1 /*
2  * Raw Video Decoder
3  * Copyright (c) 2001 Fabrice Bellard
4  *
5  * This file is part of FFmpeg.
6  *
7  * FFmpeg is free software; you can redistribute it and/or
8  * modify it under the terms of the GNU Lesser General Public
9  * License as published by the Free Software Foundation; either
10  * version 2.1 of the License, or (at your option) any later version.
11  *
12  * FFmpeg is distributed in the hope that it will be useful,
13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15  * Lesser General Public License for more details.
16  *
17  * You should have received a copy of the GNU Lesser General Public
18  * License along with FFmpeg; if not, write to the Free Software
19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20  */
21 
22 /**
23  * @file
24  * Raw Video Decoder
25  */
26 
27 #include "avcodec.h"
28 #include "bswapdsp.h"
29 #include "decode.h"
30 #include "get_bits.h"
31 #include "internal.h"
32 #include "raw.h"
33 #include "libavutil/avassert.h"
34 #include "libavutil/buffer.h"
35 #include "libavutil/common.h"
36 #include "libavutil/intreadwrite.h"
37 #include "libavutil/imgutils.h"
38 #include "libavutil/opt.h"
39 
40 typedef struct RawVideoContext {
41     AVClass *av_class;
42     AVBufferRef *palette;
43     int frame_size;  /* size of the frame in bytes */
44     int flip;
45     int is_1_2_4_8_bpp; // 1, 2, 4 and 8 bpp in avi/mov, 1 and 8 bpp in nut
46     int is_mono;
47     int is_pal8;
48     int is_nut_mono;
49     int is_nut_pal8;
50     int is_yuv2;
51     int is_lt_16bpp; // 16bpp pixfmt and bits_per_coded_sample < 16
52     int tff;
53 
54     BswapDSPContext bbdsp;
55     void *bitstream_buf;
56     unsigned int bitstream_buf_size;
57 } RawVideoContext;
58 
59 static const AVOption options[]={
60 {"top", "top field first", offsetof(RawVideoContext, tff), AV_OPT_TYPE_BOOL, {.i64 = -1}, -1, 1, AV_OPT_FLAG_DECODING_PARAM|AV_OPT_FLAG_VIDEO_PARAM},
61 {NULL}
62 };
63 
64 static const AVClass rawdec_class = {
65     .class_name = "rawdec",
66     .option     = options,
67     .version    = LIBAVUTIL_VERSION_INT,
68 };
69 
raw_init_decoder(AVCodecContext * avctx)70 static av_cold int raw_init_decoder(AVCodecContext *avctx)
71 {
72     RawVideoContext *context = avctx->priv_data;
73     const AVPixFmtDescriptor *desc;
74 
75     ff_bswapdsp_init(&context->bbdsp);
76 
77     if (   avctx->codec_tag == MKTAG('r','a','w',' ')
78         || avctx->codec_tag == MKTAG('N','O','1','6'))
79         avctx->pix_fmt = avpriv_find_pix_fmt(avpriv_pix_fmt_bps_mov,
80                                       avctx->bits_per_coded_sample);
81     else if (avctx->codec_tag == MKTAG('W', 'R', 'A', 'W'))
82         avctx->pix_fmt = avpriv_find_pix_fmt(avpriv_pix_fmt_bps_avi,
83                                       avctx->bits_per_coded_sample);
84     else if (avctx->codec_tag && (avctx->codec_tag & 0xFFFFFF) != MKTAG('B','I','T', 0))
85         avctx->pix_fmt = avpriv_find_pix_fmt(ff_raw_pix_fmt_tags, avctx->codec_tag);
86     else if (avctx->pix_fmt == AV_PIX_FMT_NONE && avctx->bits_per_coded_sample)
87         avctx->pix_fmt = avpriv_find_pix_fmt(avpriv_pix_fmt_bps_avi,
88                                       avctx->bits_per_coded_sample);
89 
90     desc = av_pix_fmt_desc_get(avctx->pix_fmt);
91     if (!desc) {
92         av_log(avctx, AV_LOG_ERROR, "Invalid pixel format.\n");
93         return AVERROR(EINVAL);
94     }
95 
96     if (desc->flags & (AV_PIX_FMT_FLAG_PAL | FF_PSEUDOPAL)) {
97         context->palette = av_buffer_alloc(AVPALETTE_SIZE);
