Messages in this thread | | | Date | Wed, 27 Jan 2010 10:02:15 +0200 | From | Pekka Enberg <> | Subject | Re: BUG at mm/slab.c:2990 with 2.6.33-rc5-tuxonice |
| |
Pedro Ribeiro kirjoitti: > 2010/1/26 Pedro Ribeiro <pedrib@gmail.com>: >> 2010/1/26 Pekka Enberg <penberg@cs.helsinki.fi>: >>> Matt Mackall wrote: >>>> On Wed, 2010-01-27 at 07:39 +1100, Nigel Cunningham wrote: >>>>> Hi. >>>>> >>>>> Pedro Ribeiro wrote: >>>>>> Hi, >>>>>> >>>>>> I hit a bug at mm/slab.c:2990 with .33-rc5. >>>>>> Unfortunately nothing more is available than a screen picture with a >>>>>> crash dump, although it is a good one. >>>>>> The bug was hit almost at the end of a hibernation cycle with >>>>>> Tux-on-Ice, while saving memory contents to an encrypted swap >>>>>> partition. >>>>>> >>>>>> The image is here http://img264.imageshack.us/img264/9634/mmslab.jpg >>>>>> (150 kb) >>>>>> >>>>>> Hopefully it is of any use for you. Please let me know if you need any >>>>>> more info. >>>>> Looks to me to be completely unrelated to TuxOnIce - at least at a first >>>>> glance. >>>>> >>>>> Ccing the slab allocator maintainers listed in MAINTAINERS. >>>> Not sure if this will do us any good, it's the second oops. >>> Looks like slab corruption to me which is usually not a slab bug but caused >>> by buggy callers. Is CONFIG_DEBUG_SLAB enabled? >>> >> I have enabled it and compiled the kernel. As soon as I hit the bug, I >> will send a photo here. >> >> Regards, >> Pedro >> > > The pic is here. > http://img43.imageshack.us/img43/3644/dsc01061ko.jpg > > There was a buttload of output before that, which I tried capturing in > video, but its too crappy to post.
Can you try passing "pause_on_oops=15" as kernel parameter? It should delay the next oops for 15 seconds so there's enough time to take a picture of the first one?
Also, you could try CONFIG_SLUB and passing "slub_debug" as kernel parameter to get nicer diagnostics of the bug.
> Do you know if/where I can get patches for KDB in .33?
Nope, sorry.
Pekka
| |