mirrors/gstreamer
mirrors/gstreamer
1
1
Fork 0
mirror of https://gitlab.freedesktop.org/gstreamer/gstreamer.git synced 2025-04-26 05:16:13 +00:00
Code Issues 1 Projects Releases Wiki Activity

v4l2videodec: support orphaning

Browse source 
Recent kernels allow REQBUFS(0) on a queue that still has buffers in
use (mmapped or exported via dmabuf), orphaning all buffers on the queue.
If this is supported, the v4l2videodec element does not have to send a
drain request downstream.
This commit is contained in:
Philipp Zabel 2019-01-24 16:12:13 +01:00 committed by Nicolas Dufresne
parent dfc20013cc
commit e2ff87732d
1 changed files with 8 additions and 7 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

15
sys/v4l2/gstv4l2videodec.c
View file

@ -233,19 +233,20 @@ gst_v4l2_video_dec_set_format (GstVideoDecoder * decoder,
gst_v4l2_video_dec_finish (decoder); gst_v4l2_video_dec_finish (decoder);
gst_v4l2_object_stop (self->v4l2output); gst_v4l2_object_stop (self->v4l2output);
/* The renegotiation flow don't blend with the base class flow. To /* The renegotiation flow don't blend with the base class flow. To properly
* properly stop the capture pool we need to reclaim our buffers, which * stop the capture pool, if the buffers can't be orphaned, we need to
* will happend through the allocation query. The allocation query is * reclaim our buffers, which will happend through the allocation query.
* triggered by gst_video_decoder_negotiate() which requires the output * The allocation query is triggered by gst_video_decoder_negotiate() which
* caps to be set, but we can't know this information as we rely on the * requires the output caps to be set, but we can't know this information
* decoder, which requires the capture queue to be stopped. * as we rely on the decoder, which requires the capture queue to be
* stopped.
* *
* To workaround this issue, we simply run an allocation query with the * To workaround this issue, we simply run an allocation query with the
* old negotiated caps in order to drain/reclaim our buffers. That breaks * old negotiated caps in order to drain/reclaim our buffers. That breaks
* the complexity and should not have much impact in performance since the * the complexity and should not have much impact in performance since the
* following allocation query will happen on a drained pipeline and won't * following allocation query will happen on a drained pipeline and won't
* block. */ * block. */
{ if (!gst_v4l2_buffer_pool_orphan (&self->v4l2capture->pool)) {
GstCaps *caps = gst_pad_get_current_caps (decoder->srcpad); GstCaps *caps = gst_pad_get_current_caps (decoder->srcpad);
if (caps) { if (caps) {
GstQuery *query = gst_query_new_allocation (caps, FALSE); GstQuery *query = gst_query_new_allocation (caps, FALSE);