| 1 | //===- BitcodeAnalyzer.cpp - Internal BitcodeAnalyzer implementation ------===// |
| 2 | // |
| 3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 4 | // See https://llvm.org/LICENSE.txt for license information. |
| 5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| 6 | // |
| 7 | //===----------------------------------------------------------------------===// |
| 8 | |
| 9 | #include "llvm/Bitcode/BitcodeAnalyzer.h" |
| 10 | #include "llvm/Bitcode/BitcodeReader.h" |
| 11 | #include "llvm/Bitcode/LLVMBitCodes.h" |
| 12 | #include "llvm/Bitstream/BitCodes.h" |
| 13 | #include "llvm/Bitstream/BitstreamReader.h" |
| 14 | #include "llvm/Support/Format.h" |
| 15 | #include "llvm/Support/SHA1.h" |
| 16 | #include <optional> |
| 17 | |
| 18 | using namespace llvm; |
| 19 | |
| 20 | static Error reportError(StringRef Message) { |
| 21 | return createStringError(EC: std::errc::illegal_byte_sequence, Fmt: Message.data()); |
| 22 | } |
| 23 | |
| 24 | /// Return a symbolic block name if known, otherwise return null. |
| 25 | static std::optional<const char *> |
| 26 | GetBlockName(unsigned BlockID, const BitstreamBlockInfo &BlockInfo, |
| 27 | CurStreamTypeType CurStreamType) { |
| 28 | // Standard blocks for all bitcode files. |
| 29 | if (BlockID < bitc::FIRST_APPLICATION_BLOCKID) { |
| 30 | if (BlockID == bitc::BLOCKINFO_BLOCK_ID) |
| 31 | return "BLOCKINFO_BLOCK" ; |
| 32 | return std::nullopt; |
| 33 | } |
| 34 | |
| 35 | // Check to see if we have a blockinfo record for this block, with a name. |
| 36 | if (const BitstreamBlockInfo::BlockInfo *Info = |
| 37 | BlockInfo.getBlockInfo(BlockID)) { |
| 38 | if (!Info->Name.empty()) |
| 39 | return Info->Name.c_str(); |
| 40 | } |
| 41 | |
| 42 | if (CurStreamType != LLVMIRBitstream) |
| 43 | return std::nullopt; |
| 44 | |
| 45 | switch (BlockID) { |
| 46 | default: |
| 47 | return std::nullopt; |
| 48 | case bitc::OPERAND_BUNDLE_TAGS_BLOCK_ID: |
| 49 | return "OPERAND_BUNDLE_TAGS_BLOCK" ; |
| 50 | case bitc::MODULE_BLOCK_ID: |
| 51 | return "MODULE_BLOCK" ; |
| 52 | case bitc::PARAMATTR_BLOCK_ID: |
| 53 | return "PARAMATTR_BLOCK" ; |
| 54 | case bitc::PARAMATTR_GROUP_BLOCK_ID: |
| 55 | return "PARAMATTR_GROUP_BLOCK_ID" ; |
| 56 | case bitc::TYPE_BLOCK_ID_NEW: |
| 57 | return "TYPE_BLOCK_ID" ; |
| 58 | case bitc::CONSTANTS_BLOCK_ID: |
| 59 | return "CONSTANTS_BLOCK" ; |
| 60 | case bitc::FUNCTION_BLOCK_ID: |
| 61 | return "FUNCTION_BLOCK" ; |
| 62 | case bitc::IDENTIFICATION_BLOCK_ID: |
| 63 | return "IDENTIFICATION_BLOCK_ID" ; |
| 64 | case bitc::VALUE_SYMTAB_BLOCK_ID: |
| 65 | return "VALUE_SYMTAB" ; |
| 66 | case bitc::METADATA_BLOCK_ID: |
| 67 | return "METADATA_BLOCK" ; |
| 68 | case bitc::METADATA_KIND_BLOCK_ID: |
| 69 | return "METADATA_KIND_BLOCK" ; |
| 70 | case bitc::METADATA_ATTACHMENT_ID: |
| 71 | return "METADATA_ATTACHMENT_BLOCK" ; |
| 72 | case bitc::USELIST_BLOCK_ID: |
| 73 | return "USELIST_BLOCK_ID" ; |
| 74 | case bitc::GLOBALVAL_SUMMARY_BLOCK_ID: |
| 75 | return "GLOBALVAL_SUMMARY_BLOCK" ; |
| 76 | case bitc::FULL_LTO_GLOBALVAL_SUMMARY_BLOCK_ID: |
| 77 | return "FULL_LTO_GLOBALVAL_SUMMARY_BLOCK" ; |
| 78 | case bitc::MODULE_STRTAB_BLOCK_ID: |
| 79 | return "MODULE_STRTAB_BLOCK" ; |
| 80 | case bitc::STRTAB_BLOCK_ID: |
| 81 | return "STRTAB_BLOCK" ; |
| 82 | case bitc::SYMTAB_BLOCK_ID: |
| 83 | return "SYMTAB_BLOCK" ; |
| 84 | } |
| 85 | } |
| 86 | |
| 87 | /// Return a symbolic code name if known, otherwise return null. |
| 88 | static std::optional<const char *> |
| 89 | GetCodeName(unsigned CodeID, unsigned BlockID, |
| 90 | const BitstreamBlockInfo &BlockInfo, |
| 91 | CurStreamTypeType CurStreamType) { |
| 92 | // Standard blocks for all bitcode files. |
| 93 | if (BlockID < bitc::FIRST_APPLICATION_BLOCKID) { |
| 94 | if (BlockID == bitc::BLOCKINFO_BLOCK_ID) { |
| 95 | switch (CodeID) { |
| 96 | default: |
| 97 | return std::nullopt; |
| 98 | case bitc::BLOCKINFO_CODE_SETBID: |
| 99 | return "SETBID" ; |
| 100 | case bitc::BLOCKINFO_CODE_BLOCKNAME: |
| 101 | return "BLOCKNAME" ; |
| 102 | case bitc::BLOCKINFO_CODE_SETRECORDNAME: |
| 103 | return "SETRECORDNAME" ; |
| 104 | } |
| 105 | } |
| 106 | return std::nullopt; |
| 107 | } |
| 108 | |
| 109 | // Check to see if we have a blockinfo record for this record, with a name. |
| 110 | if (const BitstreamBlockInfo::BlockInfo *Info = |
| 111 | BlockInfo.getBlockInfo(BlockID)) { |
| 112 | for (const std::pair<unsigned, std::string> &RN : Info->RecordNames) |
| 113 | if (RN.first == CodeID) |
| 114 | return RN.second.c_str(); |
| 115 | } |
| 116 | |
| 117 | if (CurStreamType != LLVMIRBitstream) |
| 118 | return std::nullopt; |
| 119 | |
| 120 | #define STRINGIFY_CODE(PREFIX, CODE) \ |
| 121 | case bitc::PREFIX##_##CODE: \ |
| 122 | return #CODE; |
| 123 | switch (BlockID) { |
| 124 | default: |
| 125 | return std::nullopt; |
| 126 | case bitc::MODULE_BLOCK_ID: |
| 127 | switch (CodeID) { |
| 128 | default: |
| 129 | return std::nullopt; |
| 130 | STRINGIFY_CODE(MODULE_CODE, VERSION) |
| 131 | STRINGIFY_CODE(MODULE_CODE, TRIPLE) |
