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 m_file_addr_to_index.Clear(); 84 m_file_addr_to_index_computed = false; 85 } 86 87 void Symtab::Dump(Stream *s, Target *target, SortOrder sort_order, 88 Mangled::NamePreference name_preference) { 89 std::lock_guard<std::recursive_mutex> guard(m_mutex); 90 91 // s->Printf("%.*p: ", (int)sizeof(void*) * 2, this); 92 s->Indent(); 93 const FileSpec &file_spec = m_objfile->GetFileSpec(); 94 const char *object_name = nullptr; 95 if (m_objfile->GetModule()) 96 object_name = m_objfile->GetModule()->GetObjectName().GetCString(); 97 98 if (file_spec) 99 s->Printf("Symtab, file = %s%s%s%s, num_symbols = %" PRIu64, 100 file_spec.GetPath().c_str(), object_name ? "(" : "", 101 object_name ? object_name : "", object_name ? ")" : "", 102 (uint64_t)m_symbols.size()); 103 else 104 s->Printf("Symtab, num_symbols = %" PRIu64 "", (uint64_t)m_symbols.size()); 105 106 if (!m_symbols.empty()) { 107 switch (sort_order) { 108 case eSortOrderNone: { 109 s->PutCString(":\n"); 110 DumpSymbolHeader(s); 111 const_iterator begin = m_symbols.begin(); 112 const_iterator end = m_symbols.end(); 113 for (const_iterator pos = m_symbols.begin(); pos != end; ++pos) { 114 s->Indent(); 115 pos->Dump(s, target, std::distance(begin, pos), name_preference); 116 } 117 } 118 break; 119 120 case eSortOrderByName: { 121 // Although we maintain a lookup by exact name map, the table isn't 122 // sorted by name. So we must make the ordered symbol list up ourselves. 123 s->PutCString(" (sorted by name):\n"); 124 DumpSymbolHeader(s); 125 126 std::multimap<llvm::StringRef, const Symbol *> name_map; 127 for (const Symbol &symbol : m_symbols) 128 name_map.emplace(symbol.GetName().GetStringRef(), &symbol); 129 130 for (const auto &name_to_symbol : name_map) { 131 const Symbol *symbol = name_to_symbol.second; 132 s->Indent(); 133 symbol->Dump(s, target, symbol - &m_symbols[0], name_preference); 134 } 135 } break; 136 137 case eSortOrderBySize: { 138 s->PutCString(" (sorted by size):\n"); 139 DumpSymbolHeader(s); 140 141 std::multimap<size_t, const Symbol *, std::greater<size_t>> size_map; 142 for (const Symbol &symbol : m_symbols) 143 size_map.emplace(symbol.GetByteSize(), &symbol); 144 145 size_t idx = 0; 146 for (const auto &size_to_symbol : size_map) { 147 const Symbol *symbol = size_to_symbol.second; 148 s->Indent(); 149 symbol->Dump(s, target, idx++, name_preference); 150 } 151 } break; 152 153 case eSortOrderByAddress: 154 s->PutCString(" (sorted by address):\n"); 155 DumpSymbolHeader(s); 156 if (!m_file_addr_to_index_computed) 157 InitAddressIndexes(); 158 const size_t num_entries = m_file_addr_to_index.GetSize(); 159 for (size_t i = 0; i < num_entries; ++i) { 160 s->Indent(); 161 const uint32_t symbol_idx = m_file_addr_to_index.GetEntryRef(i).data; 162 m_symbols[symbol_idx].Dump(s, target, symbol_idx, name_preference); 163 } 164 break; 165 } 166 } else { 167 s->PutCString("\n"); 168 } 169 } 170 171 void Symtab::Dump(Stream *s, Target *target, std::vector<uint32_t> &indexes, 172 Mangled::NamePreference name_preference) const { 173 std::lock_guard<std::recursive_mutex> guard(m_mutex); 174 175 const size_t num_symbols = GetNumSymbols(); 176 // s->Printf("%.*p: ", (int)sizeof(void*) * 2, this); 177 s->Indent(); 178 s->Printf("Symtab %" PRIu64 " symbol indexes (%" PRIu64 " symbols total):\n", 179 (uint64_t)indexes.size(), (uint64_t)m_symbols.size()); 180 s->IndentMore(); 181 182 if (!indexes.empty()) { 183 std::vector<uint32_t>::const_iterator pos; 184 std::vector<uint32_t>::const_iterator end = indexes.end(); 185 DumpSymbolHeader(s); 186 for (pos = indexes.begin(); pos != end; ++pos) { 187 size_t idx = *pos; 188 if (idx < num_symbols) { 189 s->Indent(); 190 m_symbols[idx].Dump(s, target, idx, name_preference); 191 } 192 } 193 } 194 s->IndentLess(); 195 } 196 197 void Symtab::DumpSymbolHeader(Stream *s) { 198 s->Indent(" Debug symbol\n"); 199 s->Indent(" |Synthetic symbol\n"); 200 s->Indent(" ||Externally Visible\n"); 201 s->Indent(" |||\n"); 202 s->Indent("Index UserID DSX Type File Address/Value Load " 203 "Address Size Flags Name\n"); 204 s->Indent("------- ------ --- --------------- ------------------ " 205 "------------------ ------------------ ---------- " 206 "----------------------------------\n"); 207 } 208 209 static int CompareSymbolID(const void *key, const void *p) { 210 const user_id_t match_uid = *(const user_id_t *)key; 211 const user_id_t symbol_uid = ((const Symbol *)p)->GetID(); 212 if (match_uid < symbol_uid) 213 return -1; 214 if (match_uid > symbol_uid) 215 return 1; 216 return 0; 217 } 218 219 Symbol *Symtab::FindSymbolByID(lldb::user_id_t symbol_uid) const { 220 std::lock_guard<std::recursive_mutex> guard(m_mutex); 221 222 Symbol *symbol = 223 (Symbol *)::bsearch(&symbol_uid, &m_symbols[0], m_symbols.size(), 224 sizeof(m_symbols[0]), CompareSymbolID); 225 return symbol; 226 } 227 228 Symbol *Symtab::SymbolAtIndex(size_t idx) { 229 // Clients should grab the mutex from this symbol table and lock it manually 230 // when calling this function to avoid performance issues. 231 if (idx < m_symbols.size()) 232 return &m_symbols[idx]; 233 return nullptr; 234 } 235 236 const Symbol *Symtab::SymbolAtIndex(size_t idx) const { 237 // Clients should grab the mutex from this symbol table and lock it manually 238 // when calling this function to avoid performance issues. 239 if (idx < m_symbols.size()) 240 return &m_symbols[idx]; 241 return nullptr; 242 } 243 244 static bool lldb_skip_name(llvm::StringRef mangled, 245 Mangled::ManglingScheme scheme) { 246 switch (scheme) { 247 case Mangled::eManglingSchemeItanium: { 248 if (mangled.size() < 3 || !mangled.starts_with("_Z")) 249 return true; 250 251 // Avoid the following types of symbols in the index. 