[alsa-devel] hdsp: potential oops in snd_hdsp_info_pref_sync_ref()

[Takashi Iwai]

At Fri, 6 Jan 2012 16:59:57 +0300,
Dan Carpenter wrote:
> 
> My static check complains that there is a potential oops here but I'm
> not sure what the correct fix is.
> 
> sound/pci/rme9652/hdsp.c
>   2631          switch (hdsp->io_type) {
>   2632          case Digiface:
>   2633          case H9652:
>   2634                  uinfo->value.enumerated.items = 6;
>   2635                  break;
>   2636          case Multiface:
>   2637                  uinfo->value.enumerated.items = 4;
>   2638                  break;
>   2639          case H9632:
>   2640                  uinfo->value.enumerated.items = 3;
>   2641                  break;
>   2642          default:
>   2643                  uinfo->value.enumerated.items = 0;
>                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> We set this to zero here.
> 
>   2644                  break;
>   2645          }
>   2646  
>   2647          if (uinfo->value.enumerated.item >= uinfo->value.enumerated.items)
>                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> enumerated.item is unsigned int so it's always greater or equal to zero.
> 
>   2648                  uinfo->value.enumerated.item = uinfo->value.enumerated.items - 1;
> 
> We set it to UINT_MAX here.
> 
>   2649          strcpy(uinfo->value.enumerated.name, texts[uinfo->value.enumerated.item]);
>                                                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> Reading UINT_MAX beyond the end of the array here will probably oops.
> 
>   2650          return 0;
> 
> There is a similar issue in snd_asihpi_cmode_info().

Right, this is an obvious bug.  The driver should return an error
immediately there.

I fixed them in the sound git tree now.

Thanks!


Takashi