Sorry, this is the kernel I use on this SL42 server:

kernel-smp-2.6.9-11.EL.XFS

On 3 other servers I use:

kernel-2.6.9-11.EL.XFS

but I haven't experienced the problem (yet) on those other three.

Michael.

> Hi,
> 
> I currently use the SL4 XFS kernel in the sl-contrib and have 
> experienced a couple of times a hard crash on my server saying:
> 
> do_IRQ: stack overflow: 284
>  [<c01078a2>] do_IRQ+0x44/0x130
> 
> After various google search on "xfs 4k stack" etc and RHEL4, it 
> turns out that RH have 4k stacks in the kernel while XFS requires 8k 
> stacks. Links similar to the following will show up in google about this:
> 
> http://www.ussg.iu.edu/hypermail/linux/kernel/0407.3/1088.html
> 
>
http://www.issociate.de/board/post/210876/SATA_RAID5_kernel_hang_on_2.6.11-1.14_FC3.html
> 
> It seems though that Red Hat also disabled the facility to allow 
> people to increase the stack size when building their own kernels (I 
> read that in one of the posts), so I'm wondering, with the people 
> that built the SL4 XFS kernels, did they enable 4k or 8k stacks when 
> compiling the kernel?
> 
> I'm at a stage now where I'm likely going to have to get rid of XFS 
> and go back to ext3 because of this 4k stack issue causing my server 
> to hard crash.
> 
> Thanks.
> 
> Michael.
------- End of Original Message -------