Re: System freezes a lot - btrfs trace in dmesg - symptom or cause?

[Qu Wenruo <quwenruo.btrfs@gmx.com>]

[-- Attachment #1.1: Type: text/plain, Size: 2527 bytes --]



On 2018年04月04日 05:35, Richard Schwab wrote:
> Hi there,
> 
> I'm currently having some issues with my system freezing a lot, from a
> few seconds up to a few minutes. I haven't figured out yet what might be
> causing this, however, coincidentally some btrfs notices appeared in my
> dmesg after I had a freeze for about 5 minutes. I'm wondering whether
> this is just another symptom of my system having some weird issues or
> whether this might actually be the cause of the freezes I've been
> experiencing.
> 
> $ uname -a
> Linux mypc 4.15.14-1-ARCH #1 SMP PREEMPT Wed Mar 28 17:34:29 UTC 2018
> x86_64 GNU/Linux
> 
> $ btrfs --version
> btrfs-progs v4.15.1
> 
> $ btrfs fi show
> Label: none  uuid: uuid1
>     Total devices 1 FS bytes used 226.60GiB
>     devid    1 size 237.37GiB used 237.31GiB path /dev/mapper/cryptroot256g
> 
> Label: none  uuid: uuid2
>     Total devices 1 FS bytes used 116.61MiB
>     devid    1 size 1.00GiB used 460.00MiB path /dev/sdb2
> 
> $ btrfs fi df /
> Data, single: total=231.28GiB, used=223.66GiB
> System, single: total=32.00MiB, used=48.00KiB
> Metadata, single: total=6.00GiB, used=2.93GiB
> GlobalReserve, single: total=512.00MiB, used=0.00B
> 
> Partial dmesg is attached as file. Please tell me if you need more than
> this, however, before these lines it only contained the startup process
> (up until 87s).

[10691.693202] btrfs-transacti D    0   447      2 0x80000000
[10691.693207] Call Trace:
[10691.693221]  ? __schedule+0x24b/0x8c0
[10691.693229]  ? enqueue_entity+0x59b/0xc70
[10691.693234]  schedule+0x32/0x90
[10691.693239]  io_schedule+0x12/0x40
[10691.693247]  __lock_page+0xf8/0x130
[10691.693252]  ? add_to_page_cache_lru+0xe0/0xe0
[10691.693309]  read_extent_buffer_pages+0xf0/0x2f0 [btrfs]
[10691.693337]  ? btrfs_alloc_root+0x30/0x30 [btrfs]
[10691.693365]  btree_read_extent_buffer_pages+0x94/0xf0 [btrfs]
[10691.693389]  read_block_for_search.isra.14+0x15a/0x2f0 [btrfs]

This means the tree block pages btrfs trying to read is being locked by
other guy.

And considering later operations also waiting on the same page, it looks
like a deadlock.

Not sure if it's something serious or just bad luck.
But looks like there is something wrong with btrfs tree locks.

Is it reproducible and what's the workload?

And if it's just bad luck, reboot the system and run "btrfs check" on
the fs is highly recommended.

Thanks,
Qu

> 
> Sincerely
> Richard Schwab
> 


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]