Re: [BUG] About "io_uring: add more uring info to fdinfo for debug"

[Jens Axboe]

On 10/28/21 3:24 PM, Eric Dumazet wrote:
> Hi
> 
> I was looking at commit 83f84356bc8f2d
> ("io_uring: add more uring info to fdinfo for debug") after receiving
> syzbot reports.
> 
> I suspect that the following :
> 
> +       for (i = cached_sq_head; i < sq_tail; i++) {
> +               unsigned int sq_idx = READ_ONCE(ctx->sq_array[i & sq_mask]);
> +
> +               if (likely(sq_idx <= sq_mask)) {
> +                       struct io_uring_sqe *sqe = &ctx->sq_sqes[sq_idx];
> +
> +                       seq_printf(m, "%5u: opcode:%d, fd:%d, flags:%x, user_data:%llu\n",
> +                                  sq_idx, sqe->opcode, sqe->fd, sqe->flags, sqe->user_data);
> +               }
> +       }
> 
> 
> Can loop around ~2^32 times if sq_tail is close to ~0U
> 
> I see various READ_ONCE(), which are probably not good enough.
> 
> At very minimum I would handling wrapping...

Thanks for reporting this. I think on top of wrapping, the loop should
just be capped at sq_entries as well. There's no point dumping more than
that, ever.

I'll take a stab at this.

-- 
Jens Axboe