blob: f0e457b70fbb2dae6b70e316cc4bad525bcb8a60
1 | /* |
2 | * RTP VP8 Depacketizer |
3 | * Copyright (c) 2010 Josh Allmann |
4 | * Copyright (c) 2012 Martin Storsjo |
5 | * |
6 | * This file is part of FFmpeg. |
7 | * |
8 | * FFmpeg is free software; you can redistribute it and/or |
9 | * modify it under the terms of the GNU Lesser General Public |
10 | * License as published by the Free Software Foundation; either |
11 | * version 2.1 of the License, or (at your option) any later version. |
12 | * |
13 | * FFmpeg is distributed in the hope that it will be useful, |
14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
16 | * Lesser General Public License for more details. |
17 | * |
18 | * You should have received a copy of the GNU Lesser General Public |
19 | * License along with FFmpeg; if not, write to the Free Software |
20 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
21 | */ |
22 | |
23 | /** |
24 | * @file |
25 | * @brief RTP support for the VP8 payload |
26 | * @author Josh Allmann <joshua.allmann@gmail.com> |
27 | * @see http://tools.ietf.org/html/draft-ietf-payload-vp8-05 |
28 | */ |
29 | |
30 | #include "libavcodec/bytestream.h" |
31 | |
32 | #include "avio_internal.h" |
33 | #include "rtpdec_formats.h" |
34 | |
35 | struct PayloadContext { |
36 | AVIOContext *data; |
37 | uint32_t timestamp; |
38 | int is_keyframe; |
39 | /* If sequence_ok is set, we keep returning data (even if we might have |
40 | * lost some data, but we haven't lost any too critical data that would |
41 | * cause the decoder to desynchronize and output random garbage). |
42 | */ |
43 | int sequence_ok; |
44 | int first_part_size; |
45 | uint16_t prev_seq; |
46 | int prev_pictureid; |
47 | int broken_frame; |
48 | /* If sequence_dirty is set, we have lost some data (critical or |
49 | * non-critical) and decoding will have some sort of artifacts, and |
50 | * we thus should request a new keyframe. |
51 | */ |
52 | int sequence_dirty; |
53 | int got_keyframe; |
54 | }; |
55 | |
56 | static int vp8_broken_sequence(AVFormatContext *ctx, PayloadContext *vp8, |
57 | const char *msg) |
58 | { |
59 | vp8->sequence_ok = 0; |
60 | av_log(ctx, AV_LOG_WARNING, "%s", msg); |
61 | ffio_free_dyn_buf(&vp8->data); |
62 | return AVERROR(EAGAIN); |
63 | } |
64 | |
65 | static int vp8_handle_packet(AVFormatContext *ctx, PayloadContext *vp8, |
66 | AVStream *st, AVPacket *pkt, uint32_t *timestamp, |
67 | const uint8_t *buf, int len, uint16_t seq, |
68 | int flags) |
69 | { |
70 | int start_partition, end_packet; |
71 | int extended_bits, part_id; |
72 | int pictureid_present = 0, tl0picidx_present = 0, tid_present = 0, |
73 | keyidx_present = 0; |
74 | int pictureid = -1, pictureid_mask = 0; |
75 | int returned_old_frame = 0; |
76 | uint32_t old_timestamp = 0; |
77 | |
78 | if (!buf) { |
79 | if (vp8->data) { |
80 | int ret = ff_rtp_finalize_packet(pkt, &vp8->data, st->index); |
81 | if (ret < 0) |
82 | return ret; |
83 | *timestamp = vp8->timestamp; |
84 | if (vp8->sequence_dirty) |
85 | pkt->flags |= AV_PKT_FLAG_CORRUPT; |
86 | return 0; |
87 | } |
88 | return AVERROR(EAGAIN); |
89 | } |
90 | |
91 | if (len < 1) |
