From 3d05f3aed5d721c2c77d20288c29ab26c6193ed5 Mon Sep 17 00:00:00 2001 From: Julia Cartwright Date: Fri, 28 Apr 2017 12:41:02 -0500 Subject: [PATCH 1/8] md/raid5: make use of spin_lock_irq over local_irq_disable + spin_lock On mainline, there is no functional difference, just less code, and symmetric lock/unlock paths. On PREEMPT_RT builds, this fixes the following warning, seen by Alexander GQ Gerasiov, due to the sleeping nature of spinlocks. BUG: sleeping function called from invalid context at kernel/locking/rtmutex.c:993 in_atomic(): 0, irqs_disabled(): 1, pid: 58, name: kworker/u12:1 CPU: 5 PID: 58 Comm: kworker/u12:1 Tainted: G W 4.9.20-rt16-stand6-686 #1 Hardware name: Supermicro SYS-5027R-WRF/X9SRW-F, BIOS 3.2a 10/28/2015 Workqueue: writeback wb_workfn (flush-253:0) Call Trace: dump_stack+0x47/0x68 ? migrate_enable+0x4a/0xf0 ___might_sleep+0x101/0x180 rt_spin_lock+0x17/0x40 add_stripe_bio+0x4e3/0x6c0 [raid456] ? preempt_count_add+0x42/0xb0 raid5_make_request+0x737/0xdd0 [raid456] Reported-by: Alexander GQ Gerasiov Tested-by: Alexander GQ Gerasiov Signed-off-by: Julia Cartwright Signed-off-by: Shaohua Li --- drivers/md/raid5.c | 17 +++++++---------- 1 file changed, 7 insertions(+), 10 deletions(-) diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 2e38cfac5b1d..3809a2192132 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -103,8 +103,7 @@ static inline void unlock_device_hash_lock(struct r5conf *conf, int hash) static inline void lock_all_device_hash_locks_irq(struct r5conf *conf) { int i; - local_irq_disable(); - spin_lock(conf->hash_locks); + spin_lock_irq(conf->hash_locks); for (i = 1; i < NR_STRIPE_HASH_LOCKS; i++) spin_lock_nest_lock(conf->hash_locks + i, conf->hash_locks); spin_lock(&conf->device_lock); @@ -114,9 +113,9 @@ static inline void unlock_all_device_hash_locks_irq(struct r5conf *conf) { int i; spin_unlock(&conf->device_lock); - for (i = NR_STRIPE_HASH_LOCKS; i; i--) - spin_unlock(conf->hash_locks + i - 1); - local_irq_enable(); + for (i = NR_STRIPE_HASH_LOCKS - 1; i; i--) + spin_unlock(conf->hash_locks + i); + spin_unlock_irq(conf->hash_locks); } /* Find first data disk in a raid6 stripe */ @@ -714,12 +713,11 @@ static bool is_full_stripe_write(struct stripe_head *sh) static void lock_two_stripes(struct stripe_head *sh1, struct stripe_head *sh2) { - local_irq_disable(); if (sh1 > sh2) { - spin_lock(&sh2->stripe_lock); + spin_lock_irq(&sh2->stripe_lock); spin_lock_nested(&sh1->stripe_lock, 1); } else { - spin_lock(&sh1->stripe_lock); + spin_lock_irq(&sh1->stripe_lock); spin_lock_nested(&sh2->stripe_lock, 1); } } @@ -727,8 +725,7 @@ static void lock_two_stripes(struct stripe_head *sh1, struct stripe_head *sh2) static void unlock_two_stripes(struct stripe_head *sh1, struct stripe_head *sh2) { spin_unlock(&sh1->stripe_lock); - spin_unlock(&sh2->stripe_lock); - local_irq_enable(); + spin_unlock_irq(&sh2->stripe_lock); } /* Only freshly new full stripe normal write stripe can be added to a batch list */ From 2214c260c72b0bd94e6c1c19bf451686212025d3 Mon Sep 17 00:00:00 2001 From: Artur Paszkiewicz Date: Mon, 8 May 2017 11:56:55 +0200 Subject: [PATCH 2/8] md: don't return -EAGAIN in md_allow_write for external metadata arrays This essentially reverts commit b5470dc5fc18 ("md: resolve external metadata handling deadlock in md_allow_write") with some adjustments. Since commit 6791875e2e53 ("md: make reconfig_mutex optional for writes to md sysfs files.") changing array_state to 'active' does not use mddev_lock() and will not cause a deadlock with md_allow_write(). This revert simplifies userspace tools that write to sysfs attributes like "stripe_cache_size" or "consistency_policy" because