xfs
[Top] [All Lists]

Re: "xfs: fix error handling for synchronous writes" revisited

To: Ajeet Yadav <ajeet.yadav.77@xxxxxxxxx>
Subject: Re: "xfs: fix error handling for synchronous writes" revisited
From: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Date: Thu, 28 Jul 2011 10:50:20 -0400
Cc: xfs@xxxxxxxxxxx, Dave Chinner <david@xxxxxxxxxxxxx>
In-reply-to: <CAB4K4y6-2odbRhqn6q7ZDOxqF9oUcj2zKvyCEV6374QGyM-ddA@xxxxxxxxxxxxxx>
References: <CAB4K4y6-2odbRhqn6q7ZDOxqF9oUcj2zKvyCEV6374QGyM-ddA@xxxxxxxxxxxxxx>
User-agent: Mutt/1.5.21 (2010-09-15)
On Thu, Jul 28, 2011 at 04:21:01PM +0530, Ajeet Yadav wrote:
> Just to recall, last year we have reported one hang issue in 2.6.30.9
> wherein we get hang if we remove the device while synchronous write was 
> underway
> It was because synchronous buffer remain locked in case of EIO error.
> 
> In response to that a fix commit: bfc60177f8ab509bc225becbb58f7e53a0e33e81
>  "xfs: fix error handling for synchronous writes" for 2.6.38 was given
> by community
> Unfortunately I did not use this patch and therefore did not test it.
> 
> Coming to present, I run the same test case again, and I got different
> hang pasted below.
> While debugging I compared with my own solution, wherein I did not
> have any issue
> Kindly review the solution patch

Given that xfs_bwrite actually does the shutdown already after
waiting for the b_iodone completion this looks fine to me.  And given
that we actually found that calling xfs_force_shutdown from inside
xfs_buf_iodone_callbacks was a major contributor the problem it looks
like an onversight that we didn't drop this call.

Can you re-send the patch with a proper changelog and signoff?

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