1 //===-- ObjectFileMachO.cpp -------------------------------------*- C++ -*-===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 10 #include "llvm/ADT/StringRef.h" 11 #include "llvm/Support/MachO.h" 12 13 #include "ObjectFileMachO.h" 14 15 #include "lldb/lldb-private-log.h" 16 #include "lldb/Core/ArchSpec.h" 17 #include "lldb/Core/DataBuffer.h" 18 #include "lldb/Core/Debugger.h" 19 #include "lldb/Core/FileSpecList.h" 20 #include "lldb/Core/Log.h" 21 #include "lldb/Core/Module.h" 22 #include "lldb/Core/ModuleSpec.h" 23 #include "lldb/Core/PluginManager.h" 24 #include "lldb/Core/RangeMap.h" 25 #include "lldb/Core/Section.h" 26 #include "lldb/Core/StreamFile.h" 27 #include "lldb/Core/StreamString.h" 28 #include "lldb/Core/Timer.h" 29 #include "lldb/Core/UUID.h" 30 #include "lldb/Host/Host.h" 31 #include "lldb/Host/FileSpec.h" 32 #include "lldb/Symbol/ClangNamespaceDecl.h" 33 #include "lldb/Symbol/DWARFCallFrameInfo.h" 34 #include "lldb/Symbol/ObjectFile.h" 35 #include "lldb/Target/Platform.h" 36 #include "lldb/Target/Process.h" 37 #include "lldb/Target/Target.h" 38 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 39 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 40 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 41 42 #if defined (__APPLE__) && defined (__arm__) 43 // GetLLDBSharedCacheUUID() needs to call dlsym() 44 #include <dlfcn.h> 45 #endif 46 47 #ifndef __APPLE__ 48 #include "Utility/UuidCompatibility.h" 49 #endif 50 51 using namespace lldb; 52 using namespace lldb_private; 53 using namespace llvm::MachO; 54 55 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 56 { 57 public: 58 RegisterContextDarwin_x86_64_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 59 RegisterContextDarwin_x86_64 (thread, 0) 60 { 61 SetRegisterDataFrom_LC_THREAD (data); 62 } 63 64 virtual void 65 InvalidateAllRegisters () 66 { 67 // Do nothing... registers are always valid... 68 } 69 70 void 71 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 72 { 73 lldb::offset_t offset = 0; 74 SetError (GPRRegSet, Read, -1); 75 SetError (FPURegSet, Read, -1); 76 SetError (EXCRegSet, Read, -1); 77 bool done = false; 78 79 while (!done) 80 { 81 int flavor = data.GetU32 (&offset); 82 if (flavor == 0) 83 done = true; 84 else 85 { 86 uint32_t i; 87 uint32_t count = data.GetU32 (&offset); 88 switch (flavor) 89 { 90 case GPRRegSet: 91 for (i=0; i<count; ++i) 92 (&gpr.rax)[i] = data.GetU64(&offset); 93 SetError (GPRRegSet, Read, 0); 94 done = true; 95 96 break; 97 case FPURegSet: 98 // TODO: fill in FPU regs.... 99 //SetError (FPURegSet, Read, -1); 100 done = true; 101 102 break; 103 case EXCRegSet: 104 exc.trapno = data.GetU32(&offset); 105 exc.err = data.GetU32(&offset); 106 exc.faultvaddr = data.GetU64(&offset); 107 SetError (EXCRegSet, Read, 0); 108 done = true; 109 break; 110 case 7: 111 case 8: 112 case 9: 113 // fancy flavors that encapsulate of the the above 114 // falvors... 115 break; 116 117 default: 118 done = true; 119 break; 120 } 121 } 122 } 123 } 124 protected: 125 virtual int 126 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 127 { 128 return 0; 129 } 130 131 virtual int 132 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 133 { 134 return 0; 135 } 136 137 virtual int 138 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 139 { 140 return 0; 141 } 142 143 virtual int 144 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 145 { 146 return 0; 147 } 148 149 virtual int 150 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 151 { 152 return 0; 153 } 154 155 virtual int 156 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 157 { 158 return 0; 159 } 160 }; 161 162 163 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 164 { 165 public: 166 RegisterContextDarwin_i386_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 167 RegisterContextDarwin_i386 (thread, 0) 168 { 169 SetRegisterDataFrom_LC_THREAD (data); 170 } 171 172 virtual void 173 InvalidateAllRegisters () 174 { 175 // Do nothing... registers are always valid... 176 } 177 178 void 179 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 180 { 181 lldb::offset_t offset = 0; 182 SetError (GPRRegSet, Read, -1); 183 SetError (FPURegSet, Read, -1); 184 SetError (EXCRegSet, Read, -1); 185 bool done = false; 186 187 while (!done) 188 { 189 int flavor = data.GetU32 (&offset); 190 if (flavor == 0) 191 done = true; 192 else 193 { 194 uint32_t i; 195 uint32_t count = data.GetU32 (&offset); 196 switch (flavor) 197 { 198 case GPRRegSet: 199 for (i=0; i<count; ++i) 200 (&gpr.eax)[i] = data.GetU32(&offset); 201 SetError (GPRRegSet, Read, 0); 202 done = true; 203 204 break; 205 case FPURegSet: 206 // TODO: fill in FPU regs.... 207 //SetError (FPURegSet, Read, -1); 208 done = true; 209 210 break; 211 case EXCRegSet: 212 exc.trapno = data.GetU32(&offset); 213 exc.err = data.GetU32(&offset); 214 exc.faultvaddr = data.GetU32(&offset); 215 SetError (EXCRegSet, Read, 0); 216 done = true; 217 break; 218 case 7: 219 case 8: 220 case 9: 221 // fancy flavors that encapsulate of the the above 222 // falvors... 