OLD | NEW |
(Empty) | |
| 1 // Copyright 2015 the V8 project authors. All rights reserved. |
| 2 // Use of this source code is governed by a BSD-style license that can be |
| 3 // found in the LICENSE file. |
| 4 |
| 5 #include "src/macro-assembler.h" |
| 6 #include "src/objects.h" |
| 7 #include "src/v8.h" |
| 8 |
| 9 #include "src/wasm/decoder.h" |
| 10 #include "src/wasm/module-decoder.h" |
| 11 |
| 12 namespace v8 { |
| 13 namespace internal { |
| 14 namespace wasm { |
| 15 |
| 16 #if DEBUG |
| 17 #define TRACE(...) \ |
| 18 do { \ |
| 19 if (FLAG_trace_wasm_decoder) PrintF(__VA_ARGS__); \ |
| 20 } while (false) |
| 21 #else |
| 22 #define TRACE(...) |
| 23 #endif |
| 24 |
| 25 |
| 26 // The main logic for decoding the bytes of a module. |
| 27 class ModuleDecoder : public Decoder { |
| 28 public: |
| 29 ModuleDecoder(Zone* zone, const byte* module_start, const byte* module_end, |
| 30 bool asm_js) |
| 31 : Decoder(module_start, module_end), module_zone(zone), asm_js_(asm_js) { |
| 32 result_.start = start_; |
| 33 if (limit_ < start_) { |
| 34 error(start_, "end is less than start"); |
| 35 limit_ = start_; |
| 36 } |
| 37 } |
| 38 |
| 39 virtual void onFirstError() { |
| 40 pc_ = limit_; // On error, terminate section decoding loop. |
| 41 } |
| 42 |
| 43 // Decodes an entire module. |
| 44 ModuleResult DecodeModule(WasmModule* module, bool verify_functions = true) { |
| 45 pc_ = start_; |
| 46 module->module_start = start_; |
| 47 module->module_end = limit_; |
| 48 module->min_mem_size_log2 = 0; |
| 49 module->max_mem_size_log2 = 0; |
| 50 module->mem_export = false; |
| 51 module->mem_external = false; |
| 52 module->globals = new std::vector<WasmGlobal>(); |
| 53 module->signatures = new std::vector<FunctionSig*>(); |
| 54 module->functions = new std::vector<WasmFunction>(); |
| 55 module->data_segments = new std::vector<WasmDataSegment>(); |
| 56 module->function_table = new std::vector<uint16_t>(); |
| 57 |
| 58 bool sections[kMaxModuleSectionCode]; |
| 59 memset(sections, 0, sizeof(sections)); |
| 60 |
| 61 // Decode the module sections. |
| 62 while (pc_ < limit_) { |
| 63 TRACE("DecodeSection\n"); |
| 64 WasmSectionDeclCode section = |
| 65 static_cast<WasmSectionDeclCode>(u8("section")); |
| 66 // Each section should appear at most once. |
| 67 if (section < kMaxModuleSectionCode) { |
| 68 CheckForPreviousSection(sections, section, false); |
| 69 sections[section] = true; |
| 70 } |
| 71 |
| 72 switch (section) { |
| 73 case kDeclEnd: |
| 74 // Terminate section decoding. |
| 75 limit_ = pc_; |
| 76 break; |
| 77 case kDeclMemory: |
| 78 module->min_mem_size_log2 = u8("min memory"); |
| 79 module->max_mem_size_log2 = u8("max memory"); |
| 80 module->mem_export = u8("export memory") != 0; |
| 81 break; |
| 82 case kDeclSignatures: { |
| 83 int length; |
| 84 uint32_t signatures_count = u32v(&length, "signatures count"); |
| 85 module->signatures->reserve(SafeReserve(signatures_count)); |
| 86 // Decode signatures. |
| 87 for (uint32_t i = 0; i < signatures_count; i++) { |
| 88 if (failed()) break; |
| 89 TRACE("DecodeSignature[%d] module+%d\n", i, |
| 90 static_cast<int>(pc_ - start_)); |
| 91 FunctionSig* s = sig(); // read function sig. |