92 | return AVERROR_INVALIDDATA; |
93 | |
94 | extended_bits = buf[0] & 0x80; |
95 | start_partition = buf[0] & 0x10; |
96 | part_id = buf[0] & 0x0f; |
97 | end_packet = flags & RTP_FLAG_MARKER; |
98 | buf++; |
99 | len--; |
100 | if (extended_bits) { |
101 | if (len < 1) |
102 | return AVERROR_INVALIDDATA; |
103 | pictureid_present = buf[0] & 0x80; |
104 | tl0picidx_present = buf[0] & 0x40; |
105 | tid_present = buf[0] & 0x20; |
106 | keyidx_present = buf[0] & 0x10; |
107 | buf++; |
108 | len--; |
109 | } |
110 | if (pictureid_present) { |
111 | if (len < 1) |
112 | return AVERROR_INVALIDDATA; |
113 | if (buf[0] & 0x80) { |
114 | if (len < 2) |
115 | return AVERROR_INVALIDDATA; |
116 | pictureid = AV_RB16(buf) & 0x7fff; |
117 | pictureid_mask = 0x7fff; |
118 | buf += 2; |
119 | len -= 2; |
120 | } else { |
121 | pictureid = buf[0] & 0x7f; |
122 | pictureid_mask = 0x7f; |
123 | buf++; |
124 | len--; |
125 | } |
126 | } |
127 | if (tl0picidx_present) { |
128 | // Ignoring temporal level zero index |
129 | buf++; |
130 | len--; |
131 | } |
132 | if (tid_present || keyidx_present) { |
133 | // Ignoring temporal layer index, layer sync bit and keyframe index |
134 | buf++; |
135 | len--; |
136 | } |
137 | if (len < 1) |
138 | return AVERROR_INVALIDDATA; |
139 | |
140 | if (start_partition && part_id == 0 && len >= 3) { |
141 | int res; |
142 | int non_key = buf[0] & 0x01; |
143 | if (!non_key) { |
144 | ffio_free_dyn_buf(&vp8->data); |
145 | // Keyframe, decoding ok again |
146 | vp8->sequence_ok = 1; |
147 | vp8->sequence_dirty = 0; |
148 | vp8->got_keyframe = 1; |
149 | } else { |
150 | int can_continue = vp8->data && !vp8->is_keyframe && |
151 | avio_tell(vp8->data) >= vp8->first_part_size; |
152 | if (!vp8->sequence_ok) |
153 | return AVERROR(EAGAIN); |
154 | if (!vp8->got_keyframe) |
155 | return vp8_broken_sequence(ctx, vp8, "Keyframe missing\n"); |
156 | if (pictureid >= 0) { |
157 | if (pictureid != ((vp8->prev_pictureid + 1) & pictureid_mask)) { |
158 | return vp8_broken_sequence(ctx, vp8, |
159 | "Missed a picture, sequence broken\n"); |
160 | } else { |
161 | if (vp8->data && !can_continue) |
162 | return vp8_broken_sequence(ctx, vp8, |
163 | "Missed a picture, sequence broken\n"); |
164 | } |
165 | } else { |
166 | uint16_t expected_seq = vp8->prev_seq + 1; |
167 | int16_t diff = seq - expected_seq; |
168 | if (vp8->data) { |
169 | // No picture id, so we can't know if missed packets |
170 | // contained any new frames. If diff == 0, we did get |
171 | // later packets from the same frame (matching timestamp), |
172 | // so we know we didn't miss any frame. If diff == 1 and |
173 | // we still have data (not flushed by the end of frame |
174 | // marker), the single missed packet must have been part |
175 | // of the same frame. |
176 | if ((diff == 0 || diff == 1) && can_continue) { |
177 | // Proceed with what we have |
178 | } else { |
179 | return vp8_broken_sequence(ctx, vp8, |
180 | "Missed too much, sequence broken\n"); |
181 | } |
182 | } else { |
183 | if (diff != 0) |
184 | return vp8_broken_sequence(ctx, vp8, |
185 | "Missed unknown data, sequence broken\n"); |
186 | } |
