Re: [syzbot] UBSAN: array-index-out-of-bounds in io_submit_sqes

[Dmitry Vyukov <dvyukov@google.com>]

On Tue, 31 May 2022 at 10:45, Jens Axboe <axboe@kernel.dk> wrote:
>
> On 5/31/22 1:55 AM, syzbot wrote:
> > Hello,
> >
> > syzbot found the following issue on:
> >
> > HEAD commit:    3b46e4e44180 Add linux-next specific files for 20220531
> > git tree:       linux-next
> > console output: https://syzkaller.appspot.com/x/log.txt?x=16e151f5f00000
> > kernel config:  https://syzkaller.appspot.com/x/.config?x=ccb8d66fc9489ef
> > dashboard link: https://syzkaller.appspot.com/bug?extid=b6c9b65b6753d333d833
> > compiler:       gcc (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.2
> >
> > Unfortunately, I don't have any reproducer for this issue yet.
> >
> > IMPORTANT: if you fix the issue, please add the following tag to the commit:
> > Reported-by: syzbot+b6c9b65b6753d333d833@syzkaller.appspotmail.com
> >
> > ================================================================================
> > ================================================================================
> > UBSAN: array-index-out-of-bounds in fs/io_uring.c:8860:19
> > index 75 is out of range for type 'io_op_def [47]'
>
> 'def' is just set here, it's not actually used after 'opcode' has been
> verified.

An interesting thing about C is that now the compiler is within its
rights to actually remove the check that is supposed to validate the
index because indexing io_op_defs[opcode] implies that opcode is
already within bounds, otherwise the program already has undefined
behavior, so removing the check is that case is also OK ;)