Re: lockdep max numbers (was Re: Crashes during boot on 2.6.30 / 2.6.31-rc, random programs)

[Joao Correia <joaomiguelcorreia@gmail.com>]

Every boot, different programs, on a fedora 11 box. If i dont touch
any of the limits mentioned, it happens every boot.

Changing the two values mentioned, a third complain appears, althouh
more infrequently. I have had  MAX_LOCKDEP_CHAINS pop up regarding
qemu-kvm, and once again on unrar, during regular box usage.

(quad core phenom 9600, asus m3a-h/hdmi board with 4gb ram).

Thank you very much for your time,
Joao Correia
Centro de Informatica
Universidade da Beira Interior
Portugal

On Thu, Jul 2, 2009 at 11:32 AM, Peter Zijlstra<a.p.zijlstra@chello.nl> wrote:
> On Thu, 2009-07-02 at 17:31 +0800, Amerigo Wang wrote:
>> On Wed, Jul 01, 2009 at 09:30:23PM +0100, Joao Correia wrote:
>> >Hello again
>> >
>> >Looks like this just opened the lid on some other limits. I just hit
>> >another thing, which hadn't shown up before i made the changes, but
>> >looks like its just another limit thats too low. This one is harder to
>> >reproduce tho.
>> >
>> >Jun 30 21:35:03 hightech kernel: BUG: MAX_LOCKDEP_CHAINS too low!
>> >Jun 30 21:35:03 hightech kernel: turning off the locking correctness validator.
>> >Jun 30 21:35:03 hightech kernel: Pid: 9379, comm: qemu-kvm Not tainted
>>
>>
>> Similar problem...
>>
>> Peter? Would like to increase these numbers?
>
>> >2.6.30-wl #3
>
>> >Jun 30 21:35:03 hightech kernel: [<c05d0c66>] debug_dma_map_sg+0xe1/0x147
>
> We recently fixed the dma debug stuff to require less lockdep resources,
> Joerg might know if that made it into .30.
>
>> >>>So i dug a little on the source and changed
>> >>>
>> >>>include/linux/sched.h
>> >>>
>> >>># define MAX_LOCK_DEPTH 48UL
>> >>>
>> >>>to
>> >>>
>> >>># define MAX_LOCK_DEPTH 96UL
>
>> >> Let's Cc: Peter to see if he would like to change this number...
>> >>
>> >> Peter?
>
> I really don't think we want to raise this, I mean, holding more that 48
> locks at the same time is somewhat pushing it, don't you think?
>
> Where do you run into this limit?
>
>