[v5,00/22] btrfs: async discard support
mbox series

Message ID cover.1575919745.git.dennis@kernel.org
Headers show
Series
  • btrfs: async discard support
Related show

Message

Dennis Zhou Dec. 9, 2019, 7:45 p.m. UTC
Hello,

Dave reported that with async discard enabled, relocation fails [1].
This could be caused by two things. First, if we unpin extents, that
means we haven't fully discarded the block group and need to let async
discard revisit it. Second, relocation removes block_groups outside of
the normal. I fixed both issues and now it successfully passes xfstests
btrfs/003.

Changes in v5:
 - Changed the rules so free space is always added as the right type
   based on discard settings (see btrfs_add_free_space()), this removes
   the need to pass around trim_state in unpin_extent_range().
 - Handled relocation block group deletion (xfstests btrfs/003)
 - When adding to the discard lists, make sure the work queue is active.
   (made all additions go through either btrfs_discard_queue_work() or
   btrfs_discard_check_filter()).
 - Added 10 sec reuse timeout for fully empty block groups.

v4 is available here: [2].

This series is on top of btrfs-devel#misc-next fbc0468e42d2 + [3].

[1] https://lore.kernel.org/linux-btrfs/20191126215204.GP2734@twin.jikos.cz/
[2] https://lore.kernel.org/linux-btrfs/cover.1574709825.git.dennis@kernel.org/
[3] https://lore.kernel.org/linux-btrfs/20191209193846.18162-1-dennis@kernel.org/

Dennis Zhou (22):
  bitmap: genericize percpu bitmap region iterators
  btrfs: rename DISCARD opt to DISCARD_SYNC
  btrfs: keep track of which extents have been discarded
  btrfs: keep track of cleanliness of the bitmap
  btrfs: add the beginning of async discard, discard workqueue
  btrfs: handle empty block_group removal
  btrfs: discard one region at a time in async discard
  btrfs: add removal calls for sysfs debug/
  btrfs: make UUID/debug have its own kobject
  btrfs: add discard sysfs directory
  btrfs: track discardable extents for async discard
  btrfs: keep track of discardable_bytes
  btrfs: calculate discard delay based on number of extents
  btrfs: add bps discard rate limit
  btrfs: limit max discard size for async discard
  btrfs: make max async discard size tunable
  btrfs: have multiple discard lists
  btrfs: only keep track of data extents for async discard
  btrfs: keep track of discard reuse stats
  btrfs: add async discard header
  btrfs: increase the metadata allowance for the free_space_cache
  btrfs: make smaller extents more likely to go into bitmaps

 fs/btrfs/Makefile           |   2 +-
 fs/btrfs/block-group.c      |  88 ++++-
 fs/btrfs/block-group.h      |  30 ++
 fs/btrfs/ctree.h            |  52 ++-
 fs/btrfs/discard.c          | 684 ++++++++++++++++++++++++++++++++++++
 fs/btrfs/discard.h          |  42 +++
 fs/btrfs/disk-io.c          |  15 +-
 fs/btrfs/extent-tree.c      |   8 +-
 fs/btrfs/free-space-cache.c | 611 +++++++++++++++++++++++++++-----
 fs/btrfs/free-space-cache.h |  41 ++-
 fs/btrfs/inode-map.c        |  13 +-
 fs/btrfs/inode.c            |   2 +-
 fs/btrfs/scrub.c            |   7 +-
 fs/btrfs/super.c            |  39 +-
 fs/btrfs/sysfs.c            | 205 ++++++++++-
 fs/btrfs/volumes.c          |   7 +
 include/linux/bitmap.h      |  35 ++
 mm/percpu.c                 |  61 +---
 18 files changed, 1789 insertions(+), 153 deletions(-)
 create mode 100644 fs/btrfs/discard.c
 create mode 100644 fs/btrfs/discard.h

Comments

David Sterba Dec. 10, 2019, 2:04 p.m. UTC | #1
On Mon, Dec 09, 2019 at 11:45:45AM -0800, Dennis Zhou wrote:
> Hello,
> 
> Dave reported that with async discard enabled, relocation fails [1].
> This could be caused by two things. First, if we unpin extents, that
> means we haven't fully discarded the block group and need to let async
> discard revisit it. Second, relocation removes block_groups outside of
> the normal. I fixed both issues and now it successfully passes xfstests
> btrfs/003.
> 
> Changes in v5:
>  - Changed the rules so free space is always added as the right type
>    based on discard settings (see btrfs_add_free_space()), this removes
>    the need to pass around trim_state in unpin_extent_range().
>  - Handled relocation block group deletion (xfstests btrfs/003)
>  - When adding to the discard lists, make sure the work queue is active.
>    (made all additions go through either btrfs_discard_queue_work() or
>    btrfs_discard_check_filter()).
>  - Added 10 sec reuse timeout for fully empty block groups.

btrfs/011 reports potential deadlock:

