Wyatt Hepler | 80c6ee5 | 2020-01-03 09:54:58 -0800 | [diff] [blame] | 1 | // Copyright 2020 The Pigweed Authors |
| 2 | // |
| 3 | // Licensed under the Apache License, Version 2.0 (the "License"); you may not |
| 4 | // use this file except in compliance with the License. You may obtain a copy of |
| 5 | // the License at |
| 6 | // |
| 7 | // https://www.apache.org/licenses/LICENSE-2.0 |
| 8 | // |
| 9 | // Unless required by applicable law or agreed to in writing, software |
| 10 | // distributed under the License is distributed on an "AS IS" BASIS, WITHOUT |
| 11 | // WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the |
| 12 | // License for the specific language governing permissions and limitations under |
| 13 | // the License. |
| 14 | |
| 15 | // This file defines the functions that encode tokenized logs at runtime. These |
| 16 | // are the only pw_tokenizer functions present in a binary that tokenizes |
| 17 | // strings. All other tokenizing code is resolved at compile time. |
| 18 | |
| 19 | #include "pw_tokenizer/tokenize.h" |
| 20 | |
| 21 | #include <algorithm> |
| 22 | #include <array> |
| 23 | #include <cstdarg> |
| 24 | #include <cstddef> |
| 25 | #include <cstring> |
| 26 | |
Wyatt Hepler | a6d5cc6 | 2020-01-17 14:15:40 -0800 | [diff] [blame] | 27 | #include "pw_polyfill/language_features.h" // static_assert |
Wyatt Hepler | 80c6ee5 | 2020-01-03 09:54:58 -0800 | [diff] [blame] | 28 | #include "pw_varint/varint.h" |
| 29 | |
Wyatt Hepler | a6d5cc6 | 2020-01-17 14:15:40 -0800 | [diff] [blame] | 30 | namespace pw { |
| 31 | namespace tokenizer { |
Wyatt Hepler | 80c6ee5 | 2020-01-03 09:54:58 -0800 | [diff] [blame] | 32 | namespace { |
| 33 | |
| 34 | // Store metadata about this compilation's string tokenization in the ELF. |
| 35 | // |
| 36 | // The tokenizer metadata will not go into the on-device executable binary code. |
| 37 | // This metadata will be present in the ELF file's .tokenizer_info section, from |
| 38 | // which the host-side tooling (Python, Java, etc.) can understand how to decode |
| 39 | // tokenized strings for the given binary. Only attributes that affect the |
| 40 | // decoding process are recorded. |
| 41 | // |
| 42 | // Tokenizer metadata is stored in an array of key-value pairs. Each Metadata |
| 43 | // object is 32 bytes: a 24-byte string and an 8-byte value. Metadata structs |
| 44 | // may be parsed in Python with the struct format '24s<Q'. |
| 45 | PW_PACKED(struct) Metadata { |
| 46 | char name[24]; // name of the metadata field |
| 47 | uint64_t value; // value of the field |
| 48 | }; |
| 49 | |
| 50 | static_assert(sizeof(Metadata) == 32); |
| 51 | |
| 52 | // Store tokenization metadata in its own section. |
| 53 | constexpr Metadata metadata[] PW_KEEP_IN_SECTION(".tokenzier_info") = { |
| 54 | {"hash_length_bytes", PW_TOKENIZER_CFG_HASH_LENGTH}, |
| 55 | {"sizeof_long", sizeof(long)}, // %l conversion specifier |
| 56 | {"sizeof_intmax_t", sizeof(intmax_t)}, // %j conversion specifier |
| 57 | {"sizeof_size_t", sizeof(size_t)}, // %z conversion specifier |
| 58 | {"sizeof_ptrdiff_t", sizeof(ptrdiff_t)}, // %t conversion specifier |
| 59 | }; |
| 60 | |
| 61 | // Declare the types as an enum for convenience. |
| 62 | enum class ArgType : uint8_t { |
| 63 | kInt = PW_TOKENIZER_ARG_TYPE_INT, |
| 64 | kInt64 = PW_TOKENIZER_ARG_TYPE_INT64, |
| 65 | kDouble = PW_TOKENIZER_ARG_TYPE_DOUBLE, |
| 66 | kString = PW_TOKENIZER_ARG_TYPE_STRING, |
| 67 | }; |
| 68 | |
| 69 | // Just to be safe, make sure these values are what we expect them to be. |
