By using the Vpp ([1]), Va drops the bit depth from the 10bpc internal
to it (P010 format, actually, a 10bits-into-a-16bits-thing version of 
NV12), to a normal I420 3xDmaBuf;  consider blitting onto a XR30 DmaBuf 
instead to avoid losing the extra bits.

[1] https://cs.chromium.org/chromium/src/media/gpu/vaapi/vaapi_video_decode_accelerator.cc?q=vaapi_video+decode&sq=package:chromium&dr=CSs&l=271