| 92 module->signatures->push_back(s); |
| 93 } |
| 94 break; |
| 95 } |
| 96 case kDeclFunctions: { |
| 97 // Functions require a signature table first. |
| 98 CheckForPreviousSection(sections, kDeclSignatures, true); |
| 99 int length; |
| 100 uint32_t functions_count = u32v(&length, "functions count"); |
| 101 module->functions->reserve(SafeReserve(functions_count)); |
| 102 // Set up module environment for verification. |
| 103 ModuleEnv menv; |
| 104 menv.module = module; |
| 105 menv.globals_area = 0; |
| 106 menv.mem_start = 0; |
| 107 menv.mem_end = 0; |
| 108 menv.function_code = nullptr; |
| 109 menv.asm_js = asm_js_; |
| 110 // Decode functions. |
| 111 for (uint32_t i = 0; i < functions_count; i++) { |
| 112 if (failed()) break; |
| 113 TRACE("DecodeFunction[%d] module+%d\n", i, |
| 114 static_cast<int>(pc_ - start_)); |
| 115 |
| 116 module->functions->push_back( |
| 117 {nullptr, 0, 0, 0, 0, 0, 0, false, false}); |
| 118 WasmFunction* function = &module->functions->back(); |
| 119 DecodeFunctionInModule(module, function, false); |
| 120 } |
| 121 if (ok() && verify_functions) { |
| 122 for (uint32_t i = 0; i < functions_count; i++) { |
| 123 if (failed()) break; |
| 124 WasmFunction* function = &module->functions->at(i); |
| 125 if (!function->external) { |
| 126 VerifyFunctionBody(i, &menv, function); |
| 127 if (result_.failed()) |
| 128 error(result_.error_pc, result_.error_msg.get()); |
| 129 } |
| 130 } |
| 131 } |
| 132 break; |
| 133 } |
| 134 case kDeclGlobals: { |
| 135 int length; |
| 136 uint32_t globals_count = u32v(&length, "globals count"); |
| 137 module->globals->reserve(SafeReserve(globals_count)); |
| 138 // Decode globals. |
| 139 for (uint32_t i = 0; i < globals_count; i++) { |
| 140 if (failed()) break; |
| 141 TRACE("DecodeGlobal[%d] module+%d\n", i, |
| 142 static_cast<int>(pc_ - start_)); |
| 143 module->globals->push_back({0, MachineType::Int32(), 0, false}); |
| 144 WasmGlobal* global = &module->globals->back(); |
| 145 DecodeGlobalInModule(global); |
| 146 } |
| 147 break; |
| 148 } |
| 149 case kDeclDataSegments: { |
| 150 int length; |
| 151 uint32_t data_segments_count = u32v(&length, "data segments count"); |
| 152 module->data_segments->reserve(SafeReserve(data_segments_count)); |
| 153 // Decode data segments. |
| 154 for (uint32_t i = 0; i < data_segments_count; i++) { |
| 155 if (failed()) break; |
| 156 TRACE("DecodeDataSegment[%d] module+%d\n", i, |
| 157 static_cast<int>(pc_ - start_)); |
| 158 module->data_segments->push_back({0, 0, 0}); |
| 159 WasmDataSegment* segment = &module->data_segments->back(); |
| 160 DecodeDataSegmentInModule(segment); |
| 161 } |
| 162 break; |
| 163 } |
| 164 case kDeclFunctionTable: { |
| 165 // An indirect function table requires functions first. |
| 166 CheckForPreviousSection(sections, kDeclFunctions, true); |
| 167 int length; |
| 168 uint32_t function_table_count = u32v(&length, "function table count"); |
| 169 module->function_table->reserve(SafeReserve(function_table_count)); |
| 170 // Decode function table. |
| 171 for (uint32_t i = 0; i < function_table_count; i++) { |
| 172 if (failed()) break; |
| 173 TRACE("DecodeFunctionTable[%d] module+%d\n", i, |
