1 //===-- llvm-bcanalyzer.cpp - Bitcode Analyzer --------------------------===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 // 10 // This tool may be invoked in the following manner: 11 // llvm-bcanalyzer [options] - Read LLVM bitcode from stdin 12 // llvm-bcanalyzer [options] x.bc - Read LLVM bitcode from the x.bc file 13 // 14 // Options: 15 // --help - Output information about command line switches 16 // --dump - Dump low-level bitcode structure in readable format 17 // 18 // This tool provides analytical information about a bitcode file. It is 19 // intended as an aid to developers of bitcode reading and writing software. It 20 // produces on std::out a summary of the bitcode file that shows various 21 // statistics about the contents of the file. By default this information is 22 // detailed and contains information about individual bitcode blocks and the 23 // functions in the module. 24 // The tool is also able to print a bitcode file in a straight forward text 25 // format that shows the containment and relationships of the information in 26 // the bitcode file (-dump option). 27 // 28 //===----------------------------------------------------------------------===// 29 30 #include "llvm/ADT/StringExtras.h" 31 #include "llvm/Bitcode/BitcodeReader.h" 32 #include "llvm/Bitcode/BitstreamReader.h" 33 #include "llvm/Bitcode/LLVMBitCodes.h" 34 #include "llvm/Support/CommandLine.h" 35 #include "llvm/Support/Format.h" 36 #include "llvm/Support/InitLLVM.h" 37 #include "llvm/Support/ManagedStatic.h" 38 #include "llvm/Support/MemoryBuffer.h" 39 #include "llvm/Support/SHA1.h" 40 #include "llvm/Support/WithColor.h" 41 #include "llvm/Support/raw_ostream.h" 42 using namespace llvm; 43 44 static cl::opt<std::string> 45 InputFilename(cl::Positional, cl::desc("<input bitcode>"), cl::init("-")); 46 47 static cl::opt<bool> Dump("dump", cl::desc("Dump low level bitcode trace")); 48 49 //===----------------------------------------------------------------------===// 50 // Bitcode specific analysis. 51 //===----------------------------------------------------------------------===// 52 53 static cl::opt<bool> NoHistogram("disable-histogram", 54 cl::desc("Do not print per-code histogram")); 55 56 static cl::opt<bool> 57 NonSymbolic("non-symbolic", 58 cl::desc("Emit numeric info in dump even if" 59 " symbolic info is available")); 60 61 static cl::opt<std::string> 62 BlockInfoFilename("block-info", 63 cl::desc("Use the BLOCK_INFO from the given file")); 64 65 static cl::opt<bool> 66 ShowBinaryBlobs("show-binary-blobs", 67 cl::desc("Print binary blobs using hex escapes")); 68 69 static cl::opt<std::string> CheckHash( 70 "check-hash", 71 cl::desc("Check module hash using the argument as a string table")); 72 73 namespace { 74 75 /// CurStreamTypeType - A type for CurStreamType 76 enum CurStreamTypeType { 77 UnknownBitstream, 78 LLVMIRBitstream, 79 ClangSerializedASTBitstream, 80 ClangSerializedDiagnosticsBitstream, 81 }; 82 83 } 84 85 /// GetBlockName - Return a symbolic block name if known, otherwise return 86 /// null. 87 static const char *GetBlockName(unsigned BlockID, 88 const BitstreamBlockInfo &BlockInfo, 89 CurStreamTypeType CurStreamType) { 90 // Standard blocks for all bitcode files. 91 if (BlockID < bitc::FIRST_APPLICATION_BLOCKID) { 92 if (BlockID == bitc::BLOCKINFO_BLOCK_ID) 93 return "BLOCKINFO_BLOCK"; 94 return nullptr; 95 } 96 97 // Check to see if we have a blockinfo record for this block, with a name. 98 if (const BitstreamBlockInfo::BlockInfo *Info = 99 BlockInfo.getBlockInfo(BlockID)) { 100 if (!Info->Name.empty()) 101 return Info->Name.c_str(); 102 } 103 104 105 if (CurStreamType != LLVMIRBitstream) return nullptr; 106 107 switch (BlockID) { 108 default: return nullptr; 109 case bitc::OPERAND_BUNDLE_TAGS_BLOCK_ID: return "OPERAND_BUNDLE_TAGS_BLOCK"; 110 case bitc::MODULE_BLOCK_ID: return "MODULE_BLOCK"; 111 case bitc::PARAMATTR_BLOCK_ID: return "PARAMATTR_BLOCK"; 112 case bitc::PARAMATTR_GROUP_BLOCK_ID: return "PARAMATTR_GROUP_BLOCK_ID"; 113 case bitc::TYPE_BLOCK_ID_NEW: return "TYPE_BLOCK_ID"; 114 case bitc::CONSTANTS_BLOCK_ID: return "CONSTANTS_BLOCK"; 115 case bitc::FUNCTION_BLOCK_ID: return "FUNCTION_BLOCK"; 116 case bitc::IDENTIFICATION_BLOCK_ID: 117 return "IDENTIFICATION_BLOCK_ID"; 118 case