| 1 | /* |
| 2 | * Copyright (C) 2018 Yusuke Suzuki <yusukesuzuki@slowstart.org>. |
| 3 | * |
| 4 | * Redistribution and use in source and binary forms, with or without |
| 5 | * modification, are permitted provided that the following conditions |
| 6 | * are met: |
| 7 | * 1. Redistributions of source code must retain the above copyright |
| 8 | * notice, this list of conditions and the following disclaimer. |
| 9 | * 2. Redistributions in binary form must reproduce the above copyright |
| 10 | * notice, this list of conditions and the following disclaimer in the |
| 11 | * documentation and/or other materials provided with the distribution. |
| 12 | * |
| 13 | * THIS SOFTWARE IS PROVIDED BY APPLE INC. ``AS IS'' AND ANY |
| 14 | * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 15 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR |
| 16 | * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL APPLE INC. OR |
| 17 | * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, |
| 18 | * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, |
| 19 | * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR |
| 20 | * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY |
| 21 | * OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
| 22 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
| 23 | * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 24 | */ |
| 25 | |
| 26 | #include "config.h" |
| 27 | #include "WasmStreamingParser.h" |
| 28 | |
| 29 | #if ENABLE(WEBASSEMBLY) |
| 30 | |
| 31 | #include "WasmModuleParser.h" |
| 32 | #include "WasmSectionParser.h" |
| 33 | #include <wtf/Optional.h> |
| 34 | #include <wtf/UnalignedAccess.h> |
| 35 | |
| 36 | namespace JSC { namespace Wasm { |
| 37 | |
| 38 | namespace WasmStreamingParserInternal { |
| 39 | static constexpr bool verbose = false; |
| 40 | } |
| 41 | |
| 42 | #define WASM_STREAMING_PARSER_FAIL_IF_HELPER_FAILS(helper) \ |
| 43 | do { \ |
| 44 | auto helperResult = helper; \ |
| 45 | if (UNLIKELY(!helperResult)) { \ |
| 46 | m_errorMessage = helperResult.error(); \ |
| 47 | return State::FatalError; \ |
| 48 | } \ |
| 49 | } while (0) |
| 50 | |
| 51 | ALWAYS_INLINE Optional<uint8_t> parseUInt7(const uint8_t* data, size_t& offset, size_t size) |
| 52 | { |
| 53 | if (offset >= size) |
| 54 | return false; |
| 55 | uint8_t result = data[offset++]; |
| 56 | if (result < 0x80) |
| 57 | return result; |
| 58 | return WTF::nullopt; |
| 59 | } |
| 60 | |
| 61 | template <typename ...Args> |
| 62 | NEVER_INLINE auto WARN_UNUSED_RETURN StreamingParser::fail(Args... args) -> State |
| 63 | { |
| 64 | using namespace FailureHelper; // See ADL comment in namespace above. |
| 65 | m_errorMessage = makeString("WebAssembly.Module doesn't parse at byte "_s , String::number(m_offset), ": "_s , makeString(args)...); |
| 66 | dataLogLnIf(WasmStreamingParserInternal::verbose, m_errorMessage); |
| 67 | return State::FatalError; |
| 68 | } |
| 69 | |
| 70 | StreamingParser::StreamingParser(ModuleInformation& info) |
| 71 | : m_info(info) |
| 72 | { |
| 73 | dataLogLnIf(WasmStreamingParserInternal::verbose, "starting validation" ); |
| 74 | } |
| 75 | |
| 76 | auto StreamingParser::(Vector<uint8_t>&& data) -> State |
