From 125bfc7cd750e68c99f1d446e2c22abea08c237f Mon Sep 17 00:00:00 2001 From: Li Nan Date: Fri, 9 Jun 2023 17:43:20 +0800 Subject: [PATCH 1/5] md/raid10: fix the condition to call bio_end_io_acct() /sys/block/[device]/queue/iostats is used to control whether to count io stat. Write 0 to it will clear queue_flags QUEUE_FLAG_IO_STAT which means iostats is disabled. If we disable iostats and later endable it, the io issued during this period will be counted incorrectly, inflight will be decreased to -1. //T1 set iostats echo 0 > /sys/block/md0/queue/iostats clear QUEUE_FLAG_IO_STAT //T2 issue io if (QUEUE_FLAG_IO_STAT) -> false bio_start_io_acct inflight++ echo 1 > /sys/block/md0/queue/iostats set QUEUE_FLAG_IO_STAT //T3 io end if (QUEUE_FLAG_IO_STAT) -> true bio_end_io_acct inflight-- -> -1 Also, if iostats is enabled while issuing io but disabled while io end, inflight will never be decreased. Fix it by checking start_time when io end. If start_time is not 0, call bio_end_io_acct(). Fixes: 528bc2cf2fcc ("md/raid10: enable io accounting") Signed-off-by: Li Nan Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20230609094320.2397604-1-linan666@huaweicloud.com --- drivers/md/raid10.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c index d0de8c9fb3cf..79067769e44b 100644 --- a/drivers/md/raid10.c +++ b/drivers/md/raid10.c @@ -325,7 +325,7 @@ static void raid_end_bio_io(struct r10bio *r10_bio) if (!test_bit(R10BIO_Uptodate, &r10_bio->state)) bio->bi_status = BLK_STS_IOERR; - if (blk_queue_io_stat(bio->bi_bdev->bd_disk->queue)) + if (r10_bio->start_time) bio_end_io_acct(bio, r10_bio->start_time); bio_endio(bio); /* From b5a99602b74bbfa655be509c615181dd95b0719e Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Fri, 16 Jun 2023 09:21:36 +0800 Subject: [PATCH 2/5] md/raid1-10: fix casting from randomized structure in raid1_submit_write() Following build error triggered while build with clang version 17.0.0 with W=1(this can't be reporduced with gcc 13.1.0): drivers/md/raid1-10.c:117:25: error: casting from randomized structure pointer type 'struct block_device *' to 'struct md_rdev *' 117 | struct md_rdev *rdev = (struct md_rdev *)bio->bi_bdev; | ^ Fix this by casting 'bio->bi_bdev' to 'void *', as it used to be. Reported-by: kernel test robot Closes: https://lore.kernel.org/oe-kbuild-all/202306142042.fmjfmTF8-lkp@intel.com/ Fixes: 8295efbe68c0 ("md/raid1-10: factor out a helper to submit normal write") Signed-off-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20230616012136.3047071-1-yukuai1@huaweicloud.com --- drivers/md/raid1-10.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/md/raid1-10.c b/drivers/md/raid1-10.c index 169ebe296f2d..3f22edec70e7 100644 --- a/drivers/md/raid1-10.c +++ b/drivers/md/raid1-10.c @@ -116,7 +116,7 @@ static void md_bio_reset_resync_pages(struct bio *bio, struct resync_pages *rp, static inline void raid1_submit_write(struct bio *bio) { - struct md_rdev *rdev = (struct md_rdev *)bio->bi_bdev; + struct md_rdev *rdev = (void *)bio->bi_bdev; bio->bi_next = NULL; bio_set_dev(bio, rdev->bdev); From a1d7671910965ca9f8f0377e7e3bfd1179fba4d8 Mon Sep 17 00:00:00 2001 From: Song Liu Date: Fri, 16 Jun 2023 22:24:04 -0700 Subject: [PATCH 3/5] md: use mddev->external to select holder in export_rdev() mdadm test "10ddf-create-fail-rebuild" triggers warnings like the following [ 215.526357] ------------[ cut here ]------------ [ 215.527243] WARNING: CPU: 18 PID: 1264 at block/bdev.c:617 