252 switch (mangled[2]) { 253 case 'G': // guard variables 254 case 'T': // virtual tables, VTT structures, typeinfo structures + names 255 case 'Z': // named local entities (if we eventually handle 256 // eSymbolTypeData, we will want this back) 257 return true; 258 259 default: 260 break; 261 } 262 263 // Include this name in the index. 264 return false; 265 } 266 267 // No filters for this scheme yet. Include all names in indexing. 268 case Mangled::eManglingSchemeMSVC: 269 case Mangled::eManglingSchemeRustV0: 270 case Mangled::eManglingSchemeD: 271 case Mangled::eManglingSchemeSwift: 272 return false; 273 274 // Don't try and demangle things we can't categorize. 275 case Mangled::eManglingSchemeNone: 276 return true; 277 } 278 llvm_unreachable("unknown scheme!"); 279 } 280 281 void Symtab::InitNameIndexes() { 282 // Protected function, no need to lock mutex... 283 if (!m_name_indexes_computed) { 284 m_name_indexes_computed = true; 285 ElapsedTime elapsed(m_objfile->GetModule()->GetSymtabIndexTime()); 286 LLDB_SCOPED_TIMER(); 287 288 // Collect all loaded language plugins. 289 std::vector<Language *> languages; 290 Language::ForEach([&languages](Language *l) { 291 languages.push_back(l); 292 return true; 293 }); 294 295 auto &name_to_index = GetNameToSymbolIndexMap(lldb::eFunctionNameTypeNone); 296 auto &basename_to_index = 297 GetNameToSymbolIndexMap(lldb::eFunctionNameTypeBase); 298 auto &method_to_index = 299 GetNameToSymbolIndexMap(lldb::eFunctionNameTypeMethod); 300 auto &selector_to_index = 301 GetNameToSymbolIndexMap(lldb::eFunctionNameTypeSelector); 302 // Create the name index vector to be able to quickly search by name 303 const size_t num_symbols = m_symbols.size(); 304 name_to_index.Reserve(num_symbols); 305 306 // The "const char *" in "class_contexts" and backlog::value_type::second 307 // must come from a ConstString::GetCString() 308 std::set<const char *> class_contexts; 309 std::vector<std::pair<NameToIndexMap::Entry, const char *>> backlog; 310 backlog.reserve(num_symbols / 2); 311 312 // Instantiation of the demangler is expensive, so better use a single one 313 // for all entries during batch processing. 314 RichManglingContext rmc; 315 for (uint32_t value = 0; value < num_symbols; ++value) { 316 Symbol *symbol = &m_symbols[value]; 317 318 // Don't let trampolines get into the lookup by name map If we ever need 319 // the trampoline symbols to be searchable by name we can remove this and 320 // then possibly add a new bool to any of the Symtab functions that 321 // lookup symbols by name to indicate if they want trampolines. We also 322 // don't want any synthetic symbols with auto generated names in the 323 // name lookups. 324 if (symbol->IsTrampoline() || symbol->IsSyntheticWithAutoGeneratedName()) 325 continue; 326 327 // If the symbol's name string matched a Mangled::ManglingScheme, it is 328 // stored in the mangled field. 329 Mangled &mangled = symbol->GetMangled(); 330 if (ConstString name = mangled.GetMangledName()) { 331 name_to_index.Append(name, value); 332 333 if (symbol->ContainsLinkerAnnotations()) { 334 // If the symbol has linker annotations, also add the version without 335 // the annotations. 336 ConstString stripped = ConstString( 337 m_objfile->StripLinkerSymbolAnnotations(name.GetStringRef())); 338 name_to_index.Append(stripped, value); 339 } 340 341 const SymbolType type = symbol->GetType(); 342 if (type == eSymbolTypeCode || type == eSymbolTypeResolver) { 343 if (mangled.GetRichManglingInfo(rmc, lldb_skip_name)) { 344 RegisterMangledNameEntry(value, class_contexts, backlog, rmc); 345 continue; 346 } 347 } 348 } 349 350 // Symbol name strings that didn't match a Mangled::ManglingScheme, are 351 // stored in the demangled field. 352 if (ConstString name = mangled.GetDemangledName()) { 353 name_to_index.Append(name, value); 354 355 if (symbol->ContainsLinkerAnnotations()) { 356 // If the symbol has linker annotations, also add the version without 357 // the annotations. 358 name = ConstString( 359 m_objfile->StripLinkerSymbolAnnotations(name.GetStringRef())); 360 name_to_index.Append(name, value); 361 } 362 363 // If the demangled name turns out to be an ObjC name, and is a category 364 // name, add the version without categories to the index too. 365 for (Language *lang : languages) { 366 for (auto variant : lang->GetMethodNameVariants(name)) { 367 if (variant.GetType() & lldb::eFunctionNameTypeSelector) 368 selector_to_index.Append(variant.GetName(), value); 369 else if (variant.GetType() & lldb::eFunctionNameTypeFull) 370 name_to_index.Append(variant.GetName(), value); 371 else if (variant.GetType() & lldb::eFunctionNameTypeMethod) 372 method_to_index.Append(variant.GetName(), value); 373 else if (variant.GetType() & lldb::eFunctionNameTypeBase) 374 basename_to_index.Append(variant.GetName(), value); 375 } 376 } 377 } 378 } 379 380 for (const auto &record : backlog) { 381 RegisterBacklogEntry(record.first, record.second, class_contexts); 382 } 383 384 name_to_index.Sort(); 385 name_to_index.SizeToFit(); 386 selector_to_index.Sort(); 387 selector_to_index.SizeToFit(); 388 basename_to_index.Sort(); 389 basename_to_index.SizeToFit(); 390 method_to_index.Sort(); 391 method_to_index.SizeToFit(); 392 } 393 } 394 395 void Symtab::RegisterMangledNameEntry( 396 uint32_t value, std::set<const char *> &class_contexts, 397 std::vector<std::pair<NameToIndexMap::Entry, const char *>> &backlog, 398 RichManglingContext &rmc) { 399 // Only register functions that have a base name. 400 llvm::StringRef base_name = rmc.ParseFunctionBaseName(); 401 if (base_name.empty()) 402 return; 403 404 // The base name will be our entry's name. 405 NameToIndexMap::Entry entry(ConstString(base_name), value); 406 llvm::StringRef decl_context = rmc.ParseFunctionDeclContextName(); 407 408 // Register functions with no context. 