| 77 | { |
| 78 | ASSERT(data.size() == moduleHeaderSize); |
| 79 | dataLogLnIf(WasmStreamingParserInternal::verbose, "header validation" ); |
| 80 | WASM_PARSER_FAIL_IF(data[0] != '\0' || data[1] != 'a' || data[2] != 's' || data[3] != 'm', "modules doesn't start with '\\0asm'" ); |
| 81 | uint32_t versionNumber = WTF::unalignedLoad<uint32_t>(data.data() + 4); |
| 82 | WASM_PARSER_FAIL_IF(versionNumber != expectedVersionNumber, "unexpected version number " , versionNumber, " expected " , expectedVersionNumber); |
| 83 | return State::SectionID; |
| 84 | } |
| 85 | |
| 86 | auto StreamingParser::parseSectionID(Vector<uint8_t>&& data) -> State |
| 87 | { |
| 88 | ASSERT(data.size() == sectionIDSize); |
| 89 | size_t offset = 0; |
| 90 | auto result = parseUInt7(data.data(), offset, data.size()); |
| 91 | WASM_PARSER_FAIL_IF(!result, "can't get section byte" ); |
| 92 | |
| 93 | Section section = Section::Custom; |
| 94 | WASM_PARSER_FAIL_IF(!decodeSection(*result, section), "invalid section" ); |
| 95 | ASSERT(section != Section::Begin); |
| 96 | WASM_PARSER_FAIL_IF(!validateOrder(m_previousKnownSection, section), "invalid section order, " , m_previousKnownSection, " followed by " , section); |
| 97 | m_section = section; |
| 98 | if (isKnownSection(section)) |
| 99 | m_previousKnownSection = section; |
| 100 | return State::SectionSize; |
| 101 | } |
| 102 | |
| 103 | auto StreamingParser::parseSectionSize(uint32_t sectionLength) -> State |
| 104 | { |
| 105 | m_sectionLength = sectionLength; |
| 106 | if (m_section == Section::Code) |
| 107 | return State::CodeSectionSize; |
| 108 | return State::SectionPayload; |
| 109 | } |
| 110 | |
| 111 | auto StreamingParser::parseCodeSectionSize(uint32_t functionCount) -> State |
| 112 | { |
| 113 | m_functionCount = functionCount; |
| 114 | m_functionIndex = 0; |
| 115 | m_codeOffset = m_offset; |
| 116 | |
| 117 | WASM_PARSER_FAIL_IF(functionCount == std::numeric_limits<uint32_t>::max(), "Code section's count is too big " , functionCount); |
| 118 | WASM_PARSER_FAIL_IF(functionCount != m_info->functions.size(), "Code section count " , functionCount, " exceeds the declared number of functions " , m_info->functions.size()); |
| 119 | |
| 120 | if (m_functionIndex == m_functionCount) { |
| 121 | WASM_PARSER_FAIL_IF((m_codeOffset + m_sectionLength) != m_nextOffset, "parsing ended before the end of " , m_section, " section" ); |
| 122 | return State::SectionID; |
| 123 | } |
| 124 | return State::FunctionSize; |
| 125 | } |
| 126 | |
| 127 | auto StreamingParser::parseFunctionSize(uint32_t functionSize) -> State |
| 128 | { |
| 129 | m_functionSize = functionSize; |
| 130 | WASM_PARSER_FAIL_IF(functionSize > maxFunctionSize, "Code function's size " , functionSize, " is too big" ); |
| 131 | return State::FunctionPayload; |
| 132 | } |
| 133 | |
| 134 | auto StreamingParser::parseFunctionPayload(Vector<uint8_t>&& data) -> State |
| 135 | { |
| 136 | auto& function = m_info->functions[m_functionIndex]; |
| 137 | function.start = m_offset; |
| 138 | function.end = m_offset + m_functionSize; |
| 139 | function.data = WTFMove(data); |
