Re: [PATCH v4] mm, hugetlb: fix racy resv_huge_pages underflow on UFFDIO_COPY
From: Mina Almasry
Date: Tue Jun 01 2021 - 14:04:44 EST
On Tue, Jun 1, 2021 at 10:09 AM Mike Kravetz <mike.kravetz@xxxxxxxxxx> wrote:
>
> On 5/31/21 7:48 PM, Mina Almasry wrote:
> > On Mon, May 31, 2021 at 5:36 PM Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
> >> On Mon, 31 May 2021 17:11:52 -0700 Mina Almasry <almasrymina@xxxxxxxxxx> wrote:
> >>> On Mon, May 31, 2021 at 4:25 PM Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
> >>>> On Thu, 27 May 2021 17:50:29 -0700 Mina Almasry <almasrymina@xxxxxxxxxx> wrote:
> >>> I've sent 2 similar patches to the list:
> >>>
> >>> 1. "[PATCH v4] mm, hugetlb: Fix simple resv_huge_pages underflow on UFFDIO_COPY"
> >>>
> >>> This one is sent to -stable and linux-mm and is a fairly simple fix.
> >>>
> >>> 2. "[PATCH v4] mm, hugetlb: fix racy resv_huge_pages underflow on UFFDIO_COPY"
> >>
> >> Ah, OK, the title of the first patch was changed, which threw me off.
> >>
> >> I'd skipped "[PATCH v4] mm, hugetlb: Fix simple resv_huge_pages
> >> underflow on UFFDIO_COPY" because Mike's comments appeared to require a
> >> v5. I applied it and made Mike's changelog suggestions. Queued for
> >> 5.13 and -stable.
> >>
> >> And I queued "[PATCH v4] mm, hugetlb: fix racy resv_huge_pages
> >> underflow on UFFDIO_COPY" for 5.14.
> >>
> >>
> >
> > Awesome, thanks! And sorry for the confusion!
> >
>
> Mina, does this patch depend on changes to restore_reserve_on_error()?
>
Yes, this patch (and only this patch) depends on your changes for
complete and correct functionality. I'm not sure what's the impact
> I am still working on those changes. It may be a few days before I can
> have something finalized.
>
> If this does depend on restore_reserve_on_error as I suspect, perhaps we
> should send these together.
I was thinking it's fine to have my fix in Andrew's tree a few days
before yours, since the race is hard to reproduce and even if the race
reproduces the userfaultfd tests still pass, so I don't see any
disastrous consequences, but I'm happy to do whatever is appropriate
here.
> --
> Mike Kravetz