Re: sound: use-after-free in hrtimer_cancel

From: Dmitry Vyukov
Date: Tue Jun 21 2016 - 13:48:36 EST


On Mon, Jun 6, 2016 at 6:39 PM, Takashi Iwai <tiwai@xxxxxxx> wrote:
> On Mon, 06 Jun 2016 18:29:25 +0200,
> Dmitry Vyukov wrote:
>>
>> On Mon, Jun 6, 2016 at 4:11 PM, Takashi Iwai <tiwai@xxxxxxx> wrote:
>> > On Sat, 04 Jun 2016 20:27:50 +0200,
>> > Dmitry Vyukov wrote:
>> >>
>> >> On Sat, Jun 4, 2016 at 8:00 PM, Dmitry Vyukov <dvyukov@xxxxxxxxxx> wrote:
>> >> > Hello,
>> >> >
>> >> > The following program triggers use-after-free:
>> >>
>> >> Forget to mention that you need to run it in a tight parallel loop. It
>> >> takes around 5 minutes to reproduce for me.
>> >
>> > Hmm, this again is a bug that is difficult to trigger... At least, I
>> > couldn't reproduce locally. How many processes are you running with
>> > stress program?
>>
>> I use a VM with 4 cores and use 20 parallel test processes.
>>
>> > It seems that there is nothing more than opening /dev/audio and does
>> > some mmap in the job. Is there any other relevant thing there?
>>
>>
>> I think poll with timeout is related. It is poll who sets hrtimer, right?
>
> If it's about snd-dummy driver, hrtimer is created at open, and
> started/stopped at PCM trigger, and removed at close.
>
> Is there any good way to decode which syscalls are executed in the
> test code?

What do you mean?
Here are the syscalls in the program:

r[2] = syscall(SYS_open, "/dev/audio", 0xa40ul, 0, 0, 0);
// r[2] is in the descriptor passed to SYS_poll
r[15] = syscall(SYS_poll, 0x2001dde8ul, 0x4ul, 0x8ul, 0, 0, 0);
r[18] = syscall(SYS_readv, r[2], 0x20dc13c0ul, 0x1ul, 0, 0, 0);
r[19] = syscall(SYS_read, r[2], 0x20dbefe0ul, 0x20ul, 0, 0, 0);