[Top] [All Lists]

Re: Kernel loop(..?) regression (& with/2.6.31-rc6)

To: Justin Piszcz <jpiszcz@xxxxxxxxxxxxxxx>
Subject: Re: Kernel loop(..?) regression (& with/2.6.31-rc6)
From: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Date: Fri, 28 Aug 2009 12:55:26 -0400
Cc: Christoph Hellwig <hch@xxxxxxxxxxxxx>, Nikanth Karthikesan <knikanth@xxxxxxx>, xfs@xxxxxxxxxxx, linux-kernel@xxxxxxxxxxxxxxx, Jens Axboe <jens.axboe@xxxxxxxxxx>
In-reply-to: <alpine.DEB.2.00.0908280720220.30806@xxxxxxxxxxxxxxxx>
References: <20090822205502.GA18904@xxxxxxxxxxxxx> <alpine.DEB.2.00.0908221659030.23463@xxxxxxxxxxxxxxxx> <20090823224504.GA19942@xxxxxxxxxxxxx> <alpine.DEB.2.00.0908240704130.9427@xxxxxxxxxxxxxxxx> <alpine.DEB.2.00.0908260949550.9396@xxxxxxxxxxxxxxxx> <20090826180234.GC14019@xxxxxxxxxxxxx> <alpine.DEB.2.00.0908261717001.1594@xxxxxxxxxxxxxxxx> <20090826212732.GA18124@xxxxxxxxxxxxx> <alpine.DEB.2.00.0908261728580.1594@xxxxxxxxxxxxxxxx> <alpine.DEB.2.00.0908280720220.30806@xxxxxxxxxxxxxxxx>
User-agent: Mutt/1.5.19 (2009-01-05)
On Fri, Aug 28, 2009 at 07:23:11AM -0400, Justin Piszcz wrote:
> If both '-o nobarrier' and the patch solves the issue, what is the next  
> action that should be taken?  Update the documentation to always use
> -o nobarrier for cryptoloop? Or get the patch reverted in mainline?

We'll have to figure out why it causes those problems and how fix them.
-o nobarrier on a loop device is extremly unsafe as you will possible
lose large amounts of log updates in case of a crash.

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