Re: WARNING at fs/btrfs/super.c:260 __btrfs_abort_transaction (error -17)

[Chris Mason <clm@fb.com>]

On Tue, Mar 24, 2015 at 6:23 PM, Sophie <just4pleisure@gmail.com> wrote:
> On 24/03/15 17:34, Chris Mason wrote:
>> 
>> 
>> On Tue, Mar 24, 2015 at 9:43 AM, Sophie Dexter 
>> <Just4pLeisure@gmail.com>
>> wrote:
>>> On 20/03/2015 15:19, Sophie Dexter wrote:
>>>> I'm given to understand that this is the right place to report a 
>>>> btrfs
>>>> problem, I apologise if not :-(
>>>> 
>>>> I have been using my router as a simple NFS NAS for around 2 years
>>>> with an ext3 formatted 2 TB Western Digital 2.5" USB Passport 
>>>> disk. I
>>>> have been slowly moving to BTRFS and thought it about time to 
>>>> convert
>>>> this disk too but unfortunately BTRFS is unreliable on my router 
>>>> :-(.
>>>> It doesn't take long for an error to happen causing a 'ro' remount.
>>>> However the disk is unreadable after the remount, both for NFS and
>>>> locally. Rebooting the router seems to be the only way to access 
>>>> the
>>>> disk again.
>>>> 
>>>> I also have a 1 GB swap partition on the disk although swap doesn't
>>>> appear to be a factor as the problem occurs whether or not swap is
>>>> enabled (this report is without swap).
>>>> 
>>>> I used my laptop to convert the fs to btrfs, not my router. My 
>>>> laptop
>>>> has Fedora 21 with 3.18 kernel and tools. No problems are found 
>>>> when I
>>>> use my laptop to check and scrub the disk (i.e. with the disk
>>>> connected directly to my laptop).
>> 
>> You have great timing, there are two reports of a very similar abort
>> with 4.0-rc5, but your report makes it clear these are not a 
>> regression
>> from 4.0-rc4.
>> 
>> Are you able to run btrfsck on this filesystem?  I'd like to check 
>> for
>> metadata inconsistencies.
>> 
>> -chris
>> 
> Hi Chris,
> 
> Haha, great timing is the secret of good comedy lol
> 
> OpenWrt has only very recently signed off the 3.18 kernel as the 
> default kernel for my router, I was using a build with 3.14 when I 
> converted my disk and saw the same problem :!: I may have posted 
> something I haven't repeated here in the OpenWrt ticket I opened:
> 
> https://dev.openwrt.org/ticket/19216
> 
> I previously checked and scrubbed the disk when the problem first 
> occurred and happily no problems were found then. Although, I had to 
> use another computer because btrfs check doesn't complete on my 
> router, the process is killed due to lack of memory (btrfs invoked 
> oom-killer) :-( Should I start another topic for this or just accept 
> that that problem is due to a lack of memory?
> 
> I have just run btrfs check again using (yet another) laptop and I 
> think everything is still OK:
> 
> # btrfs check /dev/sdb1
> Checking filesystem on /dev/sdb1
> UUID: ########-####-####-####-############
> checking extents
> checking free space cache
> checking fs roots
> checking csums
> checking root refs
> found 930516788539 bytes used err is 0
> total csum bytes: 1234353920
> total tree bytes: 1458515968
> total fs tree bytes: 54571008
> total extent tree bytes: 66936832
> btree space waste bytes: 73372568
> file data blocks allocated: 1264250781696
>  referenced 1264250781696
> Btrfs v3.14.1
> # uname -a
> Linux ######-########-#### 3.16.0-31-generic #43-Ubuntu SMP Tue Mar 
> 10 17:37:36 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux

Sophie, can you please grab the latest btrfs progs from git:

git://git.kernel.org/pub/scm/linux/kernel/git/kdave/btrfs-progs.git

And try with that btrfsck?

The other image that is reproducing this has an error in the free space 
cache, so I'd like to confirm if you're hitting the same problem.

-chris