it removes the need for special handling for external metadata arrays, checking for EAGAIN and retrying the write. Signed-off-by: Artur Paszkiewicz Signed-off-by: Shaohua Li --- drivers/md/md.c | 20 ++++++++------------ drivers/md/md.h | 2 +- drivers/md/raid1.c | 9 +++------ drivers/md/raid5.c | 12 +++--------- 4 files changed, 15 insertions(+), 28 deletions(-) diff --git a/drivers/md/md.c b/drivers/md/md.c index 82f798be964f..10367ffe92e3 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -8022,18 +8022,15 @@ EXPORT_SYMBOL(md_write_end); * may proceed without blocking. It is important to call this before * attempting a GFP_KERNEL allocation while holding the mddev lock. * Must be called with mddev_lock held. - * - * In the ->external case MD_SB_CHANGE_PENDING can not be cleared until mddev->lock - * is dropped, so return -EAGAIN after notifying userspace. */ -int md_allow_write(struct mddev *mddev) +void md_allow_write(struct mddev *mddev) { if (!mddev->pers) - return 0; + return; if (mddev->ro) - return 0; + return; if (!mddev->pers->sync_request) - return 0; + return; spin_lock(&mddev->lock); if (mddev->in_sync) { @@ -8046,13 +8043,12 @@ int md_allow_write(struct mddev *mddev) spin_unlock(&mddev->lock); md_update_sb(mddev, 0); sysfs_notify_dirent_safe(mddev->sysfs_state); + /* wait for the dirty state to be recorded in the metadata */ + wait_event(mddev->sb_wait, + !test_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags) && + !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)); } else spin_unlock(&mddev->lock); - - if (test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)) - return -EAGAIN; - else - return 0; } EXPORT_SYMBOL_GPL(md_allow_write); diff --git a/drivers/md/md.h b/drivers/md/md.h index 4e75d121bfcc..11f15146ce51 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h @@ -665,7 +665,7 @@ extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size, bool metadata_op); extern void md_do_sync(struct md_thread *thread); extern void md_new_event(struct mddev *mddev); -extern int md_allow_write(struct mddev *mddev); +extern void md_allow_write(struct mddev *mddev); extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev); extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors); extern int md_check_no_bitmap(struct mddev *mddev); diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c index 7ed59351fe97..7c1f73398800 100644 --- a/drivers/md/raid1.c +++ b/drivers/md/raid1.c @@ -3197,7 +3197,7 @@ static int raid1_reshape(struct mddev *mddev) struct r1conf *conf = mddev->private; int cnt, raid_disks; unsigned long flags; - int d, d2, err; + int d, d2; /* Cannot change chunk_size, layout, or level */ if (mddev->chunk_sectors != mddev->new_chunk_sectors || @@ -3209,11 +3209,8 @@ static int raid1_reshape(struct mddev *mddev) return -EINVAL; } - if (!mddev_is_clustered(mddev)) { - err = md_allow_write(mddev); - if (err) - return err; - } + if (!mddev_is_clustered(mddev)) + md_allow_write(mddev); raid_disks = mddev->raid_disks + mddev->delta_disks; diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 3809a2192132..f8055a7abb4b 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -2309,14 +2309,12 @@ static int resize_stripes(struct r5conf *conf, int newsize) struct stripe_head *osh, *nsh; LIST_HEAD(newstripes); struct disk_info *ndisks; - int err; + int err = 0; struct kmem_cache *sc; int i; int hash, cnt; - err = md_allow_write(conf->mddev); - if (err) - return err; + md_allow_write(conf->mddev); /* Step 1 */ sc = kmem_cache_create(conf->cache_name[1-conf->active_name], @@ -6310,7 +6308,6 @@ int raid5_set_cache_size(struct mddev *mddev, int size) { struct