Re: Repeatable ext4 oops with 3.6.0 (regression)

["Holger Hoffstätte" <holger.hoffstaette@googlemail.com>]

(dear -mm: please see
http://thread.gmane.org/gmane.comp.file-systems.ext4/34665 for the
origins of this oops)

On 10/04/12 19:34, Theodore Ts'o wrote:
> On Thu, Oct 04, 2012 at 05:31:41PM +0200, Holger Hoffstatte wrote:
> 
>> So armed with multiple running shells I finally managed to save the dmesg
>> to NFS. It doesn't get any more complete than this and again shows the
>> ext4 stacktrace from before. So maybe it really is generic kmem corruption
>> and ext4 looking at symlinks/inodes is just the victim.
> 
> That certainly seems to be the case.  As near as I can tell from the

Good to know. Unfortunately I'm still at a loss why apparently only
gthumb can trigger this; I had not noticed any other problems with 3.6.0
before that (ran it for half a day, desktop use).

> So it's very likely that the crash in __kmalloc() is probably caused
> by the internal slab/slub data structures getting scrambled.

For giggles I rebuilt with SLAB instead of SLUB, but no luck; same
segfault and delayed oopsie. I also collected an strace, but I cannot
really see anything out of the ordinary - it starts, loads things,
traverses directories and then segfaults.

I've put the earlier full dmesg and the strace into
http://hoho.dyndns.org/~holger/ext4-oops-3.6.0/ - maybe it helps someone
else.

Any suggestions for memory debugging? I saw several options in the
kernel config (under "Kernel Hacking") but was not sure what to enable.

Holger

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>