223 break; 224 225 default: 226 done = true; 227 break; 228 } 229 } 230 } 231 } 232 protected: 233 virtual int 234 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 235 { 236 return 0; 237 } 238 239 virtual int 240 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 241 { 242 return 0; 243 } 244 245 virtual int 246 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 247 { 248 return 0; 249 } 250 251 virtual int 252 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 253 { 254 return 0; 255 } 256 257 virtual int 258 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 259 { 260 return 0; 261 } 262 263 virtual int 264 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 265 { 266 return 0; 267 } 268 }; 269 270 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm 271 { 272 public: 273 RegisterContextDarwin_arm_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 274 RegisterContextDarwin_arm (thread, 0) 275 { 276 SetRegisterDataFrom_LC_THREAD (data); 277 } 278 279 virtual void 280 InvalidateAllRegisters () 281 { 282 // Do nothing... registers are always valid... 283 } 284 285 void 286 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 287 { 288 lldb::offset_t offset = 0; 289 SetError (GPRRegSet, Read, -1); 290 SetError (FPURegSet, Read, -1); 291 SetError (EXCRegSet, Read, -1); 292 bool done = false; 293 294 while (!done) 295 { 296 int flavor = data.GetU32 (&offset); 297 uint32_t count = data.GetU32 (&offset); 298 lldb::offset_t next_thread_state = offset + (count * 4); 299 switch (flavor) 300 { 301 case GPRRegSet: 302 for (uint32_t i=0; i<count; ++i) 303 { 304 gpr.r[i] = data.GetU32(&offset); 305 } 306 307 // Note that gpr.cpsr is also copied by the above loop; this loop technically extends 308 // one element past the end of the gpr.r[] array. 309 310 SetError (GPRRegSet, Read, 0); 311 offset = next_thread_state; 312 break; 313 314 case FPURegSet: 315 { 316 uint8_t *fpu_reg_buf = (uint8_t*) &fpu.floats.s[0]; 317 const int fpu_reg_buf_size = sizeof (fpu.floats); 318 if (data.ExtractBytes (offset, fpu_reg_buf_size, eByteOrderLittle, fpu_reg_buf) == fpu_reg_buf_size) 319 { 320 offset += fpu_reg_buf_size; 321 fpu.fpscr = data.GetU32(&offset); 322 SetError (FPURegSet, Read, 0); 323 } 324 else 325 { 326 done = true; 327 } 328 } 329 offset = next_thread_state; 330 break; 331 332 case EXCRegSet: 333 if (count == 3) 334 { 335 exc.exception = data.GetU32(&offset); 336 exc.fsr = data.GetU32(&offset); 337 exc.far = data.GetU32(&offset); 338 SetError (EXCRegSet, Read, 0); 339 } 340 done = true; 341 offset = next_thread_state; 342 break; 343 344 // Unknown register set flavor, stop trying to parse. 345 default: 346 done = true; 347 } 348 } 349 } 350 protected: 351 virtual int 352 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 353 { 354 return -1; 355 } 356 357 virtual int 358 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 359 { 360 return -1; 361 } 362 363 virtual int 364 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 365 { 366 return -1; 367 } 368 369 virtual int 370 DoReadDBG (lldb::tid_t tid, int flavor, DBG &dbg) 371 { 372 return -1; 373 } 374 375 virtual int 376 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 377 { 378 return 0; 379 } 380 381 virtual int 382 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 383 { 384 return 0; 385 } 386 387 virtual int 388 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 389 { 390 return 0; 391 } 392 393 virtual int 394 DoWriteDBG (lldb::tid_t tid, int flavor, const DBG &dbg) 395 { 396 return -1; 397 } 398 }; 399 400 static uint32_t 401 MachHeaderSizeFromMagic(uint32_t magic) 402 { 403 switch (magic) 404 { 405 case MH_MAGIC: 406 case MH_CIGAM: 407 return sizeof(struct mach_header); 408 409 case MH_MAGIC_64: 410 case MH_CIGAM_64: 411 return sizeof(struct mach_header_64); 412 break; 413 414 default: 415 break; 416 } 417 return 0; 418 } 419 420 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 421 422 void 423 ObjectFileMachO::Initialize() 424 { 425 PluginManager::RegisterPlugin (GetPluginNameStatic(), 426 GetPluginDescriptionStatic(), 427 CreateInstance, 428 CreateMemoryInstance, 429 GetModuleSpecifications); 430 } 431 432 void 433 ObjectFileMachO::Terminate() 434 { 435 PluginManager::UnregisterPlugin (CreateInstance); 436 } 437 438 439 lldb_private::ConstString 440 ObjectFileMachO::GetPluginNameStatic() 441 { 442 static ConstString g_name("mach-o"); 443 return g_name; 444 } 445 446 const char * 447 ObjectFileMachO::GetPluginDescriptionStatic() 448 { 449 return "Mach-o object file reader (32 and 64 bit)"; 450 } 451 452 ObjectFile * 453 ObjectFileMachO::CreateInstance (const lldb::ModuleSP &module_sp, 454 DataBufferSP& data_sp, 455 lldb::offset_t data_offset, 456 const FileSpec* file, 457 lldb::offset_t file_offset, 458 lldb::offset_t length) 459 { 460 if (!data_sp) 461 { 462 data_sp = file->MemoryMapFileContents(file_offset, length); 463 data_offset = 0; 464 } 465 466 if (ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 467 { 468 // Update the data to contain the entire file if it doesn't already 469 if (data_sp->GetByteSize() < length) 470 { 471 data_sp = file->MemoryMapFileContents(file_offset, length); 472 data_offset = 0; 473 } 474 std::unique_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module_sp, data_sp, data_offset, file, file_offset, length)); 475 if (objfile_ap.get() && objfile_ap->ParseHeader()) 476 return objfile_ap.release(); 477 } 478 return NULL; 479 } 480 481 ObjectFile * 482 ObjectFileMachO::CreateMemoryInstance (const lldb::ModuleSP &module_sp, 483 DataBufferSP& data_sp, 484 const ProcessSP &process_sp, 485 lldb::addr_t header_addr) 486 { 487 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) 488 { 489 std::unique_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module_sp, data_sp, process_sp, header_addr)); 490 if (objfile_ap.get() && objfile_ap->ParseHeader()) 491 return objfile_ap.release(); 492 } 493 return NULL; 494 } 495 496 size_t 497 ObjectFileMachO::GetModuleSpecifications (const lldb_private::FileSpec& file, 498 lldb::DataBufferSP& data_sp, 499 lldb::offset_t data_offset, 500 lldb::offset_t file_offset, 501 lldb::offset_t length, 502 lldb_private::ModuleSpecList &specs) 503 { 504 const size_t initial_count = specs.GetSize(); 505 506 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) 507 { 508 DataExtractor data; 509 data.SetData(data_sp); 510 llvm::MachO::mach_header header; 511 if (ParseHeader (data, &data_offset, header)) 512 { 513 if (header.sizeofcmds >= data_sp->GetByteSize()) 514 { 515 data_sp = file.ReadFileContents(file_offset, header.sizeofcmds); 516 data.SetData(data_sp); 517 data_offset = MachHeaderSizeFromMagic(header.magic); 518 } 519 if (data_sp) 520 { 521 ModuleSpec spec; 522 spec.GetFileSpec() = file; 523 spec.GetArchitecture().SetArchitecture(eArchTypeMachO, 524 header.cputype, 525 header.cpusubtype); 526 if (header.filetype == MH_PRELOAD) // 0x5u 527 { 528 // Set OS to "unknown" - this is a standalone binary with no dyld et al 529 spec.GetArchitecture().GetTriple().setOS (llvm::Triple::UnknownOS); 530 } 531 if (spec.GetArchitecture().IsValid()) 532 { 533 GetUUID (header, data, data_offset, spec.GetUUID()); 534 specs.Append(spec); 535 } 536 } 537 } 538 } 539 return specs.GetSize() - initial_count; 540 } 541 542 543 544 const ConstString & 545 ObjectFileMachO::GetSegmentNameTEXT() 546 { 547 static ConstString g_segment_name_TEXT ("__TEXT"); 548 return g_segment_name_TEXT; 549 } 550 551 const ConstString & 552 ObjectFileMachO::GetSegmentNameDATA() 553 { 554 static ConstString g_segment_name_DATA ("__DATA"); 555 return g_segment_name_DATA; 556 } 557 558 const ConstString & 559 ObjectFileMachO::GetSegmentNameOBJC() 560 { 561 static ConstString g_segment_name_OBJC ("__OBJC"); 562 return g_segment_name_OBJC; 563 } 564 565 const ConstString & 566 ObjectFileMachO::GetSegmentNameLINKEDIT() 567 { 568 static ConstString g_section_name_LINKEDIT ("__LINKEDIT"); 569 return g_section_name_LINKEDIT; 570 } 571 572 const ConstString & 573 ObjectFileMachO::GetSectionNameEHFrame() 574 { 575 static ConstString g_section_name_eh_frame ("__eh_frame"); 576 return g_section_name_eh_frame; 577 } 578 579 bool 580 ObjectFileMachO::MagicBytesMatch (DataBufferSP& data_sp, 581 lldb::addr_t data_offset, 582 lldb::addr_t data_length) 583 { 584 DataExtractor data; 585 data.SetData (data_sp, data_offset, data_length); 586 lldb::offset_t offset = 0; 587 uint32_t magic = data.GetU32(&offset); 588 return MachHeaderSizeFromMagic(magic) != 0; 589 } 590 591 592 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 593 DataBufferSP& data_sp, 594 lldb::offset_t data_offset, 595 const FileSpec* file, 596 lldb::offset_t file_offset, 597 lldb::offset_t length) : 598 ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 599 m_mach_segments(), 600 m_mach_sections(), 601 m_entry_point_address(), 602 m_thread_context_offsets(), 