linux-kernel - Re: [External] Re: [PATCH v5 3/5] mm: hugetlb: fix a race between freeing and dissolving the page

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <CAMZfGtW4N-CMKJUWOxQXCz+8kCLw9Hg1P3aG9nST91=18g-CvQ@mail.gmail.com>
Date:   Fri, 15 Jan 2021 00:16:04 +0800
From:   Muchun Song <songmuchun@...edance.com>
To:     Michal Hocko <mhocko@...e.com>
Cc:     Mike Kravetz <mike.kravetz@...cle.com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Naoya Horiguchi <n-horiguchi@...jp.nec.com>,
        Andi Kleen <ak@...ux.intel.com>,
        Linux Memory Management List <linux-mm@...ck.org>,
        LKML <linux-kernel@...r.kernel.org>,
        linux- stable <stable@...r.kernel.org>
Subject: Re: [External] Re: [PATCH v5 3/5] mm: hugetlb: fix a race between
 freeing and dissolving the page

On Thu, Jan 14, 2021 at 11:38 PM Michal Hocko <mhocko@...e.com> wrote:
>
> On Thu 14-01-21 21:47:36, Muchun Song wrote:
> > On Thu, Jan 14, 2021 at 9:20 PM Michal Hocko <mhocko@...e.com> wrote:
> [...]
> > > > @@ -1770,6 +1789,28 @@ int dissolve_free_huge_page(struct page *page)
> > > >               int nid = page_to_nid(head);
> > > >               if (h->free_huge_pages - h->resv_huge_pages == 0)
> > > >                       goto out;
> > > > +
> > > > +             /*
> > > > +              * We should make sure that the page is already on the free list
> > > > +              * when it is dissolved.
> > > > +              */
> > > > +             if (unlikely(!PageHugeFreed(head))) {
> > > > +                     spin_unlock(&hugetlb_lock);
> > > > +
> > > > +                     /*
> > > > +                      * Theoretically, we should return -EBUSY when we
> > > > +                      * encounter this race. In fact, we have a chance
> > > > +                      * to successfully dissolve the page if we do a
> > > > +                      * retry. Because the race window is quite small.
> > > > +                      * If we seize this opportunity, it is an optimization
> > > > +                      * for increasing the success rate of dissolving page.
> > > > +                      */
> > > > +                     while (PageHeadHuge(head) && !PageHugeFreed(head))
> > > > +                             cond_resched();
> > >
> > > Sorry, I should have raised that when replying to the previous version
> > > already but we have focused more on other things. Is there any special
> > > reason that you didn't simply
> > >         if (!PageHugeFreed(head)) {
> > >                 spin_unlock(&hugetlb_lock);
> > >                 cond_resched();
> > >                 goto retry;
> > >         }
> > >
> > > This would be less code and a very slight advantage would be that the
> > > waiter might get blocked on the spin lock while the concurrent freeing
> > > is happening. But maybe you wanted to avoid exactly this contention?
> > > Please put your thinking into the changelog.
> >
> > I want to avoid the lock contention. I will add this reason
> > to the changelog. Thanks.
>
> Please also explain why it matters and whether an unintended contention
> is a real problem.

I have no idea about this, it is just my opinion.
I will follow your suggestion.

> --
> Michal Hocko
> SUSE Labs