Re: slub error in fs/sysfs/bin.c related code

[Sasha Levin]

On 02/28/2013 11:27 AM, Greg Kroah-Hartman wrote:
> On Thu, Feb 28, 2013 at 11:03:58AM -0500, Sasha Levin wrote:
>> Hi Greg,
>>
>> While fuzzing with trinity inside a KVM tools guest, running latest -next kernel
>> I got the following spew.
>>
>> The open() and release() callbacks in the traces below point to fs/sysfs/bin.c.
> 
> Any hint as to which sysfs binary file you were accessing when this
> happened?

No, sorry - the trace only shows the allocation and freeing, I'm not sure
which file it was.

>> [  719.288925] =============================================================================
>> [  719.290663] BUG kmalloc-4096 (Tainted: G        W   ): Redzone overwritten
>> [  719.291764] -----------------------------------------------------------------------------
>> [  719.291764]
>> [  719.294527] Disabling lock debugging due to kernel taint
>> [  719.294527] INFO: 0xffff88006d1a5520-0xffff88006d1a5520. First byte 0x0 instead of 0xcc
>> [  719.294527] INFO: Allocated in open+0xb8/0x190 age=18922 cpu=1 pid=7095
>> [  719.294527] 	__slab_alloc+0x622/0x6d0
>> [  719.294527] 	kmem_cache_alloc_trace+0x123/0x2c0
>> [  719.294527] 	open+0xb8/0x190
>> [  719.294527] 	do_dentry_open+0x229/0x330
>> [  719.294527] 	finish_open+0x54/0x70
>> [  719.294527] 	do_last+0x56c/0x790
>> [  719.294527] 	path_openat+0xbe/0x490
>> [  719.294527] 	do_filp_open+0x44/0xa0
>> [  719.294527] 	do_sys_open+0x133/0x1d0
>> [  719.294527] 	sys_open+0x1c/0x20
>> [  719.294527] 	tracesys+0xdd/0xe2
>> [  719.294527] INFO: Freed in seq_release+0x18/0x30 age=18947 cpu=0 pid=31203
>> [  719.294527] 	__slab_free+0x3c/0x590
>> [  719.294527] 	kfree+0x2cb/0x2e0
>> [  719.294527] 	seq_release+0x18/0x30
>> [  719.294527] 	single_release+0x24/0x40
>> [  719.294527] 	proc_reg_release+0xed/0x110
>> [  719.294527] 	__fput+0x122/0x2d0
>> [  719.294527] 	____fput+0x9/0x10
>> [  719.294527] 	task_work_run+0xbe/0x100
>> [  719.294527] 	do_notify_resume+0x7e/0xa0
>> [  719.294527] 	int_signal+0x12/0x17
>> [  719.294527] INFO: Slab 0xffffea0001b46800 objects=7 used=7 fp=0x          (null) flags=0x1ffc0000004081
>> [  719.294527] INFO: Object 0xffff88006d1a4520 @offset=17696 fp=0x          (null)
> 
> Hm, where is sysfs in this traceback?  I don't see it mentioned anywhere
> in this report, what am I missing?

The call trace at the end of the spew:
[  719.294527] Call Trace:
[  719.294527]  [<ffffffff81260202>] print_trailer+0x132/0x140
[  719.294527]  [<ffffffff81260641>] check_bytes_and_report+0xe1/0x130
[  719.294527]  [<ffffffff812626e2>] check_object+0x52/0x220
[  719.294527]  [<ffffffff81263a13>] free_debug_processing+0x173/0x300
[  719.294527]  [<ffffffff8130c9a2>] ? release+0x72/0x90
[  719.294527]  [<ffffffff8130c9a2>] ? release+0x72/0x90
[  719.294527]  [<ffffffff812674ec>] __slab_free+0x3c/0x590
[  719.294527]  [<ffffffff8118541d>] ? trace_hardirqs_on+0xd/0x10
[  719.294527]  [<ffffffff83d7a284>] ? _raw_spin_unlock_irqrestore+0x94/0xc0
[  719.294527]  [<ffffffff81a3898f>] ? __debug_check_no_obj_freed+0x15f/0x220
[  719.294527]  [<ffffffff83d75e89>] ? mutex_unlock+0x9/0x10
[  719.294527]  [<ffffffff8130c9a2>] ? release+0x72/0x90
[  719.294527]  [<ffffffff8130c9a2>] ? release+0x72/0x90
[  719.294527]  [<ffffffff81267d0b>] kfree+0x2cb/0x2e0
[  719.294527]  [<ffffffff8130c9a2>] release+0x72/0x90
[  719.294527]  [<ffffffff81289592>] __fput+0x122/0x2d0
[  719.294527]  [<ffffffff812897a9>] ____fput+0x9/0x10
[  719.294527]  [<ffffffff81135bbe>] task_work_run+0xbe/0x100
[  719.294527]  [<ffffffff81114781>] do_exit+0x311/0x510
[  719.294527]  [<ffffffff81114a21>] do_group_exit+0xa1/0xe0
[  719.294527]  [<ffffffff81114a72>] sys_exit_group+0x12/0x20
[  719.294527]  [<ffffffff83d833d0>] tracesys+0xdd/0xe2

Specifically, this:

[  719.294527]  [<ffffffff8130c9a2>] release+0x72/0x90

Is the release() in fs/sysfs/bin.c

I've confirmed it by also checking the sizes of the open() and
release() sizes and actual code at those positions mentioned
in the backtrace, and they match the backtrace.


Thanks,
Sasha
--
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/