603 m_thread_context_offsets_valid(false) 604 { 605 ::memset (&m_header, 0, sizeof(m_header)); 606 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 607 } 608 609 ObjectFileMachO::ObjectFileMachO (const lldb::ModuleSP &module_sp, 610 lldb::DataBufferSP& header_data_sp, 611 const lldb::ProcessSP &process_sp, 612 lldb::addr_t header_addr) : 613 ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 614 m_mach_segments(), 615 m_mach_sections(), 616 m_entry_point_address(), 617 m_thread_context_offsets(), 618 m_thread_context_offsets_valid(false) 619 { 620 ::memset (&m_header, 0, sizeof(m_header)); 621 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 622 } 623 624 ObjectFileMachO::~ObjectFileMachO() 625 { 626 } 627 628 bool 629 ObjectFileMachO::ParseHeader (DataExtractor &data, 630 lldb::offset_t *data_offset_ptr, 631 llvm::MachO::mach_header &header) 632 { 633 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 634 // Leave magic in the original byte order 635 header.magic = data.GetU32(data_offset_ptr); 636 bool can_parse = false; 637 bool is_64_bit = false; 638 switch (header.magic) 639 { 640 case MH_MAGIC: 641 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 642 data.SetAddressByteSize(4); 643 can_parse = true; 644 break; 645 646 case MH_MAGIC_64: 647 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 648 data.SetAddressByteSize(8); 649 can_parse = true; 650 is_64_bit = true; 651 break; 652 653 case MH_CIGAM: 654 data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 655 data.SetAddressByteSize(4); 656 can_parse = true; 657 break; 658 659 case MH_CIGAM_64: 660 data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 661 data.SetAddressByteSize(8); 662 is_64_bit = true; 663 can_parse = true; 664 break; 665 666 default: 667 break; 668 } 669 670 if (can_parse) 671 { 672 data.GetU32(data_offset_ptr, &header.cputype, 6); 673 if (is_64_bit) 674 *data_offset_ptr += 4; 675 return true; 676 } 677 else 678 { 679 memset(&header, 0, sizeof(header)); 680 } 681 return false; 682 } 683 684 bool 685 ObjectFileMachO::ParseHeader () 686 { 687 ModuleSP module_sp(GetModule()); 688 if (module_sp) 689 { 690 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 691 bool can_parse = false; 692 lldb::offset_t offset = 0; 693 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 694 // Leave magic in the original byte order 695 m_header.magic = m_data.GetU32(&offset); 696 switch (m_header.magic) 697 { 698 case MH_MAGIC: 699 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 700 m_data.SetAddressByteSize(4); 701 can_parse = true; 702 break; 703 704 case MH_MAGIC_64: 705 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 706 m_data.SetAddressByteSize(8); 707 can_parse = true; 708 break; 709 710 case MH_CIGAM: 711 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 712 m_data.SetAddressByteSize(4); 713 can_parse = true; 714 break; 715 716 case MH_CIGAM_64: 717 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 718 m_data.SetAddressByteSize(8); 719 can_parse = true; 720 break; 721 722 default: 723 break; 724 } 725 726 if (can_parse) 727 { 728 m_data.GetU32(&offset, &m_header.cputype, 6); 729 730 ArchSpec mach_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 731 732 // Check if the module has a required architecture 733 const ArchSpec &module_arch = module_sp->GetArchitecture(); 734 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 735 return false; 736 737 if (SetModulesArchitecture (mach_arch)) 738 { 739 const size_t header_and_lc_size = m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 740 if (m_data.GetByteSize() < header_and_lc_size) 741 { 742 DataBufferSP data_sp; 743 ProcessSP process_sp (m_process_wp.lock()); 744 if (process_sp) 745 { 746 data_sp = ReadMemory (process_sp, m_memory_addr, header_and_lc_size); 747 } 748 else 749 { 750 // Read in all only the load command data from the file on disk 751 data_sp = m_file.ReadFileContents(m_file_offset, header_and_lc_size); 752 if (data_sp->GetByteSize() != header_and_lc_size) 753 return false; 754 } 755 if (data_sp) 756 m_data.SetData (data_sp); 757 } 758 } 759 return true; 760 } 761 else 762 { 763 memset(&m_header, 0, sizeof(struct mach_header)); 764 } 765 } 766 return false; 767 } 768 769 770 ByteOrder 771 ObjectFileMachO::GetByteOrder () const 772 { 773 return m_data.GetByteOrder (); 774 } 775 776 bool 777 ObjectFileMachO::IsExecutable() const 778 { 779 return m_header.filetype == MH_EXECUTE; 780 } 781 782 uint32_t 783 ObjectFileMachO::GetAddressByteSize () const 784 { 785 return m_data.GetAddressByteSize (); 786 } 787 788 AddressClass 789 ObjectFileMachO::GetAddressClass (lldb::addr_t file_addr) 790 { 791 Symtab *symtab = GetSymtab(); 792 if (symtab) 793 { 794 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 795 if (symbol) 796 { 797 if (symbol->ValueIsAddress()) 798 { 799 SectionSP section_sp (symbol->GetAddress().GetSection()); 800 if (section_sp) 801 { 802 const lldb::SectionType section_type = section_sp->GetType(); 803 switch (section_type) 804 { 805 case eSectionTypeInvalid: return eAddressClassUnknown; 806 case eSectionTypeCode: 807 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) 808 { 809 // For ARM we have a bit in the n_desc field of the symbol 810 // that tells us ARM/Thumb which is bit 0x0008. 