| 132 | STRINGIFY_CODE(MODULE_CODE, DATALAYOUT) |
| 133 | STRINGIFY_CODE(MODULE_CODE, ASM) |
| 134 | STRINGIFY_CODE(MODULE_CODE, SECTIONNAME) |
| 135 | STRINGIFY_CODE(MODULE_CODE, DEPLIB) // Deprecated, present in old bitcode |
| 136 | STRINGIFY_CODE(MODULE_CODE, GLOBALVAR) |
| 137 | STRINGIFY_CODE(MODULE_CODE, FUNCTION) |
| 138 | STRINGIFY_CODE(MODULE_CODE, ALIAS) |
| 139 | STRINGIFY_CODE(MODULE_CODE, GCNAME) |
| 140 | STRINGIFY_CODE(MODULE_CODE, COMDAT) |
| 141 | STRINGIFY_CODE(MODULE_CODE, VSTOFFSET) |
| 142 | STRINGIFY_CODE(MODULE_CODE, METADATA_VALUES_UNUSED) |
| 143 | STRINGIFY_CODE(MODULE_CODE, SOURCE_FILENAME) |
| 144 | STRINGIFY_CODE(MODULE_CODE, HASH) |
| 145 | } |
| 146 | case bitc::IDENTIFICATION_BLOCK_ID: |
| 147 | switch (CodeID) { |
| 148 | default: |
| 149 | return std::nullopt; |
| 150 | STRINGIFY_CODE(IDENTIFICATION_CODE, STRING) |
| 151 | STRINGIFY_CODE(IDENTIFICATION_CODE, EPOCH) |
| 152 | } |
| 153 | case bitc::PARAMATTR_BLOCK_ID: |
| 154 | switch (CodeID) { |
| 155 | default: |
| 156 | return std::nullopt; |
| 157 | // FIXME: Should these be different? |
| 158 | case bitc::PARAMATTR_CODE_ENTRY_OLD: |
| 159 | return "ENTRY" ; |
| 160 | case bitc::PARAMATTR_CODE_ENTRY: |
| 161 | return "ENTRY" ; |
| 162 | } |
| 163 | case bitc::PARAMATTR_GROUP_BLOCK_ID: |
| 164 | switch (CodeID) { |
| 165 | default: |
| 166 | return std::nullopt; |
| 167 | case bitc::PARAMATTR_GRP_CODE_ENTRY: |
| 168 | return "ENTRY" ; |
| 169 | } |
| 170 | case bitc::TYPE_BLOCK_ID_NEW: |
| 171 | switch (CodeID) { |
| 172 | default: |
| 173 | return std::nullopt; |
| 174 | STRINGIFY_CODE(TYPE_CODE, NUMENTRY) |
| 175 | STRINGIFY_CODE(TYPE_CODE, VOID) |
| 176 | STRINGIFY_CODE(TYPE_CODE, FLOAT) |
| 177 | STRINGIFY_CODE(TYPE_CODE, DOUBLE) |
| 178 | STRINGIFY_CODE(TYPE_CODE, LABEL) |
| 179 | STRINGIFY_CODE(TYPE_CODE, OPAQUE) |
| 180 | STRINGIFY_CODE(TYPE_CODE, INTEGER) |
| 181 | STRINGIFY_CODE(TYPE_CODE, POINTER) |
| 182 | STRINGIFY_CODE(TYPE_CODE, HALF) |
| 183 | STRINGIFY_CODE(TYPE_CODE, ARRAY) |
| 184 | STRINGIFY_CODE(TYPE_CODE, VECTOR) |
| 185 | STRINGIFY_CODE(TYPE_CODE, X86_FP80) |
| 186 | STRINGIFY_CODE(TYPE_CODE, FP128) |
| 187 | STRINGIFY_CODE(TYPE_CODE, PPC_FP128) |
| 188 | STRINGIFY_CODE(TYPE_CODE, METADATA) |
| 189 | STRINGIFY_CODE(TYPE_CODE, X86_MMX) |
| 190 | STRINGIFY_CODE(TYPE_CODE, STRUCT_ANON) |
| 191 | STRINGIFY_CODE(TYPE_CODE, STRUCT_NAME) |
| 192 | STRINGIFY_CODE(TYPE_CODE, STRUCT_NAMED) |
| 193 | STRINGIFY_CODE(TYPE_CODE, FUNCTION) |
| 194 | STRINGIFY_CODE(TYPE_CODE, TOKEN) |
| 195 | STRINGIFY_CODE(TYPE_CODE, BFLOAT) |
| 196 | } |
| 197 | |
| 198 | case bitc::CONSTANTS_BLOCK_ID: |
| 199 | switch (CodeID) { |
| 200 | default: |
| 201 | return std::nullopt; |
| 202 | STRINGIFY_CODE(CST_CODE, SETTYPE) |
| 203 | STRINGIFY_CODE(CST_CODE, NULL) |
| 204 | STRINGIFY_CODE(CST_CODE, UNDEF) |
| 205 | STRINGIFY_CODE(CST_CODE, INTEGER) |
| 206 | STRINGIFY_CODE(CST_CODE, WIDE_INTEGER) |
| 207 | STRINGIFY_CODE(CST_CODE, FLOAT) |
| 208 | STRINGIFY_CODE(CST_CODE, AGGREGATE) |
| 209 | STRINGIFY_CODE(CST_CODE, STRING) |
| 210 | STRINGIFY_CODE(CST_CODE, CSTRING) |
| 211 | STRINGIFY_CODE(CST_CODE, CE_BINOP) |
| 212 | STRINGIFY_CODE(CST_CODE, CE_CAST) |
| 213 | STRINGIFY_CODE(CST_CODE, CE_GEP) |
| 214 | STRINGIFY_CODE(CST_CODE, CE_INBOUNDS_GEP) |
| 215 | STRINGIFY_CODE(CST_CODE, CE_SELECT) |
| 216 | STRINGIFY_CODE(CST_CODE, CE_EXTRACTELT) |
| 217 | STRINGIFY_CODE(CST_CODE, CE_INSERTELT) |
| 218 | STRINGIFY_CODE(CST_CODE, CE_SHUFFLEVEC) |
| 219 | STRINGIFY_CODE(CST_CODE, CE_CMP) |
| 220 | STRINGIFY_CODE(CST_CODE, INLINEASM) |
| 221 | STRINGIFY_CODE(CST_CODE, CE_SHUFVEC_EX) |
| 222 | STRINGIFY_CODE(CST_CODE, CE_UNOP) |
| 223 | STRINGIFY_CODE(CST_CODE, DSO_LOCAL_EQUIVALENT) |
| 224 | STRINGIFY_CODE(CST_CODE, NO_CFI_VALUE) |
| 225 | STRINGIFY_CODE(CST_CODE, PTRAUTH) |
| 226 | case bitc::CST_CODE_BLOCKADDRESS: |
| 227 | return "CST_CODE_BLOCKADDRESS" ; |
| 228 | STRINGIFY_CODE(CST_CODE, DATA) |
| 229 | } |
| 230 | case bitc::FUNCTION_BLOCK_ID: |
| 231 | switch (CodeID) { |
| 232 | default: |
| 233 | return std::nullopt; |
| 234 | STRINGIFY_CODE(FUNC_CODE, DECLAREBLOCKS) |
| 235 | STRINGIFY_CODE(FUNC_CODE, INST_BINOP) |
| 236 | STRINGIFY_CODE(FUNC_CODE, INST_CAST) |
| 237 | STRINGIFY_CODE(FUNC_CODE, INST_GEP_OLD) |
| 238 | STRINGIFY_CODE(FUNC_CODE, INST_INBOUNDS_GEP_OLD) |
| 239 | STRINGIFY_CODE(FUNC_CODE, INST_SELECT) |
| 240 | STRINGIFY_CODE(FUNC_CODE, INST_EXTRACTELT) |
| 241 | STRINGIFY_CODE(FUNC_CODE, INST_INSERTELT) |
| 242 | STRINGIFY_CODE(FUNC_CODE, INST_SHUFFLEVEC) |
| 243 | STRINGIFY_CODE(FUNC_CODE, INST_CMP) |
| 244 | STRINGIFY_CODE(FUNC_CODE, INST_RET) |
| 245 | STRINGIFY_CODE(FUNC_CODE, INST_BR) |
| 246 | STRINGIFY_CODE(FUNC_CODE, INST_SWITCH) |
| 247 | STRINGIFY_CODE(FUNC_CODE, INST_INVOKE) |
| 248 | STRINGIFY_CODE(FUNC_CODE, INST_UNOP) |
| 249 | STRINGIFY_CODE(FUNC_CODE, INST_UNREACHABLE) |
| 250 | STRINGIFY_CODE(FUNC_CODE, INST_CLEANUPRET) |
| 251 | STRINGIFY_CODE(FUNC_CODE, INST_CATCHRET) |
| 252 | STRINGIFY_CODE(FUNC_CODE, INST_CATCHPAD) |
| 253 | STRINGIFY_CODE(FUNC_CODE, INST_PHI) |
