1 //===-- Symtab.cpp --------------------------------------------------------===// 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 <map> 10 #include <set> 11 12 #include "lldb/Core/DataFileCache.h" 13 #include "lldb/Core/Module.h" 14 #include "lldb/Core/RichManglingContext.h" 15 #include "lldb/Core/Section.h" 16 #include "lldb/Symbol/ObjectFile.h" 17 #include "lldb/Symbol/Symbol.h" 18 #include "lldb/Symbol/SymbolContext.h" 19 #include "lldb/Symbol/Symtab.h" 20 #include "lldb/Target/Language.h" 21 #include "lldb/Utility/DataEncoder.h" 22 #include "lldb/Utility/Endian.h" 23 #include "lldb/Utility/RegularExpression.h" 24 #include "lldb/Utility/Stream.h" 25 #include "lldb/Utility/Timer.h" 26 27 #include "llvm/ADT/ArrayRef.h" 28 #include "llvm/ADT/StringRef.h" 29 #include "llvm/Support/DJB.h" 30 31 using namespace lldb; 32 using namespace lldb_private; 33 34 Symtab::Symtab(ObjectFile *objfile) 35 : m_objfile(objfile), m_symbols(), m_file_addr_to_index(*this), 36 m_name_to_symbol_indices(), m_mutex(), 37 m_file_addr_to_index_computed(false), m_name_indexes_computed(false), 38 m_loaded_from_cache(false), m_saved_to_cache(false) { 39 m_name_to_symbol_indices.emplace(std::make_pair( 40 lldb::eFunctionNameTypeNone, UniqueCStringMap<uint32_t>())); 41 m_name_to_symbol_indices.emplace(std::make_pair( 42 lldb::eFunctionNameTypeBase, UniqueCStringMap<uint32_t>())); 43 m_name_to_symbol_indices.emplace(std::make_pair( 44 lldb::eFunctionNameTypeMethod, UniqueCStringMap<uint32_t>())); 45 m_name_to_symbol_indices.emplace(std::make_pair( 46 lldb::eFunctionNameTypeSelector, UniqueCStringMap<uint32_t>())); 47 } 48 49 Symtab::~Symtab() = default; 50 51 void Symtab::Reserve(size_t count) { 52 // Clients should grab the mutex from this symbol table and lock it manually 53 // when calling this function to avoid performance issues. 54 m_symbols.reserve(count); 55 } 56 57 Symbol *Symtab::Resize(size_t count) { 58 // Clients should grab the mutex from this symbol table and lock it manually 59 // when calling this function to avoid performance issues. 60 m_symbols.resize(count); 61 return m_symbols.empty() ? nullptr : &m_symbols[0]; 62 } 63 64 uint32_t Symtab::AddSymbol(const Symbol &symbol) { 65 // Clients should grab the mutex from this symbol table and lock it manually 66 // when calling this function to avoid performance issues. 67 uint32_t symbol_idx = m_symbols.size(); 68 auto &name_to_index = GetNameToSymbolIndexMap(lldb::eFunctionNameTypeNone); 69 name_to_index.Clear(); 70 m_file_addr_to_index.Clear(); 71 m_symbols.push_back(symbol); 72 m_file_addr_to_index_computed = false; 73 m_name_indexes_computed = false; 74 return symbol_idx; 75 } 76 77 size_t Symtab::GetNumSymbols() const { 78 std::lock_guard<std::recursive_mutex> guard(m_mutex); 79 return m_symbols.size(); 80 } 81 82 void Symtab::SectionFileAddressesChanged() { 83 auto &name_to_index = GetNameToSymbolIndexMap(lldb::eFunctionNameTypeNone); 84 name_to_index.Clear(); 85 m_file_addr_to_index_computed = false; 86 } 87 88 void Symtab::Dump(Stream *s, Target *target, SortOrder sort_order, 89 Mangled::NamePreference name_preference) { 90 std::lock_guard<std::recursive_mutex> guard(m_mutex); 91 92 // s->Printf("%.*p: ", (int)sizeof(void*) * 2, this); 93 s->Indent(); 94 const FileSpec &file_spec = m_objfile->GetFileSpec(); 95 const char *object_name = nullptr; 96 if (m_objfile->GetModule()) 97 object_name = m_objfile->GetModule()->GetObjectName().GetCString(); 98 99 if (file_spec) 100 s->Printf("Symtab, file = %s%s%s%s, num_symbols = %" PRIu64, 101 file_spec.GetPath().c_str(), object_name ? "(" : "", 102 object_name ? object_name : "", object_name ? ")" : "", 103 (uint64_t)m_symbols.size()); 104 else 105 s->Printf("Symtab, num_symbols = %" PRIu64 "", (uint64_t)m_symbols.size()); 106 107 if (!m_symbols.empty()) { 108 switch (sort_order) { 109 case eSortOrderNone: { 110 s->PutCString(":\n"); 111 DumpSymbolHeader(s); 112 const_iterator begin = m_symbols.begin(); 113 const_iterator end = m_symbols.end(); 114 for (const_iterator pos = m_symbols.begin(); pos != end; ++pos) { 115 s->Indent(); 116 pos->Dump(s, target, std::distance(begin, pos), name_preference); 117 } 118 } 119 break; 120 121 case eSortOrderByName: { 122 // Although we maintain a lookup by exact name map, the table isn't 123 // sorted by name. So we must make the ordered symbol list up ourselves. 124 s->PutCString(" (sorted by name):\n"); 125 DumpSymbolHeader(s); 126 127 std::multimap<llvm::StringRef, const Symbol *> name_map; 128 for (const_iterator pos = m_symbols.begin(), end = m_symbols.end(); 129 pos != end; ++pos) { 130 const char *name = pos->GetName().AsCString(); 131 if (name && name[0]) 132 name_map.insert(std::make_pair(name, &(*pos))); 133 } 134 135 for (const auto &name_to_symbol : name_map) { 136 const Symbol *symbol = name_to_symbol.second; 137 s->Indent(); 138 symbol->Dump(s, target, symbol - &m_symbols[0], name_preference); 139 } 140 } break; 141 142 case eSortOrderByAddress: 143 s->PutCString(" (sorted by address):\n"); 144 DumpSymbolHeader(s); 145 if (!m_file_addr_to_index_computed) 146 InitAddressIndexes(); 147 const size_t num_entries = m_file_addr_to_index.GetSize(); 148 for (size_t i = 0; i < num_entries; ++i) { 149 s->Indent(); 150 const uint32_t symbol_idx = m_file_addr_to_index.GetEntryRef(i).data; 151 m_symbols[symbol_idx].Dump(s, target, symbol_idx, name_preference); 152 } 153 break; 154 } 155 } else { 156 s->PutCString("\n"); 157 } 158 } 159 160 void Symtab::Dump(Stream *s, Target *target, std::vector<uint32_t> &indexes, 161 Mangled::NamePreference name_preference) const { 162 std::lock_guard<std::recursive_mutex> guard(m_mutex); 163 164 const size_t num_symbols = GetNumSymbols(); 165 // s->Printf("%.