queue-6.1/md-raid10-prevent-soft-lockup-while-flush-writes.patch

   1 From 5036866ba258e5dd80caf93c4ce92bf13eef7d82 Mon Sep 17 00:00:00 2001
   2 From: Sasha Levin <sashal@kernel.org>
   3 Date: Mon, 29 May 2023 21:11:00 +0800
   4 Subject: md/raid10: prevent soft lockup while flush writes
   5
   6 From: Yu Kuai <yukuai3@huawei.com>
   7
   8 [ Upstream commit 010444623e7f4da6b4a4dd603a7da7469981e293 ]
   9
  10 Currently, there is no limit for raid1/raid10 plugged bio. While flushing
  11 writes, raid1 has cond_resched() while raid10 doesn't, and too many
  12 writes can cause soft lockup.
  13
  14 Follow up soft lockup can be triggered easily with writeback test for
  15 raid10 with ramdisks:
  16
  17 watchdog: BUG: soft lockup - CPU#10 stuck for 27s! [md0_raid10:1293]
  18 Call Trace:
  19  <TASK>
  20  call_rcu+0x16/0x20
  21  put_object+0x41/0x80
  22  __delete_object+0x50/0x90
  23  delete_object_full+0x2b/0x40
  24  kmemleak_free+0x46/0xa0
  25  slab_free_freelist_hook.constprop.0+0xed/0x1a0
  26  kmem_cache_free+0xfd/0x300
  27  mempool_free_slab+0x1f/0x30
  28  mempool_free+0x3a/0x100
  29  bio_free+0x59/0x80
  30  bio_put+0xcf/0x2c0
  31  free_r10bio+0xbf/0xf0
  32  raid_end_bio_io+0x78/0xb0
  33  one_write_done+0x8a/0xa0
  34  raid10_end_write_request+0x1b4/0x430
  35  bio_endio+0x175/0x320
  36  brd_submit_bio+0x3b9/0x9b7 [brd]
  37  __submit_bio+0x69/0xe0
  38  submit_bio_noacct_nocheck+0x1e6/0x5a0
  39  submit_bio_noacct+0x38c/0x7e0
  40  flush_pending_writes+0xf0/0x240
  41  raid10d+0xac/0x1ed0
  42
  43 Fix the problem by adding cond_resched() to raid10 like what raid1 did.
  44
  45 Note that unlimited plugged bio still need to be optimized, for example,
  46 in the case of lots of dirty pages writeback, this will take lots of
  47 memory and io will spend a long time in plug, hence io latency is bad.
  48
  49 Signed-off-by: Yu Kuai <yukuai3@huawei.com>
  50 Signed-off-by: Song Liu <song@kernel.org>
  51 Link: https://lore.kernel.org/r/20230529131106.2123367-2-yukuai1@huaweicloud.com
  52 Signed-off-by: Sasha Levin <sashal@kernel.org>
  53 ---
  54  drivers/md/raid10.c | 2 ++
  55  1 file changed, 2 insertions(+)
  56
  57 diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
  58 index 7b318e7e8d459..009f7ffe4e10c 100644
  59 --- a/drivers/md/raid10.c
  60 +++ b/drivers/md/raid10.c
  61 @@ -920,6 +920,7 @@ static void flush_pending_writes(struct r10conf *conf)
  62
  63                         raid1_submit_write(bio);
  64                         bio = next;
  65 +                       cond_resched();
  66                 }
  67                 blk_finish_plug(&plug);
  68         } else
  69 @@ -1130,6 +1131,7 @@ static void raid10_unplug(struct blk_plug_cb *cb, bool from_schedule)
  70
  71                 raid1_submit_write(bio);
  72                 bio = next;
  73 +               cond_resched();
  74         }
  75         kfree(plug);
  76  }
  77 --
  78 2.43.0
  79