98         if (!context->palette)
99             return AVERROR(ENOMEM);
100 #if FF_API_PSEUDOPAL
101         if (desc->flags & AV_PIX_FMT_FLAG_PSEUDOPAL)
102             avpriv_set_systematic_pal2((uint32_t*)context->palette->data, avctx->pix_fmt);
103 #endif
104         else {
105             memset(context->palette->data, 0, AVPALETTE_SIZE);
106             if (avctx->bits_per_coded_sample == 1)
107                 memset(context->palette->data, 0xff, 4);
108         }
109     }
110 
111     if ((avctx->extradata_size >= 9 &&
112          !memcmp(avctx->extradata + avctx->extradata_size - 9, "BottomUp", 9)) ||
113         avctx->codec_tag == MKTAG('c','y','u','v') ||
114         avctx->codec_tag == MKTAG(3, 0, 0, 0) ||
115         avctx->codec_tag == MKTAG('W','R','A','W'))
116         context->flip = 1;
117 
118     if (avctx->pix_fmt == AV_PIX_FMT_MONOWHITE ||
119         avctx->pix_fmt == AV_PIX_FMT_MONOBLACK)
120         context->is_mono = 1;
121     else if (avctx->pix_fmt == AV_PIX_FMT_PAL8)
122         context->is_pal8 = 1;
123 
124     if (avctx->codec_tag == MKTAG('B','1','W','0') ||
125         avctx->codec_tag == MKTAG('B','0','W','1'))
126         context->is_nut_mono = 1;
127     else if (avctx->codec_tag == MKTAG('P','A','L',8))
128         context->is_nut_pal8 = 1;
129 
130     if (avctx->codec_tag == AV_RL32("yuv2") &&
131         avctx->pix_fmt   == AV_PIX_FMT_YUYV422)
132         context->is_yuv2 = 1;
133 
134     return 0;
135 }
136 
flip(AVCodecContext * avctx,AVFrame * frame)137 static void flip(AVCodecContext *avctx, AVFrame *frame)
138 {
139     frame->data[0]     += frame->linesize[0] * (avctx->height - 1);
140     frame->linesize[0] *= -1;
141 }
142 
143 /*
144  * Scale sample to 16-bit resolution
145  */
146 #define SCALE16(x, bits) (((x) << (16 - (bits))) | ((x) >> (2 * (bits) - 16)))
147 
148 /**
149  * Scale buffer to 16 bits per coded sample resolution
150  */
151 #define MKSCALE16(name, r16, w16) \
152 static void name(AVCodecContext *avctx, uint8_t * dst, const uint8_t *buf, int buf_size, int packed) \
153 { \
154     int i; \
155     if (!packed) { \
156         for (i = 0; i + 1 < buf_size; i += 2) \
157             w16(dst + i, SCALE16(r16(buf + i), avctx->bits_per_coded_sample)); \
158     } else { \
159         GetBitContext gb; \
160         init_get_bits(&gb, buf, buf_size * 8); \
161         for (i = 0; i < avctx->width * avctx->height; i++) { \
162             int sample = get_bits(&gb, avctx->bits_per_coded_sample); \
163             w16(dst + i*2, SCALE16(sample, avctx->bits_per_coded_sample)); \
164         } \
165    } \
166 }
167 
MKSCALE16(scale16be,AV_RB16,AV_WB16)168 MKSCALE16(scale16be, AV_RB16, AV_WB16)
169 MKSCALE16(scale16le, AV_RL16, AV_WL16)
170 
171 static int raw_decode(AVCodecContext *avctx, void *data, int *got_frame,
172                       AVPacket *avpkt)
173 {
174     const AVPixFmtDescriptor *desc;
175     RawVideoContext *context       = avctx->priv_data;
176     const uint8_t *buf             = avpkt->data;
177     int buf_size                   = avpkt->size;
178     int linesize_align             = 4;
179     int stride;
180     int res, len;
181     int need_copy;
182 
183     AVFrame   *frame   = data;
184 
185     if (avctx->width <= 0) {
186         av_log(avctx, AV_LOG_ERROR, "width is not set\n");
187         return AVERROR_INVALIDDATA;
188     }
189     if (avctx->height <= 0) {
190         av_log(avctx, AV_LOG_ERROR, "height is not set\n");
191         return AVERROR_INVALIDDATA;
