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 (frame->nb_samples < avctx->frame_size &&
178 !avctx->internal->last_audio_frame) {
179 ret = pad_last_frame(avctx, &padded_frame, frame);
180 if (ret < 0)
181 goto end;
182
183 frame = padded_frame;
184 avctx->internal->last_audio_frame = 1;
185 }
186
187 if (frame->nb_samples != avctx->frame_size) {
188 av_log(avctx, AV_LOG_ERROR, "nb_samples (%d) != frame_size (%d) (avcodec_encode_audio2)\n", frame->nb_samples, avctx->frame_size);
189 ret = AVERROR(EINVAL);
190 goto end;
191 }
192 }
193 }
194
195 av_assert0(avctx->codec->encode2);
196
197 ret = avctx->codec->encode2(avctx, avpkt, frame, got_packet_ptr);
198 if (!ret) {
199 if (*got_packet_ptr) {
200 if (!(avctx->codec->capabilities & AV_CODEC_CAP_DELAY)) {
201 if (avpkt->pts == AV_NOPTS_VALUE)
202 avpkt->pts = frame->pts;
203 if (!avpkt->duration)
204 avpkt->duration = ff_samples_to_time_base(avctx,
205 frame->nb_samples);
206 }
207 avpkt->dts = avpkt->pts;
208 } else {
209 avpkt->size = 0;
210 }
211 }
212 if (avpkt->data && avpkt->data == avctx->internal->byte_buffer) {
213 needs_realloc = 0;
214 if (user_pkt.data) {
215 if (user_pkt.size >= avpkt->size) {
216 memcpy(user_pkt.data, avpkt->data, avpkt->size);
217 } else {
218 av_log(avctx, AV_LOG_ERROR, "Provided packet is too small, needs to be %d\n", avpkt->size);
219 avpkt->size = user_pkt.size;
220 ret = -1;
221 }
222 avpkt->buf = user_pkt.buf;
223 avpkt->data = user_pkt.data;
224 } else if (!avpkt->buf) {
225 ret = av_packet_make_refcounted(avpkt);
226 if (ret < 0)
227 goto end;
228 }
229 }
230
231 if (!ret) {
232 if (needs_realloc && avpkt->data) {
233 ret = av_buffer_realloc(&avpkt->buf, avpkt->size + AV_INPUT_BUFFER_PADDING_SIZE);
234 if (ret >= 0)
235 avpkt->data = avpkt->buf->data;
236 }
237 if (frame)
238 avctx->frame_number++;
239 }
240
241 if (ret < 0 || !*got_packet_ptr) {
242 av_packet_unref(avpkt);
243 goto end;
244 }
245
246 /* NOTE: if we add any audio encoders which output non-keyframe packets,
247 * this needs to be moved to the encoders, but for now we can do it
248 * here to simplify things */
249 avpkt->flags |= AV_PKT_FLAG_KEY;
250
251 end:
252 av_frame_free(&padded_frame);
253 av_free(extended_frame);
254
255 return ret;
256 }
257
avcodec_encode_video2(AVCodecContext * avctx,AVPacket * avpkt,const AVFrame * frame,int * got_packet_ptr)258 int attribute_align_arg avcodec_encode_video2(AVCodecContext *avctx,
259 AVPacket *avpkt,
260 const AVFrame *frame,
261 int *got_packet_ptr)
262 {
263 int ret;
264 AVPacket user_pkt = *avpkt;
265 int needs_realloc = !user_pkt.data;
266
267 *got_packet_ptr = 0;
268
269 if (!avctx->codec->encode2) {
270 av_log(avctx, AV_LOG_ERROR, "This encoder requires using the avcodec_send_frame() API.\n");
271 return AVERROR(ENOSYS);
272 }
273
274 if(CONFIG_FRAME_THREAD_ENCODER &&
275 avctx->internal->frame_thread_encoder && (avctx->active_thread_type&FF_THREAD_FRAME))
276 return ff_thread_video_encode_frame(avctx, avpkt, frame, got_packet_ptr);
277
278 if ((avctx->flags&AV_CODEC_FLAG_PASS1) && avctx->stats_out)
279 avctx->stats_out[0] = '\0';
280
281 if (!(avctx->codec->capabilities & AV_CODEC_CAP_DELAY) && !frame) {
282 av_packet_unref(avpkt);
283 return 0;
284 }
285
286 if (av_image_check_size2(avctx->width, avctx->height, avctx->max_pixels, AV_PIX_FMT_NONE, 0, avctx))
287 return AVERROR(EINVAL);
288
289 if (frame && frame->format == AV_PIX_FMT_NONE)
290 av_log(avctx, AV_LOG_WARNING, "AVFrame.format is not set\n");
291 if (frame && (frame->width == 0 || frame->height == 0))
292 av_log(avctx, AV_LOG_WARNING, "AVFrame.width or height is not set\n");
293
294 av_assert0(avctx->codec->encode2);
295
296 ret = avctx->codec->encode2(avctx, avpkt, frame, got_packet_ptr);
297 av_assert0(ret <= 0);
298
299 emms_c();
300
301 if (avpkt->data && avpkt->data == avctx->internal->byte_buffer) {
302 needs_realloc = 0;
303 if (user_pkt.data) {
304 if (user_pkt.size >= avpkt->size) {
305 memcpy(user_pkt.data, avpkt->data, avpkt->size);
306 } else {
307 av_log(avctx, AV_LOG_ERROR, "Provided packet is too small, needs to be %d\n", avpkt->size);
308 avpkt->size = user_pkt.size;
309 ret = -1;
310 }
