[FFmpeg-devel] [PATCH v2 02/12] avcodec/mediacodecdec: don't break out if both input and output port return try again

Zhao Zhili quinkblack at foxmail.com
Sun Nov 20 08:49:37 EET 2022


From: Zhao Zhili <zhilizhao at tencent.com>

At the beginning of decoding, if we feed mediacodec too fast, the
input port will return try again. It takes some time for mediacodec
to consume bitstream and output frame. So the output port also return
try again. It possible that mediacodec_receive_frame doesn't consume
any AVPacket and no AVFrame is output. Then both avcodec_send_packet()
and avcodec_receive_frame() return EAGAIN, which shouldn't happen.

This bug can be produced with decoding benchmark on Pixel 3.

Signed-off-by: Zhao Zhili <zhilizhao at tencent.com>
---
 libavcodec/mediacodecdec.c | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/libavcodec/mediacodecdec.c b/libavcodec/mediacodecdec.c
index 322b448d27..2c66f38541 100644
--- a/libavcodec/mediacodecdec.c
+++ b/libavcodec/mediacodecdec.c
@@ -444,7 +444,16 @@ static int mediacodec_receive_frame(AVCodecContext *avctx, AVFrame *frame)
             index = ff_AMediaCodec_dequeueInputBuffer(s->ctx->codec, 0);
             if (index < 0) {
                 /* no space, block for an output frame to appear */
-                return ff_mediacodec_dec_receive(avctx, s->ctx, frame, true);
+                ret = ff_mediacodec_dec_receive(avctx, s->ctx, frame, true);
+                /* Try again if both input port and output port return EAGAIN.
+                 * If no data is consumed and no frame in output, it can make
+                 * both avcodec_send_packet() and avcodec_receive_frame()
+                 * return EAGAIN, which violate the design.
+                 */
+                if (ff_AMediaCodec_infoTryAgainLater(s->ctx->codec, index) &&
+                    ret == AVERROR(EAGAIN))
+                    continue;
+                return ret;
             }
             s->ctx->current_input_buffer = index;
         }
-- 
2.25.1



More information about the ffmpeg-devel mailing list