st/vdpau: Fix multithreading

The vdpau state tracker allows multiple threads access to the same gallium
context simultaneously. We can fix this either by locking the same mutex
each time the context is used or by using a different gallium context for
each mutex domain. Here we do the latter, although I'm not sure that's really
the best option.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Acked-by: Sinclair Yeh <syeh@vmware.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
This commit is contained in:
Thomas Hellstrom 2016-04-01 07:31:18 +02:00
parent bcc9fd378d
commit f1e5dfbe3c

View file

@ -68,7 +68,6 @@ vlVdpDecoderCreate(VdpDevice device,
if (!dev)
return VDP_STATUS_INVALID_HANDLE;
pipe = dev->context;
screen = dev->vscreen->pscreen;
pipe_mutex_lock(dev->mutex);
@ -123,6 +122,12 @@ vlVdpDecoderCreate(VdpDevice device,
templat.level = u_get_h264_level(templat.width, templat.height,
&templat.max_references);
pipe = screen->context_create(screen, dev->vscreen, 0);
if (!pipe) {
ret = VDP_STATUS_RESOURCES;
goto error_context;
}
vldecoder->decoder = pipe->create_video_codec(pipe, &templat);
if (!vldecoder->decoder) {
@ -145,6 +150,8 @@ error_handle:
vldecoder->decoder->destroy(vldecoder->decoder);
error_decoder:
pipe->destroy(pipe);
error_context:
pipe_mutex_unlock(dev->mutex);
DeviceReference(&vldecoder->device, NULL);
FREE(vldecoder);
@ -158,15 +165,18 @@ VdpStatus
vlVdpDecoderDestroy(VdpDecoder decoder)
{
vlVdpDecoder *vldecoder;
struct pipe_context *pipe;
vldecoder = (vlVdpDecoder *)vlGetDataHTAB(decoder);
if (!vldecoder)
return VDP_STATUS_INVALID_HANDLE;
pipe = vldecoder->decoder->context;
pipe_mutex_lock(vldecoder->mutex);
vldecoder->decoder->destroy(vldecoder->decoder);
pipe_mutex_unlock(vldecoder->mutex);
pipe_mutex_destroy(vldecoder->mutex);
pipe->destroy(pipe);
vlRemoveDataHTAB(decoder);
DeviceReference(&vldecoder->device, NULL);
@ -674,10 +684,20 @@ vlVdpDecoderRender(VdpDecoder decoder,
if (ret != VDP_STATUS_OK)
return ret;
/*
* Since we use separate contexts for the two mutex domains, we need
* to flush to make sure rendering operations happen in order.
* In particular, so that a frame is rendered before it is presented.
*/
pipe_mutex_lock(vldecoder->device->mutex);
vldecoder->device->context->flush(vldecoder->device->context, NULL, 0);
pipe_mutex_unlock(vldecoder->device->mutex);
pipe_mutex_lock(vldecoder->mutex);
dec->begin_frame(dec, vlsurf->video_buffer, &desc.base);
dec->decode_bitstream(dec, vlsurf->video_buffer, &desc.base, bitstream_buffer_count, buffers, sizes);
dec->end_frame(dec, vlsurf->video_buffer, &desc.base);
dec->flush(dec);
pipe_mutex_unlock(vldecoder->mutex);
return ret;
}