| 140 | dataLogLnIf(WasmStreamingParserInternal::verbose, "Processing function starting at: " , function.start, " and ending at: " , function.end); |
| 141 | ++m_functionIndex; |
| 142 | if (m_functionIndex == m_functionCount) { |
| 143 | WASM_PARSER_FAIL_IF((m_codeOffset + m_sectionLength) != (m_offset + m_functionSize), "parsing ended before the end of " , m_section, " section" ); |
| 144 | return State::SectionID; |
| 145 | } |
| 146 | return State::FunctionSize; |
| 147 | } |
| 148 | |
| 149 | auto StreamingParser::parseSectionPayload(Vector<uint8_t>&& data) -> State |
| 150 | { |
| 151 | SectionParser parser(data.data(), data.size(), m_offset, m_info.get()); |
| 152 | switch (m_section) { |
| 153 | #define WASM_SECTION_PARSE(NAME, ID, DESCRIPTION) \ |
| 154 | case Section::NAME: { \ |
| 155 | WASM_STREAMING_PARSER_FAIL_IF_HELPER_FAILS(parser.parse ## NAME()); \ |
| 156 | break; \ |
| 157 | } |
| 158 | FOR_EACH_KNOWN_WASM_SECTION(WASM_SECTION_PARSE) |
| 159 | #undef WASM_SECTION_PARSE |
| 160 | |
| 161 | case Section::Custom: { |
| 162 | WASM_STREAMING_PARSER_FAIL_IF_HELPER_FAILS(parser.parseCustom()); |
| 163 | break; |
| 164 | } |
| 165 | |
| 166 | case Section::Begin: { |
| 167 | RELEASE_ASSERT_NOT_REACHED(); |
| 168 | break; |
| 169 | } |
| 170 | } |
| 171 | |
| 172 | WASM_PARSER_FAIL_IF(parser.length() != parser.offset(), "parsing ended before the end of " , m_section, " section" ); |
| 173 | |
| 174 | return State::SectionID; |
| 175 | } |
| 176 | |
| 177 | auto StreamingParser::consume(const uint8_t* bytes, size_t bytesSize, size_t& offsetInBytes, size_t requiredSize) -> Optional<Vector<uint8_t>> |
| 178 | { |
| 179 | if (m_remaining.size() == requiredSize) { |
| 180 | Vector<uint8_t> result = WTFMove(m_remaining); |
| 181 | m_nextOffset += requiredSize; |
| 182 | return result; |
| 183 | } |
| 184 | |
| 185 | if (m_remaining.size() > requiredSize) { |
| 186 | Vector<uint8_t> result(requiredSize); |
| 187 | memcpy(result.data(), m_remaining.data(), requiredSize); |
| 188 | m_remaining.remove(0, requiredSize); |
| 189 | m_nextOffset += requiredSize; |
| 190 | return result; |
| 191 | } |
| 192 | |
| 193 | ASSERT(m_remaining.size() < requiredSize); |
| 194 | size_t bytesRemainingSize = bytesSize - offsetInBytes; |
| 195 | size_t totalDataSize = m_remaining.size() + bytesRemainingSize; |
| 196 | if (totalDataSize < requiredSize) { |
| 197 | m_remaining.append(bytes + offsetInBytes, bytesRemainingSize); |
| 198 | offsetInBytes = bytesSize; |
| 199 | return WTF::nullopt; |
| 200 | } |
| 201 | |
| 202 | size_t usedSize = requiredSize - m_remaining.size(); |
| 203 | m_remaining.append(bytes + offsetInBytes, usedSize); |
| 204 | offsetInBytes += usedSize; |
| 205 | Vector<uint8_t> result = WTFMove(m_remaining); |
| 206 | m_nextOffset += requiredSize; |
| 207 | return result; |
| 208 | } |
| 209 | |
| 210 | auto StreamingParser::consumeVarUInt32(const uint8_t* bytes, size_t bytesSize, size_t& offsetInBytes, IsEndOfStream isEndOfStream) -> Expected<uint32_t, State> |
| 211 | { |
| 212 | constexpr size_t maxSize = WTF::LEBDecoder::maxByteLength<uint32_t>(); |
| 213 | size_t bytesRemainingSize = bytesSize - offsetInBytes; |
| 214 | size_t totalDataSize = m_remaining.size() + bytesRemainingSize; |