| 254 | STRINGIFY_CODE(FUNC_CODE, INST_ALLOCA) |
| 255 | STRINGIFY_CODE(FUNC_CODE, INST_LOAD) |
| 256 | STRINGIFY_CODE(FUNC_CODE, INST_VAARG) |
| 257 | STRINGIFY_CODE(FUNC_CODE, INST_STORE) |
| 258 | STRINGIFY_CODE(FUNC_CODE, INST_EXTRACTVAL) |
| 259 | STRINGIFY_CODE(FUNC_CODE, INST_INSERTVAL) |
| 260 | STRINGIFY_CODE(FUNC_CODE, INST_CMP2) |
| 261 | STRINGIFY_CODE(FUNC_CODE, INST_VSELECT) |
| 262 | STRINGIFY_CODE(FUNC_CODE, DEBUG_LOC_AGAIN) |
| 263 | STRINGIFY_CODE(FUNC_CODE, INST_CALL) |
| 264 | STRINGIFY_CODE(FUNC_CODE, DEBUG_LOC) |
| 265 | STRINGIFY_CODE(FUNC_CODE, INST_GEP) |
| 266 | STRINGIFY_CODE(FUNC_CODE, OPERAND_BUNDLE) |
| 267 | STRINGIFY_CODE(FUNC_CODE, INST_FENCE) |
| 268 | STRINGIFY_CODE(FUNC_CODE, INST_ATOMICRMW) |
| 269 | STRINGIFY_CODE(FUNC_CODE, INST_LOADATOMIC) |
| 270 | STRINGIFY_CODE(FUNC_CODE, INST_STOREATOMIC) |
| 271 | STRINGIFY_CODE(FUNC_CODE, INST_CMPXCHG) |
| 272 | STRINGIFY_CODE(FUNC_CODE, INST_CALLBR) |
| 273 | STRINGIFY_CODE(FUNC_CODE, BLOCKADDR_USERS) |
| 274 | STRINGIFY_CODE(FUNC_CODE, DEBUG_RECORD_DECLARE) |
| 275 | STRINGIFY_CODE(FUNC_CODE, DEBUG_RECORD_VALUE) |
| 276 | STRINGIFY_CODE(FUNC_CODE, DEBUG_RECORD_ASSIGN) |
| 277 | STRINGIFY_CODE(FUNC_CODE, DEBUG_RECORD_VALUE_SIMPLE) |
| 278 | STRINGIFY_CODE(FUNC_CODE, DEBUG_RECORD_LABEL) |
| 279 | } |
| 280 | case bitc::VALUE_SYMTAB_BLOCK_ID: |
| 281 | switch (CodeID) { |
| 282 | default: |
| 283 | return std::nullopt; |
| 284 | STRINGIFY_CODE(VST_CODE, ENTRY) |
| 285 | STRINGIFY_CODE(VST_CODE, BBENTRY) |
| 286 | STRINGIFY_CODE(VST_CODE, FNENTRY) |
| 287 | STRINGIFY_CODE(VST_CODE, COMBINED_ENTRY) |
| 288 | } |
| 289 | case bitc::MODULE_STRTAB_BLOCK_ID: |
| 290 | switch (CodeID) { |
| 291 | default: |
| 292 | return std::nullopt; |
| 293 | STRINGIFY_CODE(MST_CODE, ENTRY) |
| 294 | STRINGIFY_CODE(MST_CODE, HASH) |
| 295 | } |
| 296 | case bitc::GLOBALVAL_SUMMARY_BLOCK_ID: |
| 297 | case bitc::FULL_LTO_GLOBALVAL_SUMMARY_BLOCK_ID: |
| 298 | switch (CodeID) { |
| 299 | default: |
| 300 | return std::nullopt; |
| 301 | STRINGIFY_CODE(FS, PERMODULE) |
| 302 | STRINGIFY_CODE(FS, PERMODULE_PROFILE) |
| 303 | STRINGIFY_CODE(FS, PERMODULE_RELBF) |
| 304 | STRINGIFY_CODE(FS, PERMODULE_GLOBALVAR_INIT_REFS) |
| 305 | STRINGIFY_CODE(FS, PERMODULE_VTABLE_GLOBALVAR_INIT_REFS) |
| 306 | STRINGIFY_CODE(FS, COMBINED) |
| 307 | STRINGIFY_CODE(FS, COMBINED_PROFILE) |
| 308 | STRINGIFY_CODE(FS, COMBINED_GLOBALVAR_INIT_REFS) |
| 309 | STRINGIFY_CODE(FS, ALIAS) |
| 310 | STRINGIFY_CODE(FS, COMBINED_ALIAS) |
| 311 | STRINGIFY_CODE(FS, COMBINED_ORIGINAL_NAME) |
| 312 | STRINGIFY_CODE(FS, VERSION) |
| 313 | STRINGIFY_CODE(FS, FLAGS) |
| 314 | STRINGIFY_CODE(FS, TYPE_TESTS) |
| 315 | STRINGIFY_CODE(FS, TYPE_TEST_ASSUME_VCALLS) |
| 316 | STRINGIFY_CODE(FS, TYPE_CHECKED_LOAD_VCALLS) |
| 317 | STRINGIFY_CODE(FS, TYPE_TEST_ASSUME_CONST_VCALL) |
| 318 | STRINGIFY_CODE(FS, TYPE_CHECKED_LOAD_CONST_VCALL) |
| 319 | STRINGIFY_CODE(FS, VALUE_GUID) |
| 320 | STRINGIFY_CODE(FS, CFI_FUNCTION_DEFS) |
| 321 | STRINGIFY_CODE(FS, CFI_FUNCTION_DECLS) |
| 322 | STRINGIFY_CODE(FS, TYPE_ID) |
| 323 | STRINGIFY_CODE(FS, TYPE_ID_METADATA) |
| 324 | STRINGIFY_CODE(FS, BLOCK_COUNT) |
| 325 | STRINGIFY_CODE(FS, PARAM_ACCESS) |
| 326 | STRINGIFY_CODE(FS, PERMODULE_CALLSITE_INFO) |
| 327 | STRINGIFY_CODE(FS, PERMODULE_ALLOC_INFO) |
| 328 | STRINGIFY_CODE(FS, COMBINED_CALLSITE_INFO) |
| 329 | STRINGIFY_CODE(FS, COMBINED_ALLOC_INFO) |
| 330 | STRINGIFY_CODE(FS, STACK_IDS) |
| 331 | STRINGIFY_CODE(FS, ALLOC_CONTEXT_IDS) |
| 332 | STRINGIFY_CODE(FS, CONTEXT_RADIX_TREE_ARRAY) |
| 333 | STRINGIFY_CODE(FS, COMBINED_ALLOC_INFO_NO_CONTEXT) |
| 334 | } |
| 335 | case bitc::METADATA_ATTACHMENT_ID: |
| 336 | switch (CodeID) { |
| 337 | default: |
| 338 | return std::nullopt; |
| 339 | STRINGIFY_CODE(METADATA, ATTACHMENT) |
| 340 | } |
| 341 | case bitc::METADATA_BLOCK_ID: |
| 342 | switch (CodeID) { |
| 343 | default: |
| 344 | return std::nullopt; |
| 345 | STRINGIFY_CODE(METADATA, STRING_OLD) |
| 346 | STRINGIFY_CODE(METADATA, VALUE) |
| 347 | STRINGIFY_CODE(METADATA, NODE) |
| 348 | STRINGIFY_CODE(METADATA, NAME) |
| 349 | STRINGIFY_CODE(METADATA, DISTINCT_NODE) |
| 350 | STRINGIFY_CODE(METADATA, KIND) // Older bitcode has it in a MODULE_BLOCK |
| 351 | STRINGIFY_CODE(METADATA, LOCATION) |
| 352 | STRINGIFY_CODE(METADATA, OLD_NODE) |
| 353 | STRINGIFY_CODE(METADATA, OLD_FN_NODE) |
| 354 | STRINGIFY_CODE(METADATA, NAMED_NODE) |
| 355 | STRINGIFY_CODE(METADATA, GENERIC_DEBUG) |
| 356 | STRINGIFY_CODE(METADATA, SUBRANGE) |
| 357 | STRINGIFY_CODE(METADATA, ENUMERATOR) |
| 358 | STRINGIFY_CODE(METADATA, BASIC_TYPE) |
| 359 | STRINGIFY_CODE(METADATA, FILE) |
| 360 | STRINGIFY_CODE(METADATA, DERIVED_TYPE) |
| 361 | STRINGIFY_CODE(METADATA, COMPOSITE_TYPE) |
| 362 | STRINGIFY_CODE(METADATA, SUBROUTINE_TYPE) |
| 363 | STRINGIFY_CODE(METADATA, COMPILE_UNIT) |
| 364 | STRINGIFY_CODE(METADATA, SUBPROGRAM) |
| 365 | STRINGIFY_CODE(METADATA, LEXICAL_BLOCK) |
| 366 | STRINGIFY_CODE(METADATA, LEXICAL_BLOCK_FILE) |
| 367 | STRINGIFY_CODE(METADATA, NAMESPACE) |
| 368 | STRINGIFY_CODE(METADATA, TEMPLATE_TYPE) |
