Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20200924175531.GH79898@xz-x1> <20200924181501.GF9916@ziepe.ca>
 <20200924183418.GJ79898@xz-x1> <20200924183953.GG9916@ziepe.ca>
 <20200924213010.GL79898@xz-x1> <CAHk-=wgz5SXKA6-uZ_BimOP1C7pHJag0ndz=tnJDAZS_Z+FrGQ@mail.gmail.com>
 <CAHk-=whDSH_MRMt80JaSwoquzt=1nQ-0n3w0aVngoWPAc10BCw@mail.gmail.com>
 <20200926004136.GJ9916@ziepe.ca> <CAHk-=wiutA_J-OfvrD8Kp3SoYcfMHUwsU7ViOH48q7QN0AQ6eg@mail.gmail.com>
 <CAHk-=wi_gd+JWj-8t8tc8cy3WZ7NMj-_1hATfH3Rt0ytUxtMpQ@mail.gmail.com>
 <20200927062337.GE2280698@unreal> <CAHk-=winqSOFsdn1ntYL13s2UuhpQQ9+GRvjWth3sA5APY4Wwg@mail.gmail.com>
In-Reply-To: <CAHk-=winqSOFsdn1ntYL13s2UuhpQQ9+GRvjWth3sA5APY4Wwg@mail.gmail.com>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Sun, 27 Sep 2020 11:45:30 -0700
Message-ID: <CAHk-=wj61s30pt8POVtKYVamYTh6h=7-_ser2Hx9sEjqeACkDA@mail.gmail.com>
Subject: Re: [PATCH 1/5] mm: Introduce mm_struct.has_pinned
To:     Leon Romanovsky <leonro@nvidia.com>
Cc:     Jason Gunthorpe <jgg@ziepe.ca>, Peter Xu <peterx@redhat.com>,
        John Hubbard <jhubbard@nvidia.com>,
        Linux-MM <linux-mm@kvack.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Andrew Morton <akpm@linux-foundation.org>,
        Jan Kara <jack@suse.cz>, Michal Hocko <mhocko@suse.com>,
        Kirill Tkhai <ktkhai@virtuozzo.com>,
        Kirill Shutemov <kirill@shutemov.name>,
        Hugh Dickins <hughd@google.com>,
        Christoph Hellwig <hch@lst.de>,
        Andrea Arcangeli <aarcange@redhat.com>,
        Oleg Nesterov <oleg@redhat.com>, Jann Horn <jannh@google.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Sun, Sep 27, 2020 at 11:16 AM Linus Torvalds
<torvalds@linux-foundation.org> wrote:
>
> Btw, I'm not convinced about the whole "turn the pte read-only and
> then back". If the fork races with another thread doing a pinning
> fast-GUP on another CPU, there are memory ordering issues etc too.
> That's not necessarily visible on x86 (the "turn read-only being a
> locked op will force serialization), but it all looks dodgy as heck.

.. looking at it more, I also think it could possibly lose the dirty
bit for the case where another CPU did a HW dirty/accessed bit update
in between the original read of the pte, and then us writing back the
writable pte again.

Us holding the page table lock means that no _software_ accesses will
happen to the PTE, but dirty/accessed bits can be modified by hardware
despite the lock.

That is, of course, a completely crazy case, and I think that since we
only do this for a COW mapping, and only do the PTE changes if the pte
was writable, the pte will always have been dirty already.

So I don't think it's an _actual_ bug, but it's another "this looks
dodgy as heck" marker. It may _work_, but it sure ain't pretty.

But despite having looked at this quite a bit, I don't see anything
that looks actively wrong, so I think the series is fine. This is more
of a note for people to perhaps think about.

                Linus