| 70 | static_assert(0b00u == static_cast<uint8_t>(ArgType::kInt)); |
| 71 | static_assert(0b01u == static_cast<uint8_t>(ArgType::kInt64)); |
| 72 | static_assert(0b10u == static_cast<uint8_t>(ArgType::kDouble)); |
| 73 | static_assert(0b11u == static_cast<uint8_t>(ArgType::kString)); |
| 74 | |
| 75 | // Buffer for encoding a tokenized string and arguments. |
| 76 | struct EncodedMessage { |
| 77 | pw_TokenizerStringToken token; |
| 78 | std::array<uint8_t, PW_TOKENIZER_CFG_ENCODING_BUFFER_SIZE_BYTES> args; |
| 79 | }; |
| 80 | |
| 81 | static_assert(offsetof(EncodedMessage, args) == sizeof(EncodedMessage::token), |
| 82 | "EncodedMessage should not have padding bytes between members"); |
| 83 | |
| 84 | size_t EncodeInt(int value, const span<uint8_t>& output) { |
| 85 | return varint::Encode(value, pw::as_writable_bytes(output)); |
| 86 | } |
| 87 | |
| 88 | size_t EncodeInt64(int64_t value, const span<uint8_t>& output) { |
| 89 | return varint::Encode(value, pw::as_writable_bytes(output)); |
| 90 | } |
| 91 | |
| 92 | size_t EncodeFloat(float value, const span<uint8_t>& output) { |
| 93 | if (output.size() < sizeof(value)) { |
| 94 | return 0; |
| 95 | } |
| 96 | std::memcpy(output.data(), &value, sizeof(value)); |
| 97 | return sizeof(value); |
| 98 | } |
| 99 | |
| 100 | size_t EncodeString(const char* string, const span<uint8_t>& output) { |
| 101 | // The top bit of the status byte indicates if the string was truncated. |
| 102 | static constexpr size_t kMaxStringLength = 0x7Fu; |
| 103 | |
| 104 | if (output.empty()) { // At least one byte is needed for the status/size. |
| 105 | return 0; |
| 106 | } |
| 107 | |
| 108 | if (string == nullptr) { |
| 109 | string = "NULL"; |
| 110 | } |
| 111 | |
| 112 | // Subtract 1 to save room for the status byte. |
| 113 | const size_t max_bytes = std::min(output.size(), kMaxStringLength) - 1; |
| 114 | |
| 115 | // Scan the string to find out how many bytes to copy. |
| 116 | size_t bytes_to_copy = 0; |
| 117 | uint8_t overflow_bit = 0; |
| 118 | |
| 119 | while (string[bytes_to_copy] != '\0') { |
| 120 | if (bytes_to_copy == max_bytes) { |
| 121 | overflow_bit = '\x80'; |
| 122 | break; |
| 123 | } |
| 124 | bytes_to_copy += 1; |
| 125 | } |
| 126 | |
| 127 | output[0] = bytes_to_copy | overflow_bit; |
| 128 | std::memcpy(output.data() + 1, string, bytes_to_copy); |
| 129 | |
| 130 | return bytes_to_copy + 1; // include the status byte in the total |
| 131 | } |
| 132 | |
| 133 | size_t EncodeArgs(pw_TokenizerArgTypes types, |
| 134 | va_list args, |
| 135 | span<uint8_t> output) { |
| 136 | size_t arg_count = types & PW_TOKENIZER_TYPE_COUNT_MASK; |
| 137 | types >>= PW_TOKENIZER_TYPE_COUNT_SIZE_BITS; |
| 138 | |
| 139 | size_t encoded_bytes = 0; |
| 140 | while (arg_count != 0u) { |
| 141 | // How many bytes were encoded; 0 indicates that there wasn't enough space. |
| 142 | size_t argument_bytes = 0; |
| 143 | |
| 144 | switch (static_cast<ArgType>(types & 0b11u)) { |
| 145 | case ArgType::kInt: |
| 146 | argument_bytes = EncodeInt(va_arg(args, int), output); |
| 147 | break; |
| 148 | case ArgType::kInt64: |
| 149 | argument_bytes = EncodeInt64(va_arg(args, int64_t), output); |
| 150 | break; |
| 151 | case ArgType::kDouble: |
| 152 | argument_bytes = |
| 153 | EncodeFloat(static_cast<float>(va_arg(args, double)), output); |
| 154 | break; |
| 155 | case ArgType::kString: |
| 156 | argument_bytes = EncodeString(va_arg(args, const char*), output); |
| 157 | break; |
| 158 | } |
| 159 | |
| 160 | // If zero bytes were encoded, the encoding buffer is full. |
