Shaka Packager SDK
mpd_utils.cc
1 // Copyright 2014 Google LLC. All rights reserved.
2 //
3 // Use of this source code is governed by a BSD-style
4 // license that can be found in the LICENSE file or at
5 // https://developers.google.com/open-source/licenses/bsd
6 
7 #include <packager/mpd/base/mpd_utils.h>
8 
9 #include <absl/flags/flag.h>
10 #include <absl/log/check.h>
11 #include <absl/log/log.h>
12 #include <absl/strings/escaping.h>
13 #include <absl/strings/numbers.h>
14 #include <absl/strings/str_format.h>
15 #include <libxml/tree.h>
16 
17 #include <packager/macros/logging.h>
18 #include <packager/media/base/fourccs.h>
19 #include <packager/media/base/language_utils.h>
20 #include <packager/media/base/protection_system_specific_info.h>
21 #include <packager/mpd/base/adaptation_set.h>
22 #include <packager/mpd/base/content_protection_element.h>
23 #include <packager/mpd/base/representation.h>
24 #include <packager/mpd/base/xml/scoped_xml_ptr.h>
25 
26 ABSL_FLAG(
27  bool,
28  use_legacy_vp9_codec_string,
29  false,
30  "Use legacy vp9 codec string 'vp9' if set to true; otherwise new style "
31  "vp09.xx.xx.xx... codec string will be used. Default to false as indicated "
32  "in https://github.com/shaka-project/shaka-packager/issues/406, all major "
33  "browsers and platforms already support the new 'vp09' codec string.");
34 
35 namespace shaka {
36 namespace {
37 
38 bool IsKeyRotationDefaultKeyId(const std::string& key_id) {
39  for (char c : key_id) {
40  if (c != '\0')
41  return false;
42  }
43  return true;
44 }
45 
46 std::string TextCodecString(const MediaInfo& media_info) {
47  CHECK(media_info.has_text_info());
48  const auto container_type = media_info.container_type();
49 
50  // Codecs are not needed when mimeType is "text/*". Having a codec would be
51  // redundant.
52  if (container_type == MediaInfo::CONTAINER_TEXT) {
53  return "";
54  }
55 
56  // DASH IOP mentions that the codec for ttml in mp4 is stpp, so override
57  // the default codec value.
58  const std::string& codec = media_info.text_info().codec();
59  if (codec == "ttml" && container_type == MediaInfo::CONTAINER_MP4) {
60  return "stpp";
61  }
62 
63  return codec;
64 }
65 
66 } // namespace
67 
68 bool HasVODOnlyFields(const MediaInfo& media_info) {
69  return media_info.has_init_range() || media_info.has_index_range() ||
70  media_info.has_media_file_url();
71 }
72 
73 bool HasLiveOnlyFields(const MediaInfo& media_info) {
74  return media_info.has_init_segment_url() ||
75  media_info.has_segment_template_url();
76 }
77 
78 void RemoveDuplicateAttributes(
79  ContentProtectionElement* content_protection_element) {
80  DCHECK(content_protection_element);
81  typedef std::map<std::string, std::string> AttributesMap;
82 
83  AttributesMap& attributes = content_protection_element->additional_attributes;
84  if (!content_protection_element->value.empty())
85  attributes.erase("value");
86 
87  if (!content_protection_element->scheme_id_uri.empty())
88  attributes.erase("schemeIdUri");
89 }
90 
91 std::string GetLanguage(const MediaInfo& media_info) {
92  std::string lang;
93  if (media_info.has_audio_info()) {
94  lang = media_info.audio_info().language();
95  } else if (media_info.has_text_info()) {
96  lang = media_info.text_info().language();
97  }
98  return LanguageToShortestForm(lang);
99 }
100 
101 std::string GetCodecs(const MediaInfo& media_info) {
102  CHECK(OnlyOneTrue(media_info.has_video_info(), media_info.has_audio_info(),
103  media_info.has_text_info()));
104 
105  if (media_info.has_video_info()) {
106  if (media_info.container_type() == MediaInfo::CONTAINER_WEBM) {
107  std::string codec = media_info.video_info().codec().substr(0, 4);
108  // media_info.video_info().codec() contains new revised codec string
109  // specified by "VPx in ISO BMFF" document, which is not compatible to
110  // old codec strings in WebM. Hack it here before all browsers support
111  // new codec strings.