r5conf *conf = mddev->private; - int err; if (size <= 16 || size > 32768) return -EINVAL; @@ -6322,10 +6319,7 @@ raid5_set_cache_size(struct mddev *mddev, int size) ; mutex_unlock(&conf->cache_size_mutex); - - err = md_allow_write(mddev); - if (err) - return err; + md_allow_write(mddev); mutex_lock(&conf->cache_size_mutex); while (size > conf->max_nr_stripes) From 29efc390b9462582ae95eb9a0b8cd17ab956afc0 Mon Sep 17 00:00:00 2001 From: Shaohua Li Date: Sun, 7 May 2017 17:36:24 -0700 Subject: [PATCH 3/8] md/md0: optimize raid0 discard handling There are complaints that raid0 discard handling is slow. Currently we divide discard request into chunks and dispatch to underlayer disks. The block layer will do merge to form big requests. This causes a lot of request split/merge and uses significant CPU time. A simple idea is to calculate the range for each raid disk for an IO request and send a discard request to raid disks, which will avoid the split/merge completely. Previously Coly tried the approach, but the implementation was too complex because of raid0 zones. This patch always split bio in zone boundary and handle bio within one zone. It simplifies the implementation a lot. Reviewed-by: NeilBrown Acked-by: Coly Li Signed-off-by: Shaohua Li --- drivers/md/raid0.c | 116 +++++++++++++++++++++++++++++++++++++++------ 1 file changed, 102 insertions(+), 14 deletions(-) diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c index 84e58596594d..d6c0bc76e837 100644 --- a/drivers/md/raid0.c +++ b/drivers/md/raid0.c @@ -385,7 +385,7 @@ static int raid0_run(struct mddev *mddev) blk_queue_max_hw_sectors(mddev->queue, mddev->chunk_sectors); blk_queue_max_write_same_sectors(mddev->queue, mddev->chunk_sectors); blk_queue_max_write_zeroes_sectors(mddev->queue, mddev->chunk_sectors); - blk_queue_max_discard_sectors(mddev->queue, mddev->chunk_sectors); + blk_queue_max_discard_sectors(mddev->queue, UINT_MAX); blk_queue_io_min(mddev->queue, mddev->chunk_sectors << 9); blk_queue_io_opt(mddev->queue, @@ -459,6 +459,95 @@ static inline int is_io_in_chunk_boundary(struct mddev *mddev, } } +static void raid0_handle_discard(struct mddev *mddev, struct bio *bio) +{ + struct r0conf *conf = mddev->private; + struct strip_zone *zone; + sector_t start = bio->bi_iter.bi_sector; + sector_t end; + unsigned int stripe_size; + sector_t first_stripe_index, last_stripe_index; + sector_t start_disk_offset; + unsigned int start_disk_index; + sector_t end_disk_offset; + unsigned int end_disk_index; + unsigned int disk; + + zone = find_zone(conf, &start); + + if (bio_end_sector(bio) > zone->zone_end) { + struct bio *split = bio_split(bio, + zone->zone_end - bio->bi_iter.bi_sector, GFP_NOIO, + mddev->bio_set); + bio_chain(split, bio); + generic_make_request(bio); + bio = split; + end = zone->zone_end; + } else + end = bio_end_sector(bio); + + if (zone != conf->strip_zone) + end = end - zone[-1].zone_end; + + /* Now start and end is the offset in zone */ + stripe_size = zone->nb_dev * mddev->chunk_sectors; + + first_stripe_index = start; + sector_div(first_stripe_index, stripe_size); + last_stripe_index = end; + sector_div(last_stripe_index, stripe_size); + + start_disk_index = (int)(start - first_stripe_index * stripe_size) / + mddev->chunk_sectors; + start_disk_offset = ((int)(start - first_stripe_index * stripe_size) % + mddev->chunk_sectors) + + first_stripe_index * mddev->chunk_sectors; + end_disk_index = (int)(end - last_stripe_index * stripe_size) / + mddev->chunk_sectors; + end_disk_offset = ((int)(end - last_stripe_index * stripe_size) % + mddev->chunk_sectors) + + last_stripe_index * mddev->chunk_sectors; + + for (disk = 0; disk < zone->nb_dev; disk++) { + sector_t dev_start, dev_end; + struct bio *discard_bio = NULL; + struct md_rdev *rdev; + + if (disk < start_disk_index) + dev_start = (first_stripe_index + 1) * + mddev->chunk_sectors; + else if (disk > start_disk_index) + dev_start = first_stripe_index * mddev->chunk_sectors; + else + dev_start = start_disk_offset; + + if (disk < end_disk_index) + dev_end = (last_stripe_index + 1) * mddev->chunk_sectors; + else if (disk > end_disk_index) + dev_end = last_stripe_index * mddev->chunk_sectors; + else + dev_end = end_disk_offset; + + if (dev_end <= dev_start) + continue; + + rdev = conf->devlist[(zone - conf->strip_zone) * + conf->strip_zone[0].nb_dev + disk]; + if (__blkdev_issue_discard(rdev->bdev, + dev_start + zone->dev_start + rdev->data_offset, + dev_end - dev_start, GFP_NOIO, 0, &discard_bio) || + !discard_bio) + continue; + bio_chain(discard_bio, bio); + if (mddev->gendisk) + trace_block_bio_remap(bdev_get_queue(rdev->bdev), + discard_bio, disk_devt(mddev->gendisk), + bio->bi_iter.bi_sector); + generic_make_request(discard_bio); + } + bio_endio(bio); +} + static void raid0_make_request(struct mddev *mddev, struct bio *bio) { struct strip_zone *zone; @@ -473,6 +562,11 @@ static void raid0_make_request(struct mddev *mddev, struct bio *bio) return; } + if (unlikely((bio_op(bio) == REQ_OP_DISCARD))) { + raid0_handle_discard(mddev, bio); + return; + } + bio_sector = bio->bi_iter.bi_sector; sector = bio_sector; chunk_sects = mddev->chunk_sectors; @@ -498,19 +592,13 @@ static void raid0_make_request(struct mddev *mddev, struct bio *bio) bio->bi_iter.bi_sector = sector + zone->dev_start + tmp_dev->data_offset; - if (unlikely((bio_op(bio) == REQ_OP_DISCARD) && - !blk_queue_discard(bdev_get_queue(bio->bi_bdev)))) { - /* Just ignore it */ - bio_endio(bio); - } else { - if (mddev->gendisk) - trace_block_bio_remap(bdev_get_queue(bio->bi_bdev), - bio, disk_devt(mddev->gendisk), - bio_sector); - mddev_check_writesame(mddev, bio); - mddev_check_write_zeroes(mddev, bio); - generic_make_request(bio); - } + if (mddev->gendisk) + trace_block_bio_remap(bdev_get_queue(bio->bi_bdev), + bio, disk_devt(mddev->gendisk), + bio_sector); + mddev_check_writesame(mddev, bio); + mddev_check_write_zeroes(mddev, bio); + generic_make_request(bio); } static void raid0_status(struct seq_file *seq, struct mddev *mddev) From bb3338d3474e0329918fda9dae2c52751731eb58 Mon Sep 17 00:00:00 2001 From: Song Liu Date: Mon, 8 May 2017 17:39:24 -0700 Subject: [PATCH 4/8] md/raid5-cache: in r5l_do_submit_io(), submit io->split_bio first In r5l_do_submit_io(), it is necessary to check io->split_bio before submit io->current_bio. This is because, endio of current_bio may free the whole IO unit, and thus change io->split_bio. Signed-off-by: Song Liu Signed-off-by: Shaohua Li --- drivers/md/raid5-cache.c | 28 +++++++++++++++++++--------- 1 file changed, 19 insertions(+), 9 deletions(-) diff --git a/drivers/md/raid5-cache.c b/drivers/md/raid5-cache.c index 26ba09282e7c..a6a62e212cd3 100644 --- a/drivers/md/raid5-cache.c +++ b/drivers/md/raid5-cache.c @@ -622,20 +622,30 @@ static void r5l_do_submit_io(struct r5l_log *log, struct r5l_io_unit *io) __r5l_set_io_unit_state(io, IO_UNIT_IO_START); spin_unlock_irqrestore(&log->io_list_lock, flags); + /* + * In case of journal device failures, submit_bio will get error + * and calls endio, then active stripes will continue write + * process. Therefore, it is not necessary to check Faulty bit + * of journal device here. + * + * We can't check split_bio after current_bio is submitted. If + * io->split_bio is null, after current_bio is submitted, current_bio + * might already be completed