192     }
193 
194     if (context->is_nut_mono)
195         stride = avctx->width / 8 + (avctx->width & 7 ? 1 : 0);
196     else if (context->is_nut_pal8)
197         stride = avctx->width;
198     else
199         stride = avpkt->size / avctx->height;
200 
201     av_log(avctx, AV_LOG_DEBUG, "PACKET SIZE: %d, STRIDE: %d\n", avpkt->size, stride);
202 
203     if (stride == 0 || avpkt->size < stride * avctx->height) {
204         av_log(avctx, AV_LOG_ERROR, "Packet too small (%d)\n", avpkt->size);
205         return AVERROR_INVALIDDATA;
206     }
207 
208     desc = av_pix_fmt_desc_get(avctx->pix_fmt);
209 
210     if ((avctx->bits_per_coded_sample == 8 || avctx->bits_per_coded_sample == 4 ||
211          avctx->bits_per_coded_sample == 2 || avctx->bits_per_coded_sample == 1 ||
212          (avctx->bits_per_coded_sample == 0 && (context->is_nut_pal8 || context->is_mono)) ) &&
213         (context->is_mono || context->is_pal8) &&
214         (!avctx->codec_tag || avctx->codec_tag == MKTAG('r','a','w',' ') ||
215                 context->is_nut_mono || context->is_nut_pal8)) {
216         context->is_1_2_4_8_bpp = 1;
217         if (context->is_mono) {
218             int row_bytes = avctx->width / 8 + (avctx->width & 7 ? 1 : 0);
219             context->frame_size = av_image_get_buffer_size(avctx->pix_fmt,
220                                                            FFALIGN(row_bytes, 16) * 8,
221                                                            avctx->height, 1);
222         } else
223             context->frame_size = av_image_get_buffer_size(avctx->pix_fmt,
224                                                            FFALIGN(avctx->width, 16),
225                                                            avctx->height, 1);
226     } else {
227         context->is_lt_16bpp = av_get_bits_per_pixel(desc) == 16 && avctx->bits_per_coded_sample > 8 && avctx->bits_per_coded_sample < 16;
228         context->frame_size = av_image_get_buffer_size(avctx->pix_fmt, avctx->width,
229                                                        avctx->height, 1);
230     }
231     if (context->frame_size < 0)
232         return context->frame_size;
233 
234     need_copy = !avpkt->buf || context->is_1_2_4_8_bpp || context->is_yuv2 || context->is_lt_16bpp;
235 
236     frame->pict_type        = AV_PICTURE_TYPE_I;
237     frame->key_frame        = 1;
238 
239     res = ff_decode_frame_props(avctx, frame);
240     if (res < 0)
241         return res;
242 
243     frame->pkt_pos      = avctx->internal->last_pkt_props->pos;
244     frame->pkt_duration = avctx->internal->last_pkt_props->duration;
245 
246     if (context->tff >= 0) {
247         frame->interlaced_frame = 1;
248         frame->top_field_first  = context->tff;
249     }
250 
251     if ((res = av_image_check_size(avctx->width, avctx->height, 0, avctx)) < 0)
252         return res;
253 
254     if (need_copy)
255         frame->buf[0] = av_buffer_alloc(FFMAX(context->frame_size, buf_size));
256     else
257         frame->buf[0] = av_buffer_ref(avpkt->buf);
258     if (!frame->buf[0])
259         return AVERROR(ENOMEM);
260 
261     // 1, 2, 4 and 8 bpp in avi/mov, 1 and 8 bpp in nut
262     if (context->is_1_2_4_8_bpp) {
263         int i, j, row_pix = 0;
264         uint8_t *dst = frame->buf[0]->data;
265         buf_size = context->frame_size - (context->is_pal8 ? AVPALETTE_SIZE : 0);
266         if (avctx->bits_per_coded_sample == 8 || context->is_nut_pal8 || context->is_mono) {
267             int pix_per_byte = context->is_mono ? 8 : 1;