bitc::VALUE_SYMTAB_BLOCK_ID: return "VALUE_SYMTAB"; 119 case bitc::METADATA_BLOCK_ID: return "METADATA_BLOCK"; 120 case bitc::METADATA_KIND_BLOCK_ID: return "METADATA_KIND_BLOCK"; 121 case bitc::METADATA_ATTACHMENT_ID: return "METADATA_ATTACHMENT_BLOCK"; 122 case bitc::USELIST_BLOCK_ID: return "USELIST_BLOCK_ID"; 123 case bitc::GLOBALVAL_SUMMARY_BLOCK_ID: 124 return "GLOBALVAL_SUMMARY_BLOCK"; 125 case bitc::FULL_LTO_GLOBALVAL_SUMMARY_BLOCK_ID: 126 return "FULL_LTO_GLOBALVAL_SUMMARY_BLOCK"; 127 case bitc::MODULE_STRTAB_BLOCK_ID: return "MODULE_STRTAB_BLOCK"; 128 case bitc::STRTAB_BLOCK_ID: return "STRTAB_BLOCK"; 129 case bitc::SYMTAB_BLOCK_ID: return "SYMTAB_BLOCK"; 130 } 131 } 132 133 /// GetCodeName - Return a symbolic code name if known, otherwise return 134 /// null. 135 static const char *GetCodeName(unsigned CodeID, unsigned BlockID, 136 const BitstreamBlockInfo &BlockInfo, 137 CurStreamTypeType CurStreamType) { 138 // Standard blocks for all bitcode files. 139 if (BlockID < bitc::FIRST_APPLICATION_BLOCKID) { 140 if (BlockID == bitc::BLOCKINFO_BLOCK_ID) { 141 switch (CodeID) { 142 default: return nullptr; 143 case bitc::BLOCKINFO_CODE_SETBID: return "SETBID"; 144 case bitc::BLOCKINFO_CODE_BLOCKNAME: return "BLOCKNAME"; 145 case bitc::BLOCKINFO_CODE_SETRECORDNAME: return "SETRECORDNAME"; 146 } 147 } 148 return nullptr; 149 } 150 151 // Check to see if we have a blockinfo record for this record, with a name. 152 if (const BitstreamBlockInfo::BlockInfo *Info = 153 BlockInfo.getBlockInfo(BlockID)) { 154 for (unsigned i = 0, e = Info->RecordNames.size(); i != e; ++i) 155 if (Info->RecordNames[i].first == CodeID) 156 return Info->RecordNames[i].second.c_str(); 157 } 158 159 160 if (CurStreamType != LLVMIRBitstream) return nullptr; 161 162 #define STRINGIFY_CODE(PREFIX, CODE) \ 163 case bitc::PREFIX##_##CODE: \ 164 return #CODE; 165 switch (BlockID) { 166 default: return nullptr; 167 case bitc::MODULE_BLOCK_ID: 168 switch (CodeID) { 169 default: return nullptr; 170 STRINGIFY_CODE(MODULE_CODE, VERSION) 171 STRINGIFY_CODE(MODULE_CODE, TRIPLE) 172 STRINGIFY_CODE(MODULE_CODE, DATALAYOUT) 173 STRINGIFY_CODE(MODULE_CODE, ASM) 174 STRINGIFY_CODE(MODULE_CODE, SECTIONNAME) 175 STRINGIFY_CODE(MODULE_CODE, DEPLIB) // FIXME: Remove in 4.0 176 STRINGIFY_CODE(MODULE_CODE, GLOBALVAR) 177 STRINGIFY_CODE(MODULE_CODE, FUNCTION) 178 STRINGIFY_CODE(MODULE_CODE, ALIAS) 179 STRINGIFY_CODE(MODULE_CODE, GCNAME) 180 STRINGIFY_CODE(MODULE_CODE, VSTOFFSET) 181 STRINGIFY_CODE(MODULE_CODE, METADATA_VALUES_UNUSED) 182 STRINGIFY_CODE(MODULE_CODE, SOURCE_FILENAME) 183 STRINGIFY_CODE(MODULE_CODE, HASH) 184 } 185 case bitc::IDENTIFICATION_BLOCK_ID: 186 switch (CodeID) { 187 default: 188 return nullptr; 189 STRINGIFY_CODE(IDENTIFICATION_CODE, STRING) 190 STRINGIFY_CODE(IDENTIFICATION_CODE, EPOCH) 191 } 192 case bitc::PARAMATTR_BLOCK_ID: 193 switch (CodeID) { 194 default: return nullptr; 195 // FIXME: Should these be different? 196 case bitc::PARAMATTR_CODE_ENTRY_OLD: return "ENTRY"; 197 case bitc::PARAMATTR_CODE_ENTRY: return "ENTRY"; 198 } 199 case bitc::PARAMATTR_GROUP_BLOCK_ID: 200 switch (CodeID) { 201 default: return nullptr; 202 case bitc::PARAMATTR_GRP_CODE_ENTRY: return "ENTRY"; 203 } 204 case bitc::TYPE_BLOCK_ID_NEW: 205 switch (CodeID) { 206 default: return nullptr; 207 STRINGIFY_CODE(TYPE_CODE, NUMENTRY) 208 STRINGIFY_CODE(TYPE_CODE, VOID) 209 STRINGIFY_CODE(TYPE_CODE, FLOAT) 210 STRINGIFY_CODE(TYPE_CODE, DOUBLE) 211 STRINGIFY_CODE(TYPE_CODE, LABEL) 212 STRINGIFY_CODE(TYPE_CODE, OPAQUE) 213 STRINGIFY_CODE(TYPE_CODE, INTEGER) 214 STRINGIFY_CODE(TYPE_CODE, POINTER) 215 STRINGIFY_CODE(TYPE_CODE, ARRAY) 216 STRINGIFY_CODE(TYPE_CODE, VECTOR) 217 STRINGIFY_CODE(TYPE_CODE, X86_FP80) 218 STRINGIFY_CODE(TYPE_CODE, FP128) 219 STRINGIFY_CODE(TYPE_CODE, PPC_FP128) 220 STRINGIFY_CODE(TYPE_CODE, METADATA) 221 STRINGIFY_CODE(TYPE_CODE, STRUCT_ANON) 222 STRINGIFY_CODE(TYPE_CODE, STRUCT_NAME) 223 STRINGIFY_CODE(TYPE_CODE, STRUCT_NAMED) 224 STRINGIFY_CODE(TYPE_CODE, FUNCTION) 225 } 226 227 case bitc::CONSTANTS_BLOCK_ID: 228 switch (CodeID) { 229 default: return nullptr; 230 STRINGIFY_CODE(CST_CODE, SETTYPE) 231 STRINGIFY_CODE(CST_CODE, NULL) 232 STRINGIFY_CODE(CST_CODE, UNDEF) 233 STRINGIFY_CODE(CST_CODE, INTEGER) 234 STRINGIFY_CODE(CST_CODE, WIDE_INTEGER) 