| 369 | STRINGIFY_CODE(METADATA, TEMPLATE_VALUE) |
| 370 | STRINGIFY_CODE(METADATA, GLOBAL_VAR) |
| 371 | STRINGIFY_CODE(METADATA, LOCAL_VAR) |
| 372 | STRINGIFY_CODE(METADATA, EXPRESSION) |
| 373 | STRINGIFY_CODE(METADATA, OBJC_PROPERTY) |
| 374 | STRINGIFY_CODE(METADATA, IMPORTED_ENTITY) |
| 375 | STRINGIFY_CODE(METADATA, MODULE) |
| 376 | STRINGIFY_CODE(METADATA, MACRO) |
| 377 | STRINGIFY_CODE(METADATA, MACRO_FILE) |
| 378 | STRINGIFY_CODE(METADATA, STRINGS) |
| 379 | STRINGIFY_CODE(METADATA, GLOBAL_DECL_ATTACHMENT) |
| 380 | STRINGIFY_CODE(METADATA, GLOBAL_VAR_EXPR) |
| 381 | STRINGIFY_CODE(METADATA, INDEX_OFFSET) |
| 382 | STRINGIFY_CODE(METADATA, INDEX) |
| 383 | STRINGIFY_CODE(METADATA, ARG_LIST) |
| 384 | } |
| 385 | case bitc::METADATA_KIND_BLOCK_ID: |
| 386 | switch (CodeID) { |
| 387 | default: |
| 388 | return std::nullopt; |
| 389 | STRINGIFY_CODE(METADATA, KIND) |
| 390 | } |
| 391 | case bitc::USELIST_BLOCK_ID: |
| 392 | switch (CodeID) { |
| 393 | default: |
| 394 | return std::nullopt; |
| 395 | case bitc::USELIST_CODE_DEFAULT: |
| 396 | return "USELIST_CODE_DEFAULT" ; |
| 397 | case bitc::USELIST_CODE_BB: |
| 398 | return "USELIST_CODE_BB" ; |
| 399 | } |
| 400 | |
| 401 | case bitc::OPERAND_BUNDLE_TAGS_BLOCK_ID: |
| 402 | switch (CodeID) { |
| 403 | default: |
| 404 | return std::nullopt; |
| 405 | case bitc::OPERAND_BUNDLE_TAG: |
| 406 | return "OPERAND_BUNDLE_TAG" ; |
| 407 | } |
| 408 | case bitc::STRTAB_BLOCK_ID: |
| 409 | switch (CodeID) { |
| 410 | default: |
| 411 | return std::nullopt; |
| 412 | case bitc::STRTAB_BLOB: |
| 413 | return "BLOB" ; |
| 414 | } |
| 415 | case bitc::SYMTAB_BLOCK_ID: |
| 416 | switch (CodeID) { |
| 417 | default: |
| 418 | return std::nullopt; |
| 419 | case bitc::SYMTAB_BLOB: |
| 420 | return "BLOB" ; |
| 421 | } |
| 422 | } |
| 423 | #undef STRINGIFY_CODE |
| 424 | } |
| 425 | |
| 426 | static void printSize(raw_ostream &OS, double Bits) { |
| 427 | OS << format(Fmt: "%.2f/%.2fB/%luW" , Vals: Bits, Vals: Bits / 8, Vals: (unsigned long)(Bits / 32)); |
| 428 | } |
| 429 | static void printSize(raw_ostream &OS, uint64_t Bits) { |
| 430 | OS << format(Fmt: "%lub/%.2fB/%luW" , Vals: (unsigned long)Bits, Vals: (double)Bits / 8, |
| 431 | Vals: (unsigned long)(Bits / 32)); |
| 432 | } |
| 433 | |
| 434 | static Expected<CurStreamTypeType> ReadSignature(BitstreamCursor &Stream) { |
| 435 | auto tryRead = [&Stream](char &Dest, size_t size) -> Error { |
| 436 | if (Expected<SimpleBitstreamCursor::word_t> MaybeWord = Stream.Read(NumBits: size)) |
| 437 | Dest = MaybeWord.get(); |
| 438 | else |
| 439 | return MaybeWord.takeError(); |
| 440 | return Error::success(); |
| 441 | }; |
| 442 | |
| 443 | char Signature[6]; |
| 444 | if (Error Err = tryRead(Signature[0], 8)) |
| 445 | return std::move(Err); |
| 446 | if (Error Err = tryRead(Signature[1], 8)) |
| 447 | return std::move(Err); |
| 448 | |
| 449 | // Autodetect the file contents, if it is one we know. |
| 450 | if (Signature[0] == 'C' && Signature[1] == 'P') { |
| 451 | if (Error Err = tryRead(Signature[2], 8)) |
| 452 | return std::move(Err); |
| 453 | if (Error Err = tryRead(Signature[3], 8)) |
| 454 | return std::move(Err); |
| 455 | if (Signature[2] == 'C' && Signature[3] == 'H') |
| 456 | return ClangSerializedASTBitstream; |
| 457 | } else if (Signature[0] == 'D' && Signature[1] == 'I') { |
| 458 | if (Error Err = tryRead(Signature[2], 8)) |
| 459 | return std::move(Err); |
| 460 | if (Error Err = tryRead(Signature[3], 8)) |
| 461 | return std::move(Err); |
| 462 | if (Signature[2] == 'A' && Signature[3] == 'G') |
| 463 | return ClangSerializedDiagnosticsBitstream; |
| 464 | } else if (Signature[0] == 'R' && Signature[1] == 'M') { |
| 465 | if (Error Err = tryRead(Signature[2], 8)) |
| 466 | return std::move(Err); |
| 467 | if (Error Err = tryRead(Signature[3], 8)) |
| 468 | return std::move(Err); |
| 469 | if (Signature[2] == 'R' && Signature[3] == 'K') |
| 470 | return LLVMBitstreamRemarks; |
| 471 | } else { |
| 472 | if (Error Err = tryRead(Signature[2], 4)) |
| 473 | return std::move(Err); |
| 474 | if (Error Err = tryRead(Signature[3], 4)) |
| 475 | return std::move(Err); |
| 476 | if (Error Err = tryRead(Signature[4], 4)) |
| 477 | return std::move(Err); |
| 478 | if (Error Err = tryRead(Signature[5], 4)) |
| 479 | return std::move(Err); |
| 480 | if (Signature[0] == 'B' && Signature[1] == 'C' && Signature[2] == 0x0 && |
| 481 | Signature[3] == 0xC && Signature[4] == 0xE && Signature[5] == 0xD) |
| 482 | return LLVMIRBitstream; |
| 483 | } |
| 484 | return UnknownBitstream; |
| 485 | } |
| 486 | |
| 487 | static Expected<CurStreamTypeType> (std::optional<BCDumpOptions> O, |
| 488 | BitstreamCursor &Stream) { |
| 489 | ArrayRef<uint8_t> Bytes = Stream.getBitcodeBytes(); |
| 490 | const unsigned char *BufPtr = (const unsigned char *)Bytes.data(); |
| 491 | const unsigned char *EndBufPtr = BufPtr + Bytes.size(); |
| 492 | |
| 493 | // If we have a wrapper header, parse it and ignore the non-bc file |
| 494 | // contents. The magic number is 0x0B17C0DE stored in little endian. |