268             for (i = 0, j = 0; j < buf_size && i<avpkt->size; i++, j++) {
269                 dst[j] = buf[i];
270                 row_pix += pix_per_byte;
271                 if (row_pix >= avctx->width) {
272                     i += stride - (i % stride) - 1;
273                     j += 16 - (j % 16) - 1;
274                     row_pix = 0;
275                 }
276             }
277         } else if (avctx->bits_per_coded_sample == 4) {
278             for (i = 0, j = 0; 2 * j + 1 < buf_size && i<avpkt->size; i++, j++) {
279                 dst[2 * j + 0] = buf[i] >> 4;
280                 dst[2 * j + 1] = buf[i] & 15;
281                 row_pix += 2;
282                 if (row_pix >= avctx->width) {
283                     i += stride - (i % stride) - 1;
284                     j += 8 - (j % 8) - 1;
285                     row_pix = 0;
286                 }
287             }
288         } else if (avctx->bits_per_coded_sample == 2) {
289             for (i = 0, j = 0; 4 * j + 3 < buf_size && i<avpkt->size; i++, j++) {
290                 dst[4 * j + 0] = buf[i] >> 6;
291                 dst[4 * j + 1] = buf[i] >> 4 & 3;
292                 dst[4 * j + 2] = buf[i] >> 2 & 3;
293                 dst[4 * j + 3] = buf[i]      & 3;
294                 row_pix += 4;
295                 if (row_pix >= avctx->width) {
296                     i += stride - (i % stride) - 1;
297                     j += 4 - (j % 4) - 1;
298                     row_pix = 0;
299                 }
300             }
301         } else {
302             av_assert0(avctx->bits_per_coded_sample == 1);
303             for (i = 0, j = 0; 8 * j + 7 < buf_size && i<avpkt->size; i++, j++) {
304                 dst[8 * j + 0] = buf[i] >> 7;
305                 dst[8 * j + 1] = buf[i] >> 6 & 1;
306                 dst[8 * j + 2] = buf[i] >> 5 & 1;
307                 dst[8 * j + 3] = buf[i] >> 4 & 1;
308                 dst[8 * j + 4] = buf[i] >> 3 & 1;
309                 dst[8 * j + 5] = buf[i] >> 2 & 1;
310                 dst[8 * j + 6] = buf[i] >> 1 & 1;
311                 dst[8 * j + 7] = buf[i]      & 1;
312                 row_pix += 8;
313                 if (row_pix >= avctx->width) {
314                     i += stride - (i % stride) - 1;
315                     j += 2 - (j % 2) - 1;
316                     row_pix = 0;
317                 }
318             }
319         }
320         linesize_align = 16;
321         buf = dst;
322     } else if (context->is_lt_16bpp) {
323         uint8_t *dst = frame->buf[0]->data;
324         int packed = (avctx->codec_tag & 0xFFFFFF) == MKTAG('B','I','T', 0);
325         int swap   =  avctx->codec_tag >> 24;
326 
327         if (packed && swap) {
328             av_fast_padded_malloc(&context->bitstream_buf, &context->bitstream_buf_size, buf_size);
329             if (!context->bitstream_buf)
330                 return AVERROR(ENOMEM);
331             if (swap == 16)
332                 context->bbdsp.bswap16_buf(context->bitstream_buf, (const uint16_t*)buf, buf_size / 2);
333             else if (swap == 32)
334                 context->bbdsp.bswap_buf(context->bitstream_buf, (const uint32_t*)buf, buf_size / 4);
335             else
336                 return AVERROR_INVALIDDATA;
337             buf = context->bitstream_buf;
338         }
339 
340         if (desc->flags & AV_PIX_FMT_FLAG_BE)
341             scale16be(avctx, dst, buf, buf_size, packed);
342         else
343             scale16le(avctx, dst, buf, buf_size, packed);