409 if (decl_context.empty()) { 410 // This has to be a basename 411 auto &basename_to_index = 412 GetNameToSymbolIndexMap(lldb::eFunctionNameTypeBase); 413 basename_to_index.Append(entry); 414 // If there is no context (no namespaces or class scopes that come before 415 // the function name) then this also could be a fullname. 416 auto &name_to_index = GetNameToSymbolIndexMap(lldb::eFunctionNameTypeNone); 417 name_to_index.Append(entry); 418 return; 419 } 420 421 // Make sure we have a pool-string pointer and see if we already know the 422 // context name. 423 const char *decl_context_ccstr = ConstString(decl_context).GetCString(); 424 auto it = class_contexts.find(decl_context_ccstr); 425 426 auto &method_to_index = 427 GetNameToSymbolIndexMap(lldb::eFunctionNameTypeMethod); 428 // Register constructors and destructors. They are methods and create 429 // declaration contexts. 430 if (rmc.IsCtorOrDtor()) { 431 method_to_index.Append(entry); 432 if (it == class_contexts.end()) 433 class_contexts.insert(it, decl_context_ccstr); 434 return; 435 } 436 437 // Register regular methods with a known declaration context. 438 if (it != class_contexts.end()) { 439 method_to_index.Append(entry); 440 return; 441 } 442 443 // Regular methods in unknown declaration contexts are put to the backlog. We 444 // will revisit them once we processed all remaining symbols. 445 backlog.push_back(std::make_pair(entry, decl_context_ccstr)); 446 } 447 448 void Symtab::RegisterBacklogEntry( 449 const NameToIndexMap::Entry &entry, const char *decl_context, 450 const std::set<const char *> &class_contexts) { 451 auto &method_to_index = 452 GetNameToSymbolIndexMap(lldb::eFunctionNameTypeMethod); 453 auto it = class_contexts.find(decl_context); 454 if (it != class_contexts.end()) { 455 method_to_index.Append(entry); 456 } else { 457 // If we got here, we have something that had a context (was inside 458 // a namespace or class) yet we don't know the entry 459 method_to_index.Append(entry); 460 auto &basename_to_index = 461 GetNameToSymbolIndexMap(lldb::eFunctionNameTypeBase); 462 basename_to_index.Append(entry); 463 } 464 } 465 466 void Symtab::PreloadSymbols() { 467 std::lock_guard<std::recursive_mutex> guard(m_mutex); 468 InitNameIndexes(); 469 } 470 471 void Symtab::AppendSymbolNamesToMap(const IndexCollection &indexes, 472 bool add_demangled, bool add_mangled, 473 NameToIndexMap &name_to_index_map) const { 474 LLDB_SCOPED_TIMER(); 475 if (add_demangled || add_mangled) { 476 std::lock_guard<std::recursive_mutex> guard(m_mutex); 477 478 // Create the name index vector to be able to quickly search by name 479 const size_t num_indexes = indexes.size(); 480 for (size_t i = 0; i < num_indexes; ++i) { 481 uint32_t value = indexes[i]; 482 assert(i < m_symbols.size()); 483 const Symbol *symbol = &m_symbols[value]; 484 485 const Mangled &mangled = symbol->GetMangled(); 486 if (add_demangled) { 487 if (ConstString name = mangled.GetDemangledName()) 488 name_to_index_map.Append(name, value); 489 } 490 491 if (add_mangled) { 492 if (ConstString name = mangled.GetMangledName()) 493 name_to_index_map.Append(name, value); 494 } 495 } 496 } 497 } 498 499 uint32_t Symtab::AppendSymbolIndexesWithType(SymbolType symbol_type, 500 std::vector<uint32_t> &indexes, 501 uint32_t start_idx, 502 uint32_t end_index) const { 503 std::lock_guard<std::recursive_mutex> guard(m_mutex); 504 505 uint32_t prev_size = indexes.size(); 506 507 const uint32_t count = std::min<uint32_t>(m_symbols.size(), end_index); 508 509 for (uint32_t i = start_idx; i < count; ++i) { 510 if (symbol_type == eSymbolTypeAny || m_symbols[i].GetType() == symbol_type) 511 indexes.push_back(i); 512 } 513 514 return indexes.size() - prev_size; 515 } 516 517 uint32_t Symtab::AppendSymbolIndexesWithTypeAndFlagsValue( 518 SymbolType symbol_type, uint32_t flags_value, 519 std::vector<uint32_t> &indexes, uint32_t start_idx, 520 uint32_t end_index) const { 521 std::lock_guard<std::recursive_mutex> guard(m_mutex); 522 523 uint32_t prev_size = indexes.size(); 524 525 const uint32_t count = std::min<uint32_t>(m_symbols.size(), end_index); 526 527 for (uint32_t i = start_idx; i < count; ++i) { 528 if ((symbol_type == eSymbolTypeAny || 529 m_symbols[i].GetType() == symbol_type) && 530 m_symbols[i].GetFlags() == flags_value) 531 indexes.push_back(i); 532 } 533 534 return indexes.size() - prev_size; 535 } 536 537 uint32_t Symtab::AppendSymbolIndexesWithType(SymbolType symbol_type, 538 Debug symbol_debug_type, 539 Visibility symbol_visibility, 540 std::vector<uint32_t> &indexes, 541 uint32_t start_idx, 542 uint32_t end_index) const { 543 std::lock_guard<std::recursive_mutex> guard(m_mutex); 544 545 uint32_t prev_size = indexes.size(); 546 547 const uint32_t count = std::min<uint32_t>(m_symbols.size(), end_index); 548 549 for (uint32_t i = start_idx; i < count; ++i) { 550 if (symbol_type == eSymbolTypeAny || 551 m_symbols[i].GetType() == symbol_type) { 552 if (CheckSymbolAtIndex(i, symbol_debug_type, symbol_visibility)) 553 indexes.push_back(i); 554 } 555 } 556 557 return indexes.size() - prev_size; 558 } 559 560 uint32_t Symtab::GetIndexForSymbol(const Symbol *symbol) const { 561 if (!m_symbols.empty()) { 562 const Symbol *first_symbol = &m_symbols[0]; 563 if (symbol >= first_symbol && symbol < first_symbol + m_symbols.size()) 564 return symbol - first_symbol; 565 } 566 return UINT32_MAX; 567 } 568 569 struct SymbolSortInfo { 570 const bool sort_by_load_addr; 571 const Symbol *symbols; 572 }; 573 574 namespace { 575 struct SymbolIndexComparator { 576 const std::vector<Symbol> &symbols; 577 std::vector<lldb::addr_t> &addr_cache; 578 579 // Getting from the symbol to the Address to the File Address involves some 580 // work. Since there are potentially many symbols here, and we're using this 581 // for sorting so we're going to be computing the address many times, cache 582 // that in addr_cache. The array passed in has to be the same size as the 583 // symbols array passed into the member variable symbols, and should be 584 // initialized with LLDB_INVALID_ADDRESS. 