1 /*
2 * generic encoding-related code
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 #include "libavutil/attributes.h"
22 #include "libavutil/avassert.h"
23 #include "libavutil/frame.h"
24 #include "libavutil/imgutils.h"
25 #include "libavutil/internal.h"
26 #include "libavutil/samplefmt.h"
27
28 #include "avcodec.h"
29 #include "frame_thread_encoder.h"
30 #include "internal.h"
31
ff_alloc_packet2(AVCodecContext * avctx,AVPacket * avpkt,int64_t size,int64_t min_size)32 int ff_alloc_packet2(AVCodecContext *avctx, AVPacket *avpkt, int64_t size, int64_t min_size)
33 {
34 if (avpkt->size < 0) {
35 av_log(avctx, AV_LOG_ERROR, "Invalid negative user packet size %d\n", avpkt->size);
36 return AVERROR(EINVAL);
37 }
38 if (size < 0 || size > INT_MAX - AV_INPUT_BUFFER_PADDING_SIZE) {
39 av_log(avctx, AV_LOG_ERROR, "Invalid minimum required packet size %"PRId64" (max allowed is %d)\n",
40 size, INT_MAX - AV_INPUT_BUFFER_PADDING_SIZE);
41 return AVERROR(EINVAL);
42 }
43
44 if (avctx && 2*min_size < size) { // FIXME The factor needs to be finetuned
45 av_assert0(!avpkt->data || avpkt->data != avctx->internal->byte_buffer);
46 if (!avpkt->data || avpkt->size < size) {
47 av_fast_padded_malloc(&avctx->internal->byte_buffer, &avctx->internal->byte_buffer_size, size);
48 avpkt->data = avctx->internal->byte_buffer;
49 avpkt->size = avctx->internal->byte_buffer_size;
50 }
51 }
52
53 if (avpkt->data) {
54 AVBufferRef *buf = avpkt->buf;
55
56 if (avpkt->size < size) {
57 av_log(avctx, AV_LOG_ERROR, "User packet is too small (%d < %"PRId64")\n", avpkt->size, size);
58 return AVERROR(EINVAL);
59 }
60
61 av_init_packet(avpkt);
62 avpkt->buf = buf;
63 avpkt->size = size;
64 return 0;
65 } else {
66 int ret = av_new_packet(avpkt, size);
67 if (ret < 0)
68 av_log(avctx, AV_LOG_ERROR, "Failed to allocate packet of size %"PRId64"\n", size);
69 return ret;
70 }
71 }
72
ff_alloc_packet(AVPacket * avpkt,int size)73 int ff_alloc_packet(AVPacket *avpkt, int size)
74 {
75 return ff_alloc_packet2(NULL, avpkt, size, 0);
76 }
77
78 /**
79 * Pad last frame with silence.
80 */
pad_last_frame(AVCodecContext * s,AVFrame ** dst,const AVFrame * src)81 static int pad_last_frame(AVCodecContext *s, AVFrame **dst, const AVFrame *src)
82 {
83 AVFrame *frame = NULL;
84 int ret;
85
86 if (!(frame = av_frame_alloc()))
87 return AVERROR(ENOMEM);
88
89 frame->format = src->format;
90 frame->channel_layout = src->channel_layout;
91 frame->channels = src->channels;
92 frame->nb_samples = s->frame_size;
93 ret = av_frame_get_buffer(frame, 32);
94 if (ret < 0)
95 goto fail;
96
97 ret = av_frame_copy_props(frame, src);
98 if (ret < 0)
99 goto fail;
100
101 if ((ret = av_samples_copy(frame->extended_data, src->extended_data, 0, 0,
102 src->nb_samples, s->channels, s->sample_fmt)) < 0)
103 goto fail;
104 if ((ret = av_samples_set_silence(frame->extended_data, src->nb_samples,
105 frame->nb_samples - src->nb_samples,
106 s->channels, s->sample_fmt)) < 0)
107 goto fail;
108
109 *dst = frame;
110
111 return 0;
112
113 fail:
114 av_frame_free(&frame);
115 return ret;
116 }
117
avcodec_encode_audio2(AVCodecContext * avctx,AVPacket * avpkt,const AVFrame * frame,int * got_packet_ptr)118 int attribute_align_arg avcodec_encode_audio2(AVCodecContext *avctx,
119 AVPacket *avpkt,
120 const AVFrame *frame,
121 int *got_packet_ptr)
122 {
123 AVFrame *extended_frame = NULL;
124 AVFrame *padded_frame = NULL;
125 int ret;
126 AVPacket user_pkt = *avpkt;
127 int needs_realloc = !user_pkt.data;
128
