[Drbd-dev] [PATCH] block: fix potential dereference null pointer

[Philipp Reisner]

Hi Lv Ruyi,

This patch does not make sense to me. A request can only get "TO_BE_SENT"
when the connection is established with the corresponding
cstate. Establishing a connection can only work if net_conf is set. net_conf
can be exchanged to a new one, but never become NULL.

Please share more details why you think this NULL check is necessary here?



On Fri, Apr 1, 2022 at 10:36 AM <cgel.zte at gmail.com> wrote:
>
> From: Lv Ruyi <lv.ruyi at zte.com.cn>
>
> rcu_dereference may return NULL, so check the returned pointer.
>
> Reported-by: Zeal Robot <zealci at zte.com.cn>
> Signed-off-by: Lv Ruyi <lv.ruyi at zte.com.cn>
> ---
>  drivers/block/drbd/drbd_req.c | 8 ++++++++
>  1 file changed, 8 insertions(+)
>
> diff --git a/drivers/block/drbd/drbd_req.c b/drivers/block/drbd/drbd_req.c
> index e1e58e91ee58..8ab6da155e2f 100644
> --- a/drivers/block/drbd/drbd_req.c
> +++ b/drivers/block/drbd/drbd_req.c
> @@ -577,6 +577,10 @@ int __req_mod(struct drbd_request *req, enum drbd_req_event what,
>                 D_ASSERT(device, !(req->rq_state & RQ_NET_MASK));
>                 rcu_read_lock();
>                 nc = rcu_dereference(connection->net_conf);
> +               if (!nc) {
> +                       rcu_read_unlock();
> +                       break;
> +               }
>                 p = nc->wire_protocol;
>                 rcu_read_unlock();
>                 req->rq_state |=
> @@ -690,6 +694,10 @@ int __req_mod(struct drbd_request *req, enum drbd_req_event what,
>                 /* close the epoch, in case it outgrew the limit */
>                 rcu_read_lock();
>                 nc = rcu_dereference(connection->net_conf);
> +               if (!nc) {
> +                       rcu_read_unlock();
> +                       break;
> +               }
>                 p = nc->max_epoch_size;
>                 rcu_read_unlock();
>                 if (connection->current_tle_writes >= p)
> --
> 2.25.1
>