[alsa-devel] INFO: rcu detected stall in io_playback_transfer

Dmitry Vyukov dvyukov at google.com
Sun Apr 8 15:09:55 CEST 2018


On Sun, Apr 8, 2018 at 2:58 PM, Tetsuo Handa
<penguin-kernel at i-love.sakura.ne.jp> wrote:
> I manually simplified the reproducer.
>
> It is quite strange that removing unshare() hides this lockup bug.
> Also, explicitly closing by "close()" hides this lockup bug.
> Triggering "fput() from do_exit()" from "different namespace"
> somehow affects this lockup bug?
>
> ----------------------------------------
> #define _GNU_SOURCE
> #include <fcntl.h>
> #include <sched.h>
> #include <unistd.h>
> #include <sys/ioctl.h>
> #include <linux/soundcard.h>
>
> int main(int argc, char *argv[])
> {
>         const int fd = open("/dev/dsp1", O_RDWR);
>         int frag = (0 << 16) | 0;
>         char buf[48] = { };
>         unshare(CLONE_NEWNS);
>         ioctl(fd, SNDCTL_DSP_SETFRAGMENT, &frag);
>         write(fd, buf, sizeof(buf));
>         return 0;
> }
> ----------------------------------------


Takashi has already fixed this. See this thread:
https://groups.google.com/forum/#!searchin/syzkaller-bugs/%22INFO$3A$20rcu$20detected$20stall$20in$20snd_pcm_oss_write3$20(2)%22%7Csort:date/syzkaller-bugs/D2xWV7WTRDk/sCw_gImmCAAJ

The fix commit includes tag this bug:
Reported-by: syzbot+4f2016cf5185da7759dc at syzkaller.appspotmail.com
But it's just that it's mentioned in the thread for another bug.


More information about the Alsa-devel mailing list