| 174 static_cast<int>(pc_ - start_)); |
| 175 uint16_t index = u16(); |
| 176 if (index >= module->functions->size()) { |
| 177 error(pc_ - 2, "invalid function index"); |
| 178 break; |
| 179 } |
| 180 module->function_table->push_back(index); |
| 181 } |
| 182 break; |
| 183 } |
| 184 default: |
| 185 error(pc_ - 1, nullptr, "unrecognized section 0x%02x", section); |
| 186 break; |
| 187 } |
| 188 } |
| 189 |
| 190 return toResult(module); |
| 191 } |
| 192 |
| 193 uint32_t SafeReserve(uint32_t count) { |
| 194 // Avoid OOM by only reserving up to a certain size. |
| 195 const uint32_t kMaxReserve = 20000; |
| 196 return count < kMaxReserve ? count : kMaxReserve; |
| 197 } |
| 198 |
| 199 void CheckForPreviousSection(bool* sections, WasmSectionDeclCode section, |
| 200 bool present) { |
| 201 if (section >= kMaxModuleSectionCode) return; |
| 202 if (sections[section] == present) return; |
| 203 const char* name = ""; |
| 204 switch (section) { |
| 205 case kDeclMemory: |
| 206 name = "memory"; |
| 207 break; |
| 208 case kDeclSignatures: |
| 209 name = "signatures"; |
| 210 break; |
| 211 case kDeclFunctions: |
| 212 name = "function declaration"; |
| 213 break; |
| 214 case kDeclGlobals: |
| 215 name = "global variable"; |
| 216 break; |
| 217 case kDeclDataSegments: |
| 218 name = "data segment"; |
| 219 break; |
| 220 case kDeclFunctionTable: |
| 221 name = "function table"; |
| 222 break; |
| 223 default: |
| 224 name = ""; |
| 225 break; |
| 226 } |
| 227 if (present) { |
| 228 error(pc_ - 1, nullptr, "required %s section missing", name); |
| 229 } else { |
| 230 error(pc_ - 1, nullptr, "%s section already present", name); |
| 231 } |
| 232 } |
| 233 |
| 234 // Decodes a single anonymous function starting at {start_}. |
| 235 FunctionResult DecodeSingleFunction(ModuleEnv* module_env, |
| 236 WasmFunction* function) { |
| 237 pc_ = start_; |
| 238 function->sig = sig(); // read signature |
| 239 function->name_offset = 0; // ---- name |
| 240 function->code_start_offset = off(pc_ + 8); // ---- code start |
| 241 function->code_end_offset = off(limit_); // ---- code end |
| 242 function->local_int32_count = u16(); // read u16 |
| 243 function->local_int64_count = u16(); // read u16 |
| 244 function->local_float32_count = u16(); // read u16 |
| 245 function->local_float64_count = u16(); // read u16 |
| 246 function->exported = false; // ---- exported |
| 247 function->external = false; // ---- external |
| 248 |
| 249 if (ok()) VerifyFunctionBody(0, module_env, function); |
| 250 |
| 251 FunctionResult result; |
| 252 result.CopyFrom(result_); // Copy error code and location. |
| 253 result.val = function; |
| 254 return result; |
| 255 } |
| 256 |
| 257 // Decodes a single function signature at {start}. |
| 258 FunctionSig* DecodeFunctionSignature(const byte* start) { |
| 259 pc_ = start; |
| 260 FunctionSig* result = sig(); |
| 261 return ok() ? result : nullptr; |
| 262 } |
| 263 |
| 264 private: |
| 265 Zone* module_zone; |
| 266 ModuleResult result_; |
| 267 bool asm_js_; |
| 268 |
| 269 uint32_t off(const byte* ptr) { return static_cast<uint32_t>(ptr - start_); } |
| 270 |
| 271 // Decodes a single global entry inside a module starting at {pc_}. |