129 *got_packet_ptr = 0;
130
131 if (!avctx->codec->encode2) {
132 av_log(avctx, AV_LOG_ERROR, "This encoder requires using the avcodec_send_frame() API.\n");
133 return AVERROR(ENOSYS);
134 }
135
136 if (!(avctx->codec->capabilities & AV_CODEC_CAP_DELAY) && !frame) {
137 av_packet_unref(avpkt);
138 return 0;
139 }
140
141 /* ensure that extended_data is properly set */
142 if (frame && !frame->extended_data) {
143 if (av_sample_fmt_is_planar(avctx->sample_fmt) &&
144 avctx->channels > AV_NUM_DATA_POINTERS) {
145 av_log(avctx, AV_LOG_ERROR, "Encoding to a planar sample format, "
146 "with more than %d channels, but extended_data is not set.\n",
147 AV_NUM_DATA_POINTERS);
148 return AVERROR(EINVAL);
149 }
150 av_log(avctx, AV_LOG_WARNING, "extended_data is not set.\n");
151
152 extended_frame = av_frame_alloc();
153 if (!extended_frame)
154 return AVERROR(ENOMEM);
155
156 memcpy(extended_frame, frame, sizeof(AVFrame));
157 extended_frame->extended_data = extended_frame->data;
158 frame = extended_frame;
159 }
160
161 /* extract audio service type metadata */
162 if (frame) {
163 AVFrameSideData *sd = av_frame_get_side_data(frame, AV_FRAME_DATA_AUDIO_SERVICE_TYPE);
164 if (sd && sd->size >= sizeof(enum AVAudioServiceType))
165 avctx->audio_service_type = *(enum AVAudioServiceType*)sd->data;
166 }
167
168 /* check for valid frame size */
169 if (frame) {
170 if (avctx->codec->capabilities & AV_CODEC_CAP_SMALL_LAST_FRAME) {
171 if (frame->nb_samples > avctx->frame_size) {
172 av_log(avctx, AV_LOG_ERROR, "more samples than frame size (avcodec_encode_audio2)\n");
173 ret = AVERROR(EINVAL);
174 goto end;
175 }
176 } else if (!(avctx->codec->capabilities & AV_CODEC_CAP_VARIABLE_FRAME_SIZE)) {
177 /* if we already got an undersized frame, that must have been the last */
178 if (avctx->internal->last_audio_frame) {
179 av_log(avctx, AV_LOG_ERROR, "frame_size (%d) was not respected for a non-last frame (avcodec_encode_audio2)\n", avctx->frame_size);
180 ret = AVERROR(EINVAL);
181 goto end;
182 }
183
184 if (frame->nb_samples < avctx->frame_size) {
185 ret = pad_last_frame(avctx, &padded_frame, frame);
186 if (ret < 0)
187 goto end;
188
189 frame = padded_frame;
190 avctx->internal->last_audio_frame = 1;
191 }
192
193 if (frame->nb_samples != avctx->frame_size) {
194 av_log(avctx, AV_LOG_ERROR, "nb_samples (%d) != frame_size (%d) (avcodec_encode_audio2)\n", frame->nb_samples, avctx->frame_size);
195 ret = AVERROR(EINVAL);
196 goto end;
197 }
198 }
199 }
200
201 av_assert0(avctx->codec->encode2);
202
203 ret = avctx->codec->encode2(avctx, avpkt, frame, got_packet_ptr);
204 if (!ret) {
205 if (*got_packet_ptr) {
206 if (!(avctx->codec->capabilities & AV_CODEC_CAP_DELAY)) {
207 if (avpkt->pts == AV_NOPTS_VALUE)
208 avpkt->pts = frame->pts;
209 if (!avpkt->duration)
210 avpkt->duration = ff_samples_to_time_base(avctx,
211 frame->nb_samples);
212 }
213 avpkt->dts = avpkt->pts;
214 } else {
215 avpkt->size = 0;
216 }
217 }
218 if (avpkt->data && avpkt->data == avctx->internal->byte_buffer) {
219 needs_realloc = 0;
220 if (user_pkt.data) {
221 if (user_pkt.size >= avpkt->size) {
222 memcpy(user_pkt.data, avpkt->data, avpkt->size);
223 } else {
224 av_log(avctx, AV_LOG_ERROR, "Provided packet is too small, needs to be %d\n", avpkt->size);
225 avpkt->size = user_pkt.size;
226 ret = -1;
227 }
228 avpkt->buf = user_pkt.buf;
229 avpkt->data = user_pkt.data;
230 } else if (!avpkt->buf) {
231 ret = av_packet_make_refcounted(avpkt);
232 if (ret < 0)
