exit(raid_exit); MODULE_LICENSE("GPL"); MODULE_ALIAS("md-personality-3"); /* RAID1 */ +MODULE_ALIAS("md-raid1"); MODULE_ALIAS("md-level-1"); diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c index 8e6f6dfddb2b..201dc7168a5f 100644 --- a/drivers/md/raid10.c +++ b/drivers/md/raid10.c @@ -2117,4 +2117,5 @@ module_init(raid_init); module_exit(raid_exit); MODULE_LICENSE("GPL"); MODULE_ALIAS("md-personality-9"); /* RAID10 */ +MODULE_ALIAS("md-raid10"); MODULE_ALIAS("md-level-10"); diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index b0cfd3ca9ca0..9cc844f455bf 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -2240,5 +2240,7 @@ module_init(raid5_init); module_exit(raid5_exit); MODULE_LICENSE("GPL"); MODULE_ALIAS("md-personality-4"); /* RAID5 */ +MODULE_ALIAS("md-raid5"); +MODULE_ALIAS("md-raid4"); MODULE_ALIAS("md-level-5"); MODULE_ALIAS("md-level-4"); diff --git a/drivers/md/raid6main.c b/drivers/md/raid6main.c index 06b32bd671a3..84dd875bb2f6 100644 --- a/drivers/md/raid6main.c +++ b/drivers/md/raid6main.c @@ -2341,4 +2341,5 @@ module_init(raid6_init); module_exit(raid6_exit); MODULE_LICENSE("GPL"); MODULE_ALIAS("md-personality-8"); /* RAID6 */ +MODULE_ALIAS("md-raid6"); MODULE_ALIAS("md-level-6"); diff --git a/include/linux/raid/md_k.h b/include/linux/raid/md_k.h index 0fb5af6d622d..686463115438 100644 --- a/include/linux/raid/md_k.h +++ b/include/linux/raid/md_k.h @@ -119,6 +119,7 @@ struct mddev_s int chunk_size; time_t ctime, utime; int level, layout; + char clevel[16]; int raid_disks; int max_disks; sector_t size; /* used size of component devices */ -- cgit v1.2.3 From 4dbcdc751cb25ffca3a8374cbc5ab6de961cc545 Mon Sep 17 00:00:00 2001 From: NeilBrown Date: Fri, 6 Jan 2006 00:20:52 -0800 Subject: [PATCH] md: count corrected read errors per drive Store this total in superblock (As appropriate), and make it available to userspace via sysfs. Signed-off-by: Neil Brown Acked-by: Greg KH Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- Documentation/md.txt | 11 +++++++++++ drivers/md/md.c | 27 ++++++++++++++++++++++++++- drivers/md/raid1.c | 2 ++ drivers/md/raid10.c | 11 ++++++++--- drivers/md/raid5.c | 3 +++ drivers/md/raid6main.c | 3 +++ include/linux/raid/md_k.h | 4 ++++ 7 files changed, 57 insertions(+), 4 deletions(-) (limited to 'include/linux/raid') diff --git a/Documentation/md.txt b/Documentation/md.txt index fd43fd2cad2f..a3eadf8e1701 100644 --- a/Documentation/md.txt +++ b/Documentation/md.txt @@ -222,6 +222,17 @@ Each directory contains: of being recoverred to This list make grow in future. + errors + An approximate count of read errors that have been detected on + this device but have not caused the device to be evicted from + the array (either because they were corrected or because they + happened while the array was read-only). When using version-1 + metadata, this value persists across restarts of the array. + + This value can be written while assembling an array thus + providing an ongoing count for arrays with metadata managed by + userspace. + An active md device will also contain and entry for each active device in the array. These are named diff --git a/drivers/md/md.c b/drivers/md/md.c index 594d8c312e6a..32a4e2311e43 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -1000,6 +1000,7 @@ static int super_1_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version) } rdev->preferred_minor = 0xffff; rdev->data_offset = le64_to_cpu(sb->data_offset); + atomic_set(&rdev->corrected_errors, le32_to_cpu(sb->cnt_corrected_read)); rdev->sb_size = le32_to_cpu(sb->max_dev) * 2 + 256; bmask = queue_hardsect_size(rdev->bdev->bd_disk->queue)-1; @@ -1139,6 +1140,8 @@ static void super_1_sync(mddev_t *mddev, mdk_rdev_t *rdev) else sb->resync_offset = cpu_to_le64(0); + sb->cnt_corrected_read = atomic_read(&rdev->corrected_errors); + if (mddev->bitmap && mddev->bitmap_file == NULL) { sb->bitmap_offset = cpu_to_le32((__u32)mddev->bitmap_offset); sb->feature_map = cpu_to_le32(MD_FEATURE_BITMAP_OFFSET); @@ -1592,9 +1595,30 @@ super_show(mdk_rdev_t *rdev, char *page) } static struct rdev_sysfs_entry rdev_super = __ATTR_RO(super); +static ssize_t +errors_show(mdk_rdev_t *rdev, char *page) +{ + return sprintf(page, "%d\n", atomic_read(&rdev->corrected_errors)); +} + +static ssize_t +errors_store(mdk_rdev_t *rdev, const char *buf, size_t len) +{ + char *e; + unsigned long n = simple_strtoul(buf, &e, 10); + if (*buf && (*e == 0 || *e == '\n')) { + atomic_set(&rdev->corrected_errors, n); + return len; + } + return -EINVAL; +} +static struct rdev_sysfs_entry rdev_errors = +__ATTR(errors, 0644, errors_show, errors_store); + static struct attribute *rdev_default_attrs[] = { &rdev_state.attr, &rdev_super.attr, + &rdev_errors.attr, NULL, }; static ssize_t @@ -1674,6 +1698,7 @@ static mdk_rdev_t *md_import_device(dev_t newdev, int super_format, int super_mi rdev->data_offset = 0; atomic_set(&rdev->nr_pending, 0); atomic_set(&rdev->read_errors, 0); + atomic_set(&rdev->corrected_errors, 0); size = rdev->bdev->bd_inode->i_size >> BLOCK_SIZE_BITS; if (!size) { @@ -4729,7 +4754,7 @@ static int set_ro(const char *val, struct kernel_param *kp) int num = simple_strtoul(val, &e, 10); if (*val && (*e == '\0' || *e == '\n')) { start_readonly = num; - return 0;; + return 0; } return -EINVAL; } diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c index 181c9616d5f1..a06ff91f27e2 100644 --- a/drivers/md/raid1.c +++ b/drivers/md/raid1.c @@ -1265,6 +1265,7 @@ static void sync_request_write(mddev_t *mddev, r1bio_t *r1_bio) if (r1_bio->bios[d]->bi_end_io != end_sync_read) continue; rdev = conf->mirrors[d].rdev; + atomic_add(s, &rdev->corrected_errors); if (sync_page_io(rdev->bdev, sect + rdev->data_offset, s<<9, @@ -1463,6 +1464,7 @@ static void raid1d(mddev_t *mddev) d = conf->raid_disks; d--; rdev = conf->mirrors[d].rdev; + atomic_add(s, &rdev->corrected_errors); if (rdev && test_bit(In_sync, &rdev->flags)) { if (sync_page_io(rdev->bdev, diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c index 201dc7168a5f..9e658e519a27 100644 --- a/drivers/md/raid10.c +++ b/drivers/md/raid10.c @@ -1122,9 +1122,13 @@ static int end_sync_read(struct bio *bio, unsigned int bytes_done, int error) if (test_bit(BIO_UPTODATE, &bio->bi_flags)) set_bit(R10BIO_Uptodate, &r10_bio->state); - else if (!test_bit(MD_RECOVERY_SYNC, &conf->mddev->recovery)) - md_error(r10_bio->mddev, - conf->mirrors[d].rdev); + else { + atomic_add(r10_bio->sectors, + &conf->mirrors[d].rdev->corrected_errors); + if (!test_bit(MD_RECOVERY_SYNC, &conf->mddev->recovery)) + md_error(r10_bio->mddev, + conf->mirrors[d].rdev); + } /* for reconstruct, we always reschedule after a read. * for resync, only after all reads @@ -1430,6 +1434,7 @@ static void raid10d(mddev_t *mddev) sl--; d = r10_bio->devs[sl].devnum; rdev = conf->mirrors[d].rdev; + atomic_add(s, &rdev->corrected_errors); if (rdev && test_bit(In_sync, &rdev->flags)) { if (sync_page_io(rdev->bdev, diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 9cc844f455bf..54f4a9847e38 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -1400,6 +1400,9 @@ static void handle_stripe(struct stripe_head *sh) bi->bi_io_vec[0].bv_offset = 0; bi->bi_size = STRIPE_SIZE; bi->bi_next = NULL; + if (rw == WRITE && + test_bit(R5_ReWrite, &sh->dev[i].flags)) + atomic_add(STRIPE_SECTORS, &rdev->corrected_errors); generic_make_request(bi); } else { if (rw == 1) diff --git a/drivers/md/raid6main.c b/drivers/md/raid6main.c index 84dd875bb2f6..8c823d686a60 100644 --- a/drivers/md/raid6main.c +++ b/drivers/md/raid6main.c @@ -1562,6 +1562,9 @@ static void handle_stripe(struct stripe_head *sh, struct page *tmp_page) bi->bi_io_vec[0].bv_offset = 0; bi->bi_size = STRIPE_SIZE; bi->bi_next = NULL; + if (rw == WRITE && + test_bit(R5_ReWrite, &sh->dev[i].flags)) + atomic_add(STRIPE_SECTORS, &rdev->corrected_errors); generic_make_request(bi); } else { if (rw == 1) diff --git a/include/linux/raid/md_k.h b/include/linux/raid/md_k.h index 686463115438..68b929c079ab 100644 --- a/include/linux/raid/md_k.h +++ b/include/linux/raid/md_k.h @@ -95,6 +95,10 @@ struct mdk_rdev_s atomic_t read_errors; /* number of consecutive read errors that * we have tried to ignore. */ + atomic_t corrected_errors; /* number of corrected read errors, + * for reporting to userspace and storing + * in superblock. + */ }; struct mddev_s -- cgit v1.2.3 From 88202a0c84e1951d6630d1d557d4801a8cc5b5ef Mon Sep 17 00:00:00 2001 From: NeilBrown Date: Fri, 6 Jan 2006 00:21:36 -0800 Subject: [PATCH] md: allow sync-speed to be controlled per-device Also export current (average) speed and status in sysfs. Signed-off-by: Neil Brown Acked-by: Greg KH Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- Documentation/md.txt | 22 ++++++++++ drivers/md/md.c | 110 +++++++++++++++++++++++++++++++++++++++++++--- include/linux/raid/md_k.h | 4 ++ 3 files changed, 131 insertions(+), 5 deletions(-) (limited to 'include/linux/raid') diff --git a/Documentation/md.txt b/Documentation/md.txt index b8d172b254f7..03a13c462cf2 100644 --- a/Documentation/md.txt +++ b/Documentation/md.txt @@ -207,6 +207,28 @@ All md devices contain: available. It will then appear at md/dev-XXX (depending on the name of the device) and further configuration is then possible. + sync_speed_min + sync_speed_max + This are similar to /proc/sys/dev/raid/speed_limit_{min,max} + however they only apply to the particular array. + If no value has been written to these, of if the word 'system' + is written, then the system-wide value is used. If a value, + in kibibytes-per-second is written, then it is used. + When the files are read, they show the currently active value + followed by "(local)" or "(system)" depending on whether it is + a locally set or system-wide value. + + sync_completed + This shows the number of sectors that have been completed of + whatever the current sync_action is, followed by the number of + sectors in total that could need to be processed. The two + numbers are separated by a '/' thus effectively showing one + value, a fraction of the process that is complete. + + sync_speed + This shows the current actual speed, in K/sec, of the current + sync_action. It is averaged over the last 30 seconds. + As component devices are added to an md array, they appear in the 'md' directory as new directories named diff --git a/drivers/md/md.c b/drivers/md/md.c index 825e235b791b..1b76fb29fb70 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -81,10 +81,22 @@ static DEFINE_SPINLOCK(pers_lock); * idle IO detection. * * you can change it via /proc/sys/dev/raid/speed_limit_min and _max. + * or /sys/block/mdX/md/sync_speed_{min,max} */ static int sysctl_speed_limit_min = 1000; static int sysctl_speed_limit_max = 200000; +static inline int speed_min(mddev_t *mddev) +{ + return mddev->sync_speed_min ? + mddev->sync_speed_min : sysctl_speed_limit_min; +} + +static inline int speed_max(mddev_t *mddev) +{ + return mddev->sync_speed_max ? + mddev->sync_speed_max : sysctl_speed_limit_max; +} static struct ctl_table_header *raid_table_header; @@ -2197,6 +2209,90 @@ md_scan_mode = __ATTR(sync_action, S_IRUGO|S_IWUSR, action_show, action_store); static struct md_sysfs_entry md_mismatches = __ATTR_RO(mismatch_cnt); +static ssize_t +sync_min_show(mddev_t *mddev, char *page) +{ + return sprintf(page, "%d (%s)\n", speed_min(mddev), + mddev->sync_speed_min ? "local": "system"); +} + +static ssize_t +sync_min_store(mddev_t *mddev, const char *buf, size_t len) +{ + int min; + char *e; + if (strncmp(buf, "system", 6)==0) { + mddev->sync_speed_min = 0; + return len; + } + min = simple_strtoul(buf, &e, 10); + if (buf == e || (*e && *e != '\n') || min <= 0) + return -EINVAL; + mddev->sync_speed_min = min; + return len; +} + +static struct md_sysfs_entry md_sync_min = +__ATTR(sync_speed_min, S_IRUGO|S_IWUSR, sync_min_show, sync_min_store); + +static ssize_t +sync_max_show(mddev_t *mddev, char *page) +{ + return sprintf(page, "%d (%s)\n", speed_max(mddev), + mddev->sync_speed_max ? "local": "system"); +} + +static ssize_t +sync_max_store(mddev_t *mddev, const char *buf, size_t len) +{ + int max; + char *e; + if (strncmp(buf, "system", 6)==0) { + mddev->sync_speed_max = 0; + return len; + } + max = simple_strtoul(buf, &e, 10); + if (buf == e || (*e && *e != '\n') || max <= 0) + return -EINVAL; + mddev->sync_speed_max = max; + return len; +} + +static struct md_sysfs_entry md_sync_max = +__ATTR(sync_speed_max, S_IRUGO|S_IWUSR, sync_max_show, sync_max_store); + + +static ssize_t +sync_speed_show(mddev_t *mddev, char *page) +{ + unsigned long resync, dt, db; + resync = (mddev->curr_resync - atomic_read(&mddev->recovery_active)); + dt = ((jiffies - mddev->resync_mark) / HZ); + if (!dt) dt++; + db = resync - (mddev->resync_mark_cnt); + return sprintf(page, "%ld\n", db/dt/2); /* K/sec */ +} + +static struct md_sysfs_entry +md_sync_speed = __ATTR_RO(sync_speed); + +static ssize_t +sync_completed_show(mddev_t *mddev, char *page) +{ + unsigned long max_blocks, resync; + + if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) + max_blocks = mddev->resync_max_sectors; + else + max_blocks = mddev->size << 1; + + resync = (mddev->curr_resync - atomic_read(&mddev->recovery_active)); + return sprintf(page, "%lu / %lu\n", resync, max_blocks); +} + +static struct md_sysfs_entry +md_sync_completed = __ATTR_RO(sync_completed); + static struct attribute *md_default_attrs[] = { &md_level.attr, &md_raid_disks.attr, @@ -2210,6 +2306,10 @@ static struct attribute *md_default_attrs[] = { static struct attribute *md_redundancy_attrs[] = { &md_scan_mode.attr, &md_mismatches.attr, + &md_sync_min.attr, + &md_sync_max.attr, + &md_sync_speed.attr, + &md_sync_completed.attr, NULL, }; static struct attribute_group md_redundancy_group = { @@ -4433,10 +4533,10 @@ static void md_do_sync(mddev_t *mddev) printk(KERN_INFO "md: syncing RAID array %s\n", mdname(mddev)); printk(KERN_INFO "md: minimum _guaranteed_ reconstruction speed:" - " %d KB/sec/disc.\n", sysctl_speed_limit_min); + " %d KB/sec/disc.\n", speed_min(mddev)); printk(KERN_INFO "md: using maximum available idle IO bandwidth " "(but not more than %d KB/sec) for reconstruction.\n", - sysctl_speed_limit_max); + speed_max(mddev)); is_mddev_idle(mddev); /* this also initializes IO event counters */ /* we don't use the checkpoint if there's a bitmap */ @@ -4477,7 +4577,7 @@ static void md_do_sync(mddev_t *mddev) skipped = 0; sectors = mddev->pers->sync_request(mddev, j, &skipped, - currspeed < sysctl_speed_limit_min); + currspeed < speed_min(mddev)); if (sectors == 0) { set_bit(MD_RECOVERY_ERR, &mddev->recovery); goto out; @@ -4542,8 +4642,8 @@ static void md_do_sync(mddev_t *mddev) currspeed = ((unsigned long)(io_sectors-mddev->resync_mark_cnt))/2 /((jiffies-mddev->resync_mark)/HZ +1) +1; - if (currspeed > sysctl_speed_limit_min) { - if ((currspeed > sysctl_speed_limit_max) || + if (currspeed > speed_min(mddev)) { + if ((currspeed > speed_max(mddev)) || !is_mddev_idle(mddev)) { msleep(500); goto repeat; diff --git a/include/linux/raid/md_k.h b/include/linux/raid/md_k.h index 68b929c079ab..617b9506c760 100644 --- a/include/linux/raid/md_k.h +++ b/include/linux/raid/md_k.h @@ -143,6 +143,10 @@ struct mddev_s sector_t resync_mismatches; /* count of sectors where * parity/replica mismatch found */ + /* if zero, use the system-wide default */ + int sync_speed_min; + int sync_speed_max; + int ok_start_degraded; /* recovery/resync flags * NEEDED: we might need to start a resync/recover -- cgit v1.2.3