811 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 812 return eAddressClassCodeAlternateISA; 813 } 814 return eAddressClassCode; 815 816 case eSectionTypeContainer: return eAddressClassUnknown; 817 case eSectionTypeData: 818 case eSectionTypeDataCString: 819 case eSectionTypeDataCStringPointers: 820 case eSectionTypeDataSymbolAddress: 821 case eSectionTypeData4: 822 case eSectionTypeData8: 823 case eSectionTypeData16: 824 case eSectionTypeDataPointers: 825 case eSectionTypeZeroFill: 826 case eSectionTypeDataObjCMessageRefs: 827 case eSectionTypeDataObjCCFStrings: 828 return eAddressClassData; 829 case eSectionTypeDebug: 830 case eSectionTypeDWARFDebugAbbrev: 831 case eSectionTypeDWARFDebugAranges: 832 case eSectionTypeDWARFDebugFrame: 833 case eSectionTypeDWARFDebugInfo: 834 case eSectionTypeDWARFDebugLine: 835 case eSectionTypeDWARFDebugLoc: 836 case eSectionTypeDWARFDebugMacInfo: 837 case eSectionTypeDWARFDebugPubNames: 838 case eSectionTypeDWARFDebugPubTypes: 839 case eSectionTypeDWARFDebugRanges: 840 case eSectionTypeDWARFDebugStr: 841 case eSectionTypeDWARFAppleNames: 842 case eSectionTypeDWARFAppleTypes: 843 case eSectionTypeDWARFAppleNamespaces: 844 case eSectionTypeDWARFAppleObjC: 845 return eAddressClassDebug; 846 case eSectionTypeEHFrame: return eAddressClassRuntime; 847 case eSectionTypeELFSymbolTable: 848 case eSectionTypeELFDynamicSymbols: 849 case eSectionTypeELFRelocationEntries: 850 case eSectionTypeELFDynamicLinkInfo: 851 case eSectionTypeOther: return eAddressClassUnknown; 852 } 853 } 854 } 855 856 const SymbolType symbol_type = symbol->GetType(); 857 switch (symbol_type) 858 { 859 case eSymbolTypeAny: return eAddressClassUnknown; 860 case eSymbolTypeAbsolute: return eAddressClassUnknown; 861 862 case eSymbolTypeCode: 863 case eSymbolTypeTrampoline: 864 case eSymbolTypeResolver: 865 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) 866 { 867 // For ARM we have a bit in the n_desc field of the symbol 868 // that tells us ARM/Thumb which is bit 0x0008. 869 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 870 return eAddressClassCodeAlternateISA; 871 } 872 return eAddressClassCode; 873 874 case eSymbolTypeData: return eAddressClassData; 875 case eSymbolTypeRuntime: return eAddressClassRuntime; 876 case eSymbolTypeException: return eAddressClassRuntime; 877 case eSymbolTypeSourceFile: return eAddressClassDebug; 878 case eSymbolTypeHeaderFile: return eAddressClassDebug; 879 case eSymbolTypeObjectFile: return eAddressClassDebug; 880 case eSymbolTypeCommonBlock: return eAddressClassDebug; 881 case eSymbolTypeBlock: return eAddressClassDebug; 882 case eSymbolTypeLocal: return eAddressClassData; 883 case eSymbolTypeParam: return eAddressClassData; 884 case eSymbolTypeVariable: return eAddressClassData; 885 case eSymbolTypeVariableType: return eAddressClassDebug; 886 case eSymbolTypeLineEntry: return eAddressClassDebug; 887 case eSymbolTypeLineHeader: return eAddressClassDebug; 888 case eSymbolTypeScopeBegin: return eAddressClassDebug; 889 case eSymbolTypeScopeEnd: return eAddressClassDebug; 890 case eSymbolTypeAdditional: return eAddressClassUnknown; 891 case eSymbolTypeCompiler: return eAddressClassDebug; 892 case eSymbolTypeInstrumentation:return eAddressClassDebug; 893 case eSymbolTypeUndefined: return eAddressClassUnknown; 894 case eSymbolTypeObjCClass: return eAddressClassRuntime; 895 case eSymbolTypeObjCMetaClass: return eAddressClassRuntime; 896 case eSymbolTypeObjCIVar: return eAddressClassRuntime; 897 case eSymbolTypeReExported: return eAddressClassRuntime; 898 } 899 } 900 } 901 return eAddressClassUnknown; 902 } 903 904 Symtab * 905 ObjectFileMachO::GetSymtab() 906 { 907 ModuleSP module_sp(GetModule()); 908 if (module_sp) 909 { 910 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 911 if (m_symtab_ap.get() == NULL) 912 { 913 m_symtab_ap.reset(new Symtab(this)); 914 Mutex::Locker symtab_locker (m_symtab_ap->GetMutex()); 915 ParseSymtab (); 916 m_symtab_ap->Finalize (); 917 } 918 } 919 return m_symtab_ap.get(); 