235 STRINGIFY_CODE(CST_CODE, FLOAT) 236 STRINGIFY_CODE(CST_CODE, AGGREGATE) 237 STRINGIFY_CODE(CST_CODE, STRING) 238 STRINGIFY_CODE(CST_CODE, CSTRING) 239 STRINGIFY_CODE(CST_CODE, CE_BINOP) 240 STRINGIFY_CODE(CST_CODE, CE_CAST) 241 STRINGIFY_CODE(CST_CODE, CE_GEP) 242 STRINGIFY_CODE(CST_CODE, CE_INBOUNDS_GEP) 243 STRINGIFY_CODE(CST_CODE, CE_SELECT) 244 STRINGIFY_CODE(CST_CODE, CE_EXTRACTELT) 245 STRINGIFY_CODE(CST_CODE, CE_INSERTELT) 246 STRINGIFY_CODE(CST_CODE, CE_SHUFFLEVEC) 247 STRINGIFY_CODE(CST_CODE, CE_CMP) 248 STRINGIFY_CODE(CST_CODE, INLINEASM) 249 STRINGIFY_CODE(CST_CODE, CE_SHUFVEC_EX) 250 case bitc::CST_CODE_BLOCKADDRESS: return "CST_CODE_BLOCKADDRESS"; 251 STRINGIFY_CODE(CST_CODE, DATA) 252 } 253 case bitc::FUNCTION_BLOCK_ID: 254 switch (CodeID) { 255 default: return nullptr; 256 STRINGIFY_CODE(FUNC_CODE, DECLAREBLOCKS) 257 STRINGIFY_CODE(FUNC_CODE, INST_BINOP) 258 STRINGIFY_CODE(FUNC_CODE, INST_CAST) 259 STRINGIFY_CODE(FUNC_CODE, INST_GEP_OLD) 260 STRINGIFY_CODE(FUNC_CODE, INST_INBOUNDS_GEP_OLD) 261 STRINGIFY_CODE(FUNC_CODE, INST_SELECT) 262 STRINGIFY_CODE(FUNC_CODE, INST_EXTRACTELT) 263 STRINGIFY_CODE(FUNC_CODE, INST_INSERTELT) 264 STRINGIFY_CODE(FUNC_CODE, INST_SHUFFLEVEC) 265 STRINGIFY_CODE(FUNC_CODE, INST_CMP) 266 STRINGIFY_CODE(FUNC_CODE, INST_RET) 267 STRINGIFY_CODE(FUNC_CODE, INST_BR) 268 STRINGIFY_CODE(FUNC_CODE, INST_SWITCH) 269 STRINGIFY_CODE(FUNC_CODE, INST_INVOKE) 270 STRINGIFY_CODE(FUNC_CODE, INST_UNREACHABLE) 271 STRINGIFY_CODE(FUNC_CODE, INST_CLEANUPRET) 272 STRINGIFY_CODE(FUNC_CODE, INST_CATCHRET) 273 STRINGIFY_CODE(FUNC_CODE, INST_CATCHPAD) 274 STRINGIFY_CODE(FUNC_CODE, INST_PHI) 275 STRINGIFY_CODE(FUNC_CODE, INST_ALLOCA) 276 STRINGIFY_CODE(FUNC_CODE, INST_LOAD) 277 STRINGIFY_CODE(FUNC_CODE, INST_VAARG) 278 STRINGIFY_CODE(FUNC_CODE, INST_STORE) 279 STRINGIFY_CODE(FUNC_CODE, INST_EXTRACTVAL) 280 STRINGIFY_CODE(FUNC_CODE, INST_INSERTVAL) 281 STRINGIFY_CODE(FUNC_CODE, INST_CMP2) 282 STRINGIFY_CODE(FUNC_CODE, INST_VSELECT) 283 STRINGIFY_CODE(FUNC_CODE, DEBUG_LOC_AGAIN) 284 STRINGIFY_CODE(FUNC_CODE, INST_CALL) 285 STRINGIFY_CODE(FUNC_CODE, DEBUG_LOC) 286 STRINGIFY_CODE(FUNC_CODE, INST_GEP) 287 STRINGIFY_CODE(FUNC_CODE, OPERAND_BUNDLE) 288 } 289 case bitc::VALUE_SYMTAB_BLOCK_ID: 290 switch (CodeID) { 291 default: return nullptr; 292 STRINGIFY_CODE(VST_CODE, ENTRY) 293 STRINGIFY_CODE(VST_CODE, BBENTRY) 294 STRINGIFY_CODE(VST_CODE, FNENTRY) 295 STRINGIFY_CODE(VST_CODE, COMBINED_ENTRY) 296 } 297 case bitc::MODULE_STRTAB_BLOCK_ID: 298 switch (CodeID) { 299 default: 300 return nullptr; 301 STRINGIFY_CODE(MST_CODE, ENTRY) 302 STRINGIFY_CODE(MST_CODE, HASH) 303 } 304 case bitc::GLOBALVAL_SUMMARY_BLOCK_ID: 305 case bitc::FULL_LTO_GLOBALVAL_SUMMARY_BLOCK_ID: 306 switch (CodeID) { 307 default: 308 return nullptr; 309 STRINGIFY_CODE(FS, PERMODULE) 310 STRINGIFY_CODE(FS, PERMODULE_PROFILE) 311 STRINGIFY_CODE(FS, PERMODULE_RELBF) 312 STRINGIFY_CODE(FS, PERMODULE_GLOBALVAR_INIT_REFS) 313 STRINGIFY_CODE(FS, COMBINED) 314 STRINGIFY_CODE(FS, COMBINED_PROFILE) 315 STRINGIFY_CODE(FS, COMBINED_GLOBALVAR_INIT_REFS) 316 STRINGIFY_CODE(FS, ALIAS) 317 STRINGIFY_CODE(FS, COMBINED_ALIAS) 318 STRINGIFY_CODE(FS, COMBINED_ORIGINAL_NAME) 319 STRINGIFY_CODE(FS, VERSION) 320 STRINGIFY_CODE(FS, FLAGS) 321 STRINGIFY_CODE(FS, TYPE_TESTS) 322 STRINGIFY_CODE(FS, TYPE_TEST_ASSUME_VCALLS) 323 STRINGIFY_CODE(FS, TYPE_CHECKED_LOAD_VCALLS) 324 STRINGIFY_CODE(FS, TYPE_TEST_ASSUME_CONST_VCALL) 325 STRINGIFY_CODE(FS, TYPE_CHECKED_LOAD_CONST_VCALL) 326 STRINGIFY_CODE(FS, VALUE_GUID) 327 STRINGIFY_CODE(FS, CFI_FUNCTION_DEFS) 328 STRINGIFY_CODE(FS, CFI_FUNCTION_DECLS) 329 STRINGIFY_CODE(FS, TYPE_ID) 330 } 331 case bitc::METADATA_ATTACHMENT_ID: 332 switch(CodeID) { 333 default:return nullptr; 334 STRINGIFY_CODE(METADATA, ATTACHMENT) 335 } 336 case bitc::METADATA_BLOCK_ID: 337 switch(CodeID) { 338 default:return nullptr; 339 STRINGIFY_CODE(METADATA, STRING_OLD) 340 STRINGIFY_CODE(METADATA, VALUE) 341 STRINGIFY_CODE(METADATA, NODE) 342 STRINGIFY_CODE(METADATA, NAME) 343 STRINGIFY_CODE(METADATA, DISTINCT_NODE) 344 STRINGIFY_CODE(METADATA, KIND) // Older bitcode has it in a MODULE_BLOCK 345 STRINGIFY_CODE(METADATA, LOCATION) 346 STRINGIFY_CODE(METADATA, OLD_NODE) 347 STRINGIFY_CODE(METADATA, OLD_FN_NODE) 348 STRINGIFY_CODE(METADATA, NAMED_NODE) 