112  if (codec == "vp08")
113  return "vp8";
114  if (absl::GetFlag(FLAGS_use_legacy_vp9_codec_string)) {
115  if (codec == "vp09")
116  return "vp9";
117  }
118  }
119  return media_info.video_info().codec();
120  }
121 
122  if (media_info.has_audio_info())
123  return media_info.audio_info().codec();
124 
125  if (media_info.has_text_info())
126  return TextCodecString(media_info);
127 
128  NOTIMPLEMENTED();
129  return "";
130 }
131 
132 std::string GetSupplementalCodecs(const MediaInfo& media_info) {
133  CHECK(OnlyOneTrue(media_info.has_video_info(), media_info.has_audio_info(),
134  media_info.has_text_info()));
135 
136  if (media_info.has_video_info() &&
137  media_info.video_info().has_supplemental_codec()) {
138  return media_info.video_info().supplemental_codec();
139  }
140  return "";
141 }
142 
143 std::string GetSupplementalProfiles(const MediaInfo& media_info) {
144  CHECK(OnlyOneTrue(media_info.has_video_info(), media_info.has_audio_info(),
145  media_info.has_text_info()));
146 
147  if (media_info.has_video_info() &&
148  media_info.video_info().has_compatible_brand()) {
149  return FourCCToString(
150  static_cast<media::FourCC>(media_info.video_info().compatible_brand()));
151  }
152  return "";
153 }
154 
155 std::string GetBaseCodec(const MediaInfo& media_info) {
156  std::string codec;
157  if (media_info.has_video_info()) {
158  codec = media_info.video_info().codec();
159  } else if (media_info.has_audio_info()) {
160  codec = media_info.audio_info().codec();
161  } else if (media_info.has_text_info()) {
162  codec = media_info.text_info().codec();
163  }
164  // Convert, for example, "mp4a.40.2" to simply "mp4a".
165  // "mp4a.40.2" and "mp4a.40.5" can exist in the same AdaptationSet.
166  size_t dot = codec.find('.');
167  if (dot != std::string::npos) {
168  codec.erase(dot);
169  }
170  return codec;
171 }
172 
173 std::string GetAdaptationSetKey(const MediaInfo& media_info,
174  bool ignore_codec) {
175  std::string key;
176 
177  if (media_info.has_video_info()) {
178  key.append("video:");
179  } else if (media_info.has_audio_info()) {
180  key.append("audio:");
181  } else if (media_info.has_text_info()) {
182  key.append(MediaInfo_TextInfo_TextType_Name(media_info.text_info().type()));
183  key.append(":");
184  } else {
185  key.append("unknown:");
186  }
187 
188  if (media_info.has_dash_label())
189  key.append(media_info.dash_label() + ":");
190 
191  key.append(MediaInfo_ContainerType_Name(media_info.container_type()));
192  if (!ignore_codec) {
193  key.append(":");
194  key.append(GetBaseCodec(media_info));
195 
196  if (GetBaseCodec(media_info).find("dvh") == 0) {
197  // Transfer characteristics for Dolby Vision (dvh1 or dvhe) must be PQ
198  // irrespective of value present in SPS VUI.
199  key.append(":");
200  key.append(std::to_string(kTransferFunctionPQ));
201  } else if (media_info.video_info().has_transfer_characteristics()) {
202  key.append(":");
203  key.append(
204  std::to_string(media_info.video_info().transfer_characteristics()));
205  }
206  }
207  key.append(":");
208  key.append(GetLanguage(media_info));
209 
210  // Trick play streams of the same original stream, but possibly with
211  // different trick_play_factors, belong to the same trick play AdaptationSet.
212  if (media_info.video_info().has_playback_rate()) {
213  key.append(":trick_play");
214  }
215 
216  if (!media_info.dash_accessibilities().empty()) {
217  key.append(":accessibility_");
218  for (const std::string& accessibility : media_info.dash_accessibilities())
219  key.append(accessibility);
220  }
221 
222  if (!media_info.dash_roles().empty()) {
223  key.append(":roles_");
224  for (const std::string& role : media_info.dash_roles())
225  key.append(role);
226  }
227 
228  return key;
229 }
230 
231 std::string FloatToXmlString(double number) {
232  // Keep up to microsecond accuracy but trim trailing 0s
233  std::string formatted = absl::StrFormat("%.6f", number);
234  size_t decimalPos = formatted.find('.');
235  if (decimalPos != std::string::npos) {
236  size_t lastNonZeroPos = formatted.find_last_not_of('0');
237  if (lastNonZeroPos >= decimalPos) {
238  formatted.erase(lastNonZeroPos + 1);
239  }
240  if (formatted.back() == '.') {
241  formatted.pop_back();
242  }
243  }
244 
245  return formatted;
246 }
247 
248 std::string SecondsToXmlDuration(double seconds) {
249  // Chrome internally uses time accurate to microseconds, which is implemented
250  // per MSE spec (https://www.w3.org/TR/media-source/).