blkdev_put+0x269/0x350 [ 215.528334] Modules linked in: [ 215.528806] CPU: 18 PID: 1264 Comm: mdmon Not tainted 6.4.0-rc2+ #768 [ 215.529863] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS [ 215.531464] RIP: 0010:blkdev_put+0x269/0x350 [ 215.532167] Code: ff ff 49 8d 7d 10 e8 56 bf b8 ff 4d 8b 65 10 49 8d bc 24 58 05 00 00 e8 05 be b8 ff 41 83 ac 24 58 05 00 00 01 e9 44 ff ff ff <0f> 0b e9 52 fe ff ff 0f 0b e9 6b fe ff ff1 [ 215.534780] RSP: 0018:ffffc900040bfbf0 EFLAGS: 00010283 [ 215.535635] RAX: ffff888174001000 RBX: ffff88810b1c3b00 RCX: ffffffff819a4061 [ 215.536645] RDX: dffffc0000000000 RSI: dffffc0000000000 RDI: ffff88810b1c3ba0 [ 215.537657] RBP: ffff88810dbde800 R08: fffffbfff0fca983 R09: fffffbfff0fca983 [ 215.538674] R10: ffffc900040bfbf0 R11: fffffbfff0fca982 R12: ffff88810b1c3b38 [ 215.539687] R13: ffff88810b1c3b10 R14: ffff88810dbdecb8 R15: ffff88810b1c3b00 [ 215.540833] FS: 00007f2aabdff700(0000) GS:ffff888dfb400000(0000) knlGS:0000000000000000 [ 215.541961] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 215.542775] CR2: 00007fa19a85d934 CR3: 000000010c076006 CR4: 0000000000370ee0 [ 215.543814] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [ 215.544840] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [ 215.545885] Call Trace: [ 215.546257] [ 215.546608] export_rdev.isra.63+0x71/0xe0 [ 215.547338] mddev_unlock+0x1b1/0x2d0 [ 215.547898] array_state_store+0x28d/0x450 [ 215.548519] md_attr_store+0xd7/0x150 [ 215.549059] ? __pfx_sysfs_kf_write+0x10/0x10 [ 215.549702] kernfs_fop_write_iter+0x1b9/0x260 [ 215.550351] vfs_write+0x491/0x760 [ 215.550863] ? __pfx_vfs_write+0x10/0x10 [ 215.551445] ? __fget_files+0x156/0x230 [ 215.552053] ksys_write+0xc0/0x160 [ 215.552570] ? __pfx_ksys_write+0x10/0x10 [ 215.553141] ? ktime_get_coarse_real_ts64+0xec/0x100 [ 215.553878] do_syscall_64+0x3a/0x90 [ 215.554403] entry_SYSCALL_64_after_hwframe+0x72/0xdc [ 215.555125] RIP: 0033:0x7f2aade11847 [ 215.555696] Code: c3 66 90 41 54 49 89 d4 55 48 89 f5 53 89 fb 48 83 ec 10 e8 1b fd ff ff 4c 89 e2 48 89 ee 89 df 41 89 c0 b8 01 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 35 44 89 c7 48 89 448 [ 215.558398] RSP: 002b:00007f2aabdfeba0 EFLAGS: 00000293 ORIG_RAX: 0000000000000001 [ 215.559516] RAX: ffffffffffffffda RBX: 0000000000000010 RCX: 00007f2aade11847 [ 215.560515] RDX: 0000000000000005 RSI: 0000000000438b8b RDI: 0000000000000010 [ 215.561512] RBP: 0000000000438b8b R08: 0000000000000000 R09: 00007f2aaecf0060 [ 215.562511] R10: 000000000e3ba40b R11: 0000000000000293 R12: 0000000000000005 [ 215.563647] R13: 0000000000000000 R14: 0000000000000001 R15: 0000000000c70750 [ 215.564693] [ 215.565029] irq event stamp: 15979 [ 215.565584] hardirqs last enabled at (15991): [] __up_console_sem+0x52/0x60 [ 215.566806] hardirqs last disabled at (16000): [] __up_console_sem+0x37/0x60 [ 215.568022] softirqs last enabled at (15716): [] __do_softirq+0x3eb/0x531 [ 215.569239] softirqs last disabled at (15711): [] irq_exit_rcu+0x115/0x160 [ 215.570434] ---[ end trace 0000000000000000 ]--- This means export_rdev() calls blkdev_put with a different holder than the one used by blkdev_get_by_dev(). This is because mddev->major_version == -2 is not a good check for external metadata. Fix this by using mddev->external instead. Also, do not clear mddev->external in md_clean(), as the flag might be used later in export_rdev(). Fixes: 2736e8eeb0cc ("block: use the holder as indication for exclusive opens") Cc: Christoph Hellwig Cc: Jens Axboe Signed-off-by: Song Liu Reviewed-by: Christoph Hellwig Link: https://lore.kernel.org/r/20230617052405.305871-1-song@kernel.org --- drivers/md/md.