Re: drivers/acpi/video.c: null pointer dereference

[Len Brown]

On Thu, 2005-03-24 at 15:37, Adrian Bunk wrote:
> The Coverity checker found the following null pointer dereference in
> drivers/acpi/video.c:
> 
> <--  snip  -->
> 
> ...
> static int
> acpi_video_switch_output(
> ...
> {
> ...
>         struct acpi_video_device *dev=NULL;
> ...
>         list_for_each_safe(node, next, &video->video_device_list) {
>                 struct acpi_video_device * dev = container_of(node,
> struct acpi_video_device, entry);
> ...
>         }
> ...
>         switch (event) {
>         case ACPI_VIDEO_NOTIFY_CYCLE:
>         case ACPI_VIDEO_NOTIFY_NEXT_OUTPUT:
>                 acpi_video_device_set_state(dev, 0);
>                 acpi_video_device_set_state(dev_next, 0x80000001);
>                 break;
>         case ACPI_VIDEO_NOTIFY_PREV_OUTPUT:
>                 acpi_video_device_set_state(dev, 0);
>                 acpi_video_device_set_state(dev_prev, 0x80000001);
> ...
> 
> <--  snip  -->
> 
> 
> Two different variables of the same name within 40 lines of code are a
> good indication that something's wrong...
> 
> 
> The outer "dev" variable is never assigned any value different from
> NULL.
> 
> acpi_video_device_set_state dereferences this variable.
> 
> 
> cu
> Adrian

Looks like we should do this:



===== drivers/acpi/video.c 1.8 vs edited =====
--- 1.8/drivers/acpi/video.c    2005-01-06 02:06:20 -05:00
+++ edited/drivers/acpi/video.c 2005-03-24 15:44:33 -05:00
@@ -1585,7 +1585,7 @@
        ACPI_FUNCTION_TRACE("acpi_video_switch_output");
 
        list_for_each_safe(node, next, &video->video_device_list) {
-               struct acpi_video_device * dev = container_of(node, struct 
acpi_video_device, entry);
+               dev = container_of(node, struct acpi_video_device, entry);
                status = acpi_video_device_get_state(dev, &state);
                if (state & 0x2){
                        dev_next = container_of(node->next, struct 
acpi_video_device, entry);