Re: [syzbot] general protection fault in __io_arm_poll_handler

[Jens Axboe <axboe@kernel.dk>]

On 5/20/22 2:41 AM, Dmitry Vyukov wrote:
> On Sat, 4 Sept 2021 at 02:49, Jens Axboe <axboe@kernel.dk> wrote:
>>
>> On 9/3/21 5:47 PM, syzbot wrote:
>>> Hello,
>>>
>>> syzbot has tested the proposed patch and the reproducer did not trigger any issue:
>>>
>>> Reported-and-tested-by: syzbot+ba74b85fa15fd7a96437@syzkaller.appspotmail.com
>>>
>>> Tested on:
>>>
>>> commit:         31efe48e io_uring: fix possible poll event lost in mul..
>>> git tree:       git://git.kernel.dk/linux-block for-5.15/io_uring
>>> kernel config:  https://syzkaller.appspot.com/x/.config?x=914bb805fa8e8da9
>>> dashboard link: https://syzkaller.appspot.com/bug?extid=ba74b85fa15fd7a96437
>>> 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.
>>
>> Dmitry, I wonder if there's a way to have syzbot know about what it's
>> testing and be able to run the pending patches for that tree? I think
>> we're up to 4 reports now that are all just fallout from the same bug,
>> and where a patch has been queued up for a few days. Since they all look
>> different, I can't fault syzbot for thinking they are different, even if
>> they have the same root cause.
>>
>> Any way we can make this situation better? I can't keep replying that we
>> should test the current branch, and it'd be a shame to have a ton of
>> dupes.
> 
> Hi Jens,
> 
> This somehow fell through the cracks, but better late than never.
> 
> We could set up a syzbot instance for the io-uring tree.
> It won't solve the problem directly, but if the branch contains both
> new development ("for-next") and fixes, it will have good chances of
> discovering issues before they reach mainline and spread to other
> trees.
> Do you think it's a good idea? Is there a branch that contains new
> development and fixes?

My for-next stuff is always in linux-next, so I think as long as that is
tested, that should be quite fine. It's _usually_ not a problem, it just
sometimes happens that a broken patch ends up triggering a bunch of
different things. And then we don't get them all attributed in a fix, or
perhaps the patch itself is fixed up (or removed) and pushed out, then
leaving the syzbot reports in limbo.

In short, I don't think we need to do anything special here for now.

-- 
Jens Axboe