920 } 921 922 bool 923 ObjectFileMachO::IsStripped () 924 { 925 if (m_dysymtab.cmd == 0) 926 { 927 ModuleSP module_sp(GetModule()); 928 if (module_sp) 929 { 930 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 931 for (uint32_t i=0; i<m_header.ncmds; ++i) 932 { 933 const lldb::offset_t load_cmd_offset = offset; 934 935 load_command lc; 936 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 937 break; 938 if (lc.cmd == LC_DYSYMTAB) 939 { 940 m_dysymtab.cmd = lc.cmd; 941 m_dysymtab.cmdsize = lc.cmdsize; 942 if (m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == NULL) 943 { 944 // Clear m_dysymtab if we were unable to read all items from the load command 945 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 946 } 947 } 948 offset = load_cmd_offset + lc.cmdsize; 949 } 950 } 951 } 952 if (m_dysymtab.cmd) 953 return m_dysymtab.nlocalsym <= 1; 954 return false; 955 } 956 957 void 958 ObjectFileMachO::CreateSections (SectionList &unified_section_list) 959 { 960 if (!m_sections_ap.get()) 961 { 962 m_sections_ap.reset(new SectionList()); 963 964 const bool is_dsym = (m_header.filetype == MH_DSYM); 965 lldb::user_id_t segID = 0; 966 lldb::user_id_t sectID = 0; 967 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 968 uint32_t i; 969 const bool is_core = GetType() == eTypeCoreFile; 970 //bool dump_sections = false; 971 ModuleSP module_sp (GetModule()); 972 // First look up any LC_ENCRYPTION_INFO load commands 973 typedef RangeArray<uint32_t, uint32_t, 8> EncryptedFileRanges; 974 EncryptedFileRanges encrypted_file_ranges; 975 encryption_info_command encryption_cmd; 976 for (i=0; i<m_header.ncmds; ++i) 977 { 978 const lldb::offset_t load_cmd_offset = offset; 979 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 980 break; 981 982 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO) 983 { 984 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) 985 { 986 if (encryption_cmd.cryptid != 0) 987 { 988 EncryptedFileRanges::Entry entry; 989 entry.SetRangeBase(encryption_cmd.cryptoff); 990 entry.SetByteSize(encryption_cmd.cryptsize); 991 encrypted_file_ranges.Append(entry); 992 } 993 } 994 } 995 offset = load_cmd_offset + encryption_cmd.cmdsize; 996 } 997 998 offset = MachHeaderSizeFromMagic(m_header.magic); 999 1000 struct segment_command_64 load_cmd; 1001 for (i=0; i<m_header.ncmds; ++i) 1002 { 1003 const lldb::offset_t load_cmd_offset = offset; 1004 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1005 break; 1006 1007 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1008 { 1009 if (m_data.GetU8(&offset, (uint8_t*)load_cmd.segname, 16)) 1010 { 1011 bool add_section = true; 1012 bool add_to_unified = true; 1013 ConstString const_segname (load_cmd.segname, std::min<size_t>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 1014 1015 SectionSP unified_section_sp(unified_section_list.FindSectionByName(const_segname)); 1016 if (is_dsym && unified_section_sp) 1017 { 1018 if (const_segname == GetSegmentNameLINKEDIT()) 1019 { 1020 // We need to keep the __LINKEDIT segment private to this object file only 1021 add_to_unified = false; 1022 } 1023 else 1024 { 1025 // This is the dSYM file and this section has already been created by 1026 // the object file, no need to create it. 1027 add_section = false; 1028 } 1029 } 1030 load_cmd.vmaddr = m_data.GetAddress(&offset); 1031 load_cmd.vmsize = m_data.GetAddress(&offset); 1032 load_cmd.fileoff = m_data.GetAddress(&offset); 1033 load_cmd.filesize = m_data.GetAddress(&offset); 1034 if (m_length != 0 && load_cmd.filesize != 0) 1035 { 1036 if (load_cmd.fileoff > m_length) 1037 { 1038 // We have a load command that says it extends past the end of hte file. This is likely 1039 // a corrupt file. We don't have any way to return an error condition here (this method 1040 // was likely invokved from something like ObjectFile::GetSectionList()) -- all we can do 1041 // is null out the SectionList vector and if a process has been set up, dump a message 1042 // to stdout. The most common case here is core file debugging with a truncated file. 1043 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1044 module_sp->ReportWarning("load command %u %s has a fileoff (0x%" PRIx64 ") that extends beyond the end of the file (0x%" PRIx64 "), ignoring this section", 1045 i, 1046 lc_segment_name, 1047 load_cmd.fileoff, 1048 m_length); 1049 1050 load_cmd.fileoff = 0; 1051 load_cmd.filesize = 0; 1052 } 1053 1054 if (load_cmd.fileoff + load_cmd.filesize > m_length) 1055 { 1056 // We have a load command that says it extends past the end of hte file. This is likely 1057 // a corrupt file. We don't have any way to return an error condition