Re: able to reproduce growfs bug on LVM(FAQ) at will

[Tupshin Harper <tupshin@tupshin.com>]

Tupshin Harper wrote:

> I am encountering the FAQ'd bug that prevents xfs_growfs from working 
> on a resized LVM volume. The workaround (umount and xfs_repair) does 
> work, but I was wondering if I could be of some assistance in tracking 
> the bug down? Since it is highly reproducible, I can gather any 
> gatherable information. FWIW, I'm running on:
> Athlon XP CPU
> Debian Sid
> Kernel 2.6.0-test3 (almost stock)
> LVM2 on dev-mapper
>
> A (mildly) interesting datapoint is that an incomplete xfs_repair (it 
> errrored out in Phase 6 because of insufficient space) still corrected 
> the problem, so some write operation that takes place in Phases 1-5 
> fixes the issue.
>
> I've included a log of a failed xfs_growfs, a xfs_repair, and a 
> successful xfs_growfs below.
>
> -Tupshin
>
Let me correct myself, and concur with Erkki. xfs_repair is absolutely 
unnecessary to fix the problem unmounting the fs, and remounting it 
allows xfs_growsfs to function correctly after an LVM resize. I was 
thrown off by the FAQ that said that xfs_repair was the way to fix it.

-Tupshin