344 
345         buf = dst;
346     } else if (need_copy) {
347         memcpy(frame->buf[0]->data, buf, buf_size);
348         buf = frame->buf[0]->data;
349     }
350 
351     if (avctx->codec_tag == MKTAG('A', 'V', '1', 'x') ||
352         avctx->codec_tag == MKTAG('A', 'V', 'u', 'p'))
353         buf += buf_size - context->frame_size;
354 
355     len = context->frame_size - (avctx->pix_fmt==AV_PIX_FMT_PAL8 ? AVPALETTE_SIZE : 0);
356     if (buf_size < len && ((avctx->codec_tag & 0xFFFFFF) != MKTAG('B','I','T', 0) || !need_copy)) {
357         av_log(avctx, AV_LOG_ERROR, "Invalid buffer size, packet size %d < expected frame_size %d\n", buf_size, len);
358         av_buffer_unref(&frame->buf[0]);
359         return AVERROR(EINVAL);
360     }
361 
362     if ((res = av_image_fill_arrays(frame->data, frame->linesize,
363                                     buf, avctx->pix_fmt,
364                                     avctx->width, avctx->height, 1)) < 0) {
365         av_buffer_unref(&frame->buf[0]);
366         return res;
367     }
368 
369     if (avctx->pix_fmt == AV_PIX_FMT_PAL8) {
370         buffer_size_t pal_size;
371         const uint8_t *pal = av_packet_get_side_data(avpkt, AV_PKT_DATA_PALETTE,
372                                                      &pal_size);
373         int ret;
374 
375         if (pal && pal_size != AVPALETTE_SIZE) {
376             av_log(avctx, AV_LOG_ERROR, "Palette size %d is wrong\n", pal_size);
377             pal = NULL;
378         }
379 
380         if (!context->palette)
381             context->palette = av_buffer_alloc(AVPALETTE_SIZE);
382         if (!context->palette) {
383             av_buffer_unref(&frame->buf[0]);
384             return AVERROR(ENOMEM);
385         }
386         ret = av_buffer_make_writable(&context->palette);
387         if (ret < 0) {
388             av_buffer_unref(&frame->buf[0]);
389             return ret;
390         }
391 
392         if (pal) {
393             memcpy(context->palette->data, pal, AVPALETTE_SIZE);
394             frame->palette_has_changed = 1;
395         } else if (context->is_nut_pal8) {
396             int vid_size = avctx->width * avctx->height;
397             int pal_size = avpkt->size - vid_size;
398 
399             if (avpkt->size > vid_size && pal_size <= AVPALETTE_SIZE) {
400                 pal = avpkt->data + vid_size;
401                 memcpy(context->palette->data, pal, pal_size);
402                 frame->palette_has_changed = 1;
403             }
404         }
405     }
406 
407     if ((avctx->pix_fmt==AV_PIX_FMT_RGB24    ||
408         avctx->pix_fmt==AV_PIX_FMT_BGR24     ||
409         avctx->pix_fmt==AV_PIX_FMT_GRAY8     ||
410         avctx->pix_fmt==AV_PIX_FMT_RGB555LE  ||
411         avctx->pix_fmt==AV_PIX_FMT_RGB555BE  ||
412         avctx->pix_fmt==AV_PIX_FMT_RGB565LE  ||
413         avctx->pix_fmt==AV_PIX_FMT_MONOWHITE ||
414         avctx->pix_fmt==AV_PIX_FMT_MONOBLACK ||
415         avctx->pix_fmt==AV_PIX_FMT_PAL8) &&
416         FFALIGN(frame->linesize[0], linesize_align) * avctx->height <= buf_size)
417         frame->linesize[0] = FFALIGN(frame->linesize[0], linesize_align);
418 
419     if (avctx->pix_fmt == AV_PIX_FMT_NV12 && avctx->codec_tag == MKTAG('N', 'V', '1', '2') &&
420         FFALIGN(frame->linesize[0], linesize_align) * avctx->height +
421         FFALIGN(frame->linesize[1], linesize_align) * ((avctx->height + 1) / 2) <= buf_size) {
422         int la0 = FFALIGN(frame->linesize[0], linesize_align);