| 495 | if (isBitcodeWrapper(BufPtr, BufEnd: EndBufPtr)) { |
| 496 | if (Bytes.size() < BWH_HeaderSize) |
| 497 | return reportError(Message: "Invalid bitcode wrapper header" ); |
| 498 | |
| 499 | if (O) { |
| 500 | unsigned Magic = support::endian::read32le(P: &BufPtr[BWH_MagicField]); |
| 501 | unsigned Version = support::endian::read32le(P: &BufPtr[BWH_VersionField]); |
| 502 | unsigned Offset = support::endian::read32le(P: &BufPtr[BWH_OffsetField]); |
| 503 | unsigned Size = support::endian::read32le(P: &BufPtr[BWH_SizeField]); |
| 504 | unsigned CPUType = support::endian::read32le(P: &BufPtr[BWH_CPUTypeField]); |
| 505 | |
| 506 | O->OS << "<BITCODE_WRAPPER_HEADER" |
| 507 | << " Magic=" << format_hex(N: Magic, Width: 10) |
| 508 | << " Version=" << format_hex(N: Version, Width: 10) |
| 509 | << " Offset=" << format_hex(N: Offset, Width: 10) |
| 510 | << " Size=" << format_hex(N: Size, Width: 10) |
| 511 | << " CPUType=" << format_hex(N: CPUType, Width: 10) << "/>\n" ; |
| 512 | } |
| 513 | |
| 514 | if (SkipBitcodeWrapperHeader(BufPtr, BufEnd&: EndBufPtr, VerifyBufferSize: true)) |
| 515 | return reportError(Message: "Invalid bitcode wrapper header" ); |
| 516 | } |
| 517 | |
| 518 | // Use the cursor modified by skipping the wrapper header. |
| 519 | Stream = BitstreamCursor(ArrayRef<uint8_t>(BufPtr, EndBufPtr)); |
| 520 | |
| 521 | return ReadSignature(Stream); |
| 522 | } |
| 523 | |
| 524 | static bool canDecodeBlob(unsigned Code, unsigned BlockID) { |
| 525 | return BlockID == bitc::METADATA_BLOCK_ID && Code == bitc::METADATA_STRINGS; |
| 526 | } |
| 527 | |
| 528 | Error BitcodeAnalyzer::decodeMetadataStringsBlob(StringRef Indent, |
| 529 | ArrayRef<uint64_t> Record, |
| 530 | StringRef Blob, |
| 531 | raw_ostream &OS) { |
| 532 | if (Blob.empty()) |
| 533 | return reportError(Message: "Cannot decode empty blob." ); |
| 534 | |
| 535 | if (Record.size() != 2) |
| 536 | return reportError( |
| 537 | Message: "Decoding metadata strings blob needs two record entries." ); |
| 538 | |
| 539 | unsigned NumStrings = Record[0]; |
| 540 | unsigned StringsOffset = Record[1]; |
| 541 | OS << " num-strings = " << NumStrings << " {\n" ; |
| 542 | |
| 543 | StringRef Lengths = Blob.slice(Start: 0, End: StringsOffset); |
| 544 | SimpleBitstreamCursor R(Lengths); |
| 545 | StringRef Strings = Blob.drop_front(N: StringsOffset); |
| 546 | do { |
| 547 | if (R.AtEndOfStream()) |
| 548 | return reportError(Message: "bad length" ); |
| 549 | |
| 550 | uint32_t Size; |
| 551 | if (Error E = R.ReadVBR(NumBits: 6).moveInto(Value&: Size)) |
| 552 | return E; |
| 553 | if (Strings.size() < Size) |
| 554 | return reportError(Message: "truncated chars" ); |
| 555 | |
| 556 | OS << Indent << " '" ; |
| 557 | OS.write_escaped(Str: Strings.slice(Start: 0, End: Size), /*hex=*/UseHexEscapes: true); |
| 558 | OS << "'\n" ; |
| 559 | Strings = Strings.drop_front(N: Size); |
| 560 | } while (--NumStrings); |
| 561 | |
| 562 | OS << Indent << " }" ; |
| 563 | return Error::success(); |
| 564 | } |
| 565 | |
| 566 | BitcodeAnalyzer::BitcodeAnalyzer(StringRef Buffer, |
| 567 | std::optional<StringRef> BlockInfoBuffer) |
| 568 | : Stream(Buffer) { |
| 569 | if (BlockInfoBuffer) |
| 570 | BlockInfoStream.emplace(args&: *BlockInfoBuffer); |
| 571 | } |
| 572 | |
| 573 | Error BitcodeAnalyzer::analyze(std::optional<BCDumpOptions> O, |
| 574 | std::optional<StringRef> CheckHash) { |
| 575 | if (Error E = analyzeHeader(O, Stream).moveInto(Value&: CurStreamType)) |
| 576 | return E; |
| 577 | |
| 578 | Stream.setBlockInfo(&BlockInfo); |
| 579 | |
| 580 | // Read block info from BlockInfoStream, if specified. |
| 581 | // The block info must be a top-level block. |
| 582 | if (BlockInfoStream) { |
| 583 | BitstreamCursor BlockInfoCursor(*BlockInfoStream); |
| 584 | if (Error E = analyzeHeader(O, Stream&: BlockInfoCursor).takeError()) |
| 585 | return E; |
| 586 | |
| 587 | while (!BlockInfoCursor.AtEndOfStream()) { |
| 588 | Expected<unsigned> MaybeCode = BlockInfoCursor.ReadCode(); |
| 589 | if (!MaybeCode) |
| 590 | return MaybeCode.takeError(); |
| 591 | if (MaybeCode.get() != bitc::ENTER_SUBBLOCK) |
| 592 | return reportError(Message: "Invalid record at top-level in block info file" ); |
| 593 | |
| 594 | Expected<unsigned> MaybeBlockID = BlockInfoCursor.ReadSubBlockID(); |
| 595 | if (!MaybeBlockID) |
| 596 | return MaybeBlockID.takeError(); |
| 597 | if (MaybeBlockID.get() == bitc::BLOCKINFO_BLOCK_ID) { |
| 598 | std::optional<BitstreamBlockInfo> NewBlockInfo; |
| 599 | if (Error E = |
| 600 | BlockInfoCursor.ReadBlockInfoBlock(/*ReadBlockInfoNames=*/true) |
| 601 | .moveInto(Value&: NewBlockInfo)) |
| 602 | return E; |
| 603 | if (!NewBlockInfo) |
| 604 | return reportError(Message: "Malformed BlockInfoBlock in block info file" ); |
| 605 | BlockInfo = std::move(*NewBlockInfo); |
| 606 | break; |
| 607 | } |
| 608 | |
| 609 | if (Error Err = BlockInfoCursor.SkipBlock()) |
| 610 | return Err; |
| 611 | } |
| 612 | } |
| 613 | |
| 614 | // Parse the top-level structure. We only allow blocks at the top-level. |
