xfs
[Top] [All Lists]

Re: Metadata CRC error upon unclean unmount

To: Fanael Linithien <fanael4@xxxxxxxxx>
Subject: Re: Metadata CRC error upon unclean unmount
From: Dave Chinner <david@xxxxxxxxxxxxx>
Date: Thu, 26 Jun 2014 09:22:22 +1000
Cc: xfs@xxxxxxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <CA+o=1OWptqFq87E0ncmYH2T6yEzeGheefE-o_VAm7-E_0Z+NJA@xxxxxxxxxxxxxx>
References: <CA+o=1OW0OXhzU+b9ACMZzg0dq=B7BSj+yPXD2Vrr9F6mWK8ruQ@xxxxxxxxxxxxxx> <20140624201946.GJ9508@dastard> <CA+o=1OVnORG0Ah3Zx8dkGzs7vtT7odRH=v12KqtLW0MP_3oHjQ@xxxxxxxxxxxxxx> <20140625012144.GK9508@dastard> <CA+o=1OXY8Ay8CSgsdiAOTacBa9800FiMKGbr0DdkE_YVpewqBw@xxxxxxxxxxxxxx> <20140625220336.GN9508@dastard> <CA+o=1OWptqFq87E0ncmYH2T6yEzeGheefE-o_VAm7-E_0Z+NJA@xxxxxxxxxxxxxx>
User-agent: Mutt/1.5.21 (2010-09-15)
On Thu, Jun 26, 2014 at 12:33:06AM +0200, Fanael Linithien wrote:
> 2014-06-26 0:03 GMT+02:00 Dave Chinner <david@xxxxxxxxxxxxx>:
> > Can you try to reproduce the AGI or AGF/freespace corruption issue
> > and post the results?
> 
> Sure, looks like the bug heard you and decided to manifest itself right now.
> 
> dmesg: https://clbin.com/g52eu (the second mount error is because I
> typo'd and tried to mount wrong partition)
> Metadata before an attempt to mount: https://clbin.com/YDZBV

Curious. Completely clean log. head/tail at 2/16292. Yet the AGF
head and a couple of btree blocks are have bad CRCs. They were on
disk *before* the system crashed.

> Metadata after an attempt to mount: https://clbin.com/2jayF

Dirty log. tail at 2/16292, head at 2/16332.

The before-mount image shows the CRC error, so it didn't come from
log recovery. It may have come from recovery in the previous
iteration, - do you still have the pre-mount images from the prior
iteration before the crash error was detected?

Cheers,

Dave.
-- 
Dave Chinner
david@xxxxxxxxxxxxx

<Prev in Thread] Current Thread [Next in Thread>