*p: ", (int)sizeof(void*) * 2, this); 166 s->Indent(); 167 s->Printf("Symtab %" PRIu64 " symbol indexes (%" PRIu64 " symbols total):\n", 168 (uint64_t)indexes.size(), (uint64_t)m_symbols.size()); 169 s->IndentMore(); 170 171 if (!indexes.empty()) { 172 std::vector<uint32_t>::const_iterator pos; 173 std::vector<uint32_t>::const_iterator end = indexes.end(); 174 DumpSymbolHeader(s); 175 for (pos = indexes.begin(); pos != end; ++pos) { 176 size_t idx = *pos; 177 if (idx < num_symbols) { 178 s->Indent(); 179 m_symbols[idx].Dump(s, target, idx, name_preference); 180 } 181 } 182 } 183 s->IndentLess(); 184 } 185 186 void Symtab::DumpSymbolHeader(Stream *s) { 187 s->Indent(" Debug symbol\n"); 188 s->Indent(" |Synthetic symbol\n"); 189 s->Indent(" ||Externally Visible\n"); 190 s->Indent(" |||\n"); 191 s->Indent("Index UserID DSX Type File Address/Value Load " 192 "Address Size Flags Name\n"); 193 s->Indent("------- ------ --- --------------- ------------------ " 194 "------------------ ------------------ ---------- " 195 "----------------------------------\n"); 196 } 197 198 static int CompareSymbolID(const void *key, const void *p) { 199 const user_id_t match_uid = *(const user_id_t *)key; 200 const user_id_t symbol_uid = ((const Symbol *)p)->GetID(); 201 if (match_uid < symbol_uid) 202 return -1; 203 if (match_uid > symbol_uid) 204 return 1; 205 return 0; 206 } 207 208 Symbol *Symtab::FindSymbolByID(lldb::user_id_t symbol_uid) const { 209 std::lock_guard<std::recursive_mutex> guard(m_mutex); 210 211 Symbol *symbol = 212 (Symbol *)::bsearch(&symbol_uid, &m_symbols[0], m_symbols.size(), 213 sizeof(m_symbols[0]), CompareSymbolID); 214 return symbol; 215 } 216 217 Symbol *Symtab::SymbolAtIndex(size_t idx) { 218 // Clients should grab the mutex from this symbol table and lock it manually 219 // when calling this function to avoid performance issues. 220 if (idx < m_symbols.size()) 221 return &m_symbols[idx]; 222 return nullptr; 223 } 224 225 const Symbol *Symtab::SymbolAtIndex(size_t idx) const { 226 // Clients should grab the mutex from this symbol table and lock it manually 227 // when calling this function to avoid performance issues. 228 if (idx < m_symbols.size()) 229 return &m_symbols[idx]; 230 return nullptr; 231 } 232 233 static bool lldb_skip_name(llvm::StringRef mangled, 234 Mangled::ManglingScheme scheme) { 235 switch (scheme) { 236 case Mangled::eManglingSchemeItanium: { 237 if (mangled.size() < 3 || !mangled.startswith("_Z")) 238 return true; 239 240 // Avoid the following types of symbols in the index. 241 switch (mangled[2]) { 242 case 'G': // guard variables 243 case 'T': // virtual tables, VTT structures, typeinfo structures + names 244 case 'Z': // named local entities (if we eventually handle 245 // eSymbolTypeData, we will want this back) 246 return true; 247 248 default: 249 break; 250 } 251 252 // Include this name in the index. 253 return false; 254 } 255 256 // No filters for this scheme yet. Include all names in indexing. 257 case Mangled::eManglingSchemeMSVC: 258 case Mangled::eManglingSchemeRustV0: 259 case Mangled::eManglingSchemeD: 260 return false; 261 262 // Don't try and demangle things we can't categorize. 263 case Mangled::eManglingSchemeNone: 264 return true; 265 } 266 llvm_unreachable("unknown scheme!"); 267 } 268 269 void Symtab::InitNameIndexes() { 270 // Protected function, no need to lock mutex... 271 if (!m_name_indexes_computed) { 272 m_name_indexes_computed = true; 273 ElapsedTime elapsed(m_objfile->GetModule()->GetSymtabIndexTime()); 274 LLDB_SCOPED_TIMER(); 275 276 // Collect all loaded language plugins. 277 std::vector<Language *> languages; 278 Language::ForEach([&languages](Language *l) { 279 languages.push_back(l); 280 return true; 281 }); 282 283 auto &name_to_index = GetNameToSymbolIndexMap(lldb::eFunctionNameTypeNone); 284 auto &basename_to_index = 285 GetNameToSymbolIndexMap(lldb::eFunctionNameTypeBase); 286 auto &method_to_index = 287 GetNameToSymbolIndexMap(lldb::eFunctionNameTypeMethod); 288 auto &selector_to_index = 289 GetNameToSymbolIndexMap(lldb::eFunctionNameTypeSelector); 290 // Create the name index vector to be able to quickly search by name 291 const size_t num_symbols = m_symbols.size(); 292 name_to_index.Reserve(num_symbols); 293 294 // The "const char *" in "class_contexts" and backlog::value_type::second 295 // must come from a ConstString::GetCString() 296 std::set<const char *> class_contexts; 297 std::vector<std::pair<NameToIndexMap::Entry, const char *>> backlog; 298 backlog.reserve(num_symbols / 2); 299 300 // Instantiation of the demangler is expensive, so better use a single one 301 // for all entries during batch processing. 302 RichManglingContext rmc; 303 for (uint32_t value = 0; value < num_symbols; ++value) { 304 Symbol *symbol = &m_symbols[value]; 305 306 // Don't let trampolines get into the lookup by name map If we ever need 307 // the trampoline symbols to be searchable by name we can remove this and 308 // then possibly add a new bool to any of the Symtab functions that 309 // lookup symbols by name to indicate if they want trampolines. We also 310 // don't want any synthetic symbols with auto generated names in the 311 // name lookups. 312 if (symbol->IsTrampoline() || symbol->IsSyntheticWithAutoGeneratedName()) 313 continue; 314 315 // If the symbol's name string matched a Mangled::ManglingScheme, it is 316 // stored in the mangled field. 317 Mangled &mangled = symbol->GetMangled(); 318 if (ConstString name = mangled.GetMangledName()) { 319 name_to_index.Append(name, value); 320 321 if (symbol->ContainsLinkerAnnotations()) { 322 // If the symbol has linker annotations, also add the version without 323 // the annotations. 324 ConstString stripped = ConstString( 325 m_objfile->StripLinkerSymbolAnnotations(name.GetStringRef())); 326 name_to_index.Append(stripped, value); 327 } 328 329 const SymbolType type = symbol->GetType(); 330 if (type == eSymbolTypeCode || type == eSymbolTypeResolver) { 331 if (mangled.DemangleWithRichManglingInfo(rmc, lldb_skip_name)) 332 RegisterMangledNameEntry(value, class_contexts, backlog, rmc); 333 } 334 } 335 336 // Symbol name strings that didn't match a Mangled::ManglingScheme, are 337 // stored in the demangled field. 