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/md/md.c b/drivers/md/md.c index cf3733c90c47..8e7cc2e69bc9 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -2458,7 +2458,7 @@ static void export_rdev(struct md_rdev *rdev, struct mddev *mddev) if (test_bit(AutoDetected, &rdev->flags)) md_autodetect_dev(rdev->bdev->bd_dev); #endif - blkdev_put(rdev->bdev, mddev->major_version == -2 ? &claim_rdev : rdev); + blkdev_put(rdev->bdev, mddev->external ? &claim_rdev : rdev); rdev->bdev = NULL; kobject_put(&rdev->kobj); } @@ -6140,7 +6140,7 @@ static void md_clean(struct mddev *mddev) mddev->resync_min = 0; mddev->resync_max = MaxSector; mddev->reshape_position = MaxSector; - mddev->external = 0; + /* we still need mddev->external in export_rdev, do not clear it yet */ mddev->persistent = 0; mddev->level = LEVEL_NONE; mddev->clevel[0] = 0; From 4934b6401a812f9fe368e7d2d091cd1d120ea262 Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Wed, 21 Jun 2023 22:29:33 +0800 Subject: [PATCH 4/5] md: fix 'delete_mutex' deadlock Commit 3ce94ce5d05a ("md: fix duplicate filename for rdev") introduce a new lock 'delete_mutex', and trigger a new deadlock: t1: remove rdev t2: sysfs writer rdev_attr_store rdev_attr_store mddev_lock state_store md_kick_rdev_from_array lock delete_mutex list_add mddev->deleting unlock delete_mutex mddev_unlock mddev_lock ... lock delete_mutex kobject_del // wait for sysfs writers to be done mddev_unlock lock delete_mutex // wait for delete_mutex, deadlock 'delete_mutex' is used to protect the list 'mddev->deleting', turns out that this list can be protected by 'reconfig_mutex' directly, and this lock can be removed. Fix this problem by removing the lock, and use 'reconfig_mutex' to protect the list. mddev_unlock() will move this list to a local list to be handled after 'reconfig_mutex' is dropped. Fixes: 3ce94ce5d05a ("md: fix duplicate filename for rdev") Signed-off-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20230621142933.1395629-1-yukuai1@huaweicloud.com --- drivers/md/md.c | 28 +++++++++------------------- drivers/md/md.h | 4 +--- 2 files changed, 10 insertions(+), 22 deletions(-) diff --git a/drivers/md/md.c b/drivers/md/md.c index 8e7cc2e69bc9..2e38ef421d69 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -643,7 +643,6 @@ void mddev_init(struct mddev *mddev) { mutex_init(&mddev->open_mutex); mutex_init(&mddev->reconfig_mutex); - mutex_init(&mddev->delete_mutex); mutex_init(&mddev->bitmap_info.mutex); INIT_LIST_HEAD(&mddev->disks); INIT_LIST_HEAD(&mddev->all_mddevs); @@ -749,26 +748,15 @@ static void mddev_free(struct mddev *mddev) static const struct attribute_group md_redundancy_group; -static void md_free_rdev(struct mddev *mddev) +void mddev_unlock(struct mddev *mddev) { struct md_rdev *rdev; struct md_rdev *tmp; + LIST_HEAD(delete); - mutex_lock(&mddev->delete_mutex); - if (list_empty(&mddev->deleting)) - goto out; + if (!list_empty(&mddev->deleting)) + list_splice_init(&mddev->deleting, &delete); - list_for_each_entry_safe(rdev, tmp, &mddev->deleting, same_set) { - list_del_init(&rdev->same_set); - kobject_del(&rdev->kobj); - export_rdev(rdev, mddev); - } -out: - mutex_unlock(&mddev->delete_mutex); -} - -void mddev_unlock(struct mddev *mddev) -{ if (mddev->to_remove) { /* These cannot be removed under