Re: repeatable slab corruption with LTP msgctl08

[Andrew Morton]

On Thu, 12 Jun 2008 21:09:10 +0200
Manfred Spraul <manfred@xxxxxxxxxxxxxxxx> wrote:

> Either someone does a set_bit() or your cpu is breaking down.

Well.  It is about ten years old.  But this is the first sign of a
problem and it's always msgctl08.

>  From looking at the the msgctl08 test: it shouldn't produce any races, 
> it just does lots of bulk msgsnd()/msgrcv() operations. Always one 
> thread sends, one thread receives on each queue. It's probably more a 
> scheduler stresstest than anything else.
> 
> Attached is a completely untested patch:
> - add 8 bytes to each slabp struct: This changes the alignment of the 
> bufctl entries.
> - add a hexdump of the redzone bytes.

OK, I'll try that this evening (eight hours hence).

I'll also try increasing /proc/sys/kernel/msgmni under 2.6.25.

> Andrew: how do you log the oops? 
> it might scroll of the screen.

netconsole-to-disk.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/