Hi Marcin,
(I'm cc'ing the ext3 developers)
On Thu, Sep 4, 2008 at 12:46 AM, Marcin Slusarz
<[email protected]> wrote:
> Sep 1 20:22:07 [kernel] [ 8541.403598] slab error in verify_redzone_free(): cache `ext3_inode_cache': memory outside object was overwritten
> Sep 1 20:22:07 [kernel] [ 8541.403613] Pid: 206, comm: kswapd0 Not tainted 2.6.27-rc5 #362
> Sep 1 20:22:07 [kernel] [ 8541.403620]
> Sep 1 20:22:07 [kernel] [ 8541.403621] Call Trace:
> Sep 1 20:22:07 [kernel] [ 8541.403635] [<ffffffff8028c034>] __slab_error+0x26/0x28
> Sep 1 20:22:07 [kernel] [ 8541.403640] [<ffffffff8028c84d>] cache_free_debugcheck+0x13a/0x1fc
> Sep 1 20:22:07 [kernel] [ 8541.403646] [<ffffffff802e966d>] ? ext3_destroy_inode+0x80/0x87
> Sep 1 20:22:07 [kernel] [ 8541.403649] [<ffffffff8028cb53>] kmem_cache_free+0x4e/0xad
> Sep 1 20:22:07 [kernel] [ 8541.403653] [<ffffffff802e966d>] ext3_destroy_inode+0x80/0x87
> Sep 1 20:22:07 [kernel] [ 8541.403657] [<ffffffff802a45e5>] destroy_inode+0x37/0x4c
> Sep 1 20:22:07 [kernel] [ 8541.403661] [<ffffffff802a4fdc>] dispose_list+0xcf/0x103
> Sep 1 20:22:07 [kernel] [ 8541.403664] [<ffffffff802a51e6>] shrink_icache_memory+0x1d6/0x208
> Sep 1 20:22:07 [kernel] [ 8541.403669] [<ffffffff802747ba>] shrink_slab+0xeb/0x181
> Sep 1 20:22:07 [kernel] [ 8541.403673] [<ffffffff80274f60>] kswapd+0x317/0x4b4
> Sep 1 20:22:07 [kernel] [ 8541.403677] [<ffffffff80273610>] ? isolate_pages_global+0x0/0x39
> Sep 1 20:22:07 [kernel] [ 8541.403684] [<ffffffff80241394>] ? autoremove_wake_function+0x0/0x3d
> Sep 1 20:22:07 [kernel] [ 8541.403688] [<ffffffff80274c49>] ? kswapd+0x0/0x4b4
> Sep 1 20:22:07 [kernel] [ 8541.403692] [<ffffffff8024127f>] kthread+0x4e/0x7b
> Sep 1 20:22:07 [kernel] [ 8541.403696] [<ffffffff8020c039>] child_rip+0xa/0x11
> Sep 1 20:22:07 [kernel] [ 8541.403701] [<ffffffff80228a10>] ? finish_task_switch+0x5f/0xa5
> Sep 1 20:22:07 [kernel] [ 8541.403705] [<ffffffff802289b1>] ? finish_task_switch+0x0/0xa5
> Sep 1 20:22:07 [kernel] [ 8541.403708] [<ffffffff8020bb83>] ? restore_args+0x0/0x30
> Sep 1 20:22:07 [kernel] [ 8541.403713] [<ffffffff80241231>] ? kthread+0x0/0x7b
> Sep 1 20:22:07 [kernel] [ 8541.403716] [<ffffffff8020c02f>] ? child_rip+0x0/0x11
> Sep 1 20:22:07 [kernel] [ 8541.403718]
> Sep 1 20:22:07 [kernel] [ 8541.403721] ffff88000000f9d8: redzone 1:0xd84156c5635688c0, redzone 2:0x43220000ff3e0001.
>
> Full log and .config is here: http://www.kadu.net/~joi/kernel/2008.09.03/
> I don't remember what I was doing while this happened - I noticed this error today...
OK, so redzone1 looks good ("RED_ACTIVE") whereas redzone2, which is
the red-zone at the end of an object, has a bogus value. So looks like
a simple buffer overflow in the ext3_inode_cachep cache. If you can
reproduce this, you could try CONFIG_SLUB which shows us the full
corruption that can give us a better clue why this is happening.
Pekka