349 STRINGIFY_CODE(METADATA, GENERIC_DEBUG) 350 STRINGIFY_CODE(METADATA, SUBRANGE) 351 STRINGIFY_CODE(METADATA, ENUMERATOR) 352 STRINGIFY_CODE(METADATA, BASIC_TYPE) 353 STRINGIFY_CODE(METADATA, FILE) 354 STRINGIFY_CODE(METADATA, DERIVED_TYPE) 355 STRINGIFY_CODE(METADATA, COMPOSITE_TYPE) 356 STRINGIFY_CODE(METADATA, SUBROUTINE_TYPE) 357 STRINGIFY_CODE(METADATA, COMPILE_UNIT) 358 STRINGIFY_CODE(METADATA, SUBPROGRAM) 359 STRINGIFY_CODE(METADATA, LEXICAL_BLOCK) 360 STRINGIFY_CODE(METADATA, LEXICAL_BLOCK_FILE) 361 STRINGIFY_CODE(METADATA, NAMESPACE) 362 STRINGIFY_CODE(METADATA, TEMPLATE_TYPE) 363 STRINGIFY_CODE(METADATA, TEMPLATE_VALUE) 364 STRINGIFY_CODE(METADATA, GLOBAL_VAR) 365 STRINGIFY_CODE(METADATA, LOCAL_VAR) 366 STRINGIFY_CODE(METADATA, EXPRESSION) 367 STRINGIFY_CODE(METADATA, OBJC_PROPERTY) 368 STRINGIFY_CODE(METADATA, IMPORTED_ENTITY) 369 STRINGIFY_CODE(METADATA, MODULE) 370 STRINGIFY_CODE(METADATA, MACRO) 371 STRINGIFY_CODE(METADATA, MACRO_FILE) 372 STRINGIFY_CODE(METADATA, STRINGS) 373 STRINGIFY_CODE(METADATA, GLOBAL_DECL_ATTACHMENT) 374 STRINGIFY_CODE(METADATA, GLOBAL_VAR_EXPR) 375 STRINGIFY_CODE(METADATA, INDEX_OFFSET) 376 STRINGIFY_CODE(METADATA, INDEX) 377 } 378 case bitc::METADATA_KIND_BLOCK_ID: 379 switch (CodeID) { 380 default: 381 return nullptr; 382 STRINGIFY_CODE(METADATA, KIND) 383 } 384 case bitc::USELIST_BLOCK_ID: 385 switch(CodeID) { 386 default:return nullptr; 387 case bitc::USELIST_CODE_DEFAULT: return "USELIST_CODE_DEFAULT"; 388 case bitc::USELIST_CODE_BB: return "USELIST_CODE_BB"; 389 } 390 391 case bitc::OPERAND_BUNDLE_TAGS_BLOCK_ID: 392 switch(CodeID) { 393 default: return nullptr; 394 case bitc::OPERAND_BUNDLE_TAG: return "OPERAND_BUNDLE_TAG"; 395 } 396 case bitc::STRTAB_BLOCK_ID: 397 switch(CodeID) { 398 default: return nullptr; 399 case bitc::STRTAB_BLOB: return "BLOB"; 400 } 401 case bitc::SYMTAB_BLOCK_ID: 402 switch(CodeID) { 403 default: return nullptr; 404 case bitc::SYMTAB_BLOB: return "BLOB"; 405 } 406 } 407 #undef STRINGIFY_CODE 408 } 409 410 struct PerRecordStats { 411 unsigned NumInstances; 412 unsigned NumAbbrev; 413 uint64_t TotalBits; 414 415 PerRecordStats() : NumInstances(0), NumAbbrev(0), TotalBits(0) {} 416 }; 417 418 struct PerBlockIDStats { 419 /// NumInstances - This the number of times this block ID has been seen. 420 unsigned NumInstances; 421 422 /// NumBits - The total size in bits of all of these blocks. 423 uint64_t NumBits; 424 425 /// NumSubBlocks - The total number of blocks these blocks contain. 426 unsigned NumSubBlocks; 427 428 /// NumAbbrevs - The total number of abbreviations. 429 unsigned NumAbbrevs; 430 431 /// NumRecords - The total number of records these blocks contain, and the 432 /// number that are abbreviated. 433 unsigned NumRecords, NumAbbreviatedRecords; 434 435 /// CodeFreq - Keep track of the number of times we see each code. 436 std::vector<PerRecordStats> CodeFreq; 437 438 PerBlockIDStats() 439 : NumInstances(0), NumBits(0), 440 NumSubBlocks(0), NumAbbrevs(0), NumRecords(0), NumAbbreviatedRecords(0) {} 441 }; 442 443 static std::map<unsigned, PerBlockIDStats> BlockIDStats; 444 445 446 447 /// ReportError - All bitcode analysis errors go through this function, making this a 448 /// good place to breakpoint if debugging. 449 static bool ReportError(const Twine &Err) { 450 WithColor::error() << Err << "\n"; 451 return true; 452 } 453 454 static bool decodeMetadataStringsBlob(StringRef Indent, 455 ArrayRef<uint64_t> Record, 456 StringRef Blob) { 457 if (Blob.empty()) 458 return true; 459 460 if (Record.size() != 2) 461 return true; 462 463 unsigned NumStrings = Record[0]; 464 unsigned StringsOffset = Record[1]; 465 outs() << " num-strings = " << NumStrings << " {\n"; 466 467 StringRef Lengths = Blob.slice(0, StringsOffset); 468 SimpleBitstreamCursor R(Lengths); 469 StringRef Strings = Blob.drop_front(StringsOffset); 470 do { 471 if (R.AtEndOfStream()) 472 return ReportError("bad length"); 473 474 unsigned Size = R.ReadVBR(6); 475 if (Strings.size() < Size) 476 return ReportError("truncated chars"); 477 478 outs() << Indent << " '"; 479 outs().write_escaped(Strings.slice(0, Size), /*hex=*/true); 480 outs() << "'\n"; 481 Strings = Strings.drop_front(Size); 482 } while (--NumStrings); 483 484 outs() << Indent << " }"; 485 