311 avpkt->buf = user_pkt.buf;
312 avpkt->data = user_pkt.data;
313 } else if (!avpkt->buf) {
314 ret = av_packet_make_refcounted(avpkt);
315 if (ret < 0)
316 return ret;
317 }
318 }
319
320 if (!ret) {
321 if (!*got_packet_ptr)
322 avpkt->size = 0;
323 else if (!(avctx->codec->capabilities & AV_CODEC_CAP_DELAY))
324 avpkt->pts = avpkt->dts = frame->pts;
325
326 if (needs_realloc && avpkt->data) {
327 ret = av_buffer_realloc(&avpkt->buf, avpkt->size + AV_INPUT_BUFFER_PADDING_SIZE);
328 if (ret >= 0)
329 avpkt->data = avpkt->buf->data;
330 }
331
332 if (frame)
333 avctx->frame_number++;
334 }
335
336 if (ret < 0 || !*got_packet_ptr)
337 av_packet_unref(avpkt);
338
339 return ret;
340 }
341
avcodec_encode_subtitle(AVCodecContext * avctx,uint8_t * buf,int buf_size,const AVSubtitle * sub)342 int avcodec_encode_subtitle(AVCodecContext *avctx, uint8_t *buf, int buf_size,
343 const AVSubtitle *sub)
344 {
345 int ret;
346 if (sub->start_display_time) {
347 av_log(avctx, AV_LOG_ERROR, "start_display_time must be 0.\n");
348 return -1;
349 }
350
351 ret = avctx->codec->encode_sub(avctx, buf, buf_size, sub);
352 avctx->frame_number++;
353 return ret;
354 }
355
do_encode(AVCodecContext * avctx,const AVFrame * frame,int * got_packet)356 static int do_encode(AVCodecContext *avctx, const AVFrame *frame, int *got_packet)
357 {
358 int ret;
359 *got_packet = 0;
360
361 av_packet_unref(avctx->internal->buffer_pkt);
362 avctx->internal->buffer_pkt_valid = 0;
363
364 if (avctx->codec_type == AVMEDIA_TYPE_VIDEO) {
365 ret = avcodec_encode_video2(avctx, avctx->internal->buffer_pkt,
366 frame, got_packet);
367 } else if (avctx->codec_type == AVMEDIA_TYPE_AUDIO) {
368 ret = avcodec_encode_audio2(avctx, avctx->internal->buffer_pkt,
369 frame, got_packet);
370 } else {
371 ret = AVERROR(EINVAL);
372 }
373
374 if (ret >= 0 && *got_packet) {
375 // Encoders must always return ref-counted buffers.
376 // Side-data only packets have no data and can be not ref-counted.
377 av_assert0(!avctx->internal->buffer_pkt->data || avctx->internal->buffer_pkt->buf);
378 avctx->internal->buffer_pkt_valid = 1;
379 ret = 0;
380 } else {
381 av_packet_unref(avctx->internal->buffer_pkt);
382 }
383
384 return ret;
385 }
386
avcodec_send_frame(AVCodecContext * avctx,const AVFrame * frame)387 int attribute_align_arg avcodec_send_frame(AVCodecContext *avctx, const AVFrame *frame)
388 {
389 if (!avcodec_is_open(avctx) || !av_codec_is_encoder(avctx->codec))
390 return AVERROR(EINVAL);
391
392 if (avctx->internal->draining)
393 return AVERROR_EOF;
394
395 if (!frame) {
396 avctx->internal->draining = 1;
397
398 if (!(avctx->codec->capabilities & AV_CODEC_CAP_DELAY))
399 return 0;
400 }
401
402 if (avctx->codec->send_frame)
403 return avctx->codec->send_frame(avctx, frame);
404
405 // Emulation via old API. Do it here instead of avcodec_receive_packet, because:
406 // 1. if the AVFrame is not refcounted, the copying will be much more
407 // expensive than copying the packet data
408 // 2. assume few users use non-refcounted AVPackets, so usually no copy is
409 // needed
410
411 if (avctx->internal->buffer_pkt_valid)
412 return AVERROR(EAGAIN);
413
414 return do_encode(avctx, frame, &(int){0});
415 }
416
avcodec_receive_packet(AVCodecContext * avctx,AVPacket * avpkt)417 int attribute_align_arg avcodec_receive_packet(AVCodecContext *avctx, AVPacket *avpkt)
418 {
419 av_packet_unref(avpkt);
420
421 if (!avcodec_is_open(avctx) || !av_codec_is_encoder(avctx->codec))
422 return AVERROR(EINVAL);
423
424 if (avctx->codec->receive_packet) {
425 if (avctx->internal->draining && !(avctx->codec->capabilities & AV_CODEC_CAP_DELAY))
426 return AVERROR_EOF;
427 return avctx->codec->receive_packet(avctx, avpkt);
428 }
429
430 // Emulation via old API.
431
432 if (!avctx->internal->buffer_pkt_valid) {
433 int got_packet;
434 int ret;
435 if (!avctx->internal->draining)
436 return AVERROR(EAGAIN);
437 ret = do_encode(avctx, NULL, &got_packet);
438 if (ret < 0)
439 return ret;
440 if (ret >= 0 && !got_packet)
441 return AVERROR_EOF;
442 }
443
444 av_packet_move_ref(avpkt, avctx->internal->buffer_pkt);
445 avctx->internal->buffer_pkt_valid = 0;
446 return 0;
447 }
448