2012-08-11 17:36:23

by Justin Piszcz

[permalink] [raw]
Subject: 3.4->(3.5.1 || 3.6-rc1) => can no longer mount 60TB ext4 filesystem

Hello,

I upgrade to each new kernel release and with 3.5.1 (from 3.4) I can no
longer mount my 60TB ext4 volume.
If I boot back to 3.4, it works fine.

Details here:
https://lkml.org/lkml/2012/8/10/205

Anything I can do besides testing each 3.5-rcX to find where the regression
lies?

Justin.


2012-08-12 13:32:07

by Daniel Mack

[permalink] [raw]
Subject: Re: 3.4->(3.5.1 || 3.6-rc1) => can no longer mount 60TB ext4 filesystem

On 11.08.2012 19:36, Justin Piszcz wrote:
> Hello,
>
> I upgrade to each new kernel release and with 3.5.1 (from 3.4) I can no
> longer mount my 60TB ext4 volume.
> If I boot back to 3.4, it works fine.
>
> Details here:
> https://lkml.org/lkml/2012/8/10/205
>
> Anything I can do besides testing each 3.5-rcX to find where the regression
> lies?

I'm not at all familiar with ext4 internals, but as always, an wasy way
is to bisect such a problem. Any change you can do that?

http://www.kernel.org/pub/software/scm/git/docs/git-bisect.html


Daniel

2012-08-12 13:52:46

by Justin Piszcz

[permalink] [raw]
Subject: Re: 3.4->(3.5.1 || 3.6-rc1) => can no longer mount 60TB ext4 filesystem

On Sun, Aug 12, 2012 at 9:31 AM, Daniel Mack <[email protected]> wrote:
> On 11.08.2012 19:36, Justin Piszcz wrote:
>> Hello,
>>
>> I upgrade to each new kernel release and with 3.5.1 (from 3.4) I can no
>> longer mount my 60TB ext4 volume.
>> If I boot back to 3.4, it works fine.
>>
>> Details here:
>> https://lkml.org/lkml/2012/8/10/205
>>
>> Anything I can do besides testing each 3.5-rcX to find where the regression
>> lies?
>
> I'm not at all familiar with ext4 internals, but as always, an wasy way
> is to bisect such a problem. Any change you can do that?
>
> http://www.kernel.org/pub/software/scm/git/docs/git-bisect.html
>
>
> Daniel
>

Hi,

Thanks, will save this for the future-- would of done this but Eric
Sandeen found the offending patch, I reverted it and I can mount the
filesystem now.

The bad commit was 8aeb00ff85ad25453765dd339b408c0087db1527 (per sandeen)

Justin.