Re: [PATCH v6 03/24] mm: Dont assume page-table invariance during faults

From: Laurent Dufour
Date: Wed Jan 17 2018 - 03:57:34 EST


On 17/01/2018 04:04, Andi Kleen wrote:
> Laurent Dufour <ldufour@xxxxxxxxxxxxxxxxxx> writes:
>
>> From: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
>>
>> One of the side effects of speculating on faults (without holding
>> mmap_sem) is that we can race with free_pgtables() and therefore we
>> cannot assume the page-tables will stick around.
>>
>> Remove the reliance on the pte pointer.
>
> This needs a lot more explanation. So why is this code not needed with
> SPF only?

Hi Andi,

This is a good question, and I should detail that more in the commit's log.

Here is my response to Balbir when he asked for:

On 10/07/2017 19:48, Laurent Dufour wrote:
> On 07/07/2017 09:07, Balbir Singh wrote:
>> On Fri, 2017-06-16 at 19:52 +0200, Laurent Dufour wrote:
>>> From: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
>>>
>>> One of the side effects of speculating on faults (without holding
>>> mmap_sem) is that we can race with free_pgtables() and therefore we
>>> cannot assume the page-tables will stick around.
>>>
>>> Remove the relyance on the pte pointer.
>> ^^ reliance
>>
>> Looking at the changelog and the code the impact is not clear.
>> It looks like after this patch we always assume the pte is not
>> the same. What is the impact of this patch?
>
> Hi Balbir,
>
> In most of the case pte_unmap_same() was returning 1, which meaning that
> do_swap_page() should do its processing.
>
> So in most of the case there will be no impact.
>
> Now regarding the case where pte_unmap_safe() was returning 0, and thus
> do_swap_page return 0 too, this happens when the page has already been
> swapped back. This may happen before do_swap_page() get called or while in
> the call to do_swap_page(). In that later case, the check done when
> swapin_readahead() returns will detect that case.
>
> The worst case would be that a page fault is occuring on 2 threads at the
> same time on the same swapped out page. In that case one thread will take
> much time looping in __read_swap_cache_async(). But in the regular page
> fault path, this is even worse since the thread would wait for semaphore to
> be released before starting anything.
>
> Cheers,
> Laurent.
>

I'll add that to the commit's log.

Thanks,
Laurent.