Re: [PATCH] cpumask: remove cpumask allocation from idle_balance

[Ingo Molnar <mingo@elte.hu>]

* Jeremy Fitzhardinge <jeremy@goop.org> wrote:

> Ingo Molnar wrote:
>> * Ingo Molnar <mingo@elte.hu> wrote:
>>
>>   
>>>> This time for sure,
>>>>       
>>> not so sure:
>>>
>>> [    4.303068] =============================================================================
>>> [    4.303332] BUG idr_layer_cache: Redzone overwritten
>>> [    4.303332] -----------------------------------------------------------------------------
>>> [    4.303332] [    4.303332] INFO: 
>>> 0xffff88003f838958-0xffff88003f83895f. First byte 0xa1 instead of 
>>> 0xbb
>>> [    4.303332] INFO: Allocated in 0xffff88003f9b94b0 age=4294878587 cpu=2 pid=0
>>> [    4.303332] INFO: Freed in 0xffff88003f8389f0 age=18446744073709463001 cpu=4294967295 pid=0
>>> [    4.303332] INFO: Slab 0xffffe200015d5340 objects=13 used=10 fp=0xffff88003f838738 flags=0x40000000000020c3
>>> [    4.303332] INFO: Object 0xffff88003f838738 @offset=1848 fp=0x(null)
>>>     
>>
>> hm, i'm seeing this even with the cpus4096 bits removed.
>>
>> It _seems_ to trigger when i add these two commits:
>>
>>   71ff49d: x86: with the last user gone, remove set_pte_present
>>   b40c757: x86/32: no need to use set_pte_present in set_pte_vaddr
>>
>> but i'm not completely sure it's related, i tried a bisection and it  
>> showed heisenbug properties.
>>
>> Might have come from upstream, that saw some IDR changes recently.
>>   
>
> Seems very unlikely; those patches only touch 32-bit only code, 
> and this looks like a 64-bit crash.

yes - that's why i think it's a Heisenbug.

	Ingo