reconfig_mutex as * an access to the files will try to take reconfig_mutex @@ -808,7 +796,11 @@ void mddev_unlock(struct mddev *mddev) } else mutex_unlock(&mddev->reconfig_mutex); - md_free_rdev(mddev); + list_for_each_entry_safe(rdev, tmp, &delete, same_set) { + list_del_init(&rdev->same_set); + kobject_del(&rdev->kobj); + export_rdev(rdev, mddev); + } md_wakeup_thread(mddev->thread); wake_up(&mddev->sb_wait); @@ -2488,9 +2480,7 @@ static void md_kick_rdev_from_array(struct md_rdev *rdev) * reconfig_mutex is held, hence it can't be called under * reconfig_mutex and it's delayed to mddev_unlock(). */ - mutex_lock(&mddev->delete_mutex); list_add(&rdev->same_set, &mddev->deleting); - mutex_unlock(&mddev->delete_mutex); } static void export_array(struct mddev *mddev) diff --git a/drivers/md/md.h b/drivers/md/md.h index bfd2306bc750..1aef86bf3fc3 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h @@ -531,11 +531,9 @@ struct mddev { /* * Temporarily store rdev that will be finally removed when - * reconfig_mutex is unlocked. + * reconfig_mutex is unlocked, protected by reconfig_mutex. */ struct list_head deleting; - /* Protect the deleting list */ - struct mutex delete_mutex; bool has_superblocks:1; bool fail_last_dev:1; From a8d5fdd4d2702d0c7ec125bd3bbce3fc589afa67 Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Wed, 21 Jun 2023 18:57:28 +0800 Subject: [PATCH 5/5] raid10: avoid spin_lock from fastpath from raid10_unplug() Commit 0c0be98bbe67 ("md/raid10: prevent unnecessary calls to wake_up() in fast path") missed one place, for example, with: fio -direct=1 -rw=write/randwrite -iodepth=1 ... Plug and unplug are called for each io, then wake_up() from raid10_unplug() will cause lock contention as well. Avoid this contention by using wake_up_barrier() instead of wake_up(), where spin_lock is not held if waitqueue is empty. Fio test script: [global] name=random reads and writes ioengine=libaio direct=1 readwrite=randrw rwmixread=70 iodepth=64 buffered=0 filename=/dev/md0 size=1G runtime=30 time_based randrepeat=0 norandommap refill_buffers ramp_time=10 bs=4k numjobs=400 group_reporting=1 [job1] Test result with ramdisk raid10(By Ali): Before this patch With this patch READ IOPS=2033k IOPS=3642k WRITE IOPS=871k IOPS=1561K By the way, in this scenario, blk_plug_cb() will be allocated and freed for each io, this seems need to be optimized as well. Reported-and-tested-by: Ali Gholami Rudi Closes: https://lore.kernel.org/all/20231606122233@laper.mirepesht/ Signed-off-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20230621105728.1268542-1-yukuai1@huaweicloud.com --- drivers/md/raid10.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c index 79067769e44b..5051149e27bb 100644 --- a/drivers/md/raid10.c +++ b/drivers/md/raid10.c @@ -1118,7 +1118,7 @@ static void raid10_unplug(struct blk_plug_cb *cb, bool from_schedule) spin_lock_irq(&conf->device_lock); bio_list_merge(&conf->pending_bio_list, &plug->pending); spin_unlock_irq(&conf->device_lock); - wake_up(&conf->wait_barrier); + wake_up_barrier(conf); md_wakeup_thread(mddev->thread); kfree(plug); return; @@ -1127,7 +1127,7 @@ static void raid10_unplug(struct blk_plug_cb *cb, bool from_schedule) /* we aren't scheduling, so we can do the write-out directly. */ bio = bio_list_get(&plug->pending); raid1_prepare_flush_writes(mddev->bitmap); - wake_up(&conf->wait_barrier); + wake_up_barrier(conf); while (bio) { /* submit pending writes */ struct bio *next = bio->bi_next;