OK, If I just enable CONFIG_DEBUG_STACKOVERFLOW I still get an Oops but no
additional info. Does this only help when using a debugger or is it meant
to spew error/warning messages to the screen?
When I enable CONFIG_DEBUG_STACK_USAGE I no longer get the Oops.
I have KDB compiled in with CONFIG_DEBUG_STACK_USAGE off so hopefully I get
an Oops I can actually work with.
Any other recommendations of what to try to help track down what is
happening.
Thanks,
Joe
-----Original Message-----
From: Joe Eiler
Sent: Wednesday, September 01, 2004 10:49 AM
To: 'Eric Sandeen'; Joe Eiler
Cc: 'linux-xfs@xxxxxxxxxxx'
Subject: RE: Oops when running xfs_fsr
Thanks for the quick reply.
I am running with 8K stacks. I read that in another response dealing with
the 2.6.5 kernel and xfs_fsr and just forgot to mention that I checked it in
another email. I tried to enable CONFIG_DEBUG_STACKOVERFLOW,
CONFIG_DEBUG_STACK_USAGE, and KDB but then everything runs great. I will
try to activate debugging stuff one at a time.
#
# Kernel hacking
#
CONFIG_DEBUG_KERNEL=y
CONFIG_EARLY_PRINTK=y
# CONFIG_DEBUG_STACKOVERFLOW is not set
# CONFIG_DEBUG_STACK_USAGE is not set
# CONFIG_DEBUG_SLAB is not set
# CONFIG_MAGIC_SYSRQ is not set
# CONFIG_DEBUG_SPINLOCK is not set
# CONFIG_DEBUG_PAGEALLOC is not set
# CONFIG_DEBUG_INFO is not set
# CONFIG_DEBUG_SPINLOCK_SLEEP is not set
# CONFIG_FRAME_POINTER is not set
# CONFIG_4KSTACKS is not set
# CONFIG_KDB is not set
CONFIG_X86_FIND_SMP_CONFIG=y
CONFIG_X86_MPPARSE=y
I took a quick look at stack usage and xfs_fsr.o didn't look too bad, at
least on a per function level
380 main
1052 gettmpname
1064 fsrall_cleanup
1200 packfile
1260 fsrallfs
4096 getnextents
4124 read_fd_bmap
7020 fsrfs
-----Original Message-----
From: Eric Sandeen [mailto:sandeen@xxxxxxx]
Sent: Wednesday, September 01, 2004 10:23 AM
To: Joe Eiler
Cc: 'linux-xfs@xxxxxxxxxxx'
Subject: Re: Oops when running xfs_fsr
Joe Eiler wrote:
> I am getting an Oops when running xfs_fsr. The only way I can get it to
> disappear is to turn on debugging in the kernel ;-) From my limited stack
> tracing abilities it doesn't look like and xfs problem but I can't
reproduce
> it anyother way than with xfs_fsr.
>
> I have reliably reproduced the problem on 3 different machines running
> 2.6.7, 2.6.8.1 (from kernel.org) Most of the rest of the system is fedora
> core2 rpms.
Are you using 4KSTACKS?
If you oops on reiserfs while running xfs userspace, something is going
quite wrong; I'd suspect corruption of some sort - and blowing a stack
might do that. Especially if you've got volume managers under it...
Try turning off CONFIG_4KSTACKS if it's on, and perhaps turn on
CONFIG_DEBUG_STACKOVERFLOW and CONFIG_DEBUG_STACK_USAGE
If you don't have 4kstacks, and you still see overflows, I may have a
patch that will help trim down stack usage during fsr.
-Eric
|