233 goto end;
234 }
235 }
236
237 if (!ret) {
238 if (needs_realloc && avpkt->data) {
239 ret = av_buffer_realloc(&avpkt->buf, avpkt->size + AV_INPUT_BUFFER_PADDING_SIZE);
240 if (ret >= 0)
241 avpkt->data = avpkt->buf->data;
242 }
243 if (frame)
244 avctx->frame_number++;
245 }
246
247 if (ret < 0 || !*got_packet_ptr) {
248 av_packet_unref(avpkt);
249 goto end;
250 }
251
252 /* NOTE: if we add any audio encoders which output non-keyframe packets,
253 * this needs to be moved to the encoders, but for now we can do it
254 * here to simplify things */
255 avpkt->flags |= AV_PKT_FLAG_KEY;
256
257 end:
258 av_frame_free(&padded_frame);
259 av_free(extended_frame);
260
261 return ret;
262 }
263
avcodec_encode_video2(AVCodecContext * avctx,AVPacket * avpkt,const AVFrame * frame,int * got_packet_ptr)264 int attribute_align_arg avcodec_encode_video2(AVCodecContext *avctx,
265 AVPacket *avpkt,
266 const AVFrame *frame,
267 int *got_packet_ptr)
268 {
269 int ret;
270 AVPacket user_pkt = *avpkt;
271 int needs_realloc = !user_pkt.data;
272
273 *got_packet_ptr = 0;
274
275 if (!avctx->codec->encode2) {
276 av_log(avctx, AV_LOG_ERROR, "This encoder requires using the avcodec_send_frame() API.\n");
277 return AVERROR(ENOSYS);
278 }
279
280 if(CONFIG_FRAME_THREAD_ENCODER &&
281 avctx->internal->frame_thread_encoder && (avctx->active_thread_type&FF_THREAD_FRAME))
282 return ff_thread_video_encode_frame(avctx, avpkt, frame, got_packet_ptr);
283
284 if ((avctx->flags&AV_CODEC_FLAG_PASS1) && avctx->stats_out)
285 avctx->stats_out[0] = '\0';
286
287 if (!(avctx->codec->capabilities & AV_CODEC_CAP_DELAY) && !frame) {
288 av_packet_unref(avpkt);
289 return 0;
290 }
291
292 if (av_image_check_size2(avctx->width, avctx->height, avctx->max_pixels, AV_PIX_FMT_NONE, 0, avctx))
293 return AVERROR(EINVAL);
294
295 if (frame && frame->format == AV_PIX_FMT_NONE)
296 av_log(avctx, AV_LOG_WARNING, "AVFrame.format is not set\n");
297 if (frame && (frame->width == 0 || frame->height == 0))
298 av_log(avctx, AV_LOG_WARNING, "AVFrame.width or height is not set\n");
299
300 av_assert0(avctx->codec->encode2);
301
302 ret = avctx->codec->encode2(avctx, avpkt, frame, got_packet_ptr);
303 av_assert0(ret <= 0);
304
305 emms_c();
306
307 if (avpkt->data && avpkt->data == avctx->internal->byte_buffer) {
308 needs_realloc = 0;
309 if (user_pkt.data) {
310 if (user_pkt.size >= avpkt->size) {
311 memcpy(user_pkt.data, avpkt->data, avpkt->size);
312 } else {
313 av_log(avctx, AV_LOG_ERROR, "Provided packet is too small, needs to be %d\n", avpkt->size);
314 avpkt->size = user_pkt.size;
315 ret = -1;
316 }
317 avpkt->buf = user_pkt.buf;
318 avpkt->data = user_pkt.data;
319 } else if (!avpkt->buf) {
320 ret = av_packet_make_refcounted(avpkt);
321 if (ret < 0)
322 return ret;
323 }
324 }
325
326 if (!ret) {
327 if (!*got_packet_ptr)
328 avpkt->size = 0;
329 else if (!(avctx->codec->capabilities & AV_CODEC_CAP_DELAY))
330 avpkt->pts = avpkt->dts = frame->pts;
331
332 if (needs_realloc && avpkt->data) {
333 ret = av_buffer_realloc(&avpkt->buf, avpkt->size + AV_INPUT_BUFFER_PADDING_SIZE);
334 if (ret >= 0)
335 avpkt->data = avpkt->buf->data;
336 }
337
338 if (frame)
339 avctx->frame_number++;
340 }
341
342 if (ret < 0 || !*got_packet_ptr)
343 av_packet_unref(avpkt);
344
345 return ret;
346 }
347
avcodec_encode_subtitle(AVCodecContext * avctx,uint8_t * buf,int buf_size,const AVSubtitle * sub)348 int avcodec_encode_subtitle(AVCodecContext *avctx, uint8_t *buf, int buf_size,
349 const AVSubtitle *sub)