and the io_unit is freed. We submit + * split_bio first to avoid the issue. + */ + if (io->split_bio) { + if (io->has_flush) + io->split_bio->bi_opf |= REQ_PREFLUSH; + if (io->has_fua) + io->split_bio->bi_opf |= REQ_FUA; + submit_bio(io->split_bio); + } + if (io->has_flush) io->current_bio->bi_opf |= REQ_PREFLUSH; if (io->has_fua) io->current_bio->bi_opf |= REQ_FUA; submit_bio(io->current_bio); - - if (!io->split_bio) - return; - - if (io->has_flush) - io->split_bio->bi_opf |= REQ_PREFLUSH; - if (io->has_fua) - io->split_bio->bi_opf |= REQ_FUA; - submit_bio(io->split_bio); } /* deferred io_unit will be dispatched here */ From 23b245c04d0ef408087430dd4d1b214a5da1eb78 Mon Sep 17 00:00:00 2001 From: Shaohua Li Date: Wed, 10 May 2017 08:47:11 -0700 Subject: [PATCH 5/8] md/raid1/10: avoid unnecessary locking If we add bios to block plugging list, locking is unnecessry, since the block unplug is guaranteed not to run at that time. Reviewed-by: NeilBrown Signed-off-by: Shaohua Li --- drivers/md/raid1.c | 7 +++---- drivers/md/raid10.c | 7 +++---- 2 files changed, 6 insertions(+), 8 deletions(-) diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c index 7c1f73398800..a17ed6218d51 100644 --- a/drivers/md/raid1.c +++ b/drivers/md/raid1.c @@ -1529,17 +1529,16 @@ static void raid1_write_request(struct mddev *mddev, struct bio *bio, plug = container_of(cb, struct raid1_plug_cb, cb); else plug = NULL; - spin_lock_irqsave(&conf->device_lock, flags); if (plug) { bio_list_add(&plug->pending, mbio); plug->pending_cnt++; } else { + spin_lock_irqsave(&conf->device_lock, flags); bio_list_add(&conf->pending_bio_list, mbio); conf->pending_count++; - } - spin_unlock_irqrestore(&conf->device_lock, flags); - if (!plug) + spin_unlock_irqrestore(&conf->device_lock, flags); md_wakeup_thread(mddev->thread); + } } r1_bio_write_done(r1_bio); diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c index 6b86a0032cf8..4343d7ff9916 100644 --- a/drivers/md/raid10.c +++ b/drivers/md/raid10.c @@ -1282,17 +1282,16 @@ static void raid10_write_one_disk(struct mddev *mddev, struct r10bio *r10_bio, plug = container_of(cb, struct raid10_plug_cb, cb); else plug = NULL; - spin_lock_irqsave(&conf->device_lock, flags); if (plug) { bio_list_add(&plug->pending, mbio); plug->pending_cnt++; } else { + spin_lock_irqsave(&conf->device_lock, flags); bio_list_add(&conf->pending_bio_list, mbio); conf->pending_count++; - } - spin_unlock_irqrestore(&conf->device_lock, flags); - if (!plug) + spin_unlock_irqrestore(&conf->device_lock, flags); md_wakeup_thread(mddev->thread); + } } static void raid10_write_request(struct mddev *mddev, struct bio *bio, From 70d466f760b351fe30b5f8c956354ddf29aa676b Mon Sep 17 00:00:00 2001 From: Song Liu Date: Thu, 11 May 2017 15:28:28 -0700 Subject: [PATCH 6/8] md/r5cache: gracefully handle journal device errors for writeback mode For the raid456 with writeback cache, when journal device failed during normal operation, it is still possible to persist all data, as all pending data is still in stripe cache. However, it is necessary to handle journal failure gracefully. During journal failures, the following logic handles the graceful shutdown of journal: 1. raid5_error() marks the device as Faulty and schedules async work log->disable_writeback_work; 2. In disable_writeback_work (r5c_disable_writeback_async), the mddev is suspended, set to write through, and then resumed. mddev_suspend() flushes all cached stripes; 3. All cached stripes need to be flushed carefully to the RAID array. This patch fixes issues within the process above: 1. In r5c_update_on_rdev_error() schedule disable_writeback_work for journal failures; 2. In r5c_disable_writeback_async(), wait for MD_SB_CHANGE_PENDING, since raid5_error() updates superblock. 