return false; 486 } 487 488 static bool decodeBlob(unsigned Code, unsigned BlockID, StringRef Indent, 489 ArrayRef<uint64_t> Record, StringRef Blob) { 490 if (BlockID != bitc::METADATA_BLOCK_ID) 491 return true; 492 if (Code != bitc::METADATA_STRINGS) 493 return true; 494 495 return decodeMetadataStringsBlob(Indent, Record, Blob); 496 } 497 498 /// ParseBlock - Read a block, updating statistics, etc. 499 static bool ParseBlock(BitstreamCursor &Stream, BitstreamBlockInfo &BlockInfo, 500 unsigned BlockID, unsigned IndentLevel, 501 CurStreamTypeType CurStreamType) { 502 std::string Indent(IndentLevel*2, ' '); 503 uint64_t BlockBitStart = Stream.GetCurrentBitNo(); 504 505 // Get the statistics for this BlockID. 506 PerBlockIDStats &BlockStats = BlockIDStats[BlockID]; 507 508 BlockStats.NumInstances++; 509 510 // BLOCKINFO is a special part of the stream. 511 bool DumpRecords = Dump; 512 if (BlockID == bitc::BLOCKINFO_BLOCK_ID) { 513 if (Dump) outs() << Indent << "<BLOCKINFO_BLOCK/>\n"; 514 Optional<BitstreamBlockInfo> NewBlockInfo = 515 Stream.ReadBlockInfoBlock(/*ReadBlockInfoNames=*/true); 516 if (!NewBlockInfo) 517 return ReportError("Malformed BlockInfoBlock"); 518 BlockInfo = std::move(*NewBlockInfo); 519 Stream.JumpToBit(BlockBitStart); 520 // It's not really interesting to dump the contents of the blockinfo block. 521 DumpRecords = false; 522 } 523 524 unsigned NumWords = 0; 525 if (Stream.EnterSubBlock(BlockID, &NumWords)) 526 return ReportError("Malformed block record"); 527 528 // Keep it for later, when we see a MODULE_HASH record 529 uint64_t BlockEntryPos = Stream.getCurrentByteNo(); 530 531 const char *BlockName = nullptr; 532 if (DumpRecords) { 533 outs() << Indent << "<"; 534 if ((BlockName = GetBlockName(BlockID, BlockInfo, CurStreamType))) 535 outs() << BlockName; 536 else 537 outs() << "UnknownBlock" << BlockID; 538 539 if (NonSymbolic && BlockName) 540 outs() << " BlockID=" << BlockID; 541 542 outs() << " NumWords=" << NumWords 543 << " BlockCodeSize=" << Stream.getAbbrevIDWidth() << ">\n"; 544 } 545 546 SmallVector<uint64_t, 64> Record; 547 548 // Keep the offset to the metadata index if seen. 549 uint64_t MetadataIndexOffset = 0; 550 551 // Read all the records for this block. 552 while (1) { 553 if (Stream.AtEndOfStream()) 554 return ReportError("Premature end of bitstream"); 555 556 uint64_t RecordStartBit = Stream.GetCurrentBitNo(); 557 558 BitstreamEntry Entry = 559 Stream.advance(BitstreamCursor::AF_DontAutoprocessAbbrevs); 560 561 switch (Entry.Kind) { 562 case BitstreamEntry::Error: 563 return ReportError("malformed bitcode file"); 564 case BitstreamEntry::EndBlock: { 565 uint64_t BlockBitEnd = Stream.GetCurrentBitNo(); 566 BlockStats.NumBits += BlockBitEnd-BlockBitStart; 567 if (DumpRecords) { 568 outs() << Indent << "</"; 569 if (BlockName) 570 outs() << BlockName << ">\n"; 571 else 572 outs() << "UnknownBlock" << BlockID << ">\n"; 573 } 574 return false; 575 } 576 577 case BitstreamEntry::SubBlock: { 578 uint64_t SubBlockBitStart = Stream.GetCurrentBitNo(); 579 if (ParseBlock(Stream, BlockInfo, Entry.ID, IndentLevel + 1, 580 CurStreamType)) 581 return true; 582 ++BlockStats.NumSubBlocks; 583 uint64_t SubBlockBitEnd = Stream.GetCurrentBitNo(); 584 585 // Don't include subblock sizes in the size of this block. 586 BlockBitStart += SubBlockBitEnd-SubBlockBitStart; 587 continue; 588 } 589 case BitstreamEntry::Record: 590 // The interesting case. 591 break; 592 } 593 594 if (Entry.ID == bitc::DEFINE_ABBREV) { 595 Stream.ReadAbbrevRecord(); 596 ++BlockStats.NumAbbrevs; 597 continue; 598 } 599 600 Record.clear(); 601 602 ++BlockStats.NumRecords; 603 604 StringRef Blob; 605 uint64_t CurrentRecordPos = Stream.GetCurrentBitNo(); 606 unsigned Code = Stream.readRecord(Entry.ID, Record, &Blob); 607 608 // Increment the # occurrences of this code. 609 if (BlockStats.CodeFreq.size() <= Code) 610 BlockStats.CodeFreq.resize(Code+1); 611 BlockStats.CodeFreq[Code].NumInstances++; 612 BlockStats.CodeFreq[Code].TotalBits += 613 Stream.GetCurrentBitNo()-RecordStartBit; 614 if (Entry.ID != bitc::UNABBREV_RECORD) { 615 BlockStats.CodeFreq[Code].NumAbbrev++; 616 ++BlockStats.NumAbbreviatedRecords; 617 } 618 619 if (DumpRecords) { 620 outs() << Indent << " <"; 621 if (const char *CodeName = 622 