Re: [PATCH 5.13] io_uring: Check current->io_uring in io_uring_cancel_sqpoll

From: Pavel Begunkov <asml.silence@gmail.com>
To: Palash Oswal <hello@oswalpalash.com>
Cc: axboe@kernel.dk, dvyukov@google.com, io-uring@vger.kernel.org,
	linux-kernel@vger.kernel.org, oswalpalash@gmail.com,
	syzbot+be51ca5a4d97f017cd50@syzkaller.appspotmail.com,
	syzkaller-bugs@googlegroups.com, stable@vger.kernel.org
Subject: Re: [PATCH 5.13] io_uring: Check current->io_uring in io_uring_cancel_sqpoll
Date: Tue, 27 Apr 2021 14:08:23 +0100	[thread overview]
Message-ID: <b62f1985-f336-d0ec-bcdc-5dac16fb760b@gmail.com> (raw)
In-Reply-To: <20210427125148.21816-1-hello@oswalpalash.com>

On 4/27/21 1:51 PM, Palash Oswal wrote:
> syzkaller identified KASAN: null-ptr-deref Write in
> io_uring_cancel_sqpoll on v5.12
> 
> io_uring_cancel_sqpoll is called by io_sq_thread before calling
> io_uring_alloc_task_context. This leads to current->io_uring being
> NULL. io_uring_cancel_sqpoll should not have to deal with threads
> where current->io_uring is NULL.
> 
> In order to cast a wider safety net, perform input sanitisation
> directly in io_uring_cancel_sqpoll and return for NULL value of
> current->io_uring.

Looks good to me, but better to add a comment why it can be ignored,
e.g. "can skip it as it couldn't have submitted requests without tctx"

Also a nit: s/current->io_uring/tctx/

> 
> Reported-by: syzbot+be51ca5a4d97f017cd50@syzkaller.appspotmail.com
> Cc: stable@vger.kernel.org
> Signed-off-by: Palash Oswal <hello@oswalpalash.com>
> ---
>  fs/io_uring.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index dff34975d86b..eccad51b7954 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -8998,6 +8998,8 @@ static void io_uring_cancel_sqpoll(struct io_ring_ctx *ctx)
>  	s64 inflight;
>  	DEFINE_WAIT(wait);
>  
> +	if (!current->io_uring)
> +		return;
>  	WARN_ON_ONCE(!sqd || ctx->sq_data->thread != current);
>  
>  	atomic_inc(&tctx->in_idle);
> 

-- 
Pavel Begunkov