3. In handle_stripe(), allow stripes with data in journal (s.injournal > 0) to make progress during log_failed; 4. In delay_towrite(), if log failed only process data in the cache (skip new writes in dev->towrite); 5. In __get_priority_stripe(), process loprio_list during journal device failures. 6. In raid5_remove_disk(), wait for all cached stripes are flushed before calling log_exit(). Signed-off-by: Song Liu Signed-off-by: Shaohua Li --- drivers/md/raid5-cache.c | 11 +++++++++-- drivers/md/raid5-log.h | 3 ++- drivers/md/raid5.c | 29 +++++++++++++++++++++++------ 3 files changed, 34 insertions(+), 9 deletions(-) diff --git a/drivers/md/raid5-cache.c b/drivers/md/raid5-cache.c index a6a62e212cd3..cc3f8442f11f 100644 --- a/drivers/md/raid5-cache.c +++ b/drivers/md/raid5-cache.c @@ -24,6 +24,7 @@ #include "md.h" #include "raid5.h" #include "bitmap.h" +#include "raid5-log.h" /* * metadata/data stored in disk with 4k size unit (a block) regardless @@ -680,6 +681,11 @@ static void r5c_disable_writeback_async(struct work_struct *work) return; pr_info("md/raid:%s: Disabling writeback cache for degraded array.\n", mdname(mddev)); + + /* wait superblock change before suspend */ + wait_event(mddev->sb_wait, + !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)); + mddev_suspend(mddev); log->r5c_journal_mode = R5C_JOURNAL_MODE_WRITE_THROUGH; mddev_resume(mddev); @@ -2983,7 +2989,7 @@ ioerr: return ret; } -void r5c_update_on_rdev_error(struct mddev *mddev) +void r5c_update_on_rdev_error(struct mddev *mddev, struct md_rdev *rdev) { struct r5conf *conf = mddev->private; struct r5l_log *log = conf->log; @@ -2991,7 +2997,8 @@ void r5c_update_on_rdev_error(struct mddev *mddev) if (!log) return; - if (raid5_calc_degraded(conf) > 0 && + if ((raid5_calc_degraded(conf) > 0 || + test_bit(Journal, &rdev->flags)) && conf->log->r5c_journal_mode == R5C_JOURNAL_MODE_WRITE_BACK) schedule_work(&log->disable_writeback_work); } diff --git a/drivers/md/raid5-log.h b/drivers/md/raid5-log.h index 27097101ccca..328d67aedda4 100644 --- a/drivers/md/raid5-log.h +++ b/drivers/md/raid5-log.h @@ -28,7 +28,8 @@ extern void r5c_flush_cache(struct r5conf *conf, int num); extern void r5c_check_stripe_cache_usage(struct r5conf *conf); extern void r5c_check_cached_full_stripe(struct r5conf *conf); extern struct md_sysfs_entry r5c_journal_mode; -extern void r5c_update_on_rdev_error(struct mddev *mddev); +extern void r5c_update_on_rdev_error(struct mddev *mddev, + struct md_rdev *rdev); extern bool r5c_big_stripe_cached(struct r5conf *conf, sector_t sect); extern struct dma_async_tx_descriptor * diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index f8055a7abb4b..0ac57a925606 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -2689,7 +2689,7 @@ static void raid5_error(struct mddev *mddev, struct md_rdev *rdev) bdevname(rdev->bdev, b), mdname(mddev), conf->raid_disks - mddev->degraded); - r5c_update_on_rdev_error(mddev); + r5c_update_on_rdev_error(mddev, rdev); } /* @@ -3050,6 +3050,11 @@ sector_t raid5_compute_blocknr(struct stripe_head *sh, int i, int previous) * When LOG_CRITICAL, stripes with injournal == 0 will be sent to * no_space_stripes list. * + * 3. during journal failure + * In journal failure, we try to flush all cached data to raid disks + * based on data in stripe cache. The array is read-only to upper + * layers, so we would skip all pending writes. + * */ static inline bool delay_towrite(struct r5conf *conf, struct r5dev *dev, @@ -3063,6 +3068,9 @@ static inline bool delay_towrite(struct r5conf *conf, if (test_bit(R5C_LOG_CRITICAL, &conf->cache_state) && s->injournal > 0) return true; + /* case 3 above */ + if (s->log_failed && s->injournal) + return true; return false; } @@ -4696,10 +4704,15 @@ static void handle_stripe(struct stripe_head *sh) " to_write=%d failed=%d failed_num=%d,%d\n", s.locked, s.uptodate, s.to_read, s.to_write, s.failed, s.failed_num[0], s.failed_num[1]); - /* check if the array has lost more than max_degraded devices and, + /* + * check if the array has lost more than max_degraded devices and, * if so, some requests might need to be failed. + * + * When journal device failed (log_failed), we will only process + * the stripe if there is data need write to raid disks */ - if (s.failed > conf->max_degraded || s.log_failed) { + if (s.failed > conf->max_degraded || + (s.log_failed && s.injournal == 0)) { sh->check_state = 0; sh->reconstruct_state = 0; break_stripe_batch_list(sh, 0); @@ -5272,8 +5285,10 @@ static struct stripe_head *__get_priority_stripe(struct r5conf *conf, int group) struct stripe_head *sh, *tmp; struct list_head *handle_list = NULL; struct r5worker_group *wg; - bool second_try = !r5c_is_writeback(conf->log); - bool try_loprio = test_bit(R5C_LOG_TIGHT, &conf->cache_state); + bool second_try = !r5c_is_writeback(conf->log) && + !r5l_log_disk_error(conf); + bool try_loprio = test_bit(R5C_LOG_TIGHT, &conf->cache_state) || + r5l_log_disk_error(conf); again: wg = NULL; @@ -7521,7 +7536,9 @@ static int raid5_remove_disk(struct mddev *mddev, struct md_rdev *rdev) * neilb: there is no locking about new writes here, * so this cannot be safe. */ - if (atomic_read(&conf->active_stripes)) { + if (atomic_read(&conf->active_stripes) || + atomic_read(&conf->r5c_cached_full_stripes) || + atomic_read(&conf->r5c_cached_partial_stripes)) { return -EBUSY; } log_exit(conf); From 5ddf0440a1a28f00f69ed2e093476bab3b60c2c3 Mon Sep 17 00:00:00 2001 From: Song Liu Date: Thu, 11 May 2017 17:03:44 -0700 Subject: [PATCH 7/8] md/r5cache: handle sync with data in write back cache Currently, sync of raid456 array cannot make progress when hitting data in writeback r5cache. This patch fixes this issue by flushing cached data of the stripe before processing the sync request. This is achived by: 1. In handle_stripe(), do not set STRIPE_SYNCING if the stripe is in write back cache; 2. In r5c_try_caching_write(), handle the stripe in sync with write through; 3. In do_release_stripe(), make stripe in sync write out and send it to the state machine. Shaohua: explictly set STRIPE_HANDLE after write out completed Signed-off-by: Song Liu Signed-off-by: Shaohua Li --- drivers/md/raid5-cache.c | 8 +++++++- drivers/md/raid5.c | 21 +++++++++++++++------ 2 files changed, 22 insertions(+), 7 deletions(-) diff --git a/drivers/md/raid5-cache.c b/drivers/md/raid5-cache.c index cc3f8442f11f..4c00bc248287 100644 --- a/drivers/md/raid5-cache.c +++ b/drivers/md/raid5-cache.c @@ -2637,8 +2637,11 @@ int r5c_try_caching_write(struct r5conf *conf, * When run in degraded mode, array is set to write-through mode. * This check helps drain pending write safely in the transition to * write-through mode. + * + * When a stripe is syncing, the write is also handled in write + * through mode. */ - if (s->failed) { + if (s->failed || test_bit(STRIPE_SYNCING, &sh->state)) { r5c_make_stripe_write_out(sh); return -EAGAIN; } @@ -2841,6 +2844,9 @@ void r5c_finish_stripe_write_out(struct r5conf *conf, } r5l_append_flush_payload(log, sh->sector); + /* stripe is flused to raid disks, we can do resync now */ + if (test_bit(STRIPE_SYNC_REQUESTED, &sh->state)) + set_bit(STRIPE_HANDLE, &sh->state); } int r5c_cache_data(struct r5l_log *log, struct stripe_head *sh) diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 0ac57a925606..9c4f7659f8b1 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -233,11 +233,15 @@ static void do_release_stripe(struct r5conf *conf, struct stripe_head *sh, if (test_bit(R5_InJournal, &sh->dev[i].flags)) injournal++; /* - * When quiesce in r5c write back, set STRIPE_HANDLE for stripes with - * data in journal, so they are not released to cached lists + * In the following cases, the stripe cannot be released to cached + * lists. Therefore, we make the stripe write out and set + * STRIPE_HANDLE: + * 1. when quiesce in r5c write back; + * 2. when resync is requested fot the stripe. */ - if (conf->quiesce && r5c_is_writeback(conf->log) && - !test_bit(STRIPE_HANDLE, &sh->state) && injournal != 0) { + if (test_bit(STRIPE_SYNC_REQUESTED, &sh->state) || + (conf->quiesce && r5c_is_writeback(conf->log) && + !test_bit(STRIPE_HANDLE, &sh->state) && injournal != 0)) { if (test_bit(STRIPE_R5C_CACHING, &sh->state)) r5c_make_stripe_write_out(sh); set_bit(STRIPE_HANDLE, &sh->state); @@ -4656,8 +4660,13 @@ static void handle_stripe(struct stripe_head *sh) if (test_bit(STRIPE_SYNC_REQUESTED, &sh->state) && !sh->batch_head) { spin_lock(&sh->stripe_lock); - /* Cannot process 'sync' concurrently with 'discard' */ - if (!test_bit(STRIPE_DISCARD, &sh->state) && + /* + * Cannot process 'sync' concurrently with 'discard'. + * Flush data in r5cache before 'sync'. + */ + if (!test_bit(STRIPE_R5C_PARTIAL_STRIPE, &sh->state) && + !test_bit(STRIPE_R5C_FULL_STRIPE, &sh->state) && + !test_bit(STRIPE_DISCARD, &sh->state) && test_and_clear_bit(STRIPE_SYNC_REQUESTED, &sh->state)) { set_bit(STRIPE_SYNCING, &sh->state); clear_bit(STRIPE_INSYNC, &sh->state); From d82dd0e34d0347be201fd274dc84cd645dccc064 Mon Sep 17 00:00:00 2001 From: Tomasz Majchrzak Date: Fri, 12 May 2017 14:26:10 +0200 Subject: [PATCH 8/8] raid1: prefer disk without bad blocks If an array consists of two drives and the first drive has the bad block, the read request to the region overlapping the bad block chooses the same disk (with bad block) as device to read from over and over and the request gets stuck. If the first disk only partially overlaps with bad block, it becomes a candidate ("best disk") for shorter range of sectors. The second disk is capable of reading the entire requested range and it is updated accordingly, however it is not recorded as a best device for the request. In the end the request is sent to the first disk to read entire range of sectors. It fails and is re-tried in a moment but with the same outcome. Actually it is quite likely scenario but it had little exposure in my test until commit 715d40b93b10 ("md/raid1: add failfast handling for reads.") removed preference for idle disk. Such scenario had been passing as second disk was always chosen when idle. Reset a candidate ("best disk") to read from if disk can read entire range. Do it only if other disk has already been chosen as a candidate for a smaller range. The head position / disk type logic will select the best disk to read from - it is fine as disk with bad block won't be considered for it. Signed-off-by: Tomasz Majchrzak Signed-off-by: Shaohua Li --- drivers/md/raid1.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c index a17ed6218d51..af5056d56878 100644 --- a/drivers/md/raid1.c +++ b/drivers/md/raid1.c @@ -666,8 +666,11 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect break; } continue; - } else + } else { + if ((sectors > best_good_sectors) && (best_disk >= 0)) + best_disk = -1; best_good_sectors = sectors; + } if (best_disk >= 0) /* At least two disks to choose from so failfast is OK */