1 /*
2 * Copyright (c) 2014 The WebRTC project authors. All Rights Reserved.
3 *
4 * Use of this source code is governed by a BSD-style license
5 * that can be found in the LICENSE file in the root of the source
6 * tree. An additional intellectual property rights grant can be found
7 * in the file PATENTS. All contributing project authors may
8 * be found in the AUTHORS file in the root of the source tree.
9 */
10
11 // Based on the WAV file format documentation at
12 // https://ccrma.stanford.edu/courses/422/projects/WaveFormat/ and
13 // http://www-mmsp.ece.mcgill.ca/Documents/AudioFormats/WAVE/WAVE.html
14
15 #include "webrtc/common_audio/wav_header.h"
16
17 #include <algorithm>
18 #include <cstring>
19 #include <limits>
20 #include <string>
21
22 #include "webrtc/base/checks.h"
23 #include "webrtc/common_audio/include/audio_util.h"
24
25 namespace webrtc {
26 namespace {
27
28 struct ChunkHeader {
29 uint32_t ID;
30 uint32_t Size;
31 };
32 static_assert(sizeof(ChunkHeader) == 8, "ChunkHeader size");
33
34 // We can't nest this definition in WavHeader, because VS2013 gives an error
35 // on sizeof(WavHeader::fmt): "error C2070: 'unknown': illegal sizeof operand".
36 struct FmtSubchunk {
37 ChunkHeader header;
38 uint16_t AudioFormat;
39 uint16_t NumChannels;
40 uint32_t SampleRate;
41 uint32_t ByteRate;
42 uint16_t BlockAlign;
43 uint16_t BitsPerSample;
44 };
45 static_assert(sizeof(FmtSubchunk) == 24, "FmtSubchunk size");
46 const uint32_t kFmtSubchunkSize = sizeof(FmtSubchunk) - sizeof(ChunkHeader);
47
48 struct WavHeader {
49 struct {
50 ChunkHeader header;
51 uint32_t Format;
52 } riff;
53 FmtSubchunk fmt;
54 struct {
55 ChunkHeader header;
56 } data;
57 };
58 static_assert(sizeof(WavHeader) == kWavHeaderSize, "no padding in header");
59
60 } // namespace
61
CheckWavParameters(size_t num_channels,int sample_rate,WavFormat format,size_t bytes_per_sample,size_t num_samples)62 bool CheckWavParameters(size_t num_channels,
63 int sample_rate,
64 WavFormat format,
65 size_t bytes_per_sample,
66 size_t num_samples) {
67 // num_channels, sample_rate, and bytes_per_sample must be positive, must fit
68 // in their respective fields, and their product must fit in the 32-bit
69 // ByteRate field.
70 if (num_channels == 0 || sample_rate <= 0 || bytes_per_sample == 0)
71 return false;
72 if (static_cast<uint64_t>(sample_rate) > std::numeric_limits<uint32_t>::max())
73 return false;
74 if (num_channels > std::numeric_limits<uint16_t>::max())
75 return false;
76 if (static_cast<uint64_t>(bytes_per_sample) * 8 >
77 std::numeric_limits<uint16_t>::max())
78 return false;
79 if (static_cast<uint64_t>(sample_rate) * num_channels * bytes_per_sample >
80 std::numeric_limits<uint32_t>::max())
81 return false;
82
83 // format and bytes_per_sample must agree.
84 switch (format) {
85 case kWavFormatPcm:
86 // Other values may be OK, but for now we're conservative:
87 if (bytes_per_sample != 1 && bytes_per_sample != 2)
88 return false;
89 break;
90 case kWavFormatALaw:
91 case kWavFormatMuLaw:
92 if (bytes_per_sample != 1)
93 return false;
94 break;
95 default:
96 return false;
97 }
98
99 // The number of bytes in the file, not counting the first ChunkHeader, must
100 // be less than 2^32; otherwise, the ChunkSize field overflows.
101 const size_t header_size = kWavHeaderSize - sizeof(ChunkHeader);
102 const size_t max_samples =
103 (std::numeric_limits<uint32_t>::max() - header_size) / bytes_per_sample;
104 if (num_samples > max_samples)
105 return false;
106
107 // Each channel must have the same number of samples.