585 // NOTE: You have to make addr_cache externally and pass it in because 586 // std::stable_sort 587 // makes copies of the comparator it is initially passed in, and you end up 588 // spending huge amounts of time copying this array... 589 590 SymbolIndexComparator(const std::vector<Symbol> &s, 591 std::vector<lldb::addr_t> &a) 592 : symbols(s), addr_cache(a) { 593 assert(symbols.size() == addr_cache.size()); 594 } 595 bool operator()(uint32_t index_a, uint32_t index_b) { 596 addr_t value_a = addr_cache[index_a]; 597 if (value_a == LLDB_INVALID_ADDRESS) { 598 value_a = symbols[index_a].GetAddressRef().GetFileAddress(); 599 addr_cache[index_a] = value_a; 600 } 601 602 addr_t value_b = addr_cache[index_b]; 603 if (value_b == LLDB_INVALID_ADDRESS) { 604 value_b = symbols[index_b].GetAddressRef().GetFileAddress(); 605 addr_cache[index_b] = value_b; 606 } 607 608 if (value_a == value_b) { 609 // The if the values are equal, use the original symbol user ID 610 lldb::user_id_t uid_a = symbols[index_a].GetID(); 611 lldb::user_id_t uid_b = symbols[index_b].GetID(); 612 if (uid_a < uid_b) 613 return true; 614 if (uid_a > uid_b) 615 return false; 616 return false; 617 } else if (value_a < value_b) 618 return true; 619 620 return false; 621 } 622 }; 623 } 624 625 void Symtab::SortSymbolIndexesByValue(std::vector<uint32_t> &indexes, 626 bool remove_duplicates) const { 627 std::lock_guard<std::recursive_mutex> guard(m_mutex); 628 LLDB_SCOPED_TIMER(); 629 // No need to sort if we have zero or one items... 630 if (indexes.size() <= 1) 631 return; 632 633 // Sort the indexes in place using std::stable_sort. 634 // NOTE: The use of std::stable_sort instead of llvm::sort here is strictly 635 // for performance, not correctness. The indexes vector tends to be "close" 636 // to sorted, which the stable sort handles better. 637 638 std::vector<lldb::addr_t> addr_cache(m_symbols.size(), LLDB_INVALID_ADDRESS); 639 640 SymbolIndexComparator comparator(m_symbols, addr_cache); 641 std::stable_sort(indexes.begin(), indexes.end(), comparator); 642 643 // Remove any duplicates if requested 644 if (remove_duplicates) { 645 auto last = std::unique(indexes.begin(), indexes.end()); 646 indexes.erase(last, indexes.end()); 647 } 648 } 649 650 uint32_t Symtab::GetNameIndexes(ConstString symbol_name, 651 std::vector<uint32_t> &indexes) { 652 auto &name_to_index = GetNameToSymbolIndexMap(lldb::eFunctionNameTypeNone); 653 const uint32_t count = name_to_index.GetValues(symbol_name, indexes); 654 if (count) 655 return count; 656 // Synthetic symbol names are not added to the name indexes, but they start 657 // with a prefix and end with a the symbol UserID. This allows users to find 658 // these symbols without having to add them to the name indexes. These 659 // queries will not happen very often since the names don't mean anything, so 660 // performance is not paramount in this case. 661 llvm::StringRef name = symbol_name.GetStringRef(); 662 // String the synthetic prefix if the name starts with it. 663 if (!name.consume_front(Symbol::GetSyntheticSymbolPrefix())) 664 return 0; // Not a synthetic symbol name 665 666 // Extract the user ID from the symbol name 667 unsigned long long uid = 0; 668 if (getAsUnsignedInteger(name, /*Radix=*/10, uid)) 669 return 0; // Failed to extract the user ID as an integer 670 Symbol *symbol = FindSymbolByID(uid); 671 if (symbol == nullptr) 672 return 0; 673 const uint32_t symbol_idx = GetIndexForSymbol(symbol); 674 if (symbol_idx == UINT32_MAX) 675 return 0; 676 indexes.push_back(symbol_idx); 677 return 1; 678 } 679 680 uint32_t Symtab::AppendSymbolIndexesWithName(ConstString symbol_name, 681 std::vector<uint32_t> &indexes) { 682 std::lock_guard<std::recursive_mutex> guard(m_mutex); 683 684 if (symbol_name) { 685 if (!m_name_indexes_computed) 686 InitNameIndexes(); 687 688 return GetNameIndexes(symbol_name, indexes); 689 } 690 return 0; 691 } 692 693 uint32_t Symtab::AppendSymbolIndexesWithName(ConstString symbol_name, 694 Debug symbol_debug_type, 695 Visibility symbol_visibility, 696 std::vector<uint32_t> &indexes) { 697 std::lock_guard<std::recursive_mutex> guard(m_mutex); 698 699 LLDB_SCOPED_TIMER(); 700 if (symbol_name) { 701 const size_t old_size = indexes.size(); 702 if (!m_name_indexes_computed) 703 InitNameIndexes(); 704 705 std::vector<uint32_t> all_name_indexes; 706 const size_t name_match_count = 707 GetNameIndexes(symbol_name, all_name_indexes); 708 for (size_t i = 0; i < name_match_count; ++i) { 709 if (CheckSymbolAtIndex(all_name_indexes[i], symbol_debug_type, 710 symbol_visibility)) 711 indexes.push_back(all_name_indexes[i]); 712 } 713 return indexes.size() - old_size; 714 } 715 return 0; 716 } 717 718 uint32_t 719 Symtab::AppendSymbolIndexesWithNameAndType(ConstString symbol_name, 720 SymbolType symbol_type, 721 std::vector<uint32_t> &indexes) { 722 std::lock_guard<std::recursive_mutex> guard(m_mutex); 723 724 if (AppendSymbolIndexesWithName(symbol_name, indexes) > 0) { 725 std::vector<uint32_t>::iterator pos = indexes.begin(); 726 while (pos != indexes.end()) { 727 if (symbol_type == eSymbolTypeAny || 728 m_symbols[*pos].GetType() == symbol_type) 729 ++pos; 730 else 731 pos = indexes.erase(pos); 732 } 733 } 734 return indexes.size(); 735 } 736 737 uint32_t Symtab::AppendSymbolIndexesWithNameAndType( 