2023-06-14 16:19:45

by Chenyuan Mi

Subject: [PATCH] drm/display/dp_mst: Fix missing check for return value of drm_atomic_get_mst_payload_state()

The drm_atomic_get_mst_payload_state() function may
return NULL, which may cause null pointer deference,
and most other callsites of drm_atomic_get_mst_payload_state()
do Null check. Add Null check for return value of
drm_atomic_get_mst_payload_state().

Found by our static analysis tool.

Signed-off-by: Chenyuan Mi <[email protected]>
---
drivers/gpu/drm/display/drm_dp_mst_topology.c | 2 ++
1 file changed, 2 insertions(+)

diff --git a/drivers/gpu/drm/display/drm_dp_mst_topology.c b/drivers/gpu/drm/display/drm_dp_mst_topology.c
index 38dab76ae69e..27f4bcf409ea 100644
--- a/drivers/gpu/drm/display/drm_dp_mst_topology.c
+++ b/drivers/gpu/drm/display/drm_dp_mst_topology.c
@@ -4434,6 +4434,8 @@ void drm_dp_mst_atomic_wait_for_dependencies(struct drm_atomic_state *state)

new_payload = drm_atomic_get_mst_payload_state(new_mst_state,
old_payload->port);
+ if (!new_payload)
+ continue;
new_payload->vc_start_slot = old_payload->vc_start_slot;
}
}
--
2.17.1

2023-06-20 19:55:30

by Lyude Paul

Subject: Re: [PATCH] drm/display/dp_mst: Fix missing check for return value of drm_atomic_get_mst_payload_state()

Hm, I suppose it's not -impossible- for this to return NULL, but if it does
return NULL that absolutely means something broke well before this point in
the code. More comments below:

On Wed, 2023-06-14 at 09:09 -0700, Chenyuan Mi wrote:
> The drm_atomic_get_mst_payload_state() function may
> return NULL, which may cause null pointer deference,
> and most other callsites of drm_atomic_get_mst_payload_state()
> do Null check. Add Null check for return value of
> drm_atomic_get_mst_payload_state().
>
> Found by our static analysis tool.
>
> Signed-off-by: Chenyuan Mi <[email protected]>
> ---
> drivers/gpu/drm/display/drm_dp_mst_topology.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/drivers/gpu/drm/display/drm_dp_mst_topology.c b/drivers/gpu/drm/display/drm_dp_mst_topology.c
> index 38dab76ae69e..27f4bcf409ea 100644
> --- a/drivers/gpu/drm/display/drm_dp_mst_topology.c
> +++ b/drivers/gpu/drm/display/drm_dp_mst_topology.c
> @@ -4434,6 +4434,8 @@ void drm_dp_mst_atomic_wait_for_dependencies(struct drm_atomic_state *state)
>
> new_payload = drm_atomic_get_mst_payload_state(new_mst_state,
> old_payload->port);
> + if (!new_payload)
> + continue;

Could you change this check to:

if (drm_WARN_ON(state->dev, !new_payload))

> new_payload->vc_start_slot = old_payload->vc_start_slot;
> }
> }

--
Cheers,
Lyude Paul (she/her)
Software Engineer at Red Hat