xfs
[Top] [All Lists]

Re: 2.4.4-xfs NFS testing

To: Andi Kleen <ak@xxxxxxx>
Subject: Re: 2.4.4-xfs NFS testing
From: Eric Whiting <ewhiting@xxxxxxxx>
Date: Thu, 03 May 2001 11:35:37 -0600
Cc: Mark Hounschell <dmarkh@xxxxxxxxxx>, utz lehmann <xfs@xxxxxxxxxx>, linux-xfs@xxxxxxxxxxx
References: <3AF0468B.70E8B919@xxxxxxxx> <20010502195501.A13116@xxxxxxxxxxxxxxxxxxx> <20010502203454.A9561@xxxxxxxxxx> <3AF137CB.5DE970F6@xxxxxxxxxx> <20010503125844.C28596@xxxxxxxxxxxxxxxxxxx>
Sender: owner-linux-xfs@xxxxxxxxxxx
I tried:
2.4.4-xfs NFS 
kernel compiled with 2.95.3 

Results:
still had severe lockup under heavy nfs loading from solaris boxes.

I'll try to get more debug info -- as well as rule out some other
issues. I read about 2.4.4 problems -- maybe I should go back to
2.4.2-xfs?

eric


Andi Kleen wrote:
> 
> On Thu, May 03, 2001 at 06:49:47AM -0400, Mark Hounschell wrote:
> > I've heard rumors on the SuSE-e list that the lastest stable release
> > 2.95.3 works
> > ok woth XFS. Haven't tried it yet. Get SuSE's rpm and try it first.
> 
> The latest XFS tree has some workaround for known 2.95 long long bugs (mainly
> division) and they seem to work somehow; but nobody knows if that really
> catched all cases or if there isn't miscompilation in more obscure code
> paths left. XFS is full of long long computation (you wanted a "64bit
> filesystem", didn't you?). Using egcs 1.1 is definitely safer, with it
> XFS has been tested a lot more.  You could also use XFS on a 64bit
> architecture like an Alpha.
> 
> -Andi

-- 
__________________________________________________________________
Eric T. Whiting                                 AMI Semiconductors

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