108 if (num_samples % num_channels != 0)
109 return false;
110
111 return true;
112 }
113
114 #ifdef WEBRTC_ARCH_LITTLE_ENDIAN
WriteLE16(uint16_t * f,uint16_t x)115 static inline void WriteLE16(uint16_t* f, uint16_t x) { *f = x; }
WriteLE32(uint32_t * f,uint32_t x)116 static inline void WriteLE32(uint32_t* f, uint32_t x) { *f = x; }
WriteFourCC(uint32_t * f,char a,char b,char c,char d)117 static inline void WriteFourCC(uint32_t* f, char a, char b, char c, char d) {
118 *f = static_cast<uint32_t>(a)
119 | static_cast<uint32_t>(b) << 8
120 | static_cast<uint32_t>(c) << 16
121 | static_cast<uint32_t>(d) << 24;
122 }
123
ReadLE16(uint16_t x)124 static inline uint16_t ReadLE16(uint16_t x) { return x; }
ReadLE32(uint32_t x)125 static inline uint32_t ReadLE32(uint32_t x) { return x; }
ReadFourCC(uint32_t x)126 static inline std::string ReadFourCC(uint32_t x) {
127 return std::string(reinterpret_cast<char*>(&x), 4);
128 }
129 #else
130 #error "Write be-to-le conversion functions"
131 #endif
132
RiffChunkSize(size_t bytes_in_payload)133 static inline uint32_t RiffChunkSize(size_t bytes_in_payload) {
134 return static_cast<uint32_t>(
135 bytes_in_payload + kWavHeaderSize - sizeof(ChunkHeader));
136 }
137
ByteRate(size_t num_channels,int sample_rate,size_t bytes_per_sample)138 static inline uint32_t ByteRate(size_t num_channels, int sample_rate,
139 size_t bytes_per_sample) {
140 return static_cast<uint32_t>(num_channels * sample_rate * bytes_per_sample);
141 }
142
BlockAlign(size_t num_channels,size_t bytes_per_sample)143 static inline uint16_t BlockAlign(size_t num_channels,
144 size_t bytes_per_sample) {
145 return static_cast<uint16_t>(num_channels * bytes_per_sample);
146 }
147
WriteWavHeader(uint8_t * buf,size_t num_channels,int sample_rate,WavFormat format,size_t bytes_per_sample,size_t num_samples)148 void WriteWavHeader(uint8_t* buf,
149 size_t num_channels,
150 int sample_rate,
151 WavFormat format,
152 size_t bytes_per_sample,
153 size_t num_samples) {
154 RTC_CHECK(CheckWavParameters(num_channels, sample_rate, format,
155 bytes_per_sample, num_samples));
156
157 WavHeader header;
158 const size_t bytes_in_payload = bytes_per_sample * num_samples;
159
160 WriteFourCC(&header.riff.header.ID, 'R', 'I', 'F', 'F');
161 WriteLE32(&header.riff.header.Size, RiffChunkSize(bytes_in_payload));
162 WriteFourCC(&header.riff.Format, 'W', 'A', 'V', 'E');
163
164 WriteFourCC(&header.fmt.header.ID, 'f', 'm', 't', ' ');
165 WriteLE32(&header.fmt.header.Size, kFmtSubchunkSize);
166 WriteLE16(&header.fmt.AudioFormat, format);
167 WriteLE16(&header.fmt.NumChannels, static_cast<uint16_t>(num_channels));
168 WriteLE32(&header.fmt.SampleRate, sample_rate);
169 WriteLE32(&header.fmt.ByteRate, ByteRate(num_channels, sample_rate,
170 bytes_per_sample));
171 WriteLE16(&header.fmt.BlockAlign, BlockAlign(num_channels, bytes_per_sample));
172 WriteLE16(&header.fmt.BitsPerSample,
173 static_cast<uint16_t>(8 * bytes_per_sample));
174
175 WriteFourCC(&header.data.header.ID, 'd', 'a', 't', 'a');
176 WriteLE32(&header.data.header.Size, static_cast<uint32_t>(bytes_in_payload));
177
178 // Do an extra copy rather than writing everything to buf directly, since buf
179 // might not be correctly aligned.
180 memcpy(buf, &header, kWavHeaderSize);
181 }
182
ReadWavHeader(ReadableWav * readable,size_t * num_channels,int * sample_rate,WavFormat * format,size_t * bytes_per_sample,size_t * num_samples)183 bool ReadWavHeader(ReadableWav* readable,
184 size_t* num_channels,
185 int* sample_rate,
186 WavFormat* format,
187 size_t* bytes_per_sample,
188 size_t* num_samples) {
189 WavHeader header;
190 if (readable->Read(&header, kWavHeaderSize - sizeof(header.data)) !=
191 kWavHeaderSize - sizeof(header.data))
192 return false;
193
194 const uint32_t fmt_size = ReadLE32(header.fmt.header.Size);
195 if (fmt_size != kFmtSubchunkSize) {
196 // There is an optional two-byte extension field permitted to be present
197 // with PCM, but which must be zero.
198 int16_t ext_size;
199 if (kFmtSubchunkSize + sizeof(ext_size) != fmt_size)
200 return false;
201 if (readable->Read(&ext_size, sizeof(ext_size)) != sizeof(ext_size))
202 return false;
203 if (ext_size != 0)
204 return false;
205 }
206 if (readable->Read(&header.data, sizeof(header.data)) != sizeof(header.data))
207 return false;
208
209 // Parse needed fields.
210 *format = static_cast<WavFormat>(ReadLE16(header.fmt.AudioFormat));
211 *num_channels = ReadLE16(header.fmt.NumChannels);
212 *sample_rate = ReadLE32(header.fmt.SampleRate);
213 *bytes_per_sample = ReadLE16(header.fmt.BitsPerSample) / 8;
214 const size_t bytes_in_payload = ReadLE32(header.data.header.Size);
215 if (*bytes_per_sample == 0)
216 return false;
217 *num_samples = bytes_in_payload / *bytes_per_sample;
218
219 // Sanity check remaining fields.
220 if (ReadFourCC(header.riff.header.ID) != "RIFF")
221 return false;
222 if (ReadFourCC(header.riff.Format) != "WAVE")
223 return false;
224 if (ReadFourCC(header.fmt.header.ID) != "fmt ")
225 return false;
226 if (ReadFourCC(header.data.header.ID) != "data")
227 return false;
228
229 if (ReadLE32(header.riff.header.Size) < RiffChunkSize(bytes_in_payload))
230 return false;
231 if (ReadLE32(header.fmt.ByteRate) !=
232 ByteRate(*num_channels, *sample_rate, *bytes_per_sample))
233 return false;
234 if (ReadLE16(header.fmt.BlockAlign) !=
235 BlockAlign(*num_channels, *bytes_per_sample))
236 return false;
237
238 return CheckWavParameters(*num_channels, *sample_rate, *format,
239 *bytes_per_sample, *num_samples);
240 }
241
242
243 } // namespace webrtc
244