338 if (ConstString name = mangled.GetDemangledName()) { 339 name_to_index.Append(name, value); 340 341 if (symbol->ContainsLinkerAnnotations()) { 342 // If the symbol has linker annotations, also add the version without 343 // the annotations. 344 name = ConstString( 345 m_objfile->StripLinkerSymbolAnnotations(name.GetStringRef())); 346 name_to_index.Append(name, value); 347 } 348 349 // If the demangled name turns out to be an ObjC name, and is a category 350 // name, add the version without categories to the index too. 351 for (Language *lang : languages) { 352 for (auto variant : lang->GetMethodNameVariants(name)) { 353 if (variant.GetType() & lldb::eFunctionNameTypeSelector) 354 selector_to_index.Append(variant.GetName(), value); 355 else if (variant.GetType() & lldb::eFunctionNameTypeFull) 356 name_to_index.Append(variant.GetName(), value); 357 else if (variant.GetType() & lldb::eFunctionNameTypeMethod) 358 method_to_index.Append(variant.GetName(), value); 359 else if (variant.GetType() & lldb::eFunctionNameTypeBase) 360 basename_to_index.Append(variant.GetName(), value); 361 } 362 } 363 } 364 } 365 366 for (const auto &record : backlog) { 367 RegisterBacklogEntry(record.first, record.second, class_contexts); 368 } 369 370 name_to_index.Sort(); 371 name_to_index.SizeToFit(); 372 selector_to_index.Sort(); 373 selector_to_index.SizeToFit(); 374 basename_to_index.Sort(); 375 basename_to_index.SizeToFit(); 376 method_to_index.Sort(); 377 method_to_index.SizeToFit(); 378 } 379 } 380 381 void Symtab::RegisterMangledNameEntry( 382 uint32_t value, std::set<const char *> &class_contexts, 383 std::vector<std::pair<NameToIndexMap::Entry, const char *>> &backlog, 384 RichManglingContext &rmc) { 385 // Only register functions that have a base name. 386 rmc.ParseFunctionBaseName(); 387 llvm::StringRef base_name = rmc.GetBufferRef(); 388 if (base_name.empty()) 389 return; 390 391 // The base name will be our entry's name. 392 NameToIndexMap::Entry entry(ConstString(base_name), value); 393 394 rmc.ParseFunctionDeclContextName(); 395 llvm::StringRef decl_context = rmc.GetBufferRef(); 396 397 // Register functions with no context. 398 if (decl_context.empty()) { 399 // This has to be a basename 400 auto &basename_to_index = 401 GetNameToSymbolIndexMap(lldb::eFunctionNameTypeBase); 402 basename_to_index.Append(entry); 403 // If there is no context (no namespaces or class scopes that come before 404 // the function name) then this also could be a fullname. 405 auto &name_to_index = GetNameToSymbolIndexMap(lldb::eFunctionNameTypeNone); 406 name_to_index.Append(entry); 407 return; 408 } 409 410 // Make sure we have a pool-string pointer and see if we already know the 411 // context name. 412 const char *decl_context_ccstr = ConstString(decl_context).GetCString(); 413 auto it = class_contexts.find(decl_context_ccstr); 414 415 auto &method_to_index = 416 GetNameToSymbolIndexMap(lldb::eFunctionNameTypeMethod); 417 // Register constructors and destructors. They are methods and create 418 // declaration contexts. 419 if (rmc.IsCtorOrDtor()) { 420 method_to_index.Append(entry); 421 if (it == class_contexts.end()) 422 class_contexts.insert(it, decl_context_ccstr); 423 return; 424 } 425 426 // Register regular methods with a known declaration context. 427 if (it != class_contexts.end()) { 428 method_to_index.Append(entry); 429 return; 430 } 431 432 // Regular methods in unknown declaration contexts are put to the backlog. We 433 // will revisit them once we processed all remaining symbols. 434 backlog.push_back(std::make_pair(entry, decl_context_ccstr)); 435 } 436 437 void Symtab::RegisterBacklogEntry( 438 const NameToIndexMap::Entry &entry, const char *decl_context, 439 const std::set<const char *> &class_contexts) { 440 auto &method_to_index = 441 GetNameToSymbolIndexMap(lldb::eFunctionNameTypeMethod); 442 auto it = class_contexts.find(decl_context); 443 if (it != class_contexts.end()) { 444 method_to_index.Append(entry); 445 } else { 446 // If we got here, we have something that had a context (was inside 447 // a namespace or class) yet we don't know the entry 448 method_to_index.Append(entry); 449 auto &basename_to_index = 450 GetNameToSymbolIndexMap(lldb::eFunctionNameTypeBase); 451 basename_to_index.Append(entry); 452 } 453 } 454 455 void Symtab::PreloadSymbols() { 456 std::lock_guard<std::recursive_mutex> guard(m_mutex); 457 InitNameIndexes(); 458 } 459 460 void Symtab::AppendSymbolNamesToMap(const IndexCollection &indexes, 461 bool add_demangled, bool add_mangled, 462 NameToIndexMap &name_to_index_map) const { 463 LLDB_SCOPED_TIMER(); 464 if (add_demangled || add_mangled) { 465 std::lock_guard<std::recursive_mutex> guard(m_mutex); 466 467 // Create the name index vector to be able to quickly search by name 468 const size_t num_indexes = indexes.size(); 469 for (size_t i = 0; i < num_indexes; ++i) { 470 uint32_t value = indexes[i]; 471 assert(i < m_symbols.size()); 472 const Symbol *symbol = &m_symbols[value]; 473 474 const Mangled &mangled = symbol->GetMangled(); 475 if (add_demangled) { 476 if (ConstString name = mangled.GetDemangledName()) 477 name_to_index_map.Append(name, value); 478 } 479 480 if (add_mangled) { 481 if (ConstString name = mangled.GetMangledName()) 482 name_to_index_map.Append(name, value); 483 } 484 } 485 } 486 } 487 488 uint32_t Symtab::AppendSymbolIndexesWithType(SymbolType symbol_type, 489 std::vector<uint32_t> &indexes, 490 uint32_t start_idx, 491 uint32_t end_index) const { 492 std::lock_guard<std::recursive_mutex> guard(m_mutex); 493 494 uint32_t prev_size = indexes.size(); 495 496 const uint32_t count = std::min<uint32_t>(m_symbols.size(), end_index); 497 498 for (uint32_t i = start_idx; i < count; ++i) { 499 if (symbol_type == eSymbolTypeAny || m_symbols[i].GetType() == symbol_type) 500 indexes.push_back(i); 501 } 502 503 return indexes.size() - prev_size; 504 } 505 506 uint32_t