On Sat, Sep 07, 2002 at 10:50:41PM +0200, Axel Thimm wrote:
> I am using Jörg Prante's kernel patch set 2.4.19-jp14 from
> http://infolinux.de/jp/, which includes among others
>
> o XFS for 2.4.19
> o rmap-13b
after some further tests to narrow the possible causes down, the non-working
combination seem to be
XFS & NVIDIA & rmap-13b
Upgrading to rmap-13c fixed the kernel Oops. Credits go to Knut J Bjuland for
pointing out to me a race condition in rmap-13b.
Why only that particular combination triggered that bug escapes me. And it
only hit applications like vmware or Samba (nmbd).
Thanks to all people for a lot of helpful remarks!
> o O1-Scheduler
> o low-latency patches
> o preempt-kernel
P.S. Jörg Prante's kernel patch set with this fix applied (rmap-13c) is really
impressive concerning performance.
> If I try to use NVIDIA's modules (the latest available are 1.0-2960), I get a
> kernel BUG at filemap.c:236 in
>
> Trace; c012cc26 <truncate_list_pages+156/1c0>
> Trace; c012ccee <truncate_inode_pages+5e/b0>
> Trace; c01b2de3 <xfs_itruncate_start+73/a0>
> Trace; c01c95e6 <xfs_setattr+7a6/d20>
> Trace; c01d6399 <linvfs_setattr+169/1c0>
> Trace; c0155efd <notify_change+6d/170>
> Trace; c013bb09 <do_truncate+159/170>
> Trace; c013bddc <sys_ftruncate+12c/150>
> Trace; c014e76f <sys_ioctl+28f/2a0>
> Trace; c0108a3b <system_call+33/38>
> (the whole Oops is attached.)
>
> This happens reproducably with VMware, but also with Samba (nmbd), which is
> pure userland. It causes severe filesystem corruption.
>
> Using the XFree86 nv driver keeps the system stable (and the NVIDIA card
> useless). Using 2.4.18-4SGI_XFS_1.1 (with the root access fix applied) and the
> NVIDIA module also works.
>
> The XFS FAQ mentions that any NVIDIA driver later than 1.0-1251 would fit. Do
> maybe the newest NVIDIA modules break XFS again?
--
Axel.Thimm@xxxxxxxxxxxxxxxxxxx
|