| 272 void DecodeGlobalInModule(WasmGlobal* global) { |
| 273 global->name_offset = string("global name"); |
| 274 global->type = mem_type(); |
| 275 global->offset = 0; |
| 276 global->exported = u8("exported") != 0; |
| 277 } |
| 278 |
| 279 // Decodes a single function entry inside a module starting at {pc_}. |
| 280 void DecodeFunctionInModule(WasmModule* module, WasmFunction* function, |
| 281 bool verify_body = true) { |
| 282 byte decl_bits = u8("function decl"); |
| 283 |
| 284 const byte* sigpos = pc_; |
| 285 function->sig_index = u16("signature index"); |
| 286 |
| 287 if (function->sig_index >= module->signatures->size()) { |
| 288 return error(sigpos, "invalid signature index"); |
| 289 } else { |
| 290 function->sig = module->signatures->at(function->sig_index); |
| 291 } |
| 292 |
| 293 TRACE(" +%d <function attributes:%s%s%s%s%s>\n", |
| 294 static_cast<int>(pc_ - start_), |
| 295 decl_bits & kDeclFunctionName ? " name" : "", |
| 296 decl_bits & kDeclFunctionImport ? " imported" : "", |
| 297 decl_bits & kDeclFunctionLocals ? " locals" : "", |
| 298 decl_bits & kDeclFunctionExport ? " exported" : "", |
| 299 (decl_bits & kDeclFunctionImport) == 0 ? " body" : ""); |
| 300 |
| 301 if (decl_bits & kDeclFunctionName) { |
| 302 function->name_offset = string("function name"); |
| 303 } |
| 304 |
| 305 function->exported = decl_bits & kDeclFunctionExport; |
| 306 |
| 307 // Imported functions have no locals or body. |
| 308 if (decl_bits & kDeclFunctionImport) { |
| 309 function->external = true; |
| 310 return; |
| 311 } |
| 312 |
| 313 if (decl_bits & kDeclFunctionLocals) { |
| 314 function->local_int32_count = u16("int32 count"); |
| 315 function->local_int64_count = u16("int64 count"); |
| 316 function->local_float32_count = u16("float32 count"); |
| 317 function->local_float64_count = u16("float64 count"); |
| 318 } |
| 319 |
| 320 uint16_t size = u16("body size"); |
| 321 if (ok()) { |
| 322 if ((pc_ + size) > limit_) { |
| 323 return error(pc_, limit_, |
| 324 "expected %d bytes for function body, fell off end", size); |
| 325 } |
| 326 function->code_start_offset = static_cast<uint32_t>(pc_ - start_); |
| 327 function->code_end_offset = function->code_start_offset + size; |
| 328 TRACE(" +%d %-20s: (%d bytes)\n", static_cast<int>(pc_ - start_), |
| 329 "function body", size); |
| 330 pc_ += size; |
| 331 } |
| 332 } |
| 333 |
| 334 // Decodes a single data segment entry inside a module starting at {pc_}. |
| 335 void DecodeDataSegmentInModule(WasmDataSegment* segment) { |
| 336 segment->dest_addr = |
| 337 u32("destination"); // TODO(titzer): check it's within the memory size. |
| 338 segment->source_offset = offset("source offset"); |
| 339 segment->source_size = |
| 340 u32("source size"); // TODO(titzer): check the size is reasonable. |
| 341 segment->init = u8("init"); |
| 342 } |
| 343 |
| 344 // Verifies the body (code) of a given function. |
| 345 void VerifyFunctionBody(uint32_t func_num, ModuleEnv* menv, |
| 346 WasmFunction* function) { |
| 347 if (FLAG_trace_wasm_decode_time) { |
| 348 // TODO(titzer): clean me up a bit. |
| 349 OFStream os(stdout); |
| 350 os << "Verifying WASM function:"; |