GetCodeName(Code, BlockID, BlockInfo, CurStreamType)) 623 outs() << CodeName; 624 else 625 outs() << "UnknownCode" << Code; 626 if (NonSymbolic && GetCodeName(Code, BlockID, BlockInfo, CurStreamType)) 627 outs() << " codeid=" << Code; 628 const BitCodeAbbrev *Abbv = nullptr; 629 if (Entry.ID != bitc::UNABBREV_RECORD) { 630 Abbv = Stream.getAbbrev(Entry.ID); 631 outs() << " abbrevid=" << Entry.ID; 632 } 633 634 for (unsigned i = 0, e = Record.size(); i != e; ++i) 635 outs() << " op" << i << "=" << (int64_t)Record[i]; 636 637 // If we found a metadata index, let's verify that we had an offset before 638 // and validate its forward reference offset was correct! 639 if (BlockID == bitc::METADATA_BLOCK_ID) { 640 if (Code == bitc::METADATA_INDEX_OFFSET) { 641 if (Record.size() != 2) 642 outs() << "(Invalid record)"; 643 else { 644 auto Offset = Record[0] + (Record[1] << 32); 645 MetadataIndexOffset = Stream.GetCurrentBitNo() + Offset; 646 } 647 } 648 if (Code == bitc::METADATA_INDEX) { 649 outs() << " (offset "; 650 if (MetadataIndexOffset == RecordStartBit) 651 outs() << "match)"; 652 else 653 outs() << "mismatch: " << MetadataIndexOffset << " vs " 654 << RecordStartBit << ")"; 655 } 656 } 657 658 // If we found a module hash, let's verify that it matches! 659 if (BlockID == bitc::MODULE_BLOCK_ID && Code == bitc::MODULE_CODE_HASH && 660 !CheckHash.empty()) { 661 if (Record.size() != 5) 662 outs() << " (invalid)"; 663 else { 664 // Recompute the hash and compare it to the one in the bitcode 665 SHA1 Hasher; 666 StringRef Hash; 667 Hasher.update(CheckHash); 668 { 669 int BlockSize = (CurrentRecordPos / 8) - BlockEntryPos; 670 auto Ptr = Stream.getPointerToByte(BlockEntryPos, BlockSize); 671 Hasher.update(ArrayRef<uint8_t>(Ptr, BlockSize)); 672 Hash = Hasher.result(); 673 } 674 SmallString<20> RecordedHash; 675 RecordedHash.resize(20); 676 int Pos = 0; 677 for (auto &Val : Record) { 678 assert(!(Val >> 32) && "Unexpected high bits set"); 679 RecordedHash[Pos++] = (Val >> 24) & 0xFF; 680 RecordedHash[Pos++] = (Val >> 16) & 0xFF; 681 RecordedHash[Pos++] = (Val >> 8) & 0xFF; 682 RecordedHash[Pos++] = (Val >> 0) & 0xFF; 683 } 684 if (Hash == RecordedHash) 685 outs() << " (match)"; 686 else 687 outs() << " (!mismatch!)"; 688 } 689 } 690 691 outs() << "/>"; 692 693 if (Abbv) { 694 for (unsigned i = 1, e = Abbv->getNumOperandInfos(); i != e; ++i) { 695 const BitCodeAbbrevOp &Op = Abbv->getOperandInfo(i); 696 if (!Op.isEncoding() || Op.getEncoding() != BitCodeAbbrevOp::Array) 697 continue; 698 assert(i + 2 == e && "Array op not second to last"); 699 std::string Str; 700 bool ArrayIsPrintable = true; 701 for (unsigned j = i - 1, je = Record.size(); j != je; ++j) { 702 if (!isPrint(static_cast<unsigned char>(Record[j]))) { 703 ArrayIsPrintable = false; 704 break; 705 } 706 Str += (char)Record[j]; 707 } 708 if (ArrayIsPrintable) 709 outs() << " record string = '" << Str << "'"; 710 break; 711 } 712 } 713 714 if (Blob.data() && decodeBlob(Code, BlockID, Indent, Record, Blob)) { 715 outs() << " blob data = "; 716 if (ShowBinaryBlobs) { 717 outs() << "'"; 718 outs().write_escaped(Blob, /*hex=*/true) << "'"; 719 } else { 720 bool BlobIsPrintable = true; 721 for (unsigned i = 0, e = Blob.size(); i != e; ++i) 722 if (!isPrint(static_cast<unsigned char>(Blob[i]))) { 723 BlobIsPrintable = false; 724 break; 725 } 726 727 if (BlobIsPrintable) 728 outs() << "'" << Blob << "'"; 729 else 730 outs() << "unprintable, " << Blob.size() << " bytes."; 731 } 732 } 733 734 outs() << "\n"; 735 } 736 737 // Make sure that we can skip the current record. 738 Stream.JumpToBit(CurrentRecordPos); 739 Stream.skipRecord(Entry.ID); 740 } 741 } 742 743 static void PrintSize(double Bits) { 744 outs() << format("%.2f/%.2fB/%luW", Bits, Bits/8,(unsigned long)(Bits/32)); 745 } 746 static void PrintSize(uint64_t Bits) { 747 outs() << format("%lub/%.2fB/%luW", (unsigned long)Bits, 748 (double)Bits/8, (unsigned long)(Bits/32)); 749 } 750 751 static CurStreamTypeType ReadSignature(BitstreamCursor &Stream) { 752 char Signature[6]; 753 Signature[0] = Stream.Read(8); 754 Signature[1] = Stream.Read(8); 755 756 // Autodetect the file contents, if it is one we know. 757 if (Signature[0] == 'C' && Signature[1] == 'P') { 758 Signature[2] = Stream.Read(8); 759 Signature[3] = Stream.Read(8); 