| 215 | if (m_remaining.size() >= maxSize) { |
| 216 | // Do nothing. |
| 217 | } else if (totalDataSize >= maxSize) { |
| 218 | size_t usedSize = maxSize - m_remaining.size(); |
| 219 | m_remaining.append(bytes + offsetInBytes, usedSize); |
| 220 | offsetInBytes += usedSize; |
| 221 | } else { |
| 222 | m_remaining.append(bytes + offsetInBytes, bytesRemainingSize); |
| 223 | offsetInBytes += bytesRemainingSize; |
| 224 | // If the given bytes are the end of the stream, we try to parse VarUInt32 |
| 225 | // with the current remaining data since VarUInt32 may not require `maxSize`. |
| 226 | if (isEndOfStream == IsEndOfStream::No) |
| 227 | return makeUnexpected(m_state); |
| 228 | } |
| 229 | |
| 230 | size_t offset = 0; |
| 231 | uint32_t result = 0; |
| 232 | if (!WTF::LEBDecoder::decodeUInt32(m_remaining.data(), m_remaining.size(), offset, result)) |
| 233 | return makeUnexpected(State::FatalError); |
| 234 | size_t consumedSize = offset; |
| 235 | m_remaining.remove(0, consumedSize); |
| 236 | m_nextOffset += consumedSize; |
| 237 | return result; |
| 238 | } |
| 239 | |
| 240 | auto StreamingParser::addBytes(const uint8_t* bytes, size_t bytesSize, IsEndOfStream isEndOfStream) -> State |
| 241 | { |
| 242 | if (m_state == State::FatalError) |
| 243 | return m_state; |
| 244 | |
| 245 | m_totalSize += bytesSize; |
| 246 | if (UNLIKELY(m_totalSize.hasOverflowed() || m_totalSize.unsafeGet() > maxModuleSize)) { |
| 247 | m_state = fail("module size is too large, maximum " , maxModuleSize); |
| 248 | return m_state; |
| 249 | } |
| 250 | |
| 251 | if (UNLIKELY(Options::useEagerWebAssemblyModuleHashing())) |
| 252 | m_hasher.addBytes(bytes, bytesSize); |
| 253 | |
| 254 | size_t offsetInBytes = 0; |
| 255 | while (true) { |
| 256 | ASSERT(offsetInBytes <= bytesSize); |
| 257 | switch (m_state) { |
| 258 | case State::ModuleHeader: { |
| 259 | auto result = consume(bytes, bytesSize, offsetInBytes, moduleHeaderSize); |
| 260 | if (!result) |
| 261 | return m_state; |
| 262 | m_state = parseModuleHeader(WTFMove(*result)); |
| 263 | break; |
| 264 | } |
| 265 | |
| 266 | case State::SectionID: { |
| 267 | auto result = consume(bytes, bytesSize, offsetInBytes, sectionIDSize); |
| 268 | if (!result) |
| 269 | return m_state; |
| 270 | m_state = parseSectionID(WTFMove(*result)); |
| 271 | break; |
| 272 | } |
| 273 | |
| 274 | case State::SectionSize: { |
| 275 | auto result = consumeVarUInt32(bytes, bytesSize, offsetInBytes, isEndOfStream); |
| 276 | if (!result) { |
| 277 | if (result.error() == State::FatalError) |
| 278 | m_state = failOnState(m_state); |
| 279 | else |
| 280 | m_state = result.error(); |
| 281 | return m_state; |
| 282 | } |
| 283 | m_state = parseSectionSize(*result); |
| 284 | break; |
| 285 | } |
| 286 | |
| 287 | case State::SectionPayload: { |
| 288 | auto result = consume(bytes, bytesSize, offsetInBytes, m_sectionLength); |
| 289 | if (!result) |
| 290 | return m_state; |
| 291 | m_state = parseSectionPayload(WTFMove(*result)); |
| 292 | break; |
| 293 | } |
| 294 | |
| 295 | case State::CodeSectionSize: { |
| 296 | auto result = consumeVarUInt32(bytes, bytesSize, offsetInBytes, isEndOfStream); |
| 297 | if (!result) { |
| 298 | if (result.error() == State::FatalError) |