| 351 if (function->name_offset > 0) { |
| 352 os << menv->module->GetName(function->name_offset); |
| 353 } |
| 354 os << std::endl; |
| 355 } |
| 356 FunctionEnv fenv; |
| 357 fenv.module = menv; |
| 358 fenv.sig = function->sig; |
| 359 fenv.local_int32_count = function->local_int32_count; |
| 360 fenv.local_int64_count = function->local_int64_count; |
| 361 fenv.local_float32_count = function->local_float32_count; |
| 362 fenv.local_float64_count = function->local_float64_count; |
| 363 fenv.SumLocals(); |
| 364 |
| 365 TreeResult result = |
| 366 VerifyWasmCode(&fenv, start_, start_ + function->code_start_offset, |
| 367 start_ + function->code_end_offset); |
| 368 if (result.failed()) { |
| 369 // Wrap the error message from the function decoder. |
| 370 std::ostringstream str; |
| 371 str << "in function #" << func_num << ": "; |
| 372 // TODO(titzer): add function name for the user? |
| 373 str << result; |
| 374 const char* raw = str.str().c_str(); |
| 375 size_t len = strlen(raw); |
| 376 char* buffer = new char[len]; |
| 377 strncpy(buffer, raw, len); |
| 378 buffer[len - 1] = 0; |
| 379 |
| 380 // Copy error code and location. |
| 381 result_.CopyFrom(result); |
| 382 result_.error_msg.Reset(buffer); |
| 383 } |
| 384 } |
| 385 |
| 386 // Reads a single 32-bit unsigned integer interpreted as an offset, checking |
| 387 // the offset is within bounds and advances. |
| 388 uint32_t offset(const char* name = nullptr) { |
| 389 uint32_t offset = u32(name ? name : "offset"); |
| 390 if (offset > (limit_ - start_)) { |
| 391 error(pc_ - sizeof(uint32_t), "offset out of bounds of module"); |
| 392 } |
| 393 return offset; |
| 394 } |
| 395 |
| 396 // Reads a single 32-bit unsigned integer interpreted as an offset into the |
| 397 // data and validating the string there and advances. |
| 398 uint32_t string(const char* name = nullptr) { |
| 399 return offset(name ? name : "string"); // TODO(titzer): validate string |
| 400 } |
| 401 |
| 402 // Reads a single 8-bit integer, interpreting it as a local type. |
| 403 LocalType local_type() { |
| 404 byte val = u8("local type"); |
| 405 LocalTypeCode t = static_cast<LocalTypeCode>(val); |
| 406 switch (t) { |
| 407 case kLocalVoid: |
| 408 return kAstStmt; |
| 409 case kLocalI32: |
| 410 return kAstI32; |
| 411 case kLocalI64: |
| 412 return kAstI64; |
| 413 case kLocalF32: |
| 414 return kAstF32; |
| 415 case kLocalF64: |
| 416 return kAstF64; |
| 417 default: |
| 418 error(pc_ - 1, "invalid local type"); |
| 419 return kAstStmt; |
| 420 } |
| 421 } |
| 422 |
| 423 // Reads a single 8-bit integer, interpreting it as a memory type. |
| 424 MachineType mem_type() { |
| 425 byte val = u8("memory type"); |
| 426 MemTypeCode t = static_cast<MemTypeCode>(val); |
| 427 switch (t) { |
| 428 case kMemI8: |
| 429 return MachineType::Int8(); |
| 430 case kMemU8: |
| 431 return MachineType::Uint8(); |
| 432 case kMemI16: |
| 433 return MachineType::Int16(); |
| 434 case kMemU16: |
| 435 return MachineType::Uint16(); |
| 436 case kMemI32: |
| 437 return MachineType::Int32(); |
| 438 case kMemU32: |
| 439 return MachineType::Uint32(); |
| 440 case kMemI64: |
| 441 return MachineType::Int64(); |
| 442 case kMemU64: |