760 if (Signature[2] == 'C' && Signature[3] == 'H') 761 return ClangSerializedASTBitstream; 762 } else if (Signature[0] == 'D' && Signature[1] == 'I') { 763 Signature[2] = Stream.Read(8); 764 Signature[3] = Stream.Read(8); 765 if (Signature[2] == 'A' && Signature[3] == 'G') 766 return ClangSerializedDiagnosticsBitstream; 767 } else { 768 Signature[2] = Stream.Read(4); 769 Signature[3] = Stream.Read(4); 770 Signature[4] = Stream.Read(4); 771 Signature[5] = Stream.Read(4); 772 if (Signature[0] == 'B' && Signature[1] == 'C' && 773 Signature[2] == 0x0 && Signature[3] == 0xC && 774 Signature[4] == 0xE && Signature[5] == 0xD) 775 return LLVMIRBitstream; 776 } 777 return UnknownBitstream; 778 } 779 780 static bool openBitcodeFile(StringRef Path, 781 std::unique_ptr<MemoryBuffer> &MemBuf, 782 BitstreamCursor &Stream, 783 CurStreamTypeType &CurStreamType) { 784 // Read the input file. 785 ErrorOr<std::unique_ptr<MemoryBuffer>> MemBufOrErr = 786 MemoryBuffer::getFileOrSTDIN(Path); 787 if (std::error_code EC = MemBufOrErr.getError()) 788 return ReportError(Twine("ReportError reading '") + Path + "': " + EC.message()); 789 MemBuf = std::move(MemBufOrErr.get()); 790 791 if (MemBuf->getBufferSize() & 3) 792 return ReportError("Bitcode stream should be a multiple of 4 bytes in length"); 793 794 const unsigned char *BufPtr = (const unsigned char *)MemBuf->getBufferStart(); 795 const unsigned char *EndBufPtr = BufPtr + MemBuf->getBufferSize(); 796 797 // If we have a wrapper header, parse it and ignore the non-bc file contents. 798 // The magic number is 0x0B17C0DE stored in little endian. 799 if (isBitcodeWrapper(BufPtr, EndBufPtr)) { 800 if (MemBuf->getBufferSize() < BWH_HeaderSize) 801 return ReportError("Invalid bitcode wrapper header"); 802 803 if (Dump) { 804 unsigned Magic = support::endian::read32le(&BufPtr[BWH_MagicField]); 805 unsigned Version = support::endian::read32le(&BufPtr[BWH_VersionField]); 806 unsigned Offset = support::endian::read32le(&BufPtr[BWH_OffsetField]); 807 unsigned Size = support::endian::read32le(&BufPtr[BWH_SizeField]); 808 unsigned CPUType = support::endian::read32le(&BufPtr[BWH_CPUTypeField]); 809 810 outs() << "<BITCODE_WRAPPER_HEADER" 811 << " Magic=" << format_hex(Magic, 10) 812 << " Version=" << format_hex(Version, 10) 813 << " Offset=" << format_hex(Offset, 10) 814 << " Size=" << format_hex(Size, 10) 815 << " CPUType=" << format_hex(CPUType, 10) << "/>\n"; 816 } 817 818 if (SkipBitcodeWrapperHeader(BufPtr, EndBufPtr, true)) 819 return ReportError("Invalid bitcode wrapper header"); 820 } 821 822 Stream = BitstreamCursor(ArrayRef<uint8_t>(BufPtr, EndBufPtr)); 823 CurStreamType = ReadSignature(Stream); 824 825 return false; 826 } 827 828 /// AnalyzeBitcode - Analyze the bitcode file specified by InputFilename. 829 static int AnalyzeBitcode() { 830 std::unique_ptr<MemoryBuffer> StreamBuffer; 831 BitstreamCursor Stream; 832 BitstreamBlockInfo BlockInfo; 833 CurStreamTypeType CurStreamType; 834 if (openBitcodeFile(InputFilename, StreamBuffer, Stream, CurStreamType)) 835 return true; 836 Stream.setBlockInfo(&BlockInfo); 837 838 // Read block info from BlockInfoFilename, if specified. 839 // The block info must be a top-level block. 840 if (!BlockInfoFilename.empty()) { 841 std::unique_ptr<MemoryBuffer> BlockInfoBuffer; 842 BitstreamCursor BlockInfoCursor; 843 CurStreamTypeType BlockInfoStreamType; 844 if (openBitcodeFile(BlockInfoFilename, BlockInfoBuffer, BlockInfoCursor, 845 BlockInfoStreamType)) 846 return true; 847 848 while (!BlockInfoCursor.AtEndOfStream()) { 849 unsigned Code = BlockInfoCursor.ReadCode(); 850 if (Code != bitc::ENTER_SUBBLOCK) 851 return ReportError("Invalid record at top-level in block info file"); 852 853 unsigned BlockID = BlockInfoCursor.ReadSubBlockID(); 854 if (BlockID == bitc::BLOCKINFO_BLOCK_ID) { 855 Optional<BitstreamBlockInfo> NewBlockInfo = 856 BlockInfoCursor.ReadBlockInfoBlock(/*ReadBlockInfoNames=*/true); 857 if (!NewBlockInfo) 858 return ReportError("Malformed BlockInfoBlock in block info file"); 859 BlockInfo = std::move(*NewBlockInfo); 860 break; 861 } 862 863 BlockInfoCursor.SkipBlock(); 864 } 865 } 866 867 unsigned NumTopBlocks = 0; 868 869 // Parse the top-level structure. We only allow blocks at the top-level. 