251  // We need a string formatter that has at least microseconds accuracy for a
252  // normal video (with duration up to 3 hours). FloatToXmlString
253  // implementation meets the requirement.
254  return absl::StrFormat("PT%sS", FloatToXmlString(seconds));
255 }
256 
257 bool GetDurationAttribute(xmlNodePtr node, float* duration) {
258  DCHECK(node);
259  DCHECK(duration);
260  static const char kDuration[] = "duration";
261  xml::scoped_xml_ptr<xmlChar> duration_value(
262  xmlGetProp(node, BAD_CAST kDuration));
263 
264  if (!duration_value)
265  return false;
266 
267  double duration_double_precision = 0.0;
268  if (!absl::SimpleAtod(reinterpret_cast<const char*>(duration_value.get()),
269  &duration_double_precision)) {
270  return false;
271  }
272 
273  *duration = static_cast<float>(duration_double_precision);
274  return true;
275 }
276 
277 bool MoreThanOneTrue(bool b1, bool b2, bool b3) {
278  return (b1 && b2) || (b2 && b3) || (b3 && b1);
279 }
280 
281 bool AtLeastOneTrue(bool b1, bool b2, bool b3) {
282  return b1 || b2 || b3;
283 }
284 
285 bool OnlyOneTrue(bool b1, bool b2, bool b3) {
286  return !MoreThanOneTrue(b1, b2, b3) && AtLeastOneTrue(b1, b2, b3);
287 }
288 
289 // Coverts binary data into human readable UUID format.
290 bool HexToUUID(const std::string& data, std::string* uuid_format) {
291  DCHECK(uuid_format);
292  const size_t kExpectedUUIDSize = 16;
293  if (data.size() != kExpectedUUIDSize) {
294  LOG(ERROR) << "UUID size is expected to be " << kExpectedUUIDSize
295  << " but is " << data.size() << " and the data in hex is "
296  << absl::BytesToHexString(data);
297  return false;
298  }
299 
300  const std::string hex_encoded =
301  absl::AsciiStrToLower(absl::BytesToHexString(data));
302  DCHECK_EQ(hex_encoded.size(), kExpectedUUIDSize * 2);
303  std::string_view all(hex_encoded);
304  // Note UUID has 5 parts separated with dashes.
305  // e.g. 123e4567-e89b-12d3-a456-426655440000
306  // These StringPieces have each part.
307  std::string_view first = all.substr(0, 8);
308  std::string_view second = all.substr(8, 4);
309  std::string_view third = all.substr(12, 4);
310  std::string_view fourth = all.substr(16, 4);
311  std::string_view fifth = all.substr(20, 12);
312 
313  // 32 hexadecimal characters with 4 hyphens.
314  const size_t kHumanReadableUUIDSize = 36;
315  uuid_format->reserve(kHumanReadableUUIDSize);
316  absl::StrAppendFormat(uuid_format, "%s-%s-%s-%s-%s", first, second, third,
317  fourth, fifth);
318  return true;
319 }
320 
321 void UpdateContentProtectionPsshHelper(
322  const std::string& drm_uuid,
323  const std::string& pssh,
324  std::list<ContentProtectionElement>* content_protection_elements) {
325  const std::string drm_uuid_schemd_id_uri_form = "urn:uuid:" + drm_uuid;
326  for (std::list<ContentProtectionElement>::iterator protection =
327  content_protection_elements->begin();
328  protection != content_protection_elements->end(); ++protection) {
329  if (protection->scheme_id_uri != drm_uuid_schemd_id_uri_form) {
330  continue;
331  }
332 
333  for (std::vector<Element>::iterator subelement =
334  protection->subelements.begin();
335  subelement != protection->subelements.end(); ++subelement) {
336  if (subelement->name == kPsshElementName) {
337  // For now, we want to remove the PSSH element because some players do
338  // not support updating pssh.