187 | } |
188 | if (vp8->data) { |
189 | vp8->sequence_dirty = 1; |
190 | if (avio_tell(vp8->data) >= vp8->first_part_size) { |
191 | int ret = ff_rtp_finalize_packet(pkt, &vp8->data, st->index); |
192 | if (ret < 0) |
193 | return ret; |
194 | pkt->flags |= AV_PKT_FLAG_CORRUPT; |
195 | returned_old_frame = 1; |
196 | old_timestamp = vp8->timestamp; |
197 | } else { |
198 | // Shouldn't happen |
199 | ffio_free_dyn_buf(&vp8->data); |
200 | } |
201 | } |
202 | } |
203 | vp8->first_part_size = (AV_RL16(&buf[1]) << 3 | buf[0] >> 5) + 3; |
204 | if ((res = avio_open_dyn_buf(&vp8->data)) < 0) |
205 | return res; |
206 | vp8->timestamp = *timestamp; |
207 | vp8->broken_frame = 0; |
208 | vp8->prev_pictureid = pictureid; |
209 | vp8->is_keyframe = !non_key; |
210 | } else { |
211 | uint16_t expected_seq = vp8->prev_seq + 1; |
212 | |
213 | if (!vp8->sequence_ok) |
214 | return AVERROR(EAGAIN); |
215 | |
216 | if (vp8->timestamp != *timestamp) { |
217 | // Missed the start of the new frame, sequence broken |
218 | return vp8_broken_sequence(ctx, vp8, |
219 | "Received no start marker; dropping frame\n"); |
220 | } |
221 | |
222 | if (seq != expected_seq) { |
223 | if (vp8->is_keyframe) { |
224 | return vp8_broken_sequence(ctx, vp8, |
225 | "Missed part of a keyframe, sequence broken\n"); |
226 | } else if (vp8->data && avio_tell(vp8->data) >= vp8->first_part_size) { |
227 | vp8->broken_frame = 1; |
228 | vp8->sequence_dirty = 1; |
229 | } else { |
230 | return vp8_broken_sequence(ctx, vp8, |
231 | "Missed part of the first partition, sequence broken\n"); |
232 | } |
233 | } |
234 | } |
235 | |
236 | if (!vp8->data) |
237 | return vp8_broken_sequence(ctx, vp8, "Received no start marker\n"); |
238 | |
239 | vp8->prev_seq = seq; |
240 | if (!vp8->broken_frame) |
241 | avio_write(vp8->data, buf, len); |
242 | |
243 | if (returned_old_frame) { |
244 | *timestamp = old_timestamp; |
245 | return end_packet ? 1 : 0; |
246 | } |
247 | |
248 | if (end_packet) { |
249 | int ret; |
250 | ret = ff_rtp_finalize_packet(pkt, &vp8->data, st->index); |
251 | if (ret < 0) |
252 | return ret; |
253 | if (vp8->sequence_dirty) |
254 | pkt->flags |= AV_PKT_FLAG_CORRUPT; |
255 | if (vp8->is_keyframe) |
256 | pkt->flags |= AV_PKT_FLAG_KEY; |
257 | return 0; |
258 | } |
259 | |
260 | return AVERROR(EAGAIN); |
261 | } |
262 | |
263 | static av_cold int vp8_init(AVFormatContext *s, int st_index, PayloadContext *vp8) |
264 | { |
265 | vp8->sequence_ok = 1; |
266 | return 0; |
267 | } |
268 | |
269 | static void vp8_close_context(PayloadContext *vp8) |
270 | { |
271 | ffio_free_dyn_buf(&vp8->data); |
272 | } |
273 | |
274 | static int vp8_need_keyframe(PayloadContext *vp8) |
275 | { |
276 | return vp8->sequence_dirty || !vp8->sequence_ok; |
277 | } |
278 | |
279 | RTPDynamicProtocolHandler ff_vp8_dynamic_handler = { |
280 | .enc_name = "VP8", |
281 | .codec_type = AVMEDIA_TYPE_VIDEO, |
282 | .codec_id = AV_CODEC_ID_VP8, |
283 | .priv_data_size = sizeof(PayloadContext), |
284 | .init = vp8_init, |
285 | .close = vp8_close_context, |
286 | .parse_packet = vp8_handle_packet, |
287 | .need_keyframe = vp8_need_keyframe, |
288 | }; |
289 |