| 161 | if (argument_bytes == 0u) { |
| 162 | break; |
| 163 | } |
| 164 | |
| 165 | output = output.subspan(argument_bytes); |
| 166 | encoded_bytes += argument_bytes; |
| 167 | |
| 168 | arg_count -= 1; |
| 169 | types >>= 2; // each argument type is encoded in two bits |
| 170 | } |
| 171 | |
| 172 | return encoded_bytes; |
| 173 | } |
| 174 | |
| 175 | } // namespace |
| 176 | |
| 177 | extern "C" { |
| 178 | |
| 179 | void pw_TokenizeToBuffer(void* buffer, |
| 180 | size_t* buffer_size_bytes, |
| 181 | pw_TokenizerStringToken token, |
| 182 | pw_TokenizerArgTypes types, |
| 183 | ...) { |
| 184 | if (*buffer_size_bytes < sizeof(token)) { |
| 185 | *buffer_size_bytes = 0; |
| 186 | return; |
| 187 | } |
| 188 | |
| 189 | std::memcpy(buffer, &token, sizeof(token)); |
| 190 | |
| 191 | va_list args; |
| 192 | va_start(args, types); |
| 193 | const size_t encoded_bytes = |
| 194 | EncodeArgs(types, |
| 195 | args, |
Wyatt Hepler | a6d5cc6 | 2020-01-17 14:15:40 -0800 | [diff] [blame] | 196 | span<uint8_t>(static_cast<uint8_t*>(buffer) + sizeof(token), |
| 197 | *buffer_size_bytes - sizeof(token))); |
Wyatt Hepler | 80c6ee5 | 2020-01-03 09:54:58 -0800 | [diff] [blame] | 198 | va_end(args); |
| 199 | |
| 200 | *buffer_size_bytes = sizeof(token) + encoded_bytes; |
| 201 | } |
| 202 | |
| 203 | void pw_TokenizeToCallback(void (*callback)(const uint8_t* encoded_message, |
| 204 | size_t size_bytes), |
| 205 | pw_TokenizerStringToken token, |
| 206 | pw_TokenizerArgTypes types, |
| 207 | ...) { |
| 208 | EncodedMessage encoded; |
| 209 | encoded.token = token; |
| 210 | |
| 211 | va_list args; |
| 212 | va_start(args, types); |
| 213 | const size_t encoded_bytes = EncodeArgs(types, args, encoded.args); |
| 214 | va_end(args); |
| 215 | |
| 216 | callback(reinterpret_cast<const uint8_t*>(&encoded), |
| 217 | sizeof(encoded.token) + encoded_bytes); |
| 218 | } |
| 219 | |
| 220 | #if PW_TOKENIZER_CFG_ENABLE_TOKENIZE_TO_GLOBAL_HANDLER |
| 221 | |
| 222 | void pw_TokenizeToGlobalHandler(pw_TokenizerStringToken token, |
| 223 | pw_TokenizerArgTypes types, |
| 224 | ...) { |
| 225 | EncodedMessage encoded; |
| 226 | encoded.token = token; |
| 227 | |
| 228 | va_list args; |
| 229 | va_start(args, types); |
| 230 | const size_t encoded_bytes = EncodeArgs(types, args, encoded.args); |
| 231 | va_end(args); |
| 232 | |
| 233 | pw_TokenizerHandleEncodedMessage(reinterpret_cast<const uint8_t*>(&encoded), |
| 234 | sizeof(encoded.token) + encoded_bytes); |
| 235 | } |
| 236 | |
| 237 | #endif // PW_TOKENIZER_CFG_ENABLE_TOKENIZE_TO_GLOBAL_HANDLER |
| 238 | |
| 239 | #if PW_TOKENIZER_CFG_ENABLE_TOKENIZE_TO_GLOBAL_HANDLER_WITH_PAYLOAD |
| 240 | |
| 241 | void pw_TokenizeToGlobalHandlerWithPayload(const pw_TokenizerPayload payload, |
| 242 | pw_TokenizerStringToken token, |
| 243 | pw_TokenizerArgTypes types, |
| 244 | ...) { |
| 245 | EncodedMessage encoded; |
| 246 | encoded.token = token; |
| 247 | |
| 248 | va_list args; |
| 249 | va_start(args, types); |
| 250 | const size_t encoded_bytes = EncodeArgs(types, args, encoded.args); |
| 251 | va_end(args); |
| 252 | |
| 253 | pw_TokenizerHandleEncodedMessageWithPayload( |
| 254 | payload, |
| 255 | reinterpret_cast<const uint8_t*>(&encoded), |
| 256 | sizeof(encoded.token) + encoded_bytes); |
| 257 | } |
| 258 | |
| 259 | #endif // PW_TOKENIZER_CFG_ENABLE_TOKENIZE_TO_GLOBAL_HANDLER_WITH_PAYLOAD |
| 260 | |
| 261 | } // extern "C" |
| 262 | |
Wyatt Hepler | a6d5cc6 | 2020-01-17 14:15:40 -0800 | [diff] [blame] | 263 | } // namespace tokenizer |
| 264 | } // namespace pw |