339  protection->subelements.erase(subelement);
340 
341  // TODO(rkuroiwa): Uncomment this and remove the line above when
342  // shaka-player supports updating PSSH.
343  // subelement->content = pssh;
344  return;
345  }
346  }
347 
348  // Reaching here means <cenc:pssh> does not exist under the
349  // ContentProtection element. Add it.
350  // TODO(rkuroiwa): Uncomment this when shaka-player supports updating PSSH.
351  // Element cenc_pssh;
352  // cenc_pssh.name = kPsshElementName;
353  // cenc_pssh.content = pssh;
354  // protection->subelements.push_back(cenc_pssh);
355  return;
356  }
357 
358  // Reaching here means that ContentProtection for the DRM does not exist.
359  // Add it.
360  ContentProtectionElement content_protection;
361  content_protection.scheme_id_uri = drm_uuid_schemd_id_uri_form;
362  // TODO(rkuroiwa): Uncomment this when shaka-player supports updating PSSH.
363  // Element cenc_pssh;
364  // cenc_pssh.name = kPsshElementName;
365  // cenc_pssh.content = pssh;
366  // content_protection.subelements.push_back(cenc_pssh);
367  content_protection_elements->push_back(content_protection);
368  return;
369 }
370 
371 namespace {
372 
373 // UUID for Marlin Adaptive Streaming Specification – Simple Profile from
374 // https://dashif.org/identifiers/content_protection/.
375 const char kMarlinUUID[] = "5e629af5-38da-4063-8977-97ffbd9902d4";
376 // String representation of media::kFairPlaySystemId.
377 const char kFairPlayUUID[] = "94ce86fb-07ff-4f43-adb8-93d2fa968ca2";
378 // String representation of media::kPlayReadySystemId.
379 const char kPlayReadyUUID[] = "9a04f079-9840-4286-ab92-e65be0885f95";
380 // It is RECOMMENDED to include the @value attribute with name and version
381 // "MSPR 2.0". See
382 // https://docs.microsoft.com/en-us/playready/specifications/mpeg-dash-playready#221-general.
383 const char kContentProtectionValueMSPR20[] = "MSPR 2.0";
384 
385 Element GenerateMarlinContentIds(const std::string& key_id) {
386  // See https://github.com/shaka-project/shaka-packager/issues/381 for details.
387  static const char kMarlinContentIdName[] = "mas:MarlinContentId";
388  static const char kMarlinContentIdPrefix[] = "urn:marlin:kid:";
389  static const char kMarlinContentIdsName[] = "mas:MarlinContentIds";
390 
391  Element marlin_content_id;
392  marlin_content_id.name = kMarlinContentIdName;
393  marlin_content_id.content =
394  kMarlinContentIdPrefix +
395  absl::AsciiStrToLower(absl::BytesToHexString(key_id));
396 
397  Element marlin_content_ids;
398  marlin_content_ids.name = kMarlinContentIdsName;
399  marlin_content_ids.subelements.push_back(marlin_content_id);
400 
401  return marlin_content_ids;
402 }
403 
404 Element GenerateCencPsshElement(const std::string& pssh) {
405  std::string base64_encoded_pssh;
406  absl::Base64Escape(std::string_view(pssh.data(), pssh.size()),
407  &base64_encoded_pssh);
408  Element cenc_pssh;
409  cenc_pssh.name = kPsshElementName;
410  cenc_pssh.content = base64_encoded_pssh;
411  return cenc_pssh;
412 }
413 
414 // Extract MS PlayReady Object from given PSSH
415 // and encode it in base64.
416 Element GenerateMsprProElement(const std::string& pssh) {
417  std::unique_ptr<media::PsshBoxBuilder> b =
419  reinterpret_cast<const uint8_t*>(pssh.data()),
420  pssh.size()
421  );
422 
423  const std::vector<uint8_t> *p_pssh = &b->pssh_data();
424  std::string base64_encoded_mspr;
425  absl::Base64Escape(
426  std::string_view(reinterpret_cast<const char*>(p_pssh->data()),
427  p_pssh->size()),
428  &base64_encoded_mspr);
429  Element mspr_pro;
430  mspr_pro.name = kMsproElementName;
431  mspr_pro.content = base64_encoded_mspr;
432  return mspr_pro;
433 }
434 
435 // Helper function. This works because Representation and AdaptationSet both
436 // have AddContentProtectionElement().
