Comment # 1
on bug 1037118
from Martin Wilck
I uploaded a btrfs-image dump of the broken file system to
l3slave.suse.de:/space/mwilck/1037118/ (506MiB; partition size was 40963 MiB).
Meanwhile I transferred the data except for older snapshots to a new btrfs
volume because (not being sure how to recover from this situation) I wanted to
avoid to be left with a totally corrupt file system, and unable to work.
In the recovery process I made some minor write accesses to the file system, so
the dumped state is not exactly the same as when this error first occured.
While I tried to copy the data, I got thousands of errors like this because I'd
forgotten to mount the volume with "noatime" or "ro":
> May 02 16:07:04 apollon.suse.de kernel: ------------[ cut here ]------------
> May 02 16:07:04 apollon.suse.de kernel: WARNING: CPU: 1 PID: 577 at ../fs/btrfs/qgroup.c:2936 btrfs_qgroup_free_meta+0x89/0x90 [btrfs]
> May 02 16:07:04 apollon.suse.de kernel: CPU: 1 PID: 577 Comm: rsync Tainted: G U 4.10.12-1-default #1
> May 02 16:07:04 apollon.suse.de kernel: Hardware name: Dell Inc. Latitude E7470/0T6HHJ, BIOS 1.9.4 08/26/2016
> May 02 16:07:04 apollon.suse.de kernel: Call Trace:
> May 02 16:07:04 apollon.suse.de kernel: dump_stack+0x5c/0x7a
> May 02 16:07:04 apollon.suse.de kernel: __warn+0xbe/0xe0
> May 02 16:07:04 apollon.suse.de kernel: btrfs_qgroup_free_meta+0x89/0x90 [btrfs]
> May 02 16:07:04 apollon.suse.de kernel: start_transaction+0x3a5/0x4a0 [btrfs]
> May 02 16:07:04 apollon.suse.de kernel: btrfs_dirty_inode+0x99/0xc0 [btrfs]
> May 02 16:07:04 apollon.suse.de kernel: touch_atime+0x89/0xb0
> May 02 16:07:04 apollon.suse.de kernel: SyS_readlink+0xd9/0x110
> May 02 16:07:04 apollon.suse.de kernel: entry_SYSCALL_64_fastpath+0x1e/0xad
> May 02 16:07:04 apollon.suse.de kernel: ---[ end trace 3a2f8149ed5137ab ]---
Another examination of the journal showed that the first occurence of this bug
occured actually before today's update. The first one that's recorded was last
week:
> Apr 24 21:16:54 apollon.suse.de kernel: CPU: 0 PID: 702 Comm: systemd-journal Tainted: G U W 4.10.9-1-default #1
> Apr 24 21:16:54 apollon.suse.de kernel: Hardware name: Dell Inc. Latitude E7470/0T6HHJ, BIOS 1.9.4 08/26/2016
> Apr 24 21:16:54 apollon.suse.de kernel: Call Trace:
> Apr 24 21:16:54 apollon.suse.de kernel: dump_stack+0x5c/0x7a
> Apr 24 21:16:54 apollon.suse.de kernel: __warn+0xbe/0xe0
> Apr 24 21:16:54 apollon.suse.de kernel: btrfs_qgroup_free_meta+0x89/0x90 [btrfs]
> Apr 24 21:16:54 apollon.suse.de kernel: start_transaction+0x3a5/0x4a0 [btrfs]
> Apr 24 21:16:54 apollon.suse.de kernel: btrfs_rename+0x184/0xd40 [btrfs]
> Apr 24 21:16:54 apollon.suse.de kernel: vfs_rename+0x6ba/0x900
> Apr 24 21:16:54 apollon.suse.de kernel: SyS_rename+0x3a6/0x3c0
> Apr 24 21:16:54 apollon.suse.de kernel: do_syscall_64+0x57/0xc0
> Apr 24 21:16:54 apollon.suse.de kernel: entry_SYSCALL64_slow_path+0x25/0x25
> Apr 24 21:16:54 apollon.suse.de kernel: ---[ end trace efd3cc31a94c61e4 ]---
However the update today was the first time that I noticed because this
resulted in a fatal error of a user command (see comment 0).
>From memory, I can't associate any relevant system configuration changes with
this date (Apr 24 21:16) except that that was the first day I ran the
4.10.9-1-default kernel (updated from 4.10.8-1-default on 2017-04-23 21:34:01).