Commit 6d82ad13 authored by Naohiro Aota's avatar Naohiro Aota Committed by David Sterba

btrfs: release correct delalloc amount in direct IO write path

Running generic/406 causes the following WARNING in btrfs_destroy_inode()
which tells there are outstanding extents left.

In btrfs_get_blocks_direct_write(), we reserve a temporary outstanding
extents with btrfs_delalloc_reserve_metadata() (or indirectly from
btrfs_delalloc_reserve_space(()). We then release the outstanding extents
with btrfs_delalloc_release_extents(). However, the "len" can be modified
in the COW case, which releases fewer outstanding extents than expected.

Fix it by calling btrfs_delalloc_release_extents() for the original length.

To reproduce the warning, the filesystem should be 1 GiB.  It's
triggering a short-write, due to not being able to allocate a large
extent and instead allocating a smaller one.

  WARNING: CPU: 0 PID: 757 at fs/btrfs/inode.c:8848 btrfs_destroy_inode+0x1e6/0x210 [btrfs]
  Modules linked in: btrfs blake2b_generic xor lzo_compress
  lzo_decompress raid6_pq zstd zstd_decompress zstd_compress xxhash zram
  zsmalloc
  CPU: 0 PID: 757 Comm: umount Not tainted 5.17.0-rc8+ #101
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS d55cb5a 04/01/2014
  RIP: 0010:btrfs_destroy_inode+0x1e6/0x210 [btrfs]
  RSP: 0018:ffffc9000327bda8 EFLAGS: 00010206
  RAX: 0000000000000000 RBX: ffff888100548b78 RCX: 0000000000000000
  RDX: 0000000000026900 RSI: 0000000000000000 RDI: ffff888100548b78
  RBP: ffff888100548940 R08: 0000000000000000 R09: ffff88810b48aba8
  R10: 0000000000000001 R11: ffff8881004eb240 R12: ffff88810b48a800
  R13: ffff88810b48ec08 R14: ffff88810b48ed00 R15: ffff888100490c68
  FS:  00007f8549ea0b80(0000) GS:ffff888237c00000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 00007f854a09e733 CR3: 000000010a2e9003 CR4: 0000000000370eb0
  DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
  DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
  Call Trace:
   <TASK>
   destroy_inode+0x33/0x70
   dispose_list+0x43/0x60
   evict_inodes+0x161/0x1b0
   generic_shutdown_super+0x2d/0x110
   kill_anon_super+0xf/0x20
   btrfs_kill_super+0xd/0x20 [btrfs]
   deactivate_locked_super+0x27/0x90
   cleanup_mnt+0x12c/0x180
   task_work_run+0x54/0x80
   exit_to_user_mode_prepare+0x152/0x160
   syscall_exit_to_user_mode+0x12/0x30
   do_syscall_64+0x42/0x80
   entry_SYSCALL_64_after_hwframe+0x44/0xae
   RIP: 0033:0x7f854a000fb7

Fixes: f0bfa76a ("btrfs: fix ENOSPC failure when attempting direct IO write into NOCOW range")
CC: stable@vger.kernel.org # 5.17
Reviewed-by: default avatarJohannes Thumshirn <johannes.thumshirn@wdc.com>
Tested-by: default avatarJohannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: default avatarFilipe Manana <fdmanana@suse.com>
Signed-off-by: default avatarNaohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: default avatarDavid Sterba <dsterba@suse.com>
parent 6d4a6b51
......@@ -7442,6 +7442,7 @@ static int btrfs_get_blocks_direct_write(struct extent_map **map,
u64 block_start, orig_start, orig_block_len, ram_bytes;
bool can_nocow = false;
bool space_reserved = false;
u64 prev_len;
int ret = 0;
/*
......@@ -7469,6 +7470,7 @@ static int btrfs_get_blocks_direct_write(struct extent_map **map,
can_nocow = true;
}
prev_len = len;
if (can_nocow) {
struct extent_map *em2;
......@@ -7498,8 +7500,6 @@ static int btrfs_get_blocks_direct_write(struct extent_map **map,
goto out;
}
} else {
const u64 prev_len = len;
/* Our caller expects us to free the input extent map. */
free_extent_map(em);
*map = NULL;
......@@ -7530,7 +7530,7 @@ static int btrfs_get_blocks_direct_write(struct extent_map **map,
* We have created our ordered extent, so we can now release our reservation
* for an outstanding extent.
*/
btrfs_delalloc_release_extents(BTRFS_I(inode), len);
btrfs_delalloc_release_extents(BTRFS_I(inode), prev_len);
/*
* Need to update the i_size under the extent lock so buffered
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment