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;
Message-ID: <0b4a5ff8-12e5-3cc7-8971-49e576444c9a@gmail.com>
Date:   Mon, 8 Nov 2021 15:29:46 +0000
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.2.0
Subject: Re: [syzbot] KASAN: stack-out-of-bounds Read in iov_iter_revert
Content-Language: en-US
To:     Lee Jones <lee.jones@linaro.org>,
        syzbot <syzbot+9671693590ef5aad8953@syzkaller.appspotmail.com>
Cc:     axboe@kernel.dk, io-uring@vger.kernel.org,
        linux-kernel@vger.kernel.org, syzkaller-bugs@googlegroups.com
References: <6f7d4c1d-f923-3ab1-c525-45316b973c72@gmail.com>
 <00000000000047f3b805c962affb@google.com> <YYLAYvFU+9cnu+4H@google.com>
From:   Pavel Begunkov <asml.silence@gmail.com>
In-Reply-To: <YYLAYvFU+9cnu+4H@google.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 11/3/21 17:01, Lee Jones wrote:
> Good afternoon Pavel,
> 
>> syzbot has tested the proposed patch and the reproducer did not trigger any issue:
>>
>> Reported-and-tested-by: syzbot+9671693590ef5aad8953@syzkaller.appspotmail.com
>>
>> Tested on:
>>
>> commit:         bff2c168 io_uring: don't retry with truncated iter
>> git tree:       https://github.com/isilence/linux.git truncate
>> kernel config:  https://syzkaller.appspot.com/x/.config?x=730106bfb5bf8ace
>> dashboard link: https://syzkaller.appspot.com/bug?extid=9671693590ef5aad8953
>> compiler:       Debian clang version 11.0.1-2, GNU ld (GNU Binutils for Debian) 2.35.1
>>
>> Note: testing is done by a robot and is best-effort only.
> 
> As you can see in the 'dashboard link' above this bug also affects
> android-5-10 which is currently based on v5.10.75.
> 
> I see that the back-port of this patch failed in v5.10.y:
> 
>    https://lore.kernel.org/stable/163152589512611@kroah.com/
> 
> And after solving the build-error by back-porting both:
> 
>    2112ff5ce0c11 iov_iter: track truncated size
>    89c2b3b749182 io_uring: reexpand under-reexpanded iters
> 
> I now see execution tripping the WARN() in iov_iter_revert():
> 
>    if (WARN_ON(unroll > MAX_RW_COUNT))
>        return
> 
> Am I missing any additional patches required to fix stable/v5.10.y?

Is it the same syz test? There was a couple more patches for
IORING_SETUP_IOPOLL, but strange if that's not the case.


fwiw, Jens decided to replace it with another mechanism shortly
after, so it may be a better idea to backport those. Jens,
what do you think?


commit 8fb0f47a9d7acf620d0fd97831b69da9bc5e22ed
Author: Jens Axboe <axboe@kernel.dk>
Date:   Fri Sep 10 11:18:36 2021 -0600

     iov_iter: add helper to save iov_iter state

commit cd65869512ab5668a5d16f789bc4da1319c435c4
Author: Jens Axboe <axboe@kernel.dk>
Date:   Fri Sep 10 11:19:14 2021 -0600

     io_uring: use iov_iter state save/restore helpers


-- 
Pavel Begunkov