| 615 | while (!Stream.AtEndOfStream()) { |
| 616 | Expected<unsigned> MaybeCode = Stream.ReadCode(); |
| 617 | if (!MaybeCode) |
| 618 | return MaybeCode.takeError(); |
| 619 | if (MaybeCode.get() != bitc::ENTER_SUBBLOCK) |
| 620 | return reportError(Message: "Invalid record at top-level" ); |
| 621 | |
| 622 | Expected<unsigned> MaybeBlockID = Stream.ReadSubBlockID(); |
| 623 | if (!MaybeBlockID) |
| 624 | return MaybeBlockID.takeError(); |
| 625 | |
| 626 | if (Error E = parseBlock(BlockID: MaybeBlockID.get(), IndentLevel: 0, O, CheckHash)) |
| 627 | return E; |
| 628 | ++NumTopBlocks; |
| 629 | } |
| 630 | |
| 631 | return Error::success(); |
| 632 | } |
| 633 | |
| 634 | void BitcodeAnalyzer::printStats(BCDumpOptions O, |
| 635 | std::optional<StringRef> Filename) { |
| 636 | uint64_t BufferSizeBits = Stream.getBitcodeBytes().size() * CHAR_BIT; |
| 637 | // Print a summary of the read file. |
| 638 | O.OS << "Summary " ; |
| 639 | if (Filename) |
| 640 | O.OS << "of " << Filename->data() << ":\n" ; |
| 641 | O.OS << " Total size: " ; |
| 642 | printSize(OS&: O.OS, Bits: BufferSizeBits); |
| 643 | O.OS << "\n" ; |
| 644 | O.OS << " Stream type: " ; |
| 645 | switch (CurStreamType) { |
| 646 | case UnknownBitstream: |
| 647 | O.OS << "unknown\n" ; |
| 648 | break; |
| 649 | case LLVMIRBitstream: |
| 650 | O.OS << "LLVM IR\n" ; |
| 651 | break; |
| 652 | case ClangSerializedASTBitstream: |
| 653 | O.OS << "Clang Serialized AST\n" ; |
| 654 | break; |
| 655 | case ClangSerializedDiagnosticsBitstream: |
| 656 | O.OS << "Clang Serialized Diagnostics\n" ; |
| 657 | break; |
| 658 | case LLVMBitstreamRemarks: |
| 659 | O.OS << "LLVM Remarks\n" ; |
| 660 | break; |
| 661 | } |
| 662 | O.OS << " # Toplevel Blocks: " << NumTopBlocks << "\n" ; |
| 663 | O.OS << "\n" ; |
| 664 | |
| 665 | // Emit per-block stats. |
| 666 | O.OS << "Per-block Summary:\n" ; |
| 667 | for (const auto &Stat : BlockIDStats) { |
| 668 | O.OS << " Block ID #" << Stat.first; |
| 669 | if (std::optional<const char *> BlockName = |
| 670 | GetBlockName(BlockID: Stat.first, BlockInfo, CurStreamType)) |
| 671 | O.OS << " (" << *BlockName << ")" ; |
| 672 | O.OS << ":\n" ; |
| 673 | |
| 674 | const PerBlockIDStats &Stats = Stat.second; |
| 675 | O.OS << " Num Instances: " << Stats.NumInstances << "\n" ; |
| 676 | O.OS << " Total Size: " ; |
| 677 | printSize(OS&: O.OS, Bits: Stats.NumBits); |
| 678 | O.OS << "\n" ; |
| 679 | double pct = (Stats.NumBits * 100.0) / BufferSizeBits; |
| 680 | O.OS << " Percent of file: " << format(Fmt: "%2.4f%%" , Vals: pct) << "\n" ; |
| 681 | if (Stats.NumInstances > 1) { |
| 682 | O.OS << " Average Size: " ; |
| 683 | printSize(OS&: O.OS, Bits: Stats.NumBits / (double)Stats.NumInstances); |
| 684 | O.OS << "\n" ; |
| 685 | O.OS << " Tot/Avg SubBlocks: " << Stats.NumSubBlocks << "/" |
| 686 | << Stats.NumSubBlocks / (double)Stats.NumInstances << "\n" ; |
| 687 | O.OS << " Tot/Avg Abbrevs: " << Stats.NumAbbrevs << "/" |
| 688 | << Stats.NumAbbrevs / (double)Stats.NumInstances << "\n" ; |
| 689 | O.OS << " Tot/Avg Records: " << Stats.NumRecords << "/" |
| 690 | << Stats.NumRecords / (double)Stats.NumInstances << "\n" ; |
| 691 | } else { |
| 692 | O.OS << " Num SubBlocks: " << Stats.NumSubBlocks << "\n" ; |
| 693 | O.OS << " Num Abbrevs: " << Stats.NumAbbrevs << "\n" ; |
| 694 | O.OS << " Num Records: " << Stats.NumRecords << "\n" ; |
| 695 | } |
| 696 | if (Stats.NumRecords) { |
| 697 | double pct = (Stats.NumAbbreviatedRecords * 100.0) / Stats.NumRecords; |
| 698 | O.OS << " Percent Abbrevs: " << format(Fmt: "%2.4f%%" , Vals: pct) << "\n" ; |
| 699 | } |
| 700 | O.OS << "\n" ; |
| 701 | |
| 702 | // Print a histogram of the codes we see. |
| 703 | if (O.Histogram && !Stats.CodeFreq.empty()) { |
| 704 | std::vector<std::pair<unsigned, unsigned>> FreqPairs; // <freq,code> |
| 705 | for (unsigned i = 0, e = Stats.CodeFreq.size(); i != e; ++i) |
| 706 | if (unsigned Freq = Stats.CodeFreq[i].NumInstances) |
| 707 | FreqPairs.push_back(x: std::make_pair(x&: Freq, y&: i)); |
| 708 | llvm::stable_sort(Range&: FreqPairs); |
| 709 | std::reverse(first: FreqPairs.begin(), last: FreqPairs.end()); |
| 710 | |
| 711 | O.OS << "\tRecord Histogram:\n" ; |
| 712 | O.OS << "\t\t Count # Bits b/Rec % Abv Record Kind\n" ; |
| 713 | for (const auto &FreqPair : FreqPairs) { |
| 714 | const PerRecordStats &RecStats = Stats.CodeFreq[FreqPair.second]; |
| 715 | |
| 716 | O.OS << format(Fmt: "\t\t%7d %9lu" , Vals: RecStats.NumInstances, |
| 717 | Vals: (unsigned long)RecStats.TotalBits); |
| 718 | |
| 719 | if (RecStats.NumInstances > 1) |
| 720 | O.OS << format(Fmt: " %9.1f" , |
| 721 | Vals: (double)RecStats.TotalBits / RecStats.NumInstances); |
| 722 | else |
| 723 | O.OS << " " ; |
| 724 | |
| 725 | if (RecStats.NumAbbrev) |
| 726 | O.OS << format(Fmt: " %7.2f" , Vals: (double)RecStats.NumAbbrev / |
| 727 | RecStats.NumInstances * 100); |
| 728 | else |
| 729 | O.OS << " " ; |
| 730 | |
| 731 | O.OS << " " ; |
| 732 | if (std::optional<const char *> CodeName = GetCodeName( |
| 733 | CodeID: FreqPair.second, BlockID: Stat.first, BlockInfo, CurStreamType)) |
| 734 | O.OS << *CodeName << "\n" ; |
| 735 | else |
| 736 | O.OS << "UnknownCode" << FreqPair.second << "\n" ; |
| 737 | } |
| 738 | O.OS << "\n" ; |
| 739 | } |
| 740 | } |
| 741 | } |
| 742 | |
| 743 | Error BitcodeAnalyzer::parseBlock(unsigned BlockID, unsigned IndentLevel, |
| 744 | std::optional<BCDumpOptions> O, |
| 745 | std::optional<StringRef> CheckHash) { |
| 746 | std::string Indent(IndentLevel * 2, ' '); |
| 747 | uint64_t BlockBitStart = Stream.GetCurrentBitNo(); |
| 748 | |
| 749 | // Get the statistics for this BlockID. |
| 750 | PerBlockIDStats &BlockStats = BlockIDStats[BlockID]; |
| 751 | |
| 752 | BlockStats.NumInstances++; |
| 753 | |
| 754 | // BLOCKINFO is a special part of the stream. |
| 755 | bool DumpRecords = O.has_value(); |
| 756 | if (BlockID == bitc::BLOCKINFO_BLOCK_ID) { |
| 757 | if (O && !O->DumpBlockinfo) |
| 758 | O->OS << Indent << "<BLOCKINFO_BLOCK/>\n" ; |
| 759 | std::optional<BitstreamBlockInfo> NewBlockInfo; |
| 760 | if (Error E = Stream.ReadBlockInfoBlock(/*ReadBlockInfoNames=*/true) |
| 761 | .moveInto(Value&: NewBlockInfo)) |
| 762 | return E; |
| 763 | if (!NewBlockInfo) |
| 764 | return reportError(Message: "Malformed BlockInfoBlock" ); |
| 765 | BlockInfo = std::move(*NewBlockInfo); |
| 766 | if (Error Err = Stream.JumpToBit(BitNo: BlockBitStart)) |
| 767 | return Err; |
| 768 | // It's not really interesting to dump the contents of the blockinfo |
| 769 | // block, so only do it if the user explicitly requests it. |
| 770 | DumpRecords = O && O->DumpBlockinfo; |
| 771 | } |
| 772 | |
| 773 | unsigned NumWords = 0; |
| 774 | if (Error Err = Stream.EnterSubBlock(BlockID, NumWordsP: &NumWords)) |
| 775 | return Err; |
| 776 | |
| 777 | // Keep it for later, when we see a MODULE_HASH record |
| 778 | uint64_t BlockEntryPos = Stream.getCurrentByteNo(); |
| 779 | |
| 780 | std::optional<const char *> BlockName; |
| 781 | if (DumpRecords) { |
| 782 | O->OS << Indent << "<" ; |
| 783 | if ((BlockName = GetBlockName(BlockID, BlockInfo, CurStreamType))) |
| 784 | O->OS << *BlockName; |
| 785 | else |
| 786 | O->OS << "UnknownBlock" << BlockID; |
| 787 | |
| 788 | if (!O->Symbolic && BlockName) |
| 789 | O->OS << " BlockID=" << BlockID; |
| 790 | |
| 791 | O->OS << " NumWords=" << NumWords |
| 792 | << " BlockCodeSize=" << Stream.getAbbrevIDWidth() << ">\n" ; |
| 793 | } |
| 794 | |
| 795 | SmallVector<uint64_t, 64> Record; |
| 796 | |
| 797 | // Keep the offset to the metadata index if seen. |
| 798 | uint64_t MetadataIndexOffset = 0; |
| 799 | |
| 800 | // Read all the records for this block. |
| 801 | while (true) { |
| 802 | if (Stream.AtEndOfStream()) |
| 803 | return reportError(Message: "Premature end of bitstream" ); |
| 804 | |
| 805 | uint64_t RecordStartBit = Stream.GetCurrentBitNo(); |
| 806 | |
| 807 | BitstreamEntry Entry; |
| 808 | if (Error E = Stream.advance(Flags: BitstreamCursor::AF_DontAutoprocessAbbrevs) |
| 809 | .moveInto(Value&: Entry)) |
| 810 | return E; |
| 811 | |
| 812 | switch (Entry.Kind) { |
| 813 | case BitstreamEntry::Error: |
| 814 | return reportError(Message: "malformed bitcode file" ); |
| 815 | case BitstreamEntry::EndBlock: { |
| 816 | uint64_t BlockBitEnd = Stream.GetCurrentBitNo(); |
| 817 | BlockStats.NumBits += BlockBitEnd - BlockBitStart; |
| 818 | if (DumpRecords) { |
| 819 | O->OS << Indent << "</" ; |
| 820 | if (BlockName) |
| 821 | O->OS << *BlockName << ">\n" ; |
| 822 | else |
| 823 | O->OS << "UnknownBlock" << BlockID << ">\n" ; |
| 824 | } |
| 825 | return Error::success(); |
| 826 | } |
| 827 | |
| 828 | case BitstreamEntry::SubBlock: { |
| 829 | uint64_t SubBlockBitStart = Stream.GetCurrentBitNo(); |
| 830 | if (Error E = parseBlock(BlockID: Entry.ID, IndentLevel: IndentLevel + 1, O, CheckHash)) |
| 831 | return E; |
| 832 | ++BlockStats.NumSubBlocks; |
| 833 | uint64_t SubBlockBitEnd = Stream.GetCurrentBitNo(); |
| 834 | |
| 835 | // Don't include subblock sizes in the size of this block. |
| 836 | BlockBitStart += SubBlockBitEnd - SubBlockBitStart; |
| 837 | continue; |
| 838 | } |
| 839 | case BitstreamEntry::Record: |
| 840 | // The interesting case. |
| 841 | break; |
| 842 | } |
| 843 | |
| 844 | if (Entry.ID == bitc::DEFINE_ABBREV) { |
| 845 | if (Error Err = Stream.ReadAbbrevRecord()) |
| 846 | return Err; |
| 847 | ++BlockStats.NumAbbrevs; |
| 848 | continue; |
| 849 | } |
| 850 | |
| 851 | Record.clear(); |
| 852 | |
| 853 | ++BlockStats.NumRecords; |
| 854 | |
| 855 | StringRef Blob; |
| 856 | uint64_t CurrentRecordPos = Stream.GetCurrentBitNo(); |
| 857 | unsigned Code; |
| 858 | if (Error E = Stream.readRecord(AbbrevID: Entry.ID, Vals&: Record, Blob: &Blob).moveInto(Value&: Code)) |
| 859 | return E; |
| 860 | |
| 861 | // Increment the # occurrences of this code. |
| 862 | if (BlockStats.CodeFreq.size() <= Code) |
| 863 | BlockStats.CodeFreq.resize(new_size: Code + 1); |
| 864 | BlockStats.CodeFreq[Code].NumInstances++; |
| 865 | BlockStats.CodeFreq[Code].TotalBits += |