Symtab::AppendSymbolIndexesWithTypeAndFlagsValue( 507 SymbolType symbol_type, uint32_t flags_value, 508 std::vector<uint32_t> &indexes, uint32_t start_idx, 509 uint32_t end_index) const { 510 std::lock_guard<std::recursive_mutex> guard(m_mutex); 511 512 uint32_t prev_size = indexes.size(); 513 514 const uint32_t count = std::min<uint32_t>(m_symbols.size(), end_index); 515 516 for (uint32_t i = start_idx; i < count; ++i) { 517 if ((symbol_type == eSymbolTypeAny || 518 m_symbols[i].GetType() == symbol_type) && 519 m_symbols[i].GetFlags() == flags_value) 520 indexes.push_back(i); 521 } 522 523 return indexes.size() - prev_size; 524 } 525 526 uint32_t Symtab::AppendSymbolIndexesWithType(SymbolType symbol_type, 527 Debug symbol_debug_type, 528 Visibility symbol_visibility, 529 std::vector<uint32_t> &indexes, 530 uint32_t start_idx, 531 uint32_t end_index) const { 532 std::lock_guard<std::recursive_mutex> guard(m_mutex); 533 534 uint32_t prev_size = indexes.size(); 535 536 const uint32_t count = std::min<uint32_t>(m_symbols.size(), end_index); 537 538 for (uint32_t i = start_idx; i < count; ++i) { 539 if (symbol_type == eSymbolTypeAny || 540 m_symbols[i].GetType() == symbol_type) { 541 if (CheckSymbolAtIndex(i, symbol_debug_type, symbol_visibility)) 542 indexes.push_back(i); 543 } 544 } 545 546 return indexes.size() - prev_size; 547 } 548 549 uint32_t Symtab::GetIndexForSymbol(const Symbol *symbol) const { 550 if (!m_symbols.empty()) { 551 const Symbol *first_symbol = &m_symbols[0]; 552 if (symbol >= first_symbol && symbol < first_symbol + m_symbols.size()) 553 return symbol - first_symbol; 554 } 555 return UINT32_MAX; 556 } 557 558 struct SymbolSortInfo { 559 const bool sort_by_load_addr; 560 const Symbol *symbols; 561 }; 562 563 namespace { 564 struct SymbolIndexComparator { 565 const std::vector<Symbol> &symbols; 566 std::vector<lldb::addr_t> &addr_cache; 567 568 // Getting from the symbol to the Address to the File Address involves some 569 // work. Since there are potentially many symbols here, and we're using this 570 // for sorting so we're going to be computing the address many times, cache 571 // that in addr_cache. The array passed in has to be the same size as the 572 // symbols array passed into the member variable symbols, and should be 573 // initialized with LLDB_INVALID_ADDRESS. 574 // NOTE: You have to make addr_cache externally and pass it in because 575 // std::stable_sort 576 // makes copies of the comparator it is initially passed in, and you end up 577 // spending huge amounts of time copying this array... 578 579 SymbolIndexComparator(const std::vector<Symbol> &s, 580 std::vector<lldb::addr_t> &a) 581 : symbols(s), addr_cache(a) { 582 assert(symbols.size() == addr_cache.size()); 583 } 584 bool operator()(uint32_t index_a, uint32_t index_b) { 585 addr_t value_a = addr_cache[index_a]; 586 if (value_a == LLDB_INVALID_ADDRESS) { 587 value_a = symbols[index_a].GetAddressRef().GetFileAddress(); 588 addr_cache[index_a] = value_a; 589 } 590 591 addr_t value_b = addr_cache[index_b]; 592 if (value_b == LLDB_INVALID_ADDRESS) { 593 value_b = symbols[index_b].GetAddressRef().GetFileAddress(); 594 addr_cache[index_b] = value_b; 595 } 596 597 if (value_a == value_b) { 598 // The if the values are equal, use the original symbol user ID 599 lldb::user_id_t uid_a = symbols[index_a].GetID(); 600 lldb::user_id_t uid_b = symbols[index_b].GetID(); 601 if (uid_a < uid_b) 602 return true; 603 if (uid_a > uid_b) 604 return false; 605 return false; 606 } else if (value_a < value_b) 607 return true; 608 609 return false; 610 } 611 }; 612 } 613 614 void Symtab::SortSymbolIndexesByValue(std::vector<uint32_t> &indexes, 615 bool remove_duplicates) const { 616 std::lock_guard<std::recursive_mutex> guard(m_mutex); 617 LLDB_SCOPED_TIMER(); 618 // No need to sort if we have zero or one items... 619 if (indexes.size() <= 1) 620 return; 621 622 // Sort the indexes in place using std::stable_sort. 623 // NOTE: The use of std::stable_sort instead of llvm::sort here is strictly 624 // for performance, not correctness. The indexes vector tends to be "close" 625 // to sorted, which the stable sort handles better. 626 627 std::vector<lldb::addr_t> addr_cache(m_symbols.size(), LLDB_INVALID_ADDRESS); 628 629 SymbolIndexComparator comparator(m_symbols, addr_cache); 630 std::stable_sort(indexes.begin(), indexes.end(), comparator); 631 632 // Remove any duplicates if requested 633 if (remove_duplicates) { 634 auto last = std::unique(indexes.begin(), indexes.end()); 635 indexes.erase(last, indexes.end()); 636 } 637 } 638 639 uint32_t Symtab::GetNameIndexes(ConstString symbol_name, 640 std::vector<uint32_t> &indexes) { 641 auto &name_to_index = GetNameToSymbolIndexMap(lldb::eFunctionNameTypeNone); 642 const uint32_t count = name_to_index.GetValues(symbol_name, indexes); 643 if (count) 644 return count; 645 // Synthetic symbol names are not added to the name indexes, but they start 646 // with a prefix and end with a the symbol UserID. This allows users to find 647 // these symbols without having to add them to the name indexes. These 648 // queries will not happen very often since the names don't mean anything, so 649 // performance is not paramount in this case. 650 llvm::StringRef name = symbol_name.GetStringRef(); 651 // String the synthetic prefix if the name starts with it. 652 if (!name.consume_front(Symbol::GetSyntheticSymbolPrefix())) 653 return 0; // Not a synthetic symbol name 654 655 // Extract the user ID from the symbol name 656 unsigned long long uid = 0; 657 if (getAsUnsignedInteger(name, /*Radix=*/10, uid)) 658 return 0; // Failed to extract the user ID as an integer 659 Symbol *symbol = FindSymbolByID(uid); 660 if (symbol == nullptr) 661 return 0; 662 const uint32_t symbol_idx = GetIndexForSymbol(symbol); 663 if (symbol_idx == UINT32_MAX) 664 return 0; 665 indexes.push_back(symbol_idx); 666 return 1; 667 } 668 669 uint32_t Symtab::AppendSymbolIndexesWithName(ConstString