Reproducible corruption issue in stable, with possible patch unmerged?

Reproducible corruption issue in stable, with possible patch unmerged?

[Diego Call.]

[repost, since the previous mail was rejected by vger.kernel.org]

I got these messages in my dmesg:

[ 2587.841376] BTRFS error (device sdc1): partial page write in btrfs
with offset 0 and length 8392704
[ 2588.814182] BTRFS critical (device sdc1): bad ordered accounting
left 0 size 4096

With no other errors before. After rebooting I got btrfs
errors and widespreading filesystem corruption (pool
of three disks created with -d raid0 and -m raid1)

I reinstalled the system (current arch linux installer, which
runs 4.2) and started to restore the backup. Surprisingly,
I got exactly the same errors again. I reinstalled again,
and after rebooting to the new system, running 4.2.5,
I tried to restore the backup and got the same messages
and corruption again. Then I tried reinstalling with
the 4.1.12 LTS version, and it all worked fine.

Now, the interesting thing is that, after googling this error, I didn't
found any interesting references except a patch sent to this
mailing list which seems to fix a problem that causes the same
error output. The patch doesn't seem to have been merged:

http://lkml.iu.edu/hypermail/linux/kernel/1412.1/03297.html

Re: Reproducible corruption issue in stable, with possible patch unmerged?

[Chris Mason]

On Sat, Nov 07, 2015 at 04:36:56PM +0100, Diego Call. wrote:
> [repost, since the previous mail was rejected by vger.kernel.org]
> 
> I got these messages in my dmesg:
> 
> [ 2587.841376] BTRFS error (device sdc1): partial page write in btrfs
> with offset 0 and length 8392704
> [ 2588.814182] BTRFS critical (device sdc1): bad ordered accounting
> left 0 size 4096
> 
> With no other errors before. After rebooting I got btrfs
> errors and widespreading filesystem corruption (pool
> of three disks created with -d raid0 and -m raid1)
> 
> I reinstalled the system (current arch linux installer, which
> runs 4.2) and started to restore the backup. Surprisingly,
> I got exactly the same errors again. I reinstalled again,
> and after rebooting to the new system, running 4.2.5,
> I tried to restore the backup and got the same messages
> and corruption again. Then I tried reinstalling with
> the 4.1.12 LTS version, and it all worked fine.
> 
> Now, the interesting thing is that, after googling this error, I didn't
> found any interesting references except a patch sent to this
> mailing list which seems to fix a problem that causes the same
> error output. The patch doesn't seem to have been merged:
> 

Good catch.  Can you describe your configuration?

-chris

Re: Reproducible corruption issue in stable, with possible patch unmerged?

[dcg]

Saturday 07 November 2015 15:17:27 Chris Mason escribió:
> > error output. The patch doesn't seem to have been merged:
> Good catch.  Can you describe your configuration?

pool of 3 spinning disks, configured with raid0 for data and raid1
for metadata

Total devices 3 FS bytes used 749.78GiB
        devid    1 size 1.81TiB used 254.00GiB path /dev/sdc1
        devid    2 size 465.76GiB used 253.01GiB path /dev/sdb1
        devid    3 size 463.76GiB used 252.01GiB path /dev/sda3

x86-64 system, 4 cores, 16 GB RAM