On Thu, Mar 31, 2022 at 03:37:06PM -0400, Nicolas Dufresne wrote: > For long term references, frame_num is set to long_term_frame_idx which > does not require wrapping. This is fixed by observation, no directly > related issue have been found yet. > > Signed-off-by: Nicolas Dufresne <nicolas.dufresne@xxxxxxxxxxxxx> > Reviewed-by: Sebastian Fricke <sebastian.fricke@xxxxxxxxxxxxx> Reviewed-by: Ezequiel Garcia <ezequiel@xxxxxxxxxxxxxxxxxxxx> > --- > drivers/media/v4l2-core/v4l2-h264.c | 4 +++- > 1 file changed, 3 insertions(+), 1 deletion(-) > > diff --git a/drivers/media/v4l2-core/v4l2-h264.c b/drivers/media/v4l2-core/v4l2-h264.c > index 5633a242520a..ac47519a9fbe 100644 > --- a/drivers/media/v4l2-core/v4l2-h264.c > +++ b/drivers/media/v4l2-core/v4l2-h264.c > @@ -57,8 +57,10 @@ v4l2_h264_init_reflist_builder(struct v4l2_h264_reflist_builder *b, > * '8.2.4.1 Decoding process for picture numbers' of the spec. > * TODO: This logic will have to be adjusted when we start > * supporting interlaced content. > + * For long term references, frame_num is set to > + * long_term_frame_idx which requires no wrapping. > */ > - if (dpb[i].frame_num > cur_frame_num) > + if (!b->refs[i].longterm && dpb[i].frame_num > cur_frame_num) > b->refs[i].frame_num = (int)dpb[i].frame_num - > max_frame_num; > else > -- > 2.34.1 >