423         frame->data[1] += (la0 - frame->linesize[0]) * avctx->height;
424         frame->linesize[0] = la0;
425         frame->linesize[1] = FFALIGN(frame->linesize[1], linesize_align);
426     }
427 
428     if ((avctx->pix_fmt == AV_PIX_FMT_PAL8 && buf_size < context->frame_size) ||
429         (desc->flags & FF_PSEUDOPAL)) {
430         frame->buf[1]  = av_buffer_ref(context->palette);
431         if (!frame->buf[1]) {
432             av_buffer_unref(&frame->buf[0]);
433             return AVERROR(ENOMEM);
434         }
435         frame->data[1] = frame->buf[1]->data;
436     }
437 
438     if (avctx->pix_fmt == AV_PIX_FMT_BGR24 &&
439         ((frame->linesize[0] + 3) & ~3) * avctx->height <= buf_size)
440         frame->linesize[0] = (frame->linesize[0] + 3) & ~3;
441 
442     if (context->flip)
443         flip(avctx, frame);
444 
445     if (avctx->codec_tag == MKTAG('Y', 'V', '1', '2') ||
446         avctx->codec_tag == MKTAG('Y', 'V', '1', '6') ||
447         avctx->codec_tag == MKTAG('Y', 'V', '2', '4') ||
448         avctx->codec_tag == MKTAG('Y', 'V', 'U', '9'))
449         FFSWAP(uint8_t *, frame->data[1], frame->data[2]);
450 
451     if (avctx->codec_tag == AV_RL32("I420") && (avctx->width+1)*(avctx->height+1) * 3/2 == buf_size) {
452         frame->data[1] = frame->data[1] +  (avctx->width+1)*(avctx->height+1) -avctx->width*avctx->height;
453         frame->data[2] = frame->data[2] + ((avctx->width+1)*(avctx->height+1) -avctx->width*avctx->height)*5/4;
454     }
455 
456     if (avctx->codec_tag == AV_RL32("yuv2") &&
457         avctx->pix_fmt   == AV_PIX_FMT_YUYV422) {
458         int x, y;
459         uint8_t *line = frame->data[0];
460         for (y = 0; y < avctx->height; y++) {
461             for (x = 0; x < avctx->width; x++)
462                 line[2 * x + 1] ^= 0x80;
463             line += frame->linesize[0];
464         }
465     }
466 
467     if (avctx->codec_tag == AV_RL32("b64a") &&
468         avctx->pix_fmt   == AV_PIX_FMT_RGBA64BE) {
469         uint8_t *dst = frame->data[0];
470         uint64_t v;
471         int x, y;
472         for (y = 0; y < avctx->height; y++) {
473             for (x = 0; x >> 3 < avctx->width; x += 8) {
474                 v = AV_RB64(&dst[x]);
475                 AV_WB64(&dst[x], v << 16 | v >> 48);
476             }
477             dst += frame->linesize[0];
478         }
479     }
480 
481     if (avctx->field_order > AV_FIELD_PROGRESSIVE) { /* we have interlaced material flagged in container */
482         frame->interlaced_frame = 1;
483         if (avctx->field_order == AV_FIELD_TT || avctx->field_order == AV_FIELD_TB)
484             frame->top_field_first = 1;
485     }
486 
487     *got_frame = 1;
488     return buf_size;
489 }
490 
raw_close_decoder(AVCodecContext * avctx)491 static av_cold int raw_close_decoder(AVCodecContext *avctx)
492 {
493     RawVideoContext *context = avctx->priv_data;
494 
495     av_buffer_unref(&context->palette);
496     av_freep(&context->bitstream_buf);
497     return 0;
498 }
499 
500 AVCodec ff_rawvideo_decoder = {
501     .name           = "rawvideo",
502     .long_name      = NULL_IF_CONFIG_SMALL("raw video"),
503     .type           = AVMEDIA_TYPE_VIDEO,
504     .id             = AV_CODEC_ID_RAWVIDEO,
505     .priv_data_size = sizeof(RawVideoContext),
506     .init           = raw_init_decoder,
507     .close          = raw_close_decoder,
508     .decode         = raw_decode,
509     .priv_class     = &rawdec_class,
510     .capabilities   = AV_CODEC_CAP_PARAM_CHANGE,
511 };
512