738 ConstString symbol_name, SymbolType symbol_type, 739 Debug symbol_debug_type, Visibility symbol_visibility, 740 std::vector<uint32_t> &indexes) { 741 std::lock_guard<std::recursive_mutex> guard(m_mutex); 742 743 if (AppendSymbolIndexesWithName(symbol_name, symbol_debug_type, 744 symbol_visibility, indexes) > 0) { 745 std::vector<uint32_t>::iterator pos = indexes.begin(); 746 while (pos != indexes.end()) { 747 if (symbol_type == eSymbolTypeAny || 748 m_symbols[*pos].GetType() == symbol_type) 749 ++pos; 750 else 751 pos = indexes.erase(pos); 752 } 753 } 754 return indexes.size(); 755 } 756 757 uint32_t Symtab::AppendSymbolIndexesMatchingRegExAndType( 758 const RegularExpression ®exp, SymbolType symbol_type, 759 std::vector<uint32_t> &indexes, Mangled::NamePreference name_preference) { 760 std::lock_guard<std::recursive_mutex> guard(m_mutex); 761 762 uint32_t prev_size = indexes.size(); 763 uint32_t sym_end = m_symbols.size(); 764 765 for (uint32_t i = 0; i < sym_end; i++) { 766 if (symbol_type == eSymbolTypeAny || 767 m_symbols[i].GetType() == symbol_type) { 768 const char *name = 769 m_symbols[i].GetMangled().GetName(name_preference).AsCString(); 770 if (name) { 771 if (regexp.Execute(name)) 772 indexes.push_back(i); 773 } 774 } 775 } 776 return indexes.size() - prev_size; 777 } 778 779 uint32_t Symtab::AppendSymbolIndexesMatchingRegExAndType( 780 const RegularExpression ®exp, SymbolType symbol_type, 781 Debug symbol_debug_type, Visibility symbol_visibility, 782 std::vector<uint32_t> &indexes, Mangled::NamePreference name_preference) { 783 std::lock_guard<std::recursive_mutex> guard(m_mutex); 784 785 uint32_t prev_size = indexes.size(); 786 uint32_t sym_end = m_symbols.size(); 787 788 for (uint32_t i = 0; i < sym_end; i++) { 789 if (symbol_type == eSymbolTypeAny || 790 m_symbols[i].GetType() == symbol_type) { 791 if (!CheckSymbolAtIndex(i, symbol_debug_type, symbol_visibility)) 792 continue; 793 794 const char *name = 795 m_symbols[i].GetMangled().GetName(name_preference).AsCString(); 796 if (name) { 797 if (regexp.Execute(name)) 798 indexes.push_back(i); 799 } 800 } 801 } 802 return indexes.size() - prev_size; 803 } 804 805 Symbol *Symtab::FindSymbolWithType(SymbolType symbol_type, 806 Debug symbol_debug_type, 807 Visibility symbol_visibility, 808 uint32_t &start_idx) { 809 std::lock_guard<std::recursive_mutex> guard(m_mutex); 810 811 const size_t count = m_symbols.size(); 812 for (size_t idx = start_idx; idx < count; ++idx) { 813 if (symbol_type == eSymbolTypeAny || 814 m_symbols[idx].GetType() == symbol_type) { 815 if (CheckSymbolAtIndex(idx, symbol_debug_type, symbol_visibility)) { 816 start_idx = idx; 817 return &m_symbols[idx]; 818 } 819 } 820 } 821 return nullptr; 822 } 823 824 void 825 Symtab::FindAllSymbolsWithNameAndType(ConstString name, 826 SymbolType symbol_type, 827 std::vector<uint32_t> &symbol_indexes) { 828 std::lock_guard<std::recursive_mutex> guard(m_mutex); 829 830 // Initialize all of the lookup by name indexes before converting NAME to a 831 // uniqued string NAME_STR below. 832 if (!m_name_indexes_computed) 833 InitNameIndexes(); 834 835 if (name) { 836 // The string table did have a string that matched, but we need to check 837 // the symbols and match the symbol_type if any was given. 838 AppendSymbolIndexesWithNameAndType(name, symbol_type, symbol_indexes); 839 } 840 } 841 842 void Symtab::FindAllSymbolsWithNameAndType( 843 ConstString name, SymbolType symbol_type, Debug symbol_debug_type, 844 Visibility symbol_visibility, std::vector<uint32_t> &symbol_indexes) { 845 std::lock_guard<std::recursive_mutex> guard(m_mutex); 846 847 LLDB_SCOPED_TIMER(); 848 // Initialize all of the lookup by name indexes before converting NAME to a 849 // uniqued string NAME_STR below. 850 if (!m_name_indexes_computed) 851 InitNameIndexes(); 852 853 if (name) { 854 // The string table did have a string that matched, but we need to check 855 // the symbols and match the symbol_type if any was given. 856 AppendSymbolIndexesWithNameAndType(name, symbol_type, symbol_debug_type, 857 symbol_visibility, symbol_indexes); 858 } 859 } 860 861 void Symtab::FindAllSymbolsMatchingRexExAndType( 862 const RegularExpression ®ex, SymbolType symbol_type, 863 Debug symbol_debug_type, Visibility symbol_visibility, 864 std::vector<uint32_t> &symbol_indexes, 865 Mangled::NamePreference name_preference) { 866 std::lock_guard<std::recursive_mutex> guard(m_mutex); 867 868 AppendSymbolIndexesMatchingRegExAndType(regex, symbol_type, symbol_debug_type, 869 symbol_visibility, symbol_indexes, 870 name_preference); 871 } 872 873 Symbol *Symtab::FindFirstSymbolWithNameAndType(ConstString name, 874 SymbolType symbol_type, 875 Debug symbol_debug_type, 876 Visibility symbol_visibility) { 877 std::lock_guard<std::recursive_mutex> guard(m_mutex); 878 LLDB_SCOPED_TIMER(); 879 if (!m_name_indexes_computed) 880 InitNameIndexes(); 881 882 if (name) { 883 std::vector<uint32_t> matching_indexes; 884 // The string table did have a string that matched, but we need to check 885 // the symbols and match the symbol_type if any was given. 886 if (AppendSymbolIndexesWithNameAndType(name, symbol_type, symbol_debug_type, 887 symbol_visibility, 888 matching_indexes)) { 889 std::vector<uint32_t>::const_iterator pos, end = matching_indexes.end(); 890 for (pos = matching_indexes.begin(); pos != end; ++pos) { 891 Symbol *symbol = SymbolAtIndex(*pos); 892 893 if (symbol->Compare(name, symbol_type)) 894 return symbol; 895 } 896 } 897 } 898 return nullptr; 899 } 900 901 typedef struct { 902 const Symtab *symtab; 903 const addr_t file_addr; 904 Symbol *match_symbol; 905 const uint32_t *match_index_ptr; 906 addr_t match_offset; 907 } SymbolSearchInfo; 908 909 // Add all the section file start address & size to the RangeVector, recusively 910 // adding any children sections. 