btrfs: qgroup: do not warn on record without old_roots populated
authorQu Wenruo <wqu@suse.com>
Tue, 10 Jan 2023 07:14:17 +0000 (15:14 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 24 Jan 2023 06:24:36 +0000 (07:24 +0100)
commitbb2c2e62539f2b63c5e0beb51501d328260c7595
tree577432ee24e8b428e8a94cada8577596f5672c89
parent34fe5b527e68c7882def3ea4c590034c5b6cb106
btrfs: qgroup: do not warn on record without old_roots populated

commit 75181406b4eafacc531ff2ee5fb032bd93317e2b upstream.

[BUG]
There are some reports from the mailing list that since v6.1 kernel, the
WARN_ON() inside btrfs_qgroup_account_extent() gets triggered during
rescan:

  WARNING: CPU: 3 PID: 6424 at fs/btrfs/qgroup.c:2756 btrfs_qgroup_account_extents+0x1ae/0x260 [btrfs]
  CPU: 3 PID: 6424 Comm: snapperd Tainted: P           OE      6.1.2-1-default #1 openSUSE Tumbleweed 05c7a1b1b61d5627475528f71f50444637b5aad7
  RIP: 0010:btrfs_qgroup_account_extents+0x1ae/0x260 [btrfs]
  Call Trace:
   <TASK>
  btrfs_commit_transaction+0x30c/0xb40 [btrfs c39c9c546c241c593f03bd6d5f39ea1b676250f6]
   ? start_transaction+0xc3/0x5b0 [btrfs c39c9c546c241c593f03bd6d5f39ea1b676250f6]
  btrfs_qgroup_rescan+0x42/0xc0 [btrfs c39c9c546c241c593f03bd6d5f39ea1b676250f6]
   btrfs_ioctl+0x1ab9/0x25c0 [btrfs c39c9c546c241c593f03bd6d5f39ea1b676250f6]
   ? __rseq_handle_notify_resume+0xa9/0x4a0
   ? mntput_no_expire+0x4a/0x240
   ? __seccomp_filter+0x319/0x4d0
   __x64_sys_ioctl+0x90/0xd0
   do_syscall_64+0x5b/0x80
   ? syscall_exit_to_user_mode+0x17/0x40
   ? do_syscall_64+0x67/0x80
  entry_SYSCALL_64_after_hwframe+0x63/0xcd
  RIP: 0033:0x7fd9b790d9bf
   </TASK>

[CAUSE]
Since commit e15e9f43c7ca ("btrfs: introduce
BTRFS_QGROUP_RUNTIME_FLAG_NO_ACCOUNTING to skip qgroup accounting"), if
our qgroup is already in inconsistent state, we will no longer do the
time-consuming backref walk.

This can leave some qgroup records without a valid old_roots ulist.
Normally this is fine, as btrfs_qgroup_account_extents() would also skip
those records if we have NO_ACCOUNTING flag set.

But there is a small window, if we have NO_ACCOUNTING flag set, and
inserted some qgroup_record without a old_roots ulist, but then the user
triggered a qgroup rescan.

During btrfs_qgroup_rescan(), we firstly clear NO_ACCOUNTING flag, then
commit current transaction.

And since we have a qgroup_record with old_roots = NULL, we trigger the
WARN_ON() during btrfs_qgroup_account_extents().

[FIX]
Unfortunately due to the introduction of NO_ACCOUNTING flag, the
assumption that every qgroup_record would have its old_roots populated
is no longer correct.

Fix the false alerts and drop the WARN_ON().

Reported-by: Lukas Straub <lukasstraub2@web.de>
Reported-by: HanatoK <summersnow9403@gmail.com>
Fixes: e15e9f43c7ca ("btrfs: introduce BTRFS_QGROUP_RUNTIME_FLAG_NO_ACCOUNTING to skip qgroup accounting")
CC: stable@vger.kernel.org # 6.1
Link: https://lore.kernel.org/linux-btrfs/2403c697-ddaf-58ad-3829-0335fc89df09@gmail.com/
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/btrfs/qgroup.c