here (this method 1058 // was likely invokved from something like ObjectFile::GetSectionList()) -- all we can do 1059 // is null out the SectionList vector and if a process has been set up, dump a message 1060 // to stdout. The most common case here is core file debugging with a truncated file. 1061 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1062 GetModule()->ReportWarning("load command %u %s has a fileoff + filesize (0x%" PRIx64 ") that extends beyond the end of the file (0x%" PRIx64 "), the segment will be truncated to match", 1063 i, 1064 lc_segment_name, 1065 load_cmd.fileoff + load_cmd.filesize, 1066 m_length); 1067 1068 // Tuncase the length 1069 load_cmd.filesize = m_length - load_cmd.fileoff; 1070 } 1071 } 1072 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1073 { 1074 1075 const bool segment_is_encrypted = (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1076 1077 // Keep a list of mach segments around in case we need to 1078 // get at data that isn't stored in the abstracted Sections. 1079 m_mach_segments.push_back (load_cmd); 1080 1081 // Use a segment ID of the segment index shifted left by 8 so they 1082 // never conflict with any of the sections. 1083 SectionSP segment_sp; 1084 if (add_section && (const_segname || is_core)) 1085 { 1086 segment_sp.reset(new Section (module_sp, // Module to which this section belongs 1087 this, // Object file to which this sections belongs 1088 ++segID << 8, // Section ID is the 1 based segment index shifted right by 8 bits as not to collide with any of the 256 section IDs that are possible 1089 const_segname, // Name of this section 1090 eSectionTypeContainer, // This section is a container of other sections. 1091 load_cmd.vmaddr, // File VM address == addresses as they are found in the object file 1092 load_cmd.vmsize, // VM size in bytes of this section 1093 load_cmd.fileoff, // Offset to the data for this section in the file 1094 load_cmd.filesize, // Size in bytes of this section as found in the the file 1095 load_cmd.flags)); // Flags for this section 1096 1097 segment_sp->SetIsEncrypted (segment_is_encrypted); 1098 m_sections_ap->AddSection(segment_sp); 1099 if (add_to_unified) 1100 unified_section_list.AddSection(segment_sp); 1101 } 1102 else if (unified_section_sp) 1103 { 1104 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) 1105 { 1106 // Check to see if the module was read from memory? 1107 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) 1108 { 1109 // We have a module that is in memory and needs to have its 1110 // file address adjusted. We need to do this because when we 1111 // load a file from memory, its addresses will be slid already, 1112 // yet the addresses in the new symbol file will still be unslid. 1113 // Since everything is stored as section offset, this shouldn't 1114 // cause any problems. 1115 1116 // Make sure we've parsed the symbol table from the 1117 // ObjectFile before we go around changing its Sections. 1118 module_sp->GetObjectFile()->GetSymtab(); 1119 // eh_frame would present the same problems but we parse that on 1120 // a per-function basis as-needed so it's more difficult to 1121 // remove its use of the Sections. Realistically, the environments 1122 // where this code path will be taken will not have eh_frame sections. 1123 1124 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1125 } 1126 } 1127 m_sections_ap->AddSection(unified_section_sp); 1128 } 1129 1130 struct section_64 sect64; 1131 ::memset (§64, 0, sizeof(sect64)); 1132 // Push a section into our mach sections for the section at 1133 // index zero (NO_SECT) if we don't have any mach sections yet... 1134 if (m_mach_sections.empty()) 1135 m_mach_sections.push_back(sect64); 1136 uint32_t segment_sect_idx; 1137 const lldb::user_id_t first_segment_sectID = sectID + 1; 1138 1139 1140 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1141 for (segment_sect_idx=0; segment_sect_idx<load_cmd.nsects; ++segment_sect_idx) 1142 { 1143 if (m_data.GetU8(&offset, (uint8_t*)sect64.sectname, sizeof(sect64.sectname)) == NULL) 1144 break; 1145 if (m_data.GetU8(&offset, (uint8_t*)sect64.segname, sizeof(sect64.segname)) == NULL) 1146 break; 1147 sect64.addr = m_data.GetAddress(&offset); 1148 sect64.size = m_data.GetAddress(&offset); 1149 1150 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1151 break; 1152 1153 // Keep a list of mach sections around in case we need to 1154 // get at data that isn't stored in the abstracted Sections. 