linux-kernel - Re: [PATCH v4] mm/gup: disallow GUP writing to file-backed mappings by default

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

Message-ID: <1b9e3406-c08e-b97c-d46f-22f36535d9e5@nvidia.com>
Date:   Wed, 26 Apr 2023 13:04:48 -0700
From:   John Hubbard <jhubbard@...dia.com>
To:     Lorenzo Stoakes <lstoakes@...il.com>,
        Mika Penttilä <mpenttil@...hat.com>
Cc:     linux-mm@...ck.org, linux-kernel@...r.kernel.org,
        Andrew Morton <akpm@...ux-foundation.org>,
        Jason Gunthorpe <jgg@...pe.ca>, Jens Axboe <axboe@...nel.dk>,
        Matthew Wilcox <willy@...radead.org>,
        Dennis Dalessandro <dennis.dalessandro@...nelisnetworks.com>,
        Leon Romanovsky <leon@...nel.org>,
        Christian Benvenuti <benve@...co.com>,
        Nelson Escobar <neescoba@...co.com>,
        Bernard Metzler <bmt@...ich.ibm.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Ingo Molnar <mingo@...hat.com>,
        Arnaldo Carvalho de Melo <acme@...nel.org>,
        Mark Rutland <mark.rutland@....com>,
        Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
        Jiri Olsa <jolsa@...nel.org>,
        Namhyung Kim <namhyung@...nel.org>,
        Ian Rogers <irogers@...gle.com>,
        Adrian Hunter <adrian.hunter@...el.com>,
        Bjorn Topel <bjorn@...nel.org>,
        Magnus Karlsson <magnus.karlsson@...el.com>,
        Maciej Fijalkowski <maciej.fijalkowski@...el.com>,
        Jonathan Lemon <jonathan.lemon@...il.com>,
        "David S . Miller" <davem@...emloft.net>,
        Eric Dumazet <edumazet@...gle.com>,
        Jakub Kicinski <kuba@...nel.org>,
        Paolo Abeni <pabeni@...hat.com>,
        Christian Brauner <brauner@...nel.org>,
        Richard Cochran <richardcochran@...il.com>,
        Alexei Starovoitov <ast@...nel.org>,
        Daniel Borkmann <daniel@...earbox.net>,
        Jesper Dangaard Brouer <hawk@...nel.org>,
        John Fastabend <john.fastabend@...il.com>,
        linux-fsdevel@...r.kernel.org, linux-perf-users@...r.kernel.org,
        netdev@...r.kernel.org, bpf@...r.kernel.org,
        Oleg Nesterov <oleg@...hat.com>,
        Jason Gunthorpe <jgg@...dia.com>, Jan Kara <jack@...e.cz>,
        "Kirill A . Shutemov" <kirill@...temov.name>,
        Pavel Begunkov <asml.silence@...il.com>,
        David Hildenbrand <david@...hat.com>
Subject: Re: [PATCH v4] mm/gup: disallow GUP writing to file-backed mappings
 by default

On 4/26/23 00:20, Lorenzo Stoakes wrote:
...
>> Could you elaborate? GUP calls handle_mm_fault() that invokes the write
>> notify the pte is made first writable. Of course, virt(pinned_page)[0] = 'x'
>> is not supposed to fault while using the kernel mapping.
>>
> 
> The issue is how dirtying works. Typically for a dirty-tracking mapping the
> kernel makes the mapping read-only, then when a write fault occurs,
> writenotify is called and the folio is marked dirty. This way the file
> system knows which files to writeback, then after writeback it 'cleans'
> them, restoring the read-only mapping and relying on the NEXT write marking
> write notifying and marking the folio dirty again.
> 
> If we GUP, _especially_ if it's long term, we run the risk of a write to
> the folio _after_ it has been cleaned and if the caller tries to do the
> 'right thing' and mark the folio dirty, it being marked dirty at an
> unexpected time which might race with other things and thus oops.
> 
> The issue is that this dirtying mechanism implicitly relies on the memory
> _only_ being accessed via user mappings, but we're providing another way to
> access that memory bypassing all that.
> 
> It's been a fundamental flaw in GUP for some time. This change tries to
> make things a little better by precluding the riskiest version of this
> which is the caller indicating that the pin is longterm (via
> FOLL_LONGTERM).
> 
> For an example of a user trying to sensibly avoid this, see io_pin_pages()
> in io_uring/rsrc.c. This is where io_uring tries to explicitly avoid this
> themselves, something that GUP should clearly be doing.
> 
> After this change, that code can be removed and we will live in a world
> where linux has a saner GUP :)

Hi Lorenzo,

This is the "missing writeup" that really helps people visualize the
problem, very nice. I think if you include the above, plus maybe a link
to Jan Kara's original report [1], in the commit description, that will
help a lot.


> 
> Of course we need to make more fundamental changes moving forward, the idea
> is this improves the situation and eliminates the need for the open-coded
> solution for io_uring which unblocks my other patch series which is also
> trying to make GUP more sane.

It is true that solving the problem has taken "a few"...years, and is still
not done, yes. :)

I'll respond to the patch with a review, as well.


[1] https://lore.kernel.org/linux-mm/20180103100430.GE4911@quack2.suse.cz/
 thanks,
-- 
John Hubbard
NVIDIA