Re: [PATCH] mm/x86: Flush lazy MMU when DEBUG_PAGEALLOC is set

["H. Peter Anvin" <hpa@zytor.com>]

Greg, policy opinion?

	-hpa

On 02/26/2013 03:57 PM, Boris Ostrovsky wrote:
> 
> ----- hpa@zytor.com wrote:
> 
>> On 02/26/2013 02:56 PM, Boris Ostrovsky wrote:
>>> When CONFIG_DEBUG_PAGEALLOC is set page table updates made by
>>> kernel_map_pages() are not made visible (via TLB flush) immediately
>> if lazy
>>> MMU is on. In environments that support lazy MMU (e.g. Xen) this may
>> lead to
>>> fatal page faults, for example, when zap_pte_range() needs to
>> allocate pages
>>> in __tlb_remove_page() -> tlb_next_batch().
>>>
>>> Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
>>> ---
>>>  arch/x86/mm/pageattr.c | 2 ++
>>>  1 file changed, 2 insertions(+)
>>>
>>> diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c
>>> index ca1f1c2..7b3216e 100644
>>> --- a/arch/x86/mm/pageattr.c
>>> +++ b/arch/x86/mm/pageattr.c
>>> @@ -1369,6 +1369,8 @@ void kernel_map_pages(struct page *page, int
>> numpages, int enable)
>>>  	 * but that can deadlock->flush only current cpu:
>>>  	 */
>>>  	__flush_tlb_all();
>>> +
>>> +	arch_flush_lazy_mmu_mode();
>>>  }
>>>  
>>>  #ifdef CONFIG_HIBERNATION
>>>
>>
>> This sounds like a critical fix, i.e. a -stable candidate.  Am I
>> correct?
> 
> I considered copying stable but then I decided that this is a debugging feature
> --- kernel_map_pages() is only defined if CONFIG_DEBUG_PAGEALLOC is set and my
> thinking was that stable kernels usually don't do this.
> 
> 
> -boris
>