[ 2233.150902] ======================================================
[ 2233.155184] WARNING: possible circular locking dependency detected
[ 2233.159746] 5.5.0-rc1-default+ #902 Not tainted
[ 2233.162553] ------------------------------------------------------
[ 2233.165116] btrfs-cleaner/3837 is trying to acquire lock:
[ 2233.166725] ffff909174ab6218 (&(&ctl->tree_lock)->rlock){+.+.}, at: btrfs_is_free_space_trimmed+0x17/0x70 [btrfs]
[ 2233.168932]
[ 2233.168932] but task is already holding lock:
[ 2233.170626] ffff909160b24c28 (&(&cache->lock)->rlock){+.+.}, at: btrfs_delete_unused_bgs+0x113/0x880 [btrfs]
[ 2233.173347]
[ 2233.173347] which lock already depends on the new lock.
[ 2233.173347]
[ 2233.176447]
[ 2233.176447] the existing dependency chain (in reverse order) is:
[ 2233.179479]
[ 2233.179479] -> #3 (&(&cache->lock)->rlock){+.+.}:
[ 2233.182365]        lock_acquire+0x95/0x1a0
[ 2233.183768]        _raw_spin_lock+0x31/0x80
[ 2233.185260]        btrfs_add_reserved_bytes+0x3c/0x440 [btrfs]
[ 2233.186583]        find_free_extent+0x63e/0xf10 [btrfs]
[ 2233.187736]        btrfs_reserve_extent+0x9b/0x180 [btrfs]
[ 2233.188923]        btrfs_alloc_tree_block+0xc1/0x350 [btrfs]
[ 2233.190150]        alloc_tree_block_no_bg_flush+0x4a/0x60 [btrfs]
[ 2233.200938]        __btrfs_cow_block+0x143/0x7a0 [btrfs]
[ 2233.202579]        btrfs_cow_block+0x15f/0x310 [btrfs]
[ 2233.204288]        commit_cowonly_roots+0x55/0x310 [btrfs]
[ 2233.206106]        btrfs_commit_transaction+0x505/0xaf0 [btrfs]
[ 2233.207801]        sync_filesystem+0x6e/0x90
[ 2233.209176]        generic_shutdown_super+0x22/0x100
[ 2233.210647]        kill_anon_super+0x14/0x30
[ 2233.212040]        btrfs_kill_super+0x12/0xa0 [btrfs]
[ 2233.213704]        deactivate_locked_super+0x2c/0x70
[ 2233.215428]        cleanup_mnt+0x100/0x160
[ 2233.216996]        task_work_run+0x90/0xc0
[ 2233.218500]        exit_to_usermode_loop+0x96/0xa0
[ 2233.220150]        do_syscall_64+0x1df/0x210
[ 2233.221654]        entry_SYSCALL_64_after_hwframe+0x49/0xbe
[ 2233.223667]
[ 2233.223667] -> #2 (&(&space_info->lock)->rlock){+.+.}:
[ 2233.226158]        lock_acquire+0x95/0x1a0
[ 2233.227665]        _raw_spin_lock+0x31/0x80
[ 2233.229209]        __btrfs_block_rsv_release+0x1a6/0x440 [btrfs]
[ 2233.231268]        btrfs_inode_rsv_release+0x4f/0x190 [btrfs]
[ 2233.232915]        btrfs_clear_delalloc_extent+0x155/0x4b0 [btrfs]
[ 2233.234643]        clear_state_bit+0x84/0x1c0 [btrfs]
[ 2233.236127]        __clear_extent_bit+0x22d/0x5b0 [btrfs]
[ 2233.237741]        clear_extent_bit+0x15/0x20 [btrfs]
[ 2233.239452]        btrfs_invalidatepage+0x29b/0x2f0 [btrfs]
[ 2233.241379]        truncate_cleanup_page+0x42/0xa0
[ 2233.243061]        truncate_inode_pages_range+0x2bf/0xb40
[ 2233.244935]        truncate_pagecache+0x44/0x60
[ 2233.246513]        btrfs_setsize+0x11f/0x4d0 [btrfs]
[ 2233.248233]        btrfs_setattr+0x5c/0xe0 [btrfs]
[ 2233.249655]        notify_change+0x283/0x415
[ 2233.251138]        do_truncate+0x76/0xd0
[ 2233.252472]        do_last+0x4a5/0x7e0
[ 2233.253775]        path_openat+0xa2/0x250
[ 2233.255239]        do_filp_open+0x91/0x100
[ 2233.256626]        do_sys_open+0x184/0x220
[ 2233.258098]        do_syscall_64+0x50/0x210
[ 2233.259539]        entry_SYSCALL_64_after_hwframe+0x49/0xbe
[ 2233.261337] -> #1 (&(&tree->lock)->rlock){+.+.}:
[ 2233.263598]        lock_acquire+0x95/0x1a0
[ 2233.265134]        _raw_spin_lock+0x31/0x80
[ 2233.266726]        find_first_extent_bit+0x32/0x150 [btrfs]
[ 2233.268354]        write_pinned_extent_entries+0xc5/0x100 [btrfs]
[ 2233.270202]        __btrfs_write_out_cache+0x16d/0x4a0 [btrfs]
