[bug] df returns 100% used for empty mixed block group filesystem

[bug] df returns 100% used for empty mixed block group filesystem

[Chris Murphy]

kernel 4.5.0 rc7
btrfs-progs-4.4.1-1

Summary: A new empty mixed block group "-M" filesystem shows up as
100% full with regular df (gnu coreutils 8.24). Pretty sure this is
not a regression, but no regression testing done.


https://bugzilla.kernel.org/show_bug.cgi?id=114281


-- 
Chris Murphy

Re: [bug] df returns 100% used for empty mixed block group filesystem

[Roman Mamedov]

[-- Attachment #1: Type: text/plain, Size: 679 bytes --]

On Thu, 10 Mar 2016 01:24:44 -0700
Chris Murphy <lists@colorremedies.com> wrote:

> kernel 4.5.0 rc7
> btrfs-progs-4.4.1-1
> 
> Summary: A new empty mixed block group "-M" filesystem shows up as
> 100% full with regular df (gnu coreutils 8.24). Pretty sure this is
> not a regression, but no regression testing done.
> 
> 
> https://bugzilla.kernel.org/show_bug.cgi?id=114281

Just guessing here, but this might be related?

  [PATCH] btrfs: statfs: report zero available if metadata are exhausted
  http://www.spinics.net/lists/linux-btrfs/msg48013.html

Maybe some maths there are wrong in case of a mixed blockgroup filesystem.

-- 
With respect,
Roman

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 181 bytes --]

Re: [bug] df returns 100% used for empty mixed block group filesystem

[David Sterba]

On Thu, Mar 10, 2016 at 01:33:34PM +0500, Roman Mamedov wrote:
> On Thu, 10 Mar 2016 01:24:44 -0700
> Chris Murphy <lists@colorremedies.com> wrote:
> 
> > kernel 4.5.0 rc7
> > btrfs-progs-4.4.1-1
> > 
> > Summary: A new empty mixed block group "-M" filesystem shows up as
> > 100% full with regular df (gnu coreutils 8.24). Pretty sure this is
> > not a regression, but no regression testing done.
> > 
> > 
> > https://bugzilla.kernel.org/show_bug.cgi?id=114281
> 
> Just guessing here, but this might be related?
> 
>   [PATCH] btrfs: statfs: report zero available if metadata are exhausted
>   http://www.spinics.net/lists/linux-btrfs/msg48013.html
> 
> Maybe some maths there are wrong in case of a mixed blockgroup filesystem.

Yes it's caused by the referenced patch.  The statfs has been broken for
mixed blockgroups but the 100% used on empty filesystem looks weird, so
it's time to fix the mixed-bg case.

Re: [bug] df returns 100% used for empty mixed block group filesystem

[David Sterba]

On Thu, Mar 10, 2016 at 06:40:54PM +0100, David Sterba wrote:
> On Thu, Mar 10, 2016 at 01:33:34PM +0500, Roman Mamedov wrote:
> > On Thu, 10 Mar 2016 01:24:44 -0700
> > Chris Murphy <lists@colorremedies.com> wrote:
> > 
> > > kernel 4.5.0 rc7
> > > btrfs-progs-4.4.1-1
> > > 
> > > Summary: A new empty mixed block group "-M" filesystem shows up as
> > > 100% full with regular df (gnu coreutils 8.24). Pretty sure this is
> > > not a regression, but no regression testing done.
> > > 
> > > 
> > > https://bugzilla.kernel.org/show_bug.cgi?id=114281
> > 
> > Just guessing here, but this might be related?
> > 
> >   [PATCH] btrfs: statfs: report zero available if metadata are exhausted
> >   http://www.spinics.net/lists/linux-btrfs/msg48013.html
> > 
> > Maybe some maths there are wrong in case of a mixed blockgroup filesystem.
> 
> Yes it's caused by the referenced patch.  The statfs has been broken for
> mixed blockgroups but the 100% used on empty filesystem looks weird, so
> it's time to fix the mixed-bg case.

JFYI, fixed by Luis de Bethencourt and patches queued for 4.7

Re: [bug] df returns 100% used for empty mixed block group filesystem

[Luis de Bethencourt]

On 27/04/16 23:55, David Sterba wrote:
> On Thu, Mar 10, 2016 at 06:40:54PM +0100, David Sterba wrote:
>> On Thu, Mar 10, 2016 at 01:33:34PM +0500, Roman Mamedov wrote:
>>> On Thu, 10 Mar 2016 01:24:44 -0700
>>> Chris Murphy <lists@colorremedies.com> wrote:
>>>
>>>> kernel 4.5.0 rc7
>>>> btrfs-progs-4.4.1-1
>>>>
>>>> Summary: A new empty mixed block group "-M" filesystem shows up as
>>>> 100% full with regular df (gnu coreutils 8.24). Pretty sure this is
>>>> not a regression, but no regression testing done.
>>>>
>>>>
>>>> https://bugzilla.kernel.org/show_bug.cgi?id=114281
>>>
>>> Just guessing here, but this might be related?
>>>
>>>   [PATCH] btrfs: statfs: report zero available if metadata are exhausted
>>>   http://www.spinics.net/lists/linux-btrfs/msg48013.html
>>>
>>> Maybe some maths there are wrong in case of a mixed blockgroup filesystem.
>>
>> Yes it's caused by the referenced patch.  The statfs has been broken for
>> mixed blockgroups but the 100% used on empty filesystem looks weird, so
>> it's time to fix the mixed-bg case.
> 
> JFYI, fixed by Luis de Bethencourt and patches queued for 4.7
> 

Good to know the issue was important to some people :)
It was fun to fix.

David,
I have been travelling the last week. I will continue with my second bug this
weekend. [0] Sorry for the delay on this.

Thanks,
Luis


[0] btrfs fi usage crash when multiple device volume contains seed device
https://bugzilla.kernel.org/show_bug.cgi?id=115851