437 template <typename ContentProtectionParent>
438 void AddContentProtectionElementsHelperTemplated(
439  const MediaInfo& media_info,
440  ContentProtectionParent* parent) {
441  DCHECK(parent);
442  if (!media_info.has_protected_content())
443  return;
444 
445  const MediaInfo::ProtectedContent& protected_content =
446  media_info.protected_content();
447 
448  // DASH MPD spec specifies a default ContentProtection element for ISO BMFF
449  // (MP4) files.
450  const bool is_mp4_container =
451  media_info.container_type() == MediaInfo::CONTAINER_MP4;
452  std::string key_id_uuid_format;
453  if (protected_content.has_default_key_id() &&
454  !IsKeyRotationDefaultKeyId(protected_content.default_key_id())) {
455  if (!HexToUUID(protected_content.default_key_id(), &key_id_uuid_format)) {
456  LOG(ERROR) << "Failed to convert default key ID into UUID format.";
457  }
458  }
459 
460  if (is_mp4_container) {
461  ContentProtectionElement mp4_content_protection;
462  mp4_content_protection.scheme_id_uri = kEncryptedMp4Scheme;
463  mp4_content_protection.value = protected_content.protection_scheme();
464  if (!key_id_uuid_format.empty()) {
465  mp4_content_protection.additional_attributes["cenc:default_KID"] =
466  key_id_uuid_format;
467  }
468 
469  parent->AddContentProtectionElement(mp4_content_protection);
470  }
471 
472  for (const auto& entry : protected_content.content_protection_entry()) {
473  if (!entry.has_uuid()) {
474  LOG(WARNING)
475  << "ContentProtectionEntry was specified but no UUID is set for "
476  << entry.name_version() << ", skipping.";
477  continue;
478  }
479 
480  ContentProtectionElement drm_content_protection;
481 
482  if (entry.has_name_version())
483  drm_content_protection.value = entry.name_version();
484 
485  if (entry.uuid() == kFairPlayUUID) {
486  VLOG(1) << "Skipping FairPlay ContentProtection element as FairPlay does "
487  "not support DASH signaling.";
488  continue;
489  } else if (entry.uuid() == kMarlinUUID) {
490  // Marlin requires its uuid to be in upper case. See #525 for details.
491  drm_content_protection.scheme_id_uri =
492  "urn:uuid:" + absl::AsciiStrToUpper(entry.uuid());
493  drm_content_protection.subelements.push_back(
494  GenerateMarlinContentIds(protected_content.default_key_id()));
495  } else {
496  drm_content_protection.scheme_id_uri = "urn:uuid:" + entry.uuid();
497  if (!entry.pssh().empty()) {
498  drm_content_protection.subelements.push_back(
499  GenerateCencPsshElement(entry.pssh()));
500  if(entry.uuid() == kPlayReadyUUID && protected_content.include_mspr_pro()) {
501  drm_content_protection.subelements.push_back(
502  GenerateMsprProElement(entry.pssh()));
503  drm_content_protection.value = kContentProtectionValueMSPR20;
504  }
505  }
506  }
507 
508  if (!key_id_uuid_format.empty() && !is_mp4_container) {
509  drm_content_protection.additional_attributes["cenc:default_KID"] =
510  key_id_uuid_format;
511  }
512 
513  parent->AddContentProtectionElement(drm_content_protection);
514  }
515 
516  if (protected_content.content_protection_entry().size() == 0) {
517  VLOG(1) << "The media is encrypted but no content protection specified "
518  << "(can happen with key rotation).";
519  }
520 }
521 } // namespace
522 
523 void AddContentProtectionElements(const MediaInfo& media_info,
524  Representation* parent) {
525  AddContentProtectionElementsHelperTemplated(media_info, parent);
526 }
527 
528 void AddContentProtectionElements(const MediaInfo& media_info,
529  AdaptationSet* parent) {
530  AddContentProtectionElementsHelperTemplated(media_info, parent);
531 }
532 
533 } // namespace shaka
static std::unique_ptr< PsshBoxBuilder > ParseFromBox(const uint8_t *data, size_t data_size)
All the methods that are virtual are virtual for mocking.
Definition: crypto_flags.cc:66
bool HexToUUID(const std::string &data, std::string *uuid_format)
Definition: mpd_utils.cc:290
std::string LanguageToShortestForm(const std::string &language)
void AddContentProtectionElements(const MediaInfo &media_info, Representation *parent)
Definition: mpd_utils.cc:523