[ 2233.272011]        btrfs_write_out_cache+0x7a/0xf0 [btrfs]
[ 2233.273581]        btrfs_write_dirty_block_groups+0x286/0x3a0 [btrfs]
[ 2233.275305]        commit_cowonly_roots+0x24f/0x310 [btrfs]
[ 2233.276808]        btrfs_commit_transaction+0x505/0xaf0 [btrfs]
[ 2233.278699]        sync_filesystem+0x6e/0x90
[ 2233.280351]        generic_shutdown_super+0x22/0x100
[ 2233.282198]        kill_anon_super+0x14/0x30
[ 2233.283907]        btrfs_kill_super+0x12/0xa0 [btrfs]
[ 2233.285774]        deactivate_locked_super+0x2c/0x70
[ 2233.287660]        cleanup_mnt+0x100/0x160
[ 2233.289326]        task_work_run+0x90/0xc0
[ 2233.291035]        exit_to_usermode_loop+0x96/0xa0
[ 2233.292580]        do_syscall_64+0x1df/0x210
[ 2233.293892]        entry_SYSCALL_64_after_hwframe+0x49/0xbe
[ 2233.295558]
[ 2233.295558] -> #0 (&(&ctl->tree_lock)->rlock){+.+.}:
[ 2233.298058]        check_prev_add+0xa2/0xa90
[ 2233.299576]        __lock_acquire+0xe97/0x1320
[ 2233.301189]        lock_acquire+0x95/0x1a0
[ 2233.302709]        _raw_spin_lock+0x31/0x80
[ 2233.304283]        btrfs_is_free_space_trimmed+0x17/0x70 [btrfs]
[ 2233.306283]        btrfs_delete_unused_bgs+0x2b3/0x880 [btrfs]
[ 2233.308376]        cleaner_kthread+0x162/0x170 [btrfs]
[ 2233.310020]        kthread+0x122/0x140
[ 2233.311285]        ret_from_fork+0x24/0x30
[ 2233.312619]
[ 2233.312619] other info that might help us debug this:
[ 2233.312619]
[ 2233.315394] Chain exists of:
[ 2233.315394]   &(&ctl->tree_lock)->rlock --> &(&space_info->lock)->rlock --> &(&cache->lock)->rlock
[ 2233.315394]
[ 2233.319913]  Possible unsafe locking scenario:
[ 2233.319913]
[ 2233.322174]        CPU0                    CPU1
[ 2233.323685]        ----                    ----
[ 2233.325156]   lock(&(&cache->lock)->rlock);
[ 2233.326601]                                lock(&(&space_info->lock)->rlock);
[ 2233.328588]                                lock(&(&cache->lock)->rlock);
[ 2233.330542]   lock(&(&ctl->tree_lock)->rlock);
[ 2233.332055]
[ 2233.332055]  *** DEADLOCK ***
[ 2233.332055]
[ 2233.334678] 3 locks held by btrfs-cleaner/3837:
[ 2233.336219]  #0: ffff90914c2bfb88 (&fs_info->delete_unused_bgs_mutex){+.+.}, at: btrfs_delete_unused_bgs+0x103/0x880 [btrfs]
[ 2233.339756]  #1: ffff909160b235c8 (&space_info->groups_sem){++++}, at: btrfs_delete_unused_bgs+0x10b/0x880 [btrfs]
[ 2233.343098]  #2: ffff909160b24c28 (&(&cache->lock)->rlock){+.+.}, at: btrfs_delete_unused_bgs+0x113/0x880 [btrfs]
[ 2233.346574]
[ 2233.346574] stack backtrace:
[ 2233.348723] CPU: 0 PID: 3837 Comm: btrfs-cleaner Not tainted 5.5.0-rc1-default+ #902
[ 2233.351558] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.12.1-0-ga5cab58-rebuilt.opensuse.org 04/01/2014
[ 2233.355122] Call Trace:
[ 2233.356104]  dump_stack+0x71/0xa0
[ 2233.357228]  check_noncircular+0x177/0x190
[ 2233.358640]  check_prev_add+0xa2/0xa90
[ 2233.359958]  ? kvm_sched_clock_read+0x14/0x30
[ 2233.361305]  __lock_acquire+0xe97/0x1320
[ 2233.362583]  lock_acquire+0x95/0x1a0
[ 2233.363853]  ? btrfs_is_free_space_trimmed+0x17/0x70 [btrfs]
[ 2233.365428]  _raw_spin_lock+0x31/0x80
[ 2233.366644]  ? btrfs_is_free_space_trimmed+0x17/0x70 [btrfs]
[ 2233.368228]  btrfs_is_free_space_trimmed+0x17/0x70 [btrfs]
[ 2233.369861]  btrfs_delete_unused_bgs+0x2b3/0x880 [btrfs]
[ 2233.371672]  ? __mutex_unlock_slowpath+0x45/0x2a0
[ 2233.373291]  cleaner_kthread+0x162/0x170 [btrfs]
[ 2233.374767]  ? __btrfs_btree_balance_dirty+0x60/0x60 [btrfs]
[ 2233.376396]  kthread+0x122/0x140
[ 2233.377525]  ? kthread_create_worker_on_cpu+0x70/0x70
[ 2233.379210]  ret_from_fork+0x24/0x30