symbol_name, 670 std::vector<uint32_t> &indexes) { 671 std::lock_guard<std::recursive_mutex> guard(m_mutex); 672 673 if (symbol_name) { 674 if (!m_name_indexes_computed) 675 InitNameIndexes(); 676 677 return GetNameIndexes(symbol_name, indexes); 678 } 679 return 0; 680 } 681 682 uint32_t Symtab::AppendSymbolIndexesWithName(ConstString symbol_name, 683 Debug symbol_debug_type, 684 Visibility symbol_visibility, 685 std::vector<uint32_t> &indexes) { 686 std::lock_guard<std::recursive_mutex> guard(m_mutex); 687 688 LLDB_SCOPED_TIMER(); 689 if (symbol_name) { 690 const size_t old_size = indexes.size(); 691 if (!m_name_indexes_computed) 692 InitNameIndexes(); 693 694 std::vector<uint32_t> all_name_indexes; 695 const size_t name_match_count = 696 GetNameIndexes(symbol_name, all_name_indexes); 697 for (size_t i = 0; i < name_match_count; ++i) { 698 if (CheckSymbolAtIndex(all_name_indexes[i], symbol_debug_type, 699 symbol_visibility)) 700 indexes.push_back(all_name_indexes[i]); 701 } 702 return indexes.size() - old_size; 703 } 704 return 0; 705 } 706 707 uint32_t 708 Symtab::AppendSymbolIndexesWithNameAndType(ConstString symbol_name, 709 SymbolType symbol_type, 710 std::vector<uint32_t> &indexes) { 711 std::lock_guard<std::recursive_mutex> guard(m_mutex); 712 713 if (AppendSymbolIndexesWithName(symbol_name, indexes) > 0) { 714 std::vector<uint32_t>::iterator pos = indexes.begin(); 715 while (pos != indexes.end()) { 716 if (symbol_type == eSymbolTypeAny || 717 m_symbols[*pos].GetType() == symbol_type) 718 ++pos; 719 else 720 pos = indexes.erase(pos); 721 } 722 } 723 return indexes.size(); 724 } 725 726 uint32_t Symtab::AppendSymbolIndexesWithNameAndType( 727 ConstString symbol_name, SymbolType symbol_type, 728 Debug symbol_debug_type, Visibility symbol_visibility, 729 std::vector<uint32_t> &indexes) { 730 std::lock_guard<std::recursive_mutex> guard(m_mutex); 731 732 if (AppendSymbolIndexesWithName(symbol_name, symbol_debug_type, 733 symbol_visibility, indexes) > 0) { 734 std::vector<uint32_t>::iterator pos = indexes.begin(); 735 while (pos != indexes.end()) { 736 if (symbol_type == eSymbolTypeAny || 737 m_symbols[*pos].GetType() == symbol_type) 738 ++pos; 739 else 740 pos = indexes.erase(pos); 741 } 742 } 743 return indexes.size(); 744 } 745 746 uint32_t Symtab::AppendSymbolIndexesMatchingRegExAndType( 747 const RegularExpression ®exp, SymbolType symbol_type, 748 std::vector<uint32_t> &indexes) { 749 std::lock_guard<std::recursive_mutex> guard(m_mutex); 750 751 uint32_t prev_size = indexes.size(); 752 uint32_t sym_end = m_symbols.size(); 753 754 for (uint32_t i = 0; i < sym_end; i++) { 755 if (symbol_type == eSymbolTypeAny || 756 m_symbols[i].GetType() == symbol_type) { 757 const char *name = m_symbols[i].GetName().AsCString(); 758 if (name) { 759 if (regexp.Execute(name)) 760 indexes.push_back(i); 761 } 762 } 763 } 764 return indexes.size() - prev_size; 765 } 766 767 uint32_t Symtab::AppendSymbolIndexesMatchingRegExAndType( 768 const RegularExpression ®exp, SymbolType symbol_type, 769 Debug symbol_debug_type, Visibility symbol_visibility, 770 std::vector<uint32_t> &indexes) { 771 std::lock_guard<std::recursive_mutex> guard(m_mutex); 772 773 uint32_t prev_size = indexes.size(); 774 uint32_t sym_end = m_symbols.size(); 775 776 for (uint32_t i = 0; i < sym_end; i++) { 777 if (symbol_type == eSymbolTypeAny || 778 m_symbols[i].GetType() == symbol_type) { 779 if (!CheckSymbolAtIndex(i, symbol_debug_type, symbol_visibility)) 780 continue; 781 782 const char *name = m_symbols[i].GetName().AsCString(); 783 if (name) { 784 if (regexp.Execute(name)) 785 indexes.push_back(i); 786 } 787 } 788 } 789 return indexes.size() - prev_size; 790 } 791 792 Symbol *Symtab::FindSymbolWithType(SymbolType symbol_type, 793 Debug symbol_debug_type, 794 Visibility symbol_visibility, 795 uint32_t &start_idx) { 796 std::lock_guard<std::recursive_mutex> guard(m_mutex); 797 798 const size_t count = m_symbols.size(); 799 for (size_t idx = start_idx; idx < count; ++idx) { 800 if (symbol_type == eSymbolTypeAny || 801 m_symbols[idx].GetType() == symbol_type) { 802 if (CheckSymbolAtIndex(idx, symbol_debug_type, symbol_visibility)) { 803 start_idx = idx; 804 return &m_symbols[idx]; 805 } 806 } 807 } 808 return nullptr; 809 } 810 811 void 812 Symtab::FindAllSymbolsWithNameAndType(ConstString name, 813 SymbolType symbol_type, 814 std::vector<uint32_t> &symbol_indexes) { 815 std::lock_guard<std::recursive_mutex> guard(m_mutex); 816 817 // Initialize all of the lookup by name indexes before converting NAME to a 818 // uniqued string NAME_STR below. 819 if (!m_name_indexes_computed) 820 InitNameIndexes(); 821 822 if (name) { 823 // The string table did have a string that matched, but we need to check 824 // the symbols and match the symbol_type if any was given. 825 AppendSymbolIndexesWithNameAndType(name, symbol_type, symbol_indexes); 826 } 827 } 828 829 void Symtab::FindAllSymbolsWithNameAndType( 830 ConstString name, SymbolType symbol_type, Debug symbol_debug_type, 831 Visibility symbol_visibility, std::vector<uint32_t> &symbol_indexes) { 832 std::lock_guard<std::recursive_mutex> guard(m_mutex); 833 834 LLDB_SCOPED_TIMER(); 835 // Initialize all of the lookup by name indexes before converting NAME to a 836 // uniqued string NAME_STR below. 837 if (!m_name_indexes_computed) 838 InitNameIndexes(); 839 840 if (name) { 841 // The string table did have a string that matched, but we need to check 842 // the symbols and match the symbol_type if any was given. 