Re: [RFC PATCH] mm: thp: implement THP reservations for anonymous memory

From: anthony . yznaga
Date: Fri Nov 09 2018 - 19:41:08 EST




On 11/09/2018 07:34 AM, Zi Yan wrote:
> On 9 Nov 2018, at 8:11, Mel Gorman wrote:
>
>> On Fri, Nov 09, 2018 at 03:13:18PM +0300, Kirill A. Shutemov wrote:
>>> On Thu, Nov 08, 2018 at 10:48:58PM -0800, Anthony Yznaga wrote:
>>>> The basic idea as outlined by Mel Gorman in [2] is:
>>>>
>>>> 1) On first fault in a sufficiently sized range, allocate a huge page
>>>> sized and aligned block of base pages. Map the base page
>>>> corresponding to the fault address and hold the rest of the pages in
>>>> reserve.
>>>> 2) On subsequent faults in the range, map the pages from the reservation.
>>>> 3) When enough pages have been mapped, promote the mapped pages and
>>>> remaining pages in the reservation to a huge page.
>>>> 4) When there is memory pressure, release the unused pages from their
>>>> reservations.
>>> I haven't yet read the patch in details, but I'm skeptical about the
>>> approach in general for few reasons:
>>>
>>> - PTE page table retracting to replace it with huge PMD entry requires
>>> down_write(mmap_sem). It makes the approach not practical for many
>>> multi-threaded workloads.
>>>
>>> I don't see a way to avoid exclusive lock here. I will be glad to
>>> be proved otherwise.
>>>
>> That problem is somewhat fundamental to the mmap_sem itself and
>> conceivably it could be alleviated by range-locking (if that gets
>> completed). The other thing to bear in mind is the timing. If the
>> promotion is in-place due to reservations, there isn't the allocation
>> overhead and the hold times *should* be short.
>>
> Is it possible to convert all these PTEs to migration entries during
> the promotion and replace them with a huge PMD entry afterwards?
> AFAIK, migrating pages does not require holding a mmap_sem.
> Basically, it will act like migrating 512 base pages to a THP without
> actually doing the page copy.
That's an interesting idea. I'll look into it.

Thanks,
Anthony

>
> --
> Best Regards
> Yan Zi