911 static void AddSectionsToRangeMap(SectionList *sectlist, 912 RangeVector<addr_t, addr_t> §ion_ranges) { 913 const int num_sections = sectlist->GetNumSections(0); 914 for (int i = 0; i < num_sections; i++) { 915 SectionSP sect_sp = sectlist->GetSectionAtIndex(i); 916 if (sect_sp) { 917 SectionList &child_sectlist = sect_sp->GetChildren(); 918 919 // If this section has children, add the children to the RangeVector. 920 // Else add this section to the RangeVector. 921 if (child_sectlist.GetNumSections(0) > 0) { 922 AddSectionsToRangeMap(&child_sectlist, section_ranges); 923 } else { 924 size_t size = sect_sp->GetByteSize(); 925 if (size > 0) { 926 addr_t base_addr = sect_sp->GetFileAddress(); 927 RangeVector<addr_t, addr_t>::Entry entry; 928 entry.SetRangeBase(base_addr); 929 entry.SetByteSize(size); 930 section_ranges.Append(entry); 931 } 932 } 933 } 934 } 935 } 936 937 void Symtab::InitAddressIndexes() { 938 // Protected function, no need to lock mutex... 939 if (!m_file_addr_to_index_computed && !m_symbols.empty()) { 940 m_file_addr_to_index_computed = true; 941 942 FileRangeToIndexMap::Entry entry; 943 const_iterator begin = m_symbols.begin(); 944 const_iterator end = m_symbols.end(); 945 for (const_iterator pos = m_symbols.begin(); pos != end; ++pos) { 946 if (pos->ValueIsAddress()) { 947 entry.SetRangeBase(pos->GetAddressRef().GetFileAddress()); 948 entry.SetByteSize(pos->GetByteSize()); 949 entry.data = std::distance(begin, pos); 950 m_file_addr_to_index.Append(entry); 951 } 952 } 953 const size_t num_entries = m_file_addr_to_index.GetSize(); 954 if (num_entries > 0) { 955 m_file_addr_to_index.Sort(); 956 957 // Create a RangeVector with the start & size of all the sections for 958 // this objfile. We'll need to check this for any FileRangeToIndexMap 959 // entries with an uninitialized size, which could potentially be a large 960 // number so reconstituting the weak pointer is busywork when it is 961 // invariant information. 962 SectionList *sectlist = m_objfile->GetSectionList(); 963 RangeVector<addr_t, addr_t> section_ranges; 964 if (sectlist) { 965 AddSectionsToRangeMap(sectlist, section_ranges); 966 section_ranges.Sort(); 967 } 968 969 // Iterate through the FileRangeToIndexMap and fill in the size for any 970 // entries that didn't already have a size from the Symbol (e.g. if we 971 // have a plain linker symbol with an address only, instead of debug info 972 // where we get an address and a size and a type, etc.) 973 for (size_t i = 0; i < num_entries; i++) { 974 FileRangeToIndexMap::Entry *entry = 975 m_file_addr_to_index.GetMutableEntryAtIndex(i); 976 if (entry->GetByteSize() == 0) { 977 addr_t curr_base_addr = entry->GetRangeBase(); 978 const RangeVector<addr_t, addr_t>::Entry *containing_section = 979 section_ranges.FindEntryThatContains(curr_base_addr); 980 981 // Use the end of the section as the default max size of the symbol 982 addr_t sym_size = 0; 983 if (containing_section) { 984 sym_size = 985 containing_section->GetByteSize() - 986 (entry->GetRangeBase() - containing_section->GetRangeBase()); 987 } 988 989 for (size_t j = i; j < num_entries; j++) { 990 FileRangeToIndexMap::Entry *next_entry = 991 m_file_addr_to_index.GetMutableEntryAtIndex(j); 992 addr_t next_base_addr = next_entry->GetRangeBase(); 993 if (next_base_addr > curr_base_addr) { 994 addr_t size_to_next_symbol = next_base_addr - curr_base_addr; 995 996 // Take the difference between this symbol and the next one as 997 // its size, if it is less than the size of the section. 998 if (sym_size == 0 || size_to_next_symbol < sym_size) { 999 sym_size = size_to_next_symbol; 1000 } 1001 break; 1002 } 1003 } 1004 1005 if (sym_size > 0) { 1006 entry->SetByteSize(sym_size); 1007 Symbol &symbol = m_symbols[entry->data]; 1008 symbol.SetByteSize(sym_size); 1009 symbol.SetSizeIsSynthesized(true); 1010 } 1011 } 1012 } 1013 1014 // Sort again in case the range size changes the ordering 1015 m_file_addr_to_index.Sort(); 1016 } 1017 } 1018 } 1019 1020 void Symtab::Finalize() { 1021 std::lock_guard<std::recursive_mutex> guard(m_mutex); 1022 // Calculate the size of symbols inside InitAddressIndexes. 1023 InitAddressIndexes(); 1024 // Shrink to fit the symbols so we don't waste memory 1025 m_symbols.shrink_to_fit(); 1026 SaveToCache(); 1027 } 1028 1029 Symbol *Symtab::FindSymbolAtFileAddress(addr_t file_addr) { 1030 std::lock_guard<std::recursive_mutex> guard(m_mutex); 1031 if (!m_file_addr_to_index_computed) 1032 InitAddressIndexes(); 1033 1034 const FileRangeToIndexMap::Entry *entry = 1035 m_file_addr_to_index.FindEntryStartsAt(file_addr); 1036 if (entry) { 1037 Symbol *symbol = SymbolAtIndex(entry->data); 1038 if (symbol->GetFileAddress() == file_addr) 1039 return symbol; 1040 } 1041 return nullptr; 1042 } 1043 1044 Symbol *Symtab::FindSymbolContainingFileAddress(addr_t file_addr) { 1045 std::lock_guard<std::recursive_mutex> guard(m_mutex); 1046 1047 if (!m_file_addr_to_index_computed) 1048 InitAddressIndexes(); 1049 1050 const FileRangeToIndexMap::Entry *entry = 1051 m_file_addr_to_index.FindEntryThatContains(file_addr); 1052 if (entry) { 1053 Symbol *symbol = SymbolAtIndex(entry->data); 1054 if (symbol->ContainsFileAddress(file_addr)) 1055 return symbol; 1056 } 1057 return nullptr; 1058 } 1059 1060 void Symtab::ForEachSymbolContainingFileAddress( 1061 addr_t file_addr, std::function<bool(Symbol *)> const &callback) { 1062 std::lock_guard<std::recursive_mutex> guard(m_mutex); 1063 1064 if (!m_file_addr_to_index_computed) 1065 InitAddressIndexes(); 1066 1067 std::vector<uint32_t> all_addr_indexes; 1068 1069 // Get all symbols with file_addr 1070 const size_t addr_match_count = 