| 443 return MachineType::Uint64(); |
| 444 case kMemF32: |
| 445 return MachineType::Float32(); |
| 446 case kMemF64: |
| 447 return MachineType::Float64(); |
| 448 default: |
| 449 error(pc_ - 1, "invalid memory type"); |
| 450 return MachineType::None(); |
| 451 } |
| 452 } |
| 453 |
| 454 // Parses an inline function signature. |
| 455 FunctionSig* sig() { |
| 456 byte count = u8("param count"); |
| 457 LocalType ret = local_type(); |
| 458 FunctionSig::Builder builder(module_zone, ret == kAstStmt ? 0 : 1, count); |
| 459 if (ret != kAstStmt) builder.AddReturn(ret); |
| 460 |
| 461 for (int i = 0; i < count; i++) { |
| 462 LocalType param = local_type(); |
| 463 if (param == kAstStmt) error(pc_ - 1, "invalid void parameter type"); |
| 464 builder.AddParam(param); |
| 465 } |
| 466 return builder.Build(); |
| 467 } |
| 468 }; |
| 469 |
| 470 |
| 471 // Helpers for nice error messages. |
| 472 class ModuleError : public ModuleResult { |
| 473 public: |
| 474 explicit ModuleError(const char* msg) { |
| 475 error_code = kError; |
| 476 size_t len = strlen(msg) + 1; |
| 477 char* result = new char[len]; |
| 478 strncpy(result, msg, len); |
| 479 result[len - 1] = 0; |
| 480 error_msg.Reset(result); |
| 481 } |
| 482 }; |
| 483 |
| 484 |
| 485 // Helpers for nice error messages. |
| 486 class FunctionError : public FunctionResult { |
| 487 public: |
| 488 explicit FunctionError(const char* msg) { |
| 489 error_code = kError; |
| 490 size_t len = strlen(msg) + 1; |
| 491 char* result = new char[len]; |
| 492 strncpy(result, msg, len); |
| 493 result[len - 1] = 0; |
| 494 error_msg.Reset(result); |
| 495 } |
| 496 }; |
| 497 |
| 498 |
| 499 ModuleResult DecodeWasmModule(Isolate* isolate, Zone* zone, |
| 500 const byte* module_start, const byte* module_end, |
| 501 bool verify_functions, bool asm_js) { |
| 502 size_t size = module_end - module_start; |
| 503 if (module_start > module_end) return ModuleError("start > end"); |
| 504 if (size >= kMaxModuleSize) return ModuleError("size > maximum module size"); |
| 505 WasmModule* module = new WasmModule(); |
| 506 ModuleDecoder decoder(zone, module_start, module_end, asm_js); |
| 507 return decoder.DecodeModule(module, verify_functions); |
| 508 } |
| 509 |
| 510 |
| 511 FunctionSig* DecodeWasmSignatureForTesting(Zone* zone, const byte* start, |
| 512 const byte* end) { |
| 513 ModuleDecoder decoder(zone, start, end, false); |
| 514 return decoder.DecodeFunctionSignature(start); |
| 515 } |
| 516 |
| 517 |
| 518 FunctionResult DecodeWasmFunction(Isolate* isolate, Zone* zone, |
| 519 ModuleEnv* module_env, |
| 520 const byte* function_start, |
| 521 const byte* function_end) { |
| 522 size_t size = function_end - function_start; |
| 523 if (function_start > function_end) return FunctionError("start > end"); |
| 524 if (size > kMaxFunctionSize) |
| 525 return FunctionError("size > maximum function size"); |
| 526 WasmFunction* function = new WasmFunction(); |
| 527 ModuleDecoder decoder(zone, function_start, function_end, false); |
| 528 return decoder.DecodeSingleFunction(module_env, function); |
| 529 } |
| 530 } // namespace wasm |
| 531 } // namespace internal |
| 532 } // namespace v8 |
OLD | NEW |