843 AppendSymbolIndexesWithNameAndType(name, symbol_type, symbol_debug_type, 844 symbol_visibility, symbol_indexes); 845 } 846 } 847 848 void Symtab::FindAllSymbolsMatchingRexExAndType( 849 const RegularExpression ®ex, SymbolType symbol_type, 850 Debug symbol_debug_type, Visibility symbol_visibility, 851 std::vector<uint32_t> &symbol_indexes) { 852 std::lock_guard<std::recursive_mutex> guard(m_mutex); 853 854 AppendSymbolIndexesMatchingRegExAndType(regex, symbol_type, symbol_debug_type, 855 symbol_visibility, symbol_indexes); 856 } 857 858 Symbol *Symtab::FindFirstSymbolWithNameAndType(ConstString name, 859 SymbolType symbol_type, 860 Debug symbol_debug_type, 861 Visibility symbol_visibility) { 862 std::lock_guard<std::recursive_mutex> guard(m_mutex); 863 LLDB_SCOPED_TIMER(); 864 if (!m_name_indexes_computed) 865 InitNameIndexes(); 866 867 if (name) { 868 std::vector<uint32_t> matching_indexes; 869 // The string table did have a string that matched, but we need to check 870 // the symbols and match the symbol_type if any was given. 871 if (AppendSymbolIndexesWithNameAndType(name, symbol_type, symbol_debug_type, 872 symbol_visibility, 873 matching_indexes)) { 874 std::vector<uint32_t>::const_iterator pos, end = matching_indexes.end(); 875 for (pos = matching_indexes.begin(); pos != end; ++pos) { 876 Symbol *symbol = SymbolAtIndex(*pos); 877 878 if (symbol->Compare(name, symbol_type)) 879 return symbol; 880 } 881 } 882 } 883 return nullptr; 884 } 885 886 typedef struct { 887 const Symtab *symtab; 888 const addr_t file_addr; 889 Symbol *match_symbol; 890 const uint32_t *match_index_ptr; 891 addr_t match_offset; 892 } SymbolSearchInfo; 893 894 // Add all the section file start address & size to the RangeVector, recusively 895 // adding any children sections. 896 static void AddSectionsToRangeMap(SectionList *sectlist, 897 RangeVector<addr_t, addr_t> §ion_ranges) { 898 const int num_sections = sectlist->GetNumSections(0); 899 for (int i = 0; i < num_sections; i++) { 900 SectionSP sect_sp = sectlist->GetSectionAtIndex(i); 901 if (sect_sp) { 902 SectionList &child_sectlist = sect_sp->GetChildren(); 903 904 // If this section has children, add the children to the RangeVector. 905 // Else add this section to the RangeVector. 906 if (child_sectlist.GetNumSections(0) > 0) { 907 AddSectionsToRangeMap(&child_sectlist, section_ranges); 908 } else { 909 size_t size = sect_sp->GetByteSize(); 910 if (size > 0) { 911 addr_t base_addr = sect_sp->GetFileAddress(); 912 RangeVector<addr_t, addr_t>::Entry entry; 913 entry.SetRangeBase(base_addr); 914 entry.SetByteSize(size); 915 section_ranges.Append(entry); 916 } 917 } 918 } 919 } 920 } 921 922 void Symtab::InitAddressIndexes() { 923 // Protected function, no need to lock mutex... 924 if (!m_file_addr_to_index_computed && !m_symbols.empty()) { 925 m_file_addr_to_index_computed = true; 926 927 FileRangeToIndexMap::Entry entry; 928 const_iterator begin = m_symbols.begin(); 929 const_iterator end = m_symbols.end(); 930 for (const_iterator pos = m_symbols.begin(); pos != end; ++pos) { 931 if (pos->ValueIsAddress()) { 932 entry.SetRangeBase(pos->GetAddressRef().GetFileAddress()); 933 entry.SetByteSize(pos->GetByteSize()); 934 entry.data = std::distance(begin, pos); 935 m_file_addr_to_index.Append(entry); 936 } 937 } 938 const size_t num_entries = m_file_addr_to_index.GetSize(); 939 if (num_entries > 0) { 940 m_file_addr_to_index.Sort(); 941 942 // Create a RangeVector with the start & size of all the sections for 943 // this objfile. We'll need to check this for any FileRangeToIndexMap 944 // entries with an uninitialized size, which could potentially be a large 945 // number so reconstituting the weak pointer is busywork when it is 946 // invariant information. 947 SectionList *sectlist = m_objfile->GetSectionList(); 948 RangeVector<addr_t, addr_t> section_ranges; 949 if (sectlist) { 950 AddSectionsToRangeMap(sectlist, section_ranges); 951 section_ranges.Sort(); 952 } 953 954 // Iterate through the FileRangeToIndexMap and fill in the size for any 955 // entries that didn't already have a size from the Symbol (e.g. if we 956 // have a plain linker symbol with an address only, instead of debug info 957 // where we get an address and a size and a type, etc.) 958 for (size_t i = 0; i < num_entries; i++) { 959 FileRangeToIndexMap::Entry *entry = 960 m_file_addr_to_index.GetMutableEntryAtIndex(i); 961 if (entry->GetByteSize() == 0) { 962 addr_t curr_base_addr = entry->GetRangeBase(); 963 const RangeVector<addr_t, addr_t>::Entry *containing_section = 964 section_ranges.FindEntryThatContains(curr_base_addr); 965 966 // Use the end of the section as the default max size of the symbol 967 addr_t sym_size = 0; 968 if (containing_section) { 969 sym_size = 970 containing_section->GetByteSize() - 971 (entry->GetRangeBase() - containing_section->GetRangeBase()); 972 } 973 974 for (size_t j = i; j < num_entries; j++) { 975 FileRangeToIndexMap::Entry *next_entry = 976 m_file_addr_to_index.GetMutableEntryAtIndex(j); 977 addr_t next_base_addr = next_entry->GetRangeBase(); 978 if (next_base_addr > curr_base_addr) { 979 addr_t size_to_next_symbol = next_base_addr - curr_base_addr; 980 981 // Take the difference between this symbol and the next one as 982 // its size, if it is less than the size of the section. 983 if (sym_size == 0 || size_to_next_symbol < sym_size) { 984 sym_size = size_to_next_symbol; 985 } 986 break; 987 } 988 } 989 990 if (sym_size > 0) { 991 entry->SetByteSize(sym_size); 992 Symbol &symbol = m_symbols[entry->data]; 993 symbol.SetByteSize(sym_size); 994 symbol.SetSizeIsSynthesized(true); 995 } 996 } 997 } 998 999 // Sort again in case the range size changes the ordering 1000 m_file_addr_to_index.Sort(); 1001 } 1002 } 1003 } 1004 1005 void Symtab::Finalize() { 1006 std::lock_guard<std::recursive_mutex> guard(m_mutex); 1007 // Calculate the size of symbols inside InitAddressIndexes. 