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;
Content-Type: text/plain;
        charset=utf-8
Mime-Version: 1.0 (Mac OS X Mail 13.4 \(3608.120.23.2.4\))
Subject: Re: [PATCH] mm/userfaultfd: fix memory corruption due to writeprotect
From:   Nadav Amit <nadav.amit@gmail.com>
In-Reply-To: <YAQAhyOFqEdjTRPJ@google.com>
Date:   Sun, 17 Jan 2021 02:13:43 -0800
Cc:     Will Deacon <will@kernel.org>,
        Laurent Dufour <ldufour@linux.vnet.ibm.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Vinayak Menon <vinmenon@codeaurora.org>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        Andy Lutomirski <luto@kernel.org>,
        Peter Xu <peterx@redhat.com>,
        Andrea Arcangeli <aarcange@redhat.com>,
        linux-mm <linux-mm@kvack.org>,
        lkml <linux-kernel@vger.kernel.org>,
        Pavel Emelyanov <xemul@openvz.org>,
        Mike Kravetz <mike.kravetz@oracle.com>,
        Mike Rapoport <rppt@linux.vnet.ibm.com>,
        stable <stable@vger.kernel.org>,
        Minchan Kim <minchan@kernel.org>, surenb@google.com,
        Mel Gorman <mgorman@suse.de>
Content-Transfer-Encoding: quoted-printable
Message-Id: <1A664155-462A-451D-A21E-D749A0ADBD09@gmail.com>
References: <bfb1cbe6-a705-469d-c95a-776624817e33@codeaurora.org>
 <0201238b-e716-2a3c-e9ea-d5294ff77525@linux.vnet.ibm.com>
 <X/3VE64nr91WCtuM@hirez.programming.kicks-ass.net>
 <ec912505-ed4d-a45d-2ed4-7586919da4de@linux.vnet.ibm.com>
 <C7D5A74C-25BF-458A-AAD9-61E484B9F225@gmail.com>
 <X/3+6ZnRCNOwhjGT@google.com>
 <2C7AE23B-ACA3-4D55-A907-AF781C5608F0@gmail.com>
 <20210112214337.GA10434@willie-the-truck> <YAO/9YVceghRYo4T@google.com>
 <85DAADF4-2537-40BD-8580-A57C201FF5F3@gmail.com>
 <YAQAhyOFqEdjTRPJ@google.com>
To:     Yu Zhao <yuzhao@google.com>
Precedence: bulk

> On Jan 17, 2021, at 1:16 AM, Yu Zhao <yuzhao@google.com> wrote:
>=20
> On Sat, Jan 16, 2021 at 11:32:22PM -0800, Nadav Amit wrote:
>>> On Jan 16, 2021, at 8:41 PM, Yu Zhao <yuzhao@google.com> wrote:
>>>=20
>>> On Tue, Jan 12, 2021 at 09:43:38PM +0000, Will Deacon wrote:
>>>> On Tue, Jan 12, 2021 at 12:38:34PM -0800, Nadav Amit wrote:
>>>>>> On Jan 12, 2021, at 11:56 AM, Yu Zhao <yuzhao@google.com> wrote:
>>>>>> On Tue, Jan 12, 2021 at 11:15:43AM -0800, Nadav Amit wrote:
>>>>>>> I will send an RFC soon for per-table deferred TLB flushes =
tracking.
>>>>>>> The basic idea is to save a generation in the page-struct that =
tracks
>>>>>>> when deferred PTE change took place, and track whenever a TLB =
flush
>>>>>>> completed. In addition, other users - such as mprotect - would =
use
>>>>>>> the tlb_gather interface.
>>>>>>>=20
>>>>>>> Unfortunately, due to limited space in page-struct this would =
only
>>>>>>> be possible for 64-bit (and my implementation is only for =
x86-64).
>>>>>>=20
>>>>>> I don't want to discourage you but I don't think this would end =
up
>>>>>> well. PPC doesn't necessarily follow one-page-struct-per-table =
rule,
>>>>>> and I've run into problems with this before while trying to do
>>>>>> something similar.
>>>>>=20
>>>>> Discourage, discourage. Better now than later.
>>>>>=20
>>>>> It will be relatively easy to extend the scheme to be per-VMA =
instead of
>>>>> per-table for architectures that prefer it this way. It does =
require
>>>>> TLB-generation tracking though, which Andy only implemented for =
x86, so I
>>>>> will focus on x86-64 right now.
>>>>=20
>>>> Can you remind me of what we're missing on arm64 in this area, =
please? I'm
>>>> happy to help get this up and running once you have something I can =
build
>>>> on.
>>>=20
>>> I noticed arm/arm64 don't support ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH.
>>> Would it be something worth pursuing? Arm has been using mm_cpumask,
>>> so it might not be too difficult I guess?
>>=20
>> [ +Mel Gorman who implemented ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH ]
>>=20
>> IIUC, there are at least two bugs in x86 implementation.
>>=20
>> First, there is a missing memory barrier in tlbbatch_add_mm() between
>> inc_mm_tlb_gen() and the read of mm_cpumask().
>=20
> In arch_tlbbatch_add_mm()? inc_mm_tlb_gen() has builtin barrier as its
> comment says -- atomic update ops that return values are also full
> memory barriers.

Yes, you are correct.

>=20
>> Second, try_to_unmap_flush() clears flush_required after flushing. =
Another
>> thread can call set_tlb_ubc_flush_pending() after the flush and =
before
>> flush_required is cleared, and the indication that a TLB flush is =
pending
>> can be lost.
>=20
> This isn't a problem either because flush_required is per thread.

Sorry, I meant mm->tlb_flush_batched . It is not per-thread.
flush_tlb_batched_pending() clears it after flush and indications that
set_tlb_ubc_flush_pending() sets in between can be lost.

>> I am working on addressing these issues among others, but, as you =
already
>> saw, I am a bit slow.
>>=20
>> On a different but related topic: Another thing that I noticed that =
Arm does
>> not do is batching TLB flushes across VMAs. Since Arm does not have =
its own
>> tlb_end_vma(), it uses the default tlb_end_vma(), which flushes each =
VMA
>> separately. Peter Zijlstra=E2=80=99s comment says that there are =
advantages in
>> flushing each VMA separately, but I am not sure it is better or =
intentional
>> (especially since x86 does not do so).
>>=20
>> I am trying to remove the arch-specific tlb_end_vma() and have a =
config
>> option to control this behavior.
>=20
> One thing worth noting is not all arm/arm64 hw versions support =
ranges.
> (system_supports_tlb_range()). But IIUC what you are trying to do, =
this
> isn't a problem.

I just wanted to get rid of arch-specific tlb_start_vma() it in order to
cleanup the code (I am doing first the VMA-deferred tracking, as you =
asked).
While I was doing that, I noticed that Arm does per-VMA TLB flushes. I =
do
not know whether it is intentional, but it seems rather easy to get this
behavior unintentionally.=