2022-10-28 13:45:47

by Anastasia Belova

[permalink] [raw]
Subject: [PATCH] vcodec: mediatek: add check for NULL for vsi->frm_bufs[vsi->new_fb_idx].buf.fb in vp9_swap_frm_bufs

If vsi->frm_bufs[vsi->new_fb_idx].buf.fb == NULL while cleaning
fb_free_list NULL-pointer is dereferenced.

Found by Linux Verification Center (linuxtesting.org) with SVACE.

Fixes: f77e89854b3e ("[media] vcodec: mediatek: Add Mediatek VP9 Video Decoder Driver")

Signed-off-by: Anastasia Belova <[email protected]>
---
drivers/media/platform/mediatek/vcodec/vdec/vdec_vp9_if.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/media/platform/mediatek/vcodec/vdec/vdec_vp9_if.c b/drivers/media/platform/mediatek/vcodec/vdec/vdec_vp9_if.c
index 70b8383f7c8e..b0679aaf6192 100644
--- a/drivers/media/platform/mediatek/vcodec/vdec/vdec_vp9_if.c
+++ b/drivers/media/platform/mediatek/vcodec/vdec/vdec_vp9_if.c
@@ -512,7 +512,7 @@ static void vp9_swap_frm_bufs(struct vdec_vp9_inst *inst)
* clean fb_free_list
*/
if (vsi->frm_bufs[vsi->new_fb_idx].ref_cnt == 0) {
- if (!vp9_is_sf_ref_fb(
+ if (vsi->frm_bufs[vsi->new_fb_idx].buf.fb != NULL && !vp9_is_sf_ref_fb(
inst, vsi->frm_bufs[vsi->new_fb_idx].buf.fb)) {
struct vdec_fb *fb;

--
2.30.2



2022-10-28 13:51:30

by Nicolas Dufresne

[permalink] [raw]
Subject: Re: [PATCH] vcodec: mediatek: add check for NULL for vsi->frm_bufs[vsi->new_fb_idx].buf.fb in vp9_swap_frm_bufs

Hi,

Le vendredi 28 octobre 2022 à 15:58 +0300, Anastasia Belova a écrit :
> If vsi->frm_bufs[vsi->new_fb_idx].buf.fb == NULL while cleaning
> fb_free_list NULL-pointer is dereferenced.
>
> Found by Linux Verification Center (linuxtesting.org) with SVACE.
>
> Fixes: f77e89854b3e ("[media] vcodec: mediatek: Add Mediatek VP9 Video Decoder Driver")
>
> Signed-off-by: Anastasia Belova <[email protected]>
> ---
> drivers/media/platform/mediatek/vcodec/vdec/vdec_vp9_if.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/media/platform/mediatek/vcodec/vdec/vdec_vp9_if.c b/drivers/media/platform/mediatek/vcodec/vdec/vdec_vp9_if.c
> index 70b8383f7c8e..b0679aaf6192 100644
> --- a/drivers/media/platform/mediatek/vcodec/vdec/vdec_vp9_if.c
> +++ b/drivers/media/platform/mediatek/vcodec/vdec/vdec_vp9_if.c
> @@ -512,7 +512,7 @@ static void vp9_swap_frm_bufs(struct vdec_vp9_inst *inst)
> * clean fb_free_list
> */
> if (vsi->frm_bufs[vsi->new_fb_idx].ref_cnt == 0) {
> - if (!vp9_is_sf_ref_fb(
> + if (vsi->frm_bufs[vsi->new_fb_idx].buf.fb != NULL && !vp9_is_sf_ref_fb(
> inst, vsi->frm_bufs[vsi->new_fb_idx].buf.fb)) {
> struct vdec_fb *fb;

Perhaps we could try and maintain some readability ? I'd suggest to move the
check into vp9_is_sf_ref_fb() as an early return.

Nicolas



Subject: Re: [PATCH] vcodec: mediatek: add check for NULL for vsi->frm_bufs[vsi->new_fb_idx].buf.fb in vp9_swap_frm_bufs

Il 28/10/22 15:02, Nicolas Dufresne ha scritto:
> Hi,
>
> Le vendredi 28 octobre 2022 à 15:58 +0300, Anastasia Belova a écrit :
>> If vsi->frm_bufs[vsi->new_fb_idx].buf.fb == NULL while cleaning
>> fb_free_list NULL-pointer is dereferenced.
>>
>> Found by Linux Verification Center (linuxtesting.org) with SVACE.
>>
>> Fixes: f77e89854b3e ("[media] vcodec: mediatek: Add Mediatek VP9 Video Decoder Driver")
>>
>> Signed-off-by: Anastasia Belova <[email protected]>
>> ---
>> drivers/media/platform/mediatek/vcodec/vdec/vdec_vp9_if.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/media/platform/mediatek/vcodec/vdec/vdec_vp9_if.c b/drivers/media/platform/mediatek/vcodec/vdec/vdec_vp9_if.c
>> index 70b8383f7c8e..b0679aaf6192 100644
>> --- a/drivers/media/platform/mediatek/vcodec/vdec/vdec_vp9_if.c
>> +++ b/drivers/media/platform/mediatek/vcodec/vdec/vdec_vp9_if.c
>> @@ -512,7 +512,7 @@ static void vp9_swap_frm_bufs(struct vdec_vp9_inst *inst)
>> * clean fb_free_list
>> */
>> if (vsi->frm_bufs[vsi->new_fb_idx].ref_cnt == 0) {
>> - if (!vp9_is_sf_ref_fb(
>> + if (vsi->frm_bufs[vsi->new_fb_idx].buf.fb != NULL && !vp9_is_sf_ref_fb(
>> inst, vsi->frm_bufs[vsi->new_fb_idx].buf.fb)) {
>> struct vdec_fb *fb;
>
> Perhaps we could try and maintain some readability ? I'd suggest to move the
> check into vp9_is_sf_ref_fb() as an early return.
>

I don't think that this will ever happen, as the check was omitted because of being
totally sure that vsi->frm_bufs[vsi->new_fb_idx].buf.fb is not NULL.

In any case, it doesn't really cost anything to add a check... in which case, I
agree with Nicolas about moving that to vp9_is_sf_ref_fb().

if (!fb)
return false
for .....

Regards,
Angelo