350 {
351 int ret;
352 if (sub->start_display_time) {
353 av_log(avctx, AV_LOG_ERROR, "start_display_time must be 0.\n");
354 return -1;
355 }
356
357 ret = avctx->codec->encode_sub(avctx, buf, buf_size, sub);
358 avctx->frame_number++;
359 return ret;
360 }
361
do_encode(AVCodecContext * avctx,const AVFrame * frame,int * got_packet)362 static int do_encode(AVCodecContext *avctx, const AVFrame *frame, int *got_packet)
363 {
364 int ret;
365 *got_packet = 0;
366
367 av_packet_unref(avctx->internal->buffer_pkt);
368 avctx->internal->buffer_pkt_valid = 0;
369
370 if (avctx->codec_type == AVMEDIA_TYPE_VIDEO) {
371 ret = avcodec_encode_video2(avctx, avctx->internal->buffer_pkt,
372 frame, got_packet);
373 } else if (avctx->codec_type == AVMEDIA_TYPE_AUDIO) {
374 ret = avcodec_encode_audio2(avctx, avctx->internal->buffer_pkt,
375 frame, got_packet);
376 } else {
377 ret = AVERROR(EINVAL);
378 }
379
380 if (ret >= 0 && *got_packet) {
381 // Encoders must always return ref-counted buffers.
382 // Side-data only packets have no data and can be not ref-counted.
383 av_assert0(!avctx->internal->buffer_pkt->data || avctx->internal->buffer_pkt->buf);
384 avctx->internal->buffer_pkt_valid = 1;
385 ret = 0;
386 } else {
387 av_packet_unref(avctx->internal->buffer_pkt);
388 }
389
390 return ret;
391 }
392
avcodec_send_frame(AVCodecContext * avctx,const AVFrame * frame)393 int attribute_align_arg avcodec_send_frame(AVCodecContext *avctx, const AVFrame *frame)
394 {
395 if (!avcodec_is_open(avctx) || !av_codec_is_encoder(avctx->codec))
396 return AVERROR(EINVAL);
397
398 if (avctx->internal->draining)
399 return AVERROR_EOF;
400
401 if (!frame) {
402 avctx->internal->draining = 1;
403
404 if (!(avctx->codec->capabilities & AV_CODEC_CAP_DELAY))
405 return 0;
406 }
407
408 if (avctx->codec->send_frame)
409 return avctx->codec->send_frame(avctx, frame);
410
411 // Emulation via old API. Do it here instead of avcodec_receive_packet, because:
412 // 1. if the AVFrame is not refcounted, the copying will be much more
413 // expensive than copying the packet data
414 // 2. assume few users use non-refcounted AVPackets, so usually no copy is
415 // needed
416
417 if (avctx->internal->buffer_pkt_valid)
418 return AVERROR(EAGAIN);
419
420 return do_encode(avctx, frame, &(int){0});
421 }
422
avcodec_receive_packet(AVCodecContext * avctx,AVPacket * avpkt)423 int attribute_align_arg avcodec_receive_packet(AVCodecContext *avctx, AVPacket *avpkt)
424 {
425 av_packet_unref(avpkt);
426
427 if (!avcodec_is_open(avctx) || !av_codec_is_encoder(avctx->codec))
428 return AVERROR(EINVAL);
429
430 if (avctx->codec->receive_packet) {
431 int ret;
432 if (avctx->internal->draining && !(avctx->codec->capabilities & AV_CODEC_CAP_DELAY))
433 return AVERROR_EOF;
434 ret = avctx->codec->receive_packet(avctx, avpkt);
435 if (!ret)
436 // Encoders must always return ref-counted buffers.
437 // Side-data only packets have no data and can be not ref-counted.
438 av_assert0(!avpkt->data || avpkt->buf);
439 return ret;
440 }
441
442 // Emulation via old API.
443
444 if (!avctx->internal->buffer_pkt_valid) {
445 int got_packet;
446 int ret;
447 if (!avctx->internal->draining)
448 return AVERROR(EAGAIN);
449 ret = do_encode(avctx, NULL, &got_packet);
450 if (ret < 0)
451 return ret;
452 if (ret >= 0 && !got_packet)
453 return AVERROR_EOF;
454 }
455
456 av_packet_move_ref(avpkt, avctx->internal->buffer_pkt);
457 avctx->internal->buffer_pkt_valid = 0;
458 return 0;
459 }
460