xfs
[Top] [All Lists]

Re: [PATCH] jfs: flush journal completely before releasing metadata ino

To: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Subject: Re: [PATCH] jfs: flush journal completely before releasing metadata inodes
From: Dave Kleikamp <dave.kleikamp@xxxxxxxxxx>
Date: Mon, 01 Aug 2011 15:04:32 -0500
Cc: jfs-discussion@xxxxxxxxxxxxxxxxxxxxx, xfs@xxxxxxxxxxx
In-reply-to: <20110801183255.GA30985@xxxxxxxxxxxxx>
References: <20110729155039.GA28306@xxxxxxxxxxxxx> <4E32D9EB.9020406@xxxxxxxxxx> <4E36F086.4070900@xxxxxxxxxx> <20110801183255.GA30985@xxxxxxxxxxxxx>
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:5.0) Gecko/20110628 Thunderbird/5.0

On 08/01/2011 01:32 PM, Christoph Hellwig wrote:
> On Mon, Aug 01, 2011 at 01:29:26PM -0500, Dave Kleikamp wrote:
>> On 07/29/2011 11:03 AM, Dave Kleikamp wrote:

>> This fixes a race during unmount. We need to not only make sure that
>> the journal is completely written, but that the metadata changes make
>> it to disk before releasing ipimap and ipbmap.
> 
> Looks sensible.  How far does a ./check -g auto get now?  Any other
> failures?

It runs through them all. A lot of tests get skipped claiming not to be
supported on jfs. A failure because jfs doesn't support fallocate and
hole-punch, but no crashes.

Shaggy

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