1008 InitAddressIndexes(); 1009 // Shrink to fit the symbols so we don't waste memory 1010 if (m_symbols.capacity() > m_symbols.size()) { 1011 collection new_symbols(m_symbols.begin(), m_symbols.end()); 1012 m_symbols.swap(new_symbols); 1013 } 1014 SaveToCache(); 1015 } 1016 1017 Symbol *Symtab::FindSymbolAtFileAddress(addr_t file_addr) { 1018 std::lock_guard<std::recursive_mutex> guard(m_mutex); 1019 if (!m_file_addr_to_index_computed) 1020 InitAddressIndexes(); 1021 1022 const FileRangeToIndexMap::Entry *entry = 1023 m_file_addr_to_index.FindEntryStartsAt(file_addr); 1024 if (entry) { 1025 Symbol *symbol = SymbolAtIndex(entry->data); 1026 if (symbol->GetFileAddress() == file_addr) 1027 return symbol; 1028 } 1029 return nullptr; 1030 } 1031 1032 Symbol *Symtab::FindSymbolContainingFileAddress(addr_t file_addr) { 1033 std::lock_guard<std::recursive_mutex> guard(m_mutex); 1034 1035 if (!m_file_addr_to_index_computed) 1036 InitAddressIndexes(); 1037 1038 const FileRangeToIndexMap::Entry *entry = 1039 m_file_addr_to_index.FindEntryThatContains(file_addr); 1040 if (entry) { 1041 Symbol *symbol = SymbolAtIndex(entry->data); 1042 if (symbol->ContainsFileAddress(file_addr)) 1043 return symbol; 1044 } 1045 return nullptr; 1046 } 1047 1048 void Symtab::ForEachSymbolContainingFileAddress( 1049 addr_t file_addr, std::function<bool(Symbol *)> const &callback) { 1050 std::lock_guard<std::recursive_mutex> guard(m_mutex); 1051 1052 if (!m_file_addr_to_index_computed) 1053 InitAddressIndexes(); 1054 1055 std::vector<uint32_t> all_addr_indexes; 1056 1057 // Get all symbols with file_addr 1058 const size_t addr_match_count = 1059 m_file_addr_to_index.FindEntryIndexesThatContain(file_addr, 1060 all_addr_indexes); 1061 1062 for (size_t i = 0; i < addr_match_count; ++i) { 1063 Symbol *symbol = SymbolAtIndex(all_addr_indexes[i]); 1064 if (symbol->ContainsFileAddress(file_addr)) { 1065 if (!callback(symbol)) 1066 break; 1067 } 1068 } 1069 } 1070 1071 void Symtab::SymbolIndicesToSymbolContextList( 1072 std::vector<uint32_t> &symbol_indexes, SymbolContextList &sc_list) { 1073 // No need to protect this call using m_mutex all other method calls are 1074 // already thread safe. 1075 1076 const bool merge_symbol_into_function = true; 1077 size_t num_indices = symbol_indexes.size(); 1078 if (num_indices > 0) { 1079 SymbolContext sc; 1080 sc.module_sp = m_objfile->GetModule(); 1081 for (size_t i = 0; i < num_indices; i++) { 1082 sc.symbol = SymbolAtIndex(symbol_indexes[i]); 1083 if (sc.symbol) 1084 sc_list.AppendIfUnique(sc, merge_symbol_into_function); 1085 } 1086 } 1087 } 1088 1089 void Symtab::FindFunctionSymbols(ConstString name, uint32_t name_type_mask, 1090 SymbolContextList &sc_list) { 1091 std::vector<uint32_t> symbol_indexes; 1092 1093 // eFunctionNameTypeAuto should be pre-resolved by a call to 1094 // Module::LookupInfo::LookupInfo() 1095 assert((name_type_mask & eFunctionNameTypeAuto) == 0); 1096 1097 if (name_type_mask & (eFunctionNameTypeBase | eFunctionNameTypeFull)) { 1098 std::vector<uint32_t> temp_symbol_indexes; 1099 FindAllSymbolsWithNameAndType(name, eSymbolTypeAny, temp_symbol_indexes); 1100 1101 unsigned temp_symbol_indexes_size = temp_symbol_indexes.size(); 1102 if (temp_symbol_indexes_size > 0) { 1103 std::lock_guard<std::recursive_mutex> guard(m_mutex); 1104 for (unsigned i = 0; i < temp_symbol_indexes_size; i++) { 1105 SymbolContext sym_ctx; 1106 sym_ctx.symbol = SymbolAtIndex(temp_symbol_indexes[i]); 1107 if (sym_ctx.symbol) { 1108 switch (sym_ctx.symbol->GetType()) { 1109 case eSymbolTypeCode: 1110 case eSymbolTypeResolver: 1111 case eSymbolTypeReExported: 1112 case eSymbolTypeAbsolute: 1113 symbol_indexes.push_back(temp_symbol_indexes[i]); 1114 break; 1115 default: 1116 break; 1117 } 1118 } 1119 } 1120 } 1121 } 1122 1123 if (!m_name_indexes_computed) 1124 InitNameIndexes(); 1125 1126 for (lldb::FunctionNameType type : 1127 {lldb::eFunctionNameTypeBase, lldb::eFunctionNameTypeMethod, 1128 lldb::eFunctionNameTypeSelector}) { 1129 if (name_type_mask & type) { 1130 auto map = GetNameToSymbolIndexMap(type); 1131 1132 const UniqueCStringMap<uint32_t>::Entry *match; 1133 for (match = map.FindFirstValueForName(name); match != nullptr; 1134 match = map.FindNextValueForName(match)) { 1135 symbol_indexes.push_back(match->value); 1136 } 1137 } 1138 } 1139 1140 if (!symbol_indexes.empty()) { 1141 llvm::sort(symbol_indexes.begin(), symbol_indexes.end()); 1142 symbol_indexes.erase( 1143 std::unique(symbol_indexes.begin(), symbol_indexes.end()), 1144 symbol_indexes.end()); 1145 SymbolIndicesToSymbolContextList(symbol_indexes, sc_list); 1146 } 1147 } 1148 1149 const Symbol *Symtab::GetParent(Symbol *child_symbol) const { 1150 uint32_t child_idx = GetIndexForSymbol(child_symbol); 1151 if (child_idx != UINT32_MAX && child_idx > 0) { 1152 for (uint32_t idx = child_idx - 1; idx != UINT32_MAX; --idx) { 1153 const Symbol *symbol = SymbolAtIndex(idx); 1154 const uint32_t sibling_idx = symbol->GetSiblingIndex(); 1155 if (sibling_idx != UINT32_MAX && sibling_idx > child_idx) 1156 return symbol; 1157 } 1158 } 1159 return nullptr; 1160 } 1161 1162 std::string Symtab::GetCacheKey() { 1163 std::string key; 1164 llvm::raw_string_ostream strm(key); 1165 // Symbol table can come from different object files for the same module. A 1166 // module can have one object file as the main executable and might have 1167 // another object file in a separate symbol file. 1168 strm << m_objfile->GetModule()->GetCacheKey() << "-symtab-" 1169 << llvm::format_hex(m_objfile->GetCacheHash(), 10); 1170 return strm.str(); 1171 } 1172 1173 void Symtab::SaveToCache() { 1174 DataFileCache *cache = Module::GetIndexCache(); 1175 if (!cache) 1176 return; // Caching is not enabled. 1177 InitNameIndexes(); // Init the name indexes so we can cache them as well. 1178 const auto byte_order = endian::InlHostByteOrder(); 1179 DataEncoder file(byte_order, /*addr_size=*/8); 1180 // Encode will return false if the symbol table's object file doesn't have 1181 // anything to make a signature from. 1182 if (Encode(file)) 1183 if (cache->SetCachedData(GetCacheKey(), file.GetData())) 1184 SetWasSavedToCache(); 1185 } 1186 1187 constexpr llvm::StringLiteral kIdentifierCStrMap("CMAP"); 1188 1189 static void EncodeCStrMap(DataEncoder &encoder, ConstStringTable &strtab, 1190 const UniqueCStringMap<uint32_t> &cstr_map) { 1191 encoder.AppendData(kIdentifierCStrMap); 1192 encoder.AppendU32(cstr_map.GetSize()); 1193 for (const auto &entry: cstr_map) { 1194 // Make sure there are no empty strings. 