xfs
[Top] [All Lists]

Re: 2.4.11 don't work yet.

To: nic@xxxxxxxxxxx, linux-xfs@xxxxxxxxxxx
Subject: Re: 2.4.11 don't work yet.
From: Seth Mos <knuffie@xxxxxxxxx>
Date: Wed, 10 Oct 2001 14:40:36 +0200
In-reply-to: <E15rI4P-0000RJ-00@pyewacket.nic.uklinux.net>
References: <4.3.2.7.2.20011010132408.02d1f3f0@pop.xs4all.nl> <4.3.2.7.2.20011010132408.02d1f3f0@pop.xs4all.nl>
Sender: owner-linux-xfs@xxxxxxxxxxx
At 13:02 10-10-2001 +0100, nic wrote:
On Wednesday 10 October 2001 12:24, Seth Mos wrote:
>
> >Software info:
> >Machine is RedHat 7.1 + glibc (and other bits) from 7.1.94
> >gcc-2.96-85
>
> Can you compile with compat-egcs and try again?
> Sometimes weird things happen with xfs and 2.96-??

I've seen this said before (most often by your good self :-) but I've always
built my 2.4.x-xfs kernels with it... including the one I'm running now.

Will try and let you know,
nic


It compiles most of the times but if you push the kernel hard enough some things budge. The mandrake kernel I am testing on is compiled with 2.96-85 and is working as one should expect.

I have uploaded a new mandrake kernel to my website which has a working "make install". The kernel is fairly stable and includes the Gigabit drivers for the box.

I am hitting it with mongo.pl with 50 processes on a 2GB dual 1.13Ghz box and it survives in a normal fashion. No hangs and no IO stalls.
The benchmark time for "create" on a ext2 fs is down from 4200 secs to 2600 secs. This is the performance of Redhat-2.4.2 vs. Mandrake-2.4.8.


Maybe someone is willing to convert the original mandrake source rpm for use on redhat/suse.

If you are trying to convert the source rpm you will have to make sure that arch/i386/boot/install.sh is not altered. Mandrake added a -a argument for their installkernel program which confuses the redhat (and possibly SuSE) installkernel utility.

Some XFS results are expected to follow soon. If people want a more up to date XFS kernel based on a -ac kernel this is probably a good choice.

Cheers

--
Seth
Every program has two purposes one for which
it was written and another for which it wasn't
I use the last kind.


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