xfs
[Top] [All Lists]

Re: REPOST : linux-2.5.5-xfs-dj1 - 2.5.7-dj2 (raid0_make_request bug)

To: Stephen Lord <lord@xxxxxxx>
Subject: Re: REPOST : linux-2.5.5-xfs-dj1 - 2.5.7-dj2 (raid0_make_request bug)
From: Andrew Morton <akpm@xxxxxxxxxx>
Date: Fri, 05 Apr 2002 10:37:41 -0800
Cc: Dave Jones <davej@xxxxxxx>, svetljo <galia@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>, linux-kernel@xxxxxxxxxxxxxxx, linux-xfs@xxxxxxxxxxx
References: <3CAD8B9D.8070902@xxxxxxxxxxxxxxxxxxxxxxxxxxxx> <20020405184103.F14828@xxxxxxx> <3CADD6CA.8010600@xxxxxxx>
Sender: owner-linux-xfs@xxxxxxxxxxx
Stephen Lord wrote:
> 
> Dave Jones wrote:
> 
> >On Fri, Apr 05, 2002 at 01:33:49PM +0200, svetljo wrote:
> > > i'm having some interesting troubles
> > > i have lvm over soft RAID-0 with LV's formated with XFS and JFS
> > > i can work with the JFS LV's,
> > >    but i can not with the XFS one's, i can not mount them ( no troubles
> > > with XFS normal partitions)
> > > so i'd like to ask is this problem with XFS or with raid or lvm
> > > and is there a way to fix it
> >
> >IIRC, this was reported a while ago, and it was something to do with
> >XFS creating too large requests that upset the raid code.
> >
> Or the raid code not handling the bio layer too well, depends on your point
> of view ;-)
> 

Stephen's point of view is correct.  RAID0 fails
in the same manner with the large pagecache BIOs
which I'm feeding it.  It fails in the same manner
with O_DIRECT on ext2.

Neil knows about it, and will get to it.  mkp has
a 2.4 request splitter which he will turn into a
2.5 BIO splitter.

As you said - it's being worked on.

-

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