870 while (!Stream.AtEndOfStream()) { 871 unsigned Code = Stream.ReadCode(); 872 if (Code != bitc::ENTER_SUBBLOCK) 873 return ReportError("Invalid record at top-level"); 874 875 unsigned BlockID = Stream.ReadSubBlockID(); 876 877 if (ParseBlock(Stream, BlockInfo, BlockID, 0, CurStreamType)) 878 return true; 879 ++NumTopBlocks; 880 } 881 882 if (Dump) outs() << "\n\n"; 883 884 uint64_t BufferSizeBits = Stream.getBitcodeBytes().size() * CHAR_BIT; 885 // Print a summary of the read file. 886 outs() << "Summary of " << InputFilename << ":\n"; 887 outs() << " Total size: "; 888 PrintSize(BufferSizeBits); 889 outs() << "\n"; 890 outs() << " Stream type: "; 891 switch (CurStreamType) { 892 case UnknownBitstream: 893 outs() << "unknown\n"; 894 break; 895 case LLVMIRBitstream: 896 outs() << "LLVM IR\n"; 897 break; 898 case ClangSerializedASTBitstream: 899 outs() << "Clang Serialized AST\n"; 900 break; 901 case ClangSerializedDiagnosticsBitstream: 902 outs() << "Clang Serialized Diagnostics\n"; 903 break; 904 } 905 outs() << " # Toplevel Blocks: " << NumTopBlocks << "\n"; 906 outs() << "\n"; 907 908 // Emit per-block stats. 909 outs() << "Per-block Summary:\n"; 910 for (std::map<unsigned, PerBlockIDStats>::iterator I = BlockIDStats.begin(), 911 E = BlockIDStats.end(); I != E; ++I) { 912 outs() << " Block ID #" << I->first; 913 if (const char *BlockName = 914 GetBlockName(I->first, BlockInfo, CurStreamType)) 915 outs() << " (" << BlockName << ")"; 916 outs() << ":\n"; 917 918 const PerBlockIDStats &Stats = I->second; 919 outs() << " Num Instances: " << Stats.NumInstances << "\n"; 920 outs() << " Total Size: "; 921 PrintSize(Stats.NumBits); 922 outs() << "\n"; 923 double pct = (Stats.NumBits * 100.0) / BufferSizeBits; 924 outs() << " Percent of file: " << format("%2.4f%%", pct) << "\n"; 925 if (Stats.NumInstances > 1) { 926 outs() << " Average Size: "; 927 PrintSize(Stats.NumBits/(double)Stats.NumInstances); 928 outs() << "\n"; 929 outs() << " Tot/Avg SubBlocks: " << Stats.NumSubBlocks << "/" 930 << Stats.NumSubBlocks/(double)Stats.NumInstances << "\n"; 931 outs() << " Tot/Avg Abbrevs: " << Stats.NumAbbrevs << "/" 932 << Stats.NumAbbrevs/(double)Stats.NumInstances << "\n"; 933 outs() << " Tot/Avg Records: " << Stats.NumRecords << "/" 934 << Stats.NumRecords/(double)Stats.NumInstances << "\n"; 935 } else { 936 outs() << " Num SubBlocks: " << Stats.NumSubBlocks << "\n"; 937 outs() << " Num Abbrevs: " << Stats.NumAbbrevs << "\n"; 938 outs() << " Num Records: " << Stats.NumRecords << "\n"; 939 } 940 if (Stats.NumRecords) { 941 double pct = (Stats.NumAbbreviatedRecords * 100.0) / Stats.NumRecords; 942 outs() << " Percent Abbrevs: " << format("%2.4f%%", pct) << "\n"; 943 } 944 outs() << "\n"; 945 946 // Print a histogram of the codes we see. 947 if (!NoHistogram && !Stats.CodeFreq.empty()) { 948 std::vector<std::pair<unsigned, unsigned> > FreqPairs; // <freq,code> 949 for (unsigned i = 0, e = Stats.CodeFreq.size(); i != e; ++i) 950 if (unsigned Freq = Stats.CodeFreq[i].NumInstances) 951 FreqPairs.push_back(std::make_pair(Freq, i)); 952 std::stable_sort(FreqPairs.begin(), FreqPairs.end()); 953 std::reverse(FreqPairs.begin(), FreqPairs.end()); 954 955 outs() << "\tRecord Histogram:\n"; 956 outs() << "\t\t Count # Bits b/Rec % Abv Record Kind\n"; 957 for (unsigned i = 0, e = FreqPairs.size(); i != e; ++i) { 958 const PerRecordStats &RecStats = Stats.CodeFreq[FreqPairs[i].second]; 959 960 outs() << format("\t\t%7d %9lu", 961 RecStats.NumInstances, 962 (unsigned long)RecStats.TotalBits); 963 964 if (RecStats.NumInstances > 1) 965 outs() << format(" %9.1f", 966 (double)RecStats.TotalBits/RecStats.NumInstances); 967 else 968 outs() << " "; 969 970 if (RecStats.NumAbbrev) 971 outs() << 972 format(" %7.2f", 973 (double)RecStats.NumAbbrev/RecStats.NumInstances*100); 974 else 975 outs() << " "; 976 977 outs() << " "; 978 if (const char *CodeName = GetCodeName(FreqPairs[i].second, I->first, 979 BlockInfo, CurStreamType)) 980 outs() << CodeName << "\n"; 981 else 982 outs() << "UnknownCode" << FreqPairs[i].second << "\n"; 983 } 984 outs() << "\n"; 985 986 } 987 } 988 return 0; 989 } 990 991 992 int main(int argc, char **argv) { 993 InitLLVM X(argc, argv); 994 cl::ParseCommandLineOptions(argc, argv, "llvm-bcanalyzer file analyzer\n"); 995 return AnalyzeBitcode(); 996 } 997