| 866 | Stream.GetCurrentBitNo() - RecordStartBit; |
| 867 | if (Entry.ID != bitc::UNABBREV_RECORD) { |
| 868 | BlockStats.CodeFreq[Code].NumAbbrev++; |
| 869 | ++BlockStats.NumAbbreviatedRecords; |
| 870 | } |
| 871 | |
| 872 | if (DumpRecords) { |
| 873 | O->OS << Indent << " <" ; |
| 874 | std::optional<const char *> CodeName = |
| 875 | GetCodeName(CodeID: Code, BlockID, BlockInfo, CurStreamType); |
| 876 | if (CodeName) |
| 877 | O->OS << *CodeName; |
| 878 | else |
| 879 | O->OS << "UnknownCode" << Code; |
| 880 | if (!O->Symbolic && CodeName) |
| 881 | O->OS << " codeid=" << Code; |
| 882 | const BitCodeAbbrev *Abbv = nullptr; |
| 883 | if (Entry.ID != bitc::UNABBREV_RECORD) { |
| 884 | Expected<const BitCodeAbbrev *> MaybeAbbv = Stream.getAbbrev(AbbrevID: Entry.ID); |
| 885 | if (!MaybeAbbv) |
| 886 | return MaybeAbbv.takeError(); |
| 887 | Abbv = MaybeAbbv.get(); |
| 888 | O->OS << " abbrevid=" << Entry.ID; |
| 889 | } |
| 890 | |
| 891 | for (unsigned i = 0, e = Record.size(); i != e; ++i) |
| 892 | O->OS << " op" << i << "=" << (int64_t)Record[i]; |
| 893 | |
| 894 | // If we found a metadata index, let's verify that we had an offset |
| 895 | // before and validate its forward reference offset was correct! |
| 896 | if (BlockID == bitc::METADATA_BLOCK_ID) { |
| 897 | if (Code == bitc::METADATA_INDEX_OFFSET) { |
| 898 | if (Record.size() != 2) |
| 899 | O->OS << "(Invalid record)" ; |
| 900 | else { |
| 901 | auto Offset = Record[0] + (Record[1] << 32); |
| 902 | MetadataIndexOffset = Stream.GetCurrentBitNo() + Offset; |
| 903 | } |
| 904 | } |
| 905 | if (Code == bitc::METADATA_INDEX) { |
| 906 | O->OS << " (offset " ; |
| 907 | if (MetadataIndexOffset == RecordStartBit) |
| 908 | O->OS << "match)" ; |
| 909 | else |
| 910 | O->OS << "mismatch: " << MetadataIndexOffset << " vs " |
| 911 | << RecordStartBit << ")" ; |
| 912 | } |
| 913 | } |
| 914 | |
| 915 | // If we found a module hash, let's verify that it matches! |
| 916 | if (BlockID == bitc::MODULE_BLOCK_ID && Code == bitc::MODULE_CODE_HASH && |
| 917 | CheckHash) { |
| 918 | if (Record.size() != 5) |
| 919 | O->OS << " (invalid)" ; |
| 920 | else { |
| 921 | // Recompute the hash and compare it to the one in the bitcode |
| 922 | SHA1 Hasher; |
| 923 | std::array<uint8_t, 20> Hash; |
| 924 | Hasher.update(Str: *CheckHash); |
| 925 | { |
| 926 | int BlockSize = (CurrentRecordPos / 8) - BlockEntryPos; |
| 927 | auto Ptr = Stream.getPointerToByte(ByteNo: BlockEntryPos, NumBytes: BlockSize); |
| 928 | Hasher.update(Data: ArrayRef<uint8_t>(Ptr, BlockSize)); |
| 929 | Hash = Hasher.result(); |
| 930 | } |
| 931 | std::array<uint8_t, 20> RecordedHash; |
| 932 | int Pos = 0; |
| 933 | for (auto &Val : Record) { |
| 934 | assert(!(Val >> 32) && "Unexpected high bits set" ); |
| 935 | support::endian::write32be(P: &RecordedHash[Pos], V: Val); |
| 936 | Pos += 4; |
| 937 | } |
| 938 | if (Hash == RecordedHash) |
| 939 | O->OS << " (match)" ; |
| 940 | else |
| 941 | O->OS << " (!mismatch!)" ; |
| 942 | } |
| 943 | } |
| 944 | |
| 945 | O->OS << "/>" ; |
| 946 | |
| 947 | if (Abbv) { |
| 948 | for (unsigned i = 1, e = Abbv->getNumOperandInfos(); i != e; ++i) { |
| 949 | const BitCodeAbbrevOp &Op = Abbv->getOperandInfo(N: i); |
| 950 | if (!Op.isEncoding() || Op.getEncoding() != BitCodeAbbrevOp::Array) |
| 951 | continue; |
| 952 | assert(i + 2 == e && "Array op not second to last" ); |
| 953 | std::string Str; |
| 954 | bool ArrayIsPrintable = true; |
| 955 | for (unsigned j = i - 1, je = Record.size(); j != je; ++j) { |
| 956 | if (!isPrint(C: static_cast<unsigned char>(Record[j]))) { |
| 957 | ArrayIsPrintable = false; |
| 958 | break; |
| 959 | } |
| 960 | Str += (char)Record[j]; |
| 961 | } |
| 962 | if (ArrayIsPrintable) |
| 963 | O->OS << " record string = '" << Str << "'" ; |
| 964 | break; |
| 965 | } |
| 966 | } |
| 967 | |
| 968 | if (Blob.data()) { |
| 969 | if (canDecodeBlob(Code, BlockID)) { |
| 970 | if (Error E = decodeMetadataStringsBlob(Indent, Record, Blob, OS&: O->OS)) |
| 971 | return E; |
| 972 | } else { |
| 973 | O->OS << " blob data = " ; |
| 974 | if (O->ShowBinaryBlobs) { |
| 975 | O->OS << "'" ; |
| 976 | O->OS.write_escaped(Str: Blob, /*hex=*/UseHexEscapes: true) << "'" ; |
| 977 | } else { |
| 978 | bool BlobIsPrintable = true; |
| 979 | for (char C : Blob) |
| 980 | if (!isPrint(C: static_cast<unsigned char>(C))) { |
| 981 | BlobIsPrintable = false; |
| 982 | break; |
| 983 | } |
| 984 | |
| 985 | if (BlobIsPrintable) |
| 986 | O->OS << "'" << Blob << "'" ; |
| 987 | else |
| 988 | O->OS << "unprintable, " << Blob.size() << " bytes." ; |
| 989 | } |
| 990 | } |
| 991 | } |
| 992 | |
| 993 | O->OS << "\n" ; |
| 994 | } |
| 995 | |
| 996 | // Make sure that we can skip the current record. |
| 997 | if (Error Err = Stream.JumpToBit(BitNo: CurrentRecordPos)) |
| 998 | return Err; |
| 999 | if (Expected<unsigned> Skipped = Stream.skipRecord(AbbrevID: Entry.ID)) |
| 1000 | ; // Do nothing. |
| 1001 | else |
| 1002 | return Skipped.takeError(); |
| 1003 | } |
| 1004 | } |
| 1005 | |
| 1006 | |