| 299 | m_state = failOnState(m_state); |
| 300 | else |
| 301 | m_state = result.error(); |
| 302 | return m_state; |
| 303 | } |
| 304 | m_state = parseCodeSectionSize(*result); |
| 305 | break; |
| 306 | } |
| 307 | |
| 308 | case State::FunctionSize: { |
| 309 | auto result = consumeVarUInt32(bytes, bytesSize, offsetInBytes, isEndOfStream); |
| 310 | if (!result) { |
| 311 | if (result.error() == State::FatalError) |
| 312 | m_state = failOnState(m_state); |
| 313 | else |
| 314 | m_state = result.error(); |
| 315 | return m_state; |
| 316 | } |
| 317 | m_state = parseFunctionSize(*result); |
| 318 | break; |
| 319 | } |
| 320 | |
| 321 | case State::FunctionPayload: { |
| 322 | auto result = consume(bytes, bytesSize, offsetInBytes, m_functionSize); |
| 323 | if (!result) |
| 324 | return m_state; |
| 325 | m_state = parseFunctionPayload(WTFMove(*result)); |
| 326 | break; |
| 327 | } |
| 328 | |
| 329 | case State::Finished: |
| 330 | case State::FatalError: |
| 331 | return m_state; |
| 332 | } |
| 333 | |
| 334 | m_offset = m_nextOffset; |
| 335 | } |
| 336 | } |
| 337 | |
| 338 | auto StreamingParser::failOnState(State) -> State |
| 339 | { |
| 340 | switch (m_state) { |
| 341 | case State::ModuleHeader: |
| 342 | return fail("expected a module of at least " , moduleHeaderSize, " bytes" ); |
| 343 | case State::SectionID: |
| 344 | return fail("can't get section byte" ); |
| 345 | case State::SectionSize: |
| 346 | return fail("can't get " , m_section, " section's length" ); |
| 347 | case State::SectionPayload: |
| 348 | return fail(m_section, " section of size " , m_sectionLength, " would overflow Module's size" ); |
| 349 | case State::CodeSectionSize: |
| 350 | return fail("can't get Code section's count" ); |
| 351 | case State::FunctionSize: |
| 352 | return fail("can't get " , m_functionIndex, "th Code function's size" ); |
| 353 | case State::FunctionPayload: |
| 354 | return fail("Code function's size " , m_functionSize, " exceeds the module's remaining size" ); |
| 355 | case State::Finished: |
| 356 | case State::FatalError: |
| 357 | return m_state; |
| 358 | } |
| 359 | return m_state; |
| 360 | } |
| 361 | |
| 362 | auto StreamingParser::finalize() -> State |
| 363 | { |
| 364 | addBytes(nullptr, 0, IsEndOfStream::Yes); |
| 365 | switch (m_state) { |
| 366 | case State::ModuleHeader: |
| 367 | case State::SectionSize: |
| 368 | case State::SectionPayload: |
| 369 | case State::CodeSectionSize: |
| 370 | case State::FunctionSize: |
| 371 | case State::FunctionPayload: |
| 372 | m_state = failOnState(m_state); |
| 373 | break; |
| 374 | |
| 375 | case State::Finished: |
| 376 | case State::FatalError: |
| 377 | break; |
| 378 | |
| 379 | case State::SectionID: |
| 380 | if (m_remaining.isEmpty()) { |
| 381 | if (UNLIKELY(Options::useEagerWebAssemblyModuleHashing())) |
| 382 | m_info->nameSection->setHash(m_hasher.computeHexDigest()); |
| 383 | m_state = State::Finished; |
| 384 | } else |
| 385 | m_state = failOnState(State::SectionID); |
| 386 | break; |
| 387 | } |
| 388 | return m_state; |
| 389 | } |
| 390 | |
| 391 | } } // namespace JSC::Wasm |
| 392 | |
| 393 | #endif // ENABLE(WEBASSEMBLY) |
| 394 | |