1071 m_file_addr_to_index.FindEntryIndexesThatContain(file_addr, 1072 all_addr_indexes); 1073 1074 for (size_t i = 0; i < addr_match_count; ++i) { 1075 Symbol *symbol = SymbolAtIndex(all_addr_indexes[i]); 1076 if (symbol->ContainsFileAddress(file_addr)) { 1077 if (!callback(symbol)) 1078 break; 1079 } 1080 } 1081 } 1082 1083 void Symtab::SymbolIndicesToSymbolContextList( 1084 std::vector<uint32_t> &symbol_indexes, SymbolContextList &sc_list) { 1085 // No need to protect this call using m_mutex all other method calls are 1086 // already thread safe. 1087 1088 const bool merge_symbol_into_function = true; 1089 size_t num_indices = symbol_indexes.size(); 1090 if (num_indices > 0) { 1091 SymbolContext sc; 1092 sc.module_sp = m_objfile->GetModule(); 1093 for (size_t i = 0; i < num_indices; i++) { 1094 sc.symbol = SymbolAtIndex(symbol_indexes[i]); 1095 if (sc.symbol) 1096 sc_list.AppendIfUnique(sc, merge_symbol_into_function); 1097 } 1098 } 1099 } 1100 1101 void Symtab::FindFunctionSymbols(ConstString name, uint32_t name_type_mask, 1102 SymbolContextList &sc_list) { 1103 std::vector<uint32_t> symbol_indexes; 1104 1105 // eFunctionNameTypeAuto should be pre-resolved by a call to 1106 // Module::LookupInfo::LookupInfo() 1107 assert((name_type_mask & eFunctionNameTypeAuto) == 0); 1108 1109 if (name_type_mask & (eFunctionNameTypeBase | eFunctionNameTypeFull)) { 1110 std::vector<uint32_t> temp_symbol_indexes; 1111 FindAllSymbolsWithNameAndType(name, eSymbolTypeAny, temp_symbol_indexes); 1112 1113 unsigned temp_symbol_indexes_size = temp_symbol_indexes.size(); 1114 if (temp_symbol_indexes_size > 0) { 1115 std::lock_guard<std::recursive_mutex> guard(m_mutex); 1116 for (unsigned i = 0; i < temp_symbol_indexes_size; i++) { 1117 SymbolContext sym_ctx; 1118 sym_ctx.symbol = SymbolAtIndex(temp_symbol_indexes[i]); 1119 if (sym_ctx.symbol) { 1120 switch (sym_ctx.symbol->GetType()) { 1121 case eSymbolTypeCode: 1122 case eSymbolTypeResolver: 1123 case eSymbolTypeReExported: 1124 case eSymbolTypeAbsolute: 1125 symbol_indexes.push_back(temp_symbol_indexes[i]); 1126 break; 1127 default: 1128 break; 1129 } 1130 } 1131 } 1132 } 1133 } 1134 1135 if (!m_name_indexes_computed) 1136 InitNameIndexes(); 1137 1138 for (lldb::FunctionNameType type : 1139 {lldb::eFunctionNameTypeBase, lldb::eFunctionNameTypeMethod, 1140 lldb::eFunctionNameTypeSelector}) { 1141 if (name_type_mask & type) { 1142 auto map = GetNameToSymbolIndexMap(type); 1143 1144 const UniqueCStringMap<uint32_t>::Entry *match; 1145 for (match = map.FindFirstValueForName(name); match != nullptr; 1146 match = map.FindNextValueForName(match)) { 1147 symbol_indexes.push_back(match->value); 1148 } 1149 } 1150 } 1151 1152 if (!symbol_indexes.empty()) { 1153 llvm::sort(symbol_indexes); 1154 symbol_indexes.erase( 1155 std::unique(symbol_indexes.begin(), symbol_indexes.end()), 1156 symbol_indexes.end()); 1157 SymbolIndicesToSymbolContextList(symbol_indexes, sc_list); 1158 } 1159 } 1160 1161 const Symbol *Symtab::GetParent(Symbol *child_symbol) const { 1162 uint32_t child_idx = GetIndexForSymbol(child_symbol); 1163 if (child_idx != UINT32_MAX && child_idx > 0) { 1164 for (uint32_t idx = child_idx - 1; idx != UINT32_MAX; --idx) { 1165 const Symbol *symbol = SymbolAtIndex(idx); 1166 const uint32_t sibling_idx = symbol->GetSiblingIndex(); 1167 if (sibling_idx != UINT32_MAX && sibling_idx > child_idx) 1168 return symbol; 1169 } 1170 } 1171 return nullptr; 1172 } 1173 1174 std::string Symtab::GetCacheKey() { 1175 std::string key; 1176 llvm::raw_string_ostream strm(key); 1177 // Symbol table can come from different object files for the same module. A 1178 // module can have one object file as the main executable and might have 1179 // another object file in a separate symbol file. 1180 strm << m_objfile->GetModule()->GetCacheKey() << "-symtab-" 1181 << llvm::format_hex(m_objfile->GetCacheHash(), 10); 1182 return strm.str(); 1183 } 1184 1185 void Symtab::SaveToCache() { 1186 DataFileCache *cache = Module::GetIndexCache(); 1187 if (!cache) 1188 return; // Caching is not enabled. 1189 InitNameIndexes(); // Init the name indexes so we can cache them as well. 1190 const auto byte_order = endian::InlHostByteOrder(); 1191 DataEncoder file(byte_order, /*addr_size=*/8); 1192 // Encode will return false if the symbol table's object file doesn't have 1193 // anything to make a signature from. 1194 if (Encode(file)) 1195 if (cache->SetCachedData(GetCacheKey(), file.GetData())) 1196 SetWasSavedToCache(); 1197 } 1198 1199 constexpr llvm::StringLiteral kIdentifierCStrMap("CMAP"); 1200 1201 static void EncodeCStrMap(DataEncoder &encoder, ConstStringTable &strtab, 1202 const UniqueCStringMap<uint32_t> &cstr_map) { 1203 encoder.AppendData(kIdentifierCStrMap); 1204 encoder.AppendU32(cstr_map.GetSize()); 1205 for (const auto &entry: cstr_map) { 1206 // Make sure there are no empty strings. 1207 assert((bool)entry.cstring); 1208 encoder.AppendU32(strtab.Add(entry.cstring)); 1209 encoder.AppendU32(entry.value); 1210 } 1211 } 1212 1213 bool DecodeCStrMap(const DataExtractor &data, lldb::offset_t *offset_ptr, 1214 const StringTableReader &strtab, 1215 UniqueCStringMap<uint32_t> &cstr_map) { 1216 llvm::StringRef identifier((const char *)data.GetData(offset_ptr, 4), 4); 1217 if (identifier != kIdentifierCStrMap) 1218 return false; 1219 const uint32_t count = data.GetU32(offset_ptr); 1220 cstr_map.Reserve(count); 1221 for (uint32_t i=0; i<count; ++i) 1222 { 1223 llvm::StringRef str(strtab.Get(data.GetU32(offset_ptr))); 1224 uint32_t value = data.GetU32(offset_ptr); 1225 // No empty strings in the name indexes in Symtab 1226 if (str.empty()) 1227 return false; 1228 cstr_map.Append(ConstString(str), value); 1229 } 1230 // We must sort the UniqueCStringMap after decoding it since it is a vector 1231 // of UniqueCStringMap::Entry objects which contain a ConstString and type T. 1232 // ConstString objects are sorted by "const char *" and then type T and 1233 // the "const char *" are point values that will depend on the order in which 1234 // ConstString objects are created and in which of the 256 string pools they 1235 // are created in. So after we decode all of the entries, we must sort the 1236 // name map to ensure name lookups succeed. If we encode and decode within 1237 // the same process we wouldn't need to sort, so unit testing didn't catch 1238 // this issue when first checked in. 1239 cstr_map.Sort(); 1240 return true; 1241 } 1242 1243 constexpr llvm::StringLiteral kIdentifierSymbolTable("SYMB"); 1244 constexpr uint32_t CURRENT_CACHE_VERSION = 1; 1245 1246 /// The encoding format for the symbol table is as follows: 1247 /// 1248 /// Signature signature; 1249 /// ConstStringTable strtab; 1250 /// Identifier four character code: 'SYMB' 1251 /// uint32_t version; 1252 /// uint32_t num_symbols; 1253 /// Symbol symbols[num_symbols]; 1254 /// uint8_t num_cstr_maps; 1255 /// UniqueCStringMap<uint32_t> cstr_maps[num_cstr_maps] 1256 bool Symtab::Encode(DataEncoder &encoder) const { 1257 // Name indexes must be computed before calling this function. 1258 assert(m_name_indexes_computed); 1259 1260 // Encode the object file's signature 1261 CacheSignature signature(m_objfile); 1262 if (!signature.Encode(encoder)) 1263 return false; 1264 ConstStringTable strtab; 1265 1266 // Encoder the symbol table into a separate encoder first. This allows us 1267 // gather all of the strings we willl need in "strtab" as we will need to 1268 // write the string table out before the symbol table. 1269 DataEncoder symtab_encoder(encoder.GetByteOrder(), 1270 encoder.GetAddressByteSize()); 1271 symtab_encoder.AppendData(kIdentifierSymbolTable); 1272 // Encode the symtab data version. 1273 symtab_encoder.AppendU32(CURRENT_CACHE_VERSION); 1274 // Encode the number of symbols. 1275 symtab_encoder.AppendU32(m_symbols.size()); 1276 // Encode the symbol data for all symbols. 1277 for (const auto &symbol: m_symbols) 1278 symbol.Encode(symtab_encoder, strtab); 1279 1280 // Emit a byte for how many C string maps we emit. We will fix this up after 1281 // we emit the C string maps since we skip emitting C string maps if they are 1282 // empty. 1283 size_t num_cmaps_offset = symtab_encoder.GetByteSize(); 1284 uint8_t num_cmaps = 0; 1285 symtab_encoder.AppendU8(0); 1286 for (const auto &pair: m_name_to_symbol_indices) { 1287 if (pair.second.IsEmpty()) 1288 continue; 1289 ++num_cmaps; 1290 symtab_encoder.AppendU8(pair.first); 1291 EncodeCStrMap(symtab_encoder, strtab, pair.second); 1292 } 1293 if (num_cmaps > 0) 1294 symtab_encoder.PutU8(num_cmaps_offset, num_cmaps); 1295 1296 // Now that all strings have been gathered, we will emit the string table. 1297 strtab.Encode(encoder); 1298 // Followed by the symbol table data. 1299 encoder.AppendData(symtab_encoder.GetData()); 1300 return true; 1301 } 1302 1303 bool Symtab::Decode(const DataExtractor &data, lldb::offset_t *offset_ptr, 1304 bool &signature_mismatch) { 1305 signature_mismatch = false; 1306 CacheSignature signature; 1307 StringTableReader strtab; 1308 { // Scope for "elapsed" object below so it can measure the time parse. 1309 ElapsedTime elapsed(m_objfile->GetModule()->GetSymtabParseTime()); 1310 if (!signature.Decode(data, offset_ptr)) 1311 return false; 1312 if (CacheSignature(m_objfile) != signature) { 1313 signature_mismatch = true; 1314 return false; 1315 } 1316 // We now decode the string table for all strings in the data cache file. 1317 if (!strtab.Decode(data, offset_ptr)) 1318 return false; 1319 1320 // And now we can decode the symbol table with string table we just decoded. 1321 llvm::StringRef identifier((const char *)data.GetData(offset_ptr, 4), 4); 1322 if (identifier != kIdentifierSymbolTable) 1323 return false; 1324 const uint32_t version = data.GetU32(offset_ptr); 1325 if (version != CURRENT_CACHE_VERSION) 1326 return false; 1327 const uint32_t num_symbols = data.GetU32(offset_ptr); 1328 if (num_symbols == 0) 1329 return true; 1330 m_symbols.resize(num_symbols); 1331 SectionList *sections = m_objfile->GetModule()->GetSectionList(); 1332 for (uint32_t i=0; i<num_symbols; ++i) { 1333 if (!m_symbols[i].Decode(data, offset_ptr, sections, strtab)) 1334 return false; 1335 } 1336 } 1337 1338 { // Scope for "elapsed" object below so it can measure the time to index. 1339 ElapsedTime elapsed(m_objfile->GetModule()->GetSymtabIndexTime()); 1340 const uint8_t num_cstr_maps = data.GetU8(offset_ptr); 1341 for (uint8_t i=0; i<num_cstr_maps; ++i) { 1342 uint8_t type = data.GetU8(offset_ptr); 1343 UniqueCStringMap<uint32_t> &cstr_map = 1344 GetNameToSymbolIndexMap((lldb::FunctionNameType)type); 1345 if (!DecodeCStrMap(data, offset_ptr, strtab, cstr_map)) 1346 return false; 1347 } 1348 m_name_indexes_computed = true; 1349 } 1350 return true; 1351 } 1352 1353 bool Symtab::LoadFromCache() { 1354 DataFileCache *cache = Module::GetIndexCache(); 1355 if (!cache) 1356 return false; 1357 1358 std::unique_ptr<llvm::MemoryBuffer> mem_buffer_up = 1359 cache->GetCachedData(GetCacheKey()); 1360 if (!mem_buffer_up) 1361 return false; 1362 DataExtractor data(mem_buffer_up->getBufferStart(), 1363 mem_buffer_up->getBufferSize(), 1364 m_objfile->GetByteOrder(), 1365 m_objfile->GetAddressByteSize()); 1366 bool signature_mismatch = false; 1367 lldb::offset_t offset = 0; 1368 const bool result = Decode(data, &offset, signature_mismatch); 1369 if (signature_mismatch) 1370 cache->RemoveCacheFile(GetCacheKey()); 1371 if (result) 1372 SetWasLoadedFromCache(); 1373 return result; 1374 } 1375