xfs
[Top] [All Lists]

Re: TAKE - acl_init

To: Andi Kleen <ak@xxxxxxx>
Subject: Re: TAKE - acl_init
From: Nathan Scott <nathans@xxxxxxx>
Date: Fri, 21 Feb 2003 08:33:27 +1100
Cc: linux-xfs@xxxxxxxxxxx, agruen@xxxxxxx
In-reply-to: <20030220075926.GA21973@xxxxxxxxxxxxx>
References: <200302182335.h1INZ7la8907219@xxxxxxxxxxxxxxxxxxxxxxx> <20030220075926.GA21973@xxxxxxxxxxxxx>
Sender: linux-xfs-bounce@xxxxxxxxxxx
User-agent: Mutt/1.5.3i
hi Andi,

On Thu, Feb 20, 2003 at 08:59:26AM +0100, Andi Kleen wrote:
> On Wed, Feb 19, 2003 at 10:35:07AM +1100, Nathan Scott wrote:
> > Fix a zero-length malloc in acl_init - causing QA failures
> > when libacl is linked with the electric fence library.
> 
> Still using ElectricFence and not valgrind for testing ? @)
> 
> valgrind will find much more bugs.

Hmmm... just tried running valgrind on xfs_repair and looks
like its going to take a bit of work to get valgrind to work
on any of our tools - we use ustat in libxfs for example, &
valgrind barfs on that as an unrecognised syscall.

The acl and attr tools make extensive use of all the xattr
syscalls, which are still relatively new, so those are likely
to have similar problems I guess.

I don't have time to hack on valgrind right now - maybe some
community-minded person out there wants to give this a go?
There's detailed instructions in the valgrind source on how
to modify the code to fix these issues...

cheers.

-- 
Nathan


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