1195 assert((bool)entry.cstring); 1196 encoder.AppendU32(strtab.Add(entry.cstring)); 1197 encoder.AppendU32(entry.value); 1198 } 1199 } 1200 1201 bool DecodeCStrMap(const DataExtractor &data, lldb::offset_t *offset_ptr, 1202 const StringTableReader &strtab, 1203 UniqueCStringMap<uint32_t> &cstr_map) { 1204 llvm::StringRef identifier((const char *)data.GetData(offset_ptr, 4), 4); 1205 if (identifier != kIdentifierCStrMap) 1206 return false; 1207 const uint32_t count = data.GetU32(offset_ptr); 1208 for (uint32_t i=0; i<count; ++i) 1209 { 1210 llvm::StringRef str(strtab.Get(data.GetU32(offset_ptr))); 1211 uint32_t value = data.GetU32(offset_ptr); 1212 // No empty strings in the name indexes in Symtab 1213 if (str.empty()) 1214 return false; 1215 cstr_map.Append(ConstString(str), value); 1216 } 1217 return true; 1218 } 1219 1220 constexpr llvm::StringLiteral kIdentifierSymbolTable("SYMB"); 1221 constexpr uint32_t CURRENT_CACHE_VERSION = 1; 1222 1223 /// The encoding format for the symbol table is as follows: 1224 /// 1225 /// Signature signature; 1226 /// ConstStringTable strtab; 1227 /// Identifier four character code: 'SYMB' 1228 /// uint32_t version; 1229 /// uint32_t num_symbols; 1230 /// Symbol symbols[num_symbols]; 1231 /// uint8_t num_cstr_maps; 1232 /// UniqueCStringMap<uint32_t> cstr_maps[num_cstr_maps] 1233 bool Symtab::Encode(DataEncoder &encoder) const { 1234 // Name indexes must be computed before calling this function. 1235 assert(m_name_indexes_computed); 1236 1237 // Encode the object file's signature 1238 CacheSignature signature(m_objfile); 1239 if (!signature.Encode(encoder)) 1240 return false; 1241 ConstStringTable strtab; 1242 1243 // Encoder the symbol table into a separate encoder first. This allows us 1244 // gather all of the strings we willl need in "strtab" as we will need to 1245 // write the string table out before the symbol table. 1246 DataEncoder symtab_encoder(encoder.GetByteOrder(), 1247 encoder.GetAddressByteSize()); 1248 symtab_encoder.AppendData(kIdentifierSymbolTable); 1249 // Encode the symtab data version. 1250 symtab_encoder.AppendU32(CURRENT_CACHE_VERSION); 1251 // Encode the number of symbols. 1252 symtab_encoder.AppendU32(m_symbols.size()); 1253 // Encode the symbol data for all symbols. 1254 for (const auto &symbol: m_symbols) 1255 symbol.Encode(symtab_encoder, strtab); 1256 1257 // Emit a byte for how many C string maps we emit. We will fix this up after 1258 // we emit the C string maps since we skip emitting C string maps if they are 1259 // empty. 1260 size_t num_cmaps_offset = symtab_encoder.GetByteSize(); 1261 uint8_t num_cmaps = 0; 1262 symtab_encoder.AppendU8(0); 1263 for (const auto &pair: m_name_to_symbol_indices) { 1264 if (pair.second.IsEmpty()) 1265 continue; 1266 ++num_cmaps; 1267 symtab_encoder.AppendU8(pair.first); 1268 EncodeCStrMap(symtab_encoder, strtab, pair.second); 1269 } 1270 if (num_cmaps > 0) 1271 symtab_encoder.PutU8(num_cmaps_offset, num_cmaps); 1272 1273 // Now that all strings have been gathered, we will emit the string table. 1274 strtab.Encode(encoder); 1275 // Followed the the symbol table data. 1276 encoder.AppendData(symtab_encoder.GetData()); 1277 return true; 1278 } 1279 1280 bool Symtab::Decode(const DataExtractor &data, lldb::offset_t *offset_ptr, 1281 bool &signature_mismatch) { 1282 signature_mismatch = false; 1283 CacheSignature signature; 1284 StringTableReader strtab; 1285 { // Scope for "elapsed" object below so it can measure the time parse. 1286 ElapsedTime elapsed(m_objfile->GetModule()->GetSymtabParseTime()); 1287 if (!signature.Decode(data, offset_ptr)) 1288 return false; 1289 if (CacheSignature(m_objfile) != signature) { 1290 signature_mismatch = true; 1291 return false; 1292 } 1293 // We now decode the string table for all strings in the data cache file. 1294 if (!strtab.Decode(data, offset_ptr)) 1295 return false; 1296 1297 // And now we can decode the symbol table with string table we just decoded. 1298 llvm::StringRef identifier((const char *)data.GetData(offset_ptr, 4), 4); 1299 if (identifier != kIdentifierSymbolTable) 1300 return false; 1301 const uint32_t version = data.GetU32(offset_ptr); 1302 if (version != CURRENT_CACHE_VERSION) 1303 return false; 1304 const uint32_t num_symbols = data.GetU32(offset_ptr); 1305 if (num_symbols == 0) 1306 return true; 1307 m_symbols.resize(num_symbols); 1308 SectionList *sections = m_objfile->GetModule()->GetSectionList(); 1309 for (uint32_t i=0; i<num_symbols; ++i) { 1310 if (!m_symbols[i].Decode(data, offset_ptr, sections, strtab)) 1311 return false; 1312 } 1313 } 1314 1315 { // Scope for "elapsed" object below so it can measure the time to index. 1316 ElapsedTime elapsed(m_objfile->GetModule()->GetSymtabIndexTime()); 1317 const uint8_t num_cstr_maps = data.GetU8(offset_ptr); 1318 for (uint8_t i=0; i<num_cstr_maps; ++i) { 1319 uint8_t type = data.GetU8(offset_ptr); 1320 UniqueCStringMap<uint32_t> &cstr_map = 1321 GetNameToSymbolIndexMap((lldb::FunctionNameType)type); 1322 if (!DecodeCStrMap(data, offset_ptr, strtab, cstr_map)) 1323 return false; 1324 } 1325 m_name_indexes_computed = true; 1326 } 1327 return true; 1328 } 1329 1330 bool Symtab::LoadFromCache() { 1331 DataFileCache *cache = Module::GetIndexCache(); 1332 if (!cache) 1333 return false; 1334 1335 std::unique_ptr<llvm::MemoryBuffer> mem_buffer_up = 1336 cache->GetCachedData(GetCacheKey()); 1337 if (!mem_buffer_up) 1338 return false; 1339 DataExtractor data(mem_buffer_up->getBufferStart(), 1340 mem_buffer_up->getBufferSize(), 1341 m_objfile->GetByteOrder(), 1342 m_objfile->GetAddressByteSize()); 1343 bool signature_mismatch = false; 1344 lldb::offset_t offset = 0; 1345 const bool result = Decode(data, &offset, signature_mismatch); 1346 if (signature_mismatch) 1347 cache->RemoveCacheFile(GetCacheKey()); 1348 if (result) 1349 SetWasLoadedFromCache(); 1350 return result; 1351 } 1352