Re: WARNING in kernfs_get

From: Tetsuo Handa
Date: Wed Oct 31 2018 - 05:56:43 EST


On 2018/09/10 16:31, syzbot wrote:
> HEAD commit:ÂÂÂ 9a5682765a2e Merge branch 'x86-urgent-for-linus' of git://..
> git tree:ÂÂÂÂÂÂ upstream
> console output: https://syzkaller.appspot.com/x/log.txt?x=1624858e400000
> kernel config:Â https://syzkaller.appspot.com/x/.config?x=8f59875069d721b6
> dashboard link: https://syzkaller.appspot.com/bug?extid=3dcb532381f98c86aeb1
> compiler:ÂÂÂÂÂÂ gcc (GCC) 8.0.1 20180413 (experimental)
> syz repro:ÂÂÂÂÂ https://syzkaller.appspot.com/x/repro.syz?x=12657f0a400000
> C reproducer:ÂÂ https://syzkaller.appspot.com/x/repro.c?x=117728ae400000

Here is a simplified reproducer.

----------
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>

int main(int argc, char *argv[])
{
int i;
for (i = 0; i < 16; i++)
if (fork() == 0)
while (1) {
int fd = open("/dev/vhci", O_WRONLY);
write(fd, "\xff\x80", 2);
close(fd);
}
while (1)
pause();
return 0;
}
----------

According to debug printk() patch,
atomic_read(&kn->active) == KN_DEACTIVATED_BIAS when we hit
WARN_ON(!atomic_read(&kn->count)) at kernfs_get().

Therefore, I guess that the caller is failing to check whether
kn is already deactivated (and hence need to recreate a new
entry by that name) before reusing existing kn.