[PATCH] media: rkvdec: hevc: cap EXT SPS RPS control counts before descriptor assembly
Detlev Casanova
detlev.casanova at collabora.com
Fri May 22 06:06:32 PDT 2026
Hi Mike,
On 5/20/26 20:57, Michael Bommarito wrote:
> On Tue, May 19, 2026 at 9:04 AM Detlev Casanova
> <detlev.casanova at collabora.com> wrote:
>> Still, did you try just changing the cap to 64 (.cfg.dims = { 64 },) ?
>> You'd need a test that sets the control from userspace though.
>>
>> It should refuse setting the control if there are more than 64 elements,
>> therefore the hevc decoder will not run any function using the count
>> values from the SPS (See rkvdec-vdpu381-hevc.c:601)
> Sure, I can test that and send a v2 for ST. My understanding is that
> we have four spots we need to check across the flow though:
>
> 1. ST count > 64
> 2. LT count > 32
ST and LT can be handled through cfg.dims and when num_negative_pics /
num_positive_pics are non-zero, the code check that ST and LT have
actually been set. If not, they can be skipped.
> 3. num_negative_pics / num_positive_pics > 16
This could be checked in the validation step of the control in v4l2-core.
> 4. delta_idx_minus1 + 1 > i
This one would need to be checked in the construction code indeed.
>
> So would you also want the same .cfg approach for the LT cap?
So yes for LT dimensions.
Detlev.
More information about the linux-arm-kernel
mailing list