1155 m_mach_sections.push_back (sect64); 1156 1157 if (add_section) 1158 { 1159 ConstString section_name (sect64.sectname, std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 1160 if (!const_segname) 1161 { 1162 // We have a segment with no name so we need to conjure up 1163 // segments that correspond to the section's segname if there 1164 // isn't already such a section. If there is such a section, 1165 // we resize the section so that it spans all sections. 1166 // We also mark these sections as fake so address matches don't 1167 // hit if they land in the gaps between the child sections. 1168 const_segname.SetTrimmedCStringWithLength(sect64.segname, sizeof(sect64.segname)); 1169 segment_sp = unified_section_list.FindSectionByName (const_segname); 1170 if (segment_sp.get()) 1171 { 1172 Section *segment = segment_sp.get(); 1173 // Grow the section size as needed. 1174 const lldb::addr_t sect64_min_addr = sect64.addr; 1175 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1176 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1177 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1178 const lldb::addr_t curr_seg_max_addr = curr_seg_min_addr + curr_seg_byte_size; 1179 if (sect64_min_addr >= curr_seg_min_addr) 1180 { 1181 const lldb::addr_t new_seg_byte_size = sect64_max_addr - curr_seg_min_addr; 1182 // Only grow the section size if needed 1183 if (new_seg_byte_size > curr_seg_byte_size) 1184 segment->SetByteSize (new_seg_byte_size); 1185 } 1186 else 1187 { 1188 // We need to change the base address of the segment and 1189 // adjust the child section offsets for all existing children. 1190 const lldb::addr_t slide_amount = sect64_min_addr - curr_seg_min_addr; 1191 segment->Slide(slide_amount, false); 1192 segment->GetChildren().Slide(-slide_amount, false); 1193 segment->SetByteSize (curr_seg_max_addr - sect64_min_addr); 1194 } 1195 1196 // Grow the section size as needed. 1197 if (sect64.offset) 1198 { 1199 const lldb::addr_t segment_min_file_offset = segment->GetFileOffset(); 1200 const lldb::addr_t segment_max_file_offset = segment_min_file_offset + segment->GetFileSize(); 1201 1202 const lldb::addr_t section_min_file_offset = sect64.offset; 1203 const lldb::addr_t section_max_file_offset = section_min_file_offset + sect64.size; 1204 const lldb::addr_t new_file_offset = std::min (section_min_file_offset, segment_min_file_offset); 1205 const lldb::addr_t new_file_size = std::max (section_max_file_offset, segment_max_file_offset) - new_file_offset; 1206 segment->SetFileOffset (new_file_offset); 1207 segment->SetFileSize (new_file_size); 1208 } 1209 } 1210 else 1211 { 1212 // Create a fake section for the section's named segment 1213 segment_sp.reset(new Section (segment_sp, // Parent section 1214 module_sp, // Module to which this section belongs 1215 this, // Object file to which this section belongs 1216 ++segID << 8, // Section ID is the 1 based segment index shifted right by 8 bits as not to collide with any of the 256 section IDs that are possible 1217 const_segname, // Name of this section 1218 eSectionTypeContainer, // This section is a container of other sections. 1219 sect64.addr, // File VM address == addresses as they are found in the object file 1220 sect64.size, // VM size in bytes of this section 1221 sect64.offset, // Offset to the data for this section in the file 1222 sect64.offset ? sect64.size : 0, // Size in bytes of this section as found in the the file 1223 load_cmd.flags)); // Flags for this section 1224 segment_sp->SetIsFake(true); 1225 1226 m_sections_ap->AddSection(segment_sp); 1227 if (add_to_unified) 1228 unified_section_list.AddSection(segment_sp); 1229 segment_sp->SetIsEncrypted (segment_is_encrypted); 1230 } 1231 } 1232 assert (segment_sp.get()); 1233 1234 uint32_t mach_sect_type = sect64.flags & SECTION_TYPE; 1235 static ConstString g_sect_name_objc_data ("__objc_data"); 1236 static ConstString g_sect_name_objc_msgrefs ("__objc_msgrefs"); 1237 static ConstString g_sect_name_objc_selrefs ("__objc_selrefs"); 1238 static ConstString g_sect_name_objc_classrefs ("__objc_classrefs"); 1239 static ConstString g_sect_name_objc_superrefs ("__objc_superrefs"); 1240 static ConstString g_sect_name_objc_const ("__objc_const"); 1241 static ConstString g_sect_name_objc_classlist ("__objc_classlist"); 1242 static ConstString g_sect_name_cfstring ("__cfstring"); 1243 1244 static ConstString g_sect_name_dwarf_debug_abbrev ("__debug_abbrev"); 1245 static ConstString g_sect_name_dwarf_debug_aranges ("__debug_aranges"); 1246 static ConstString g_sect_name_dwarf_debug_frame ("__debug_frame"); 1247 static ConstString g_sect_name_dwarf_debug_info ("__debug_info"); 1248 static ConstString g_sect_name_dwarf_debug_line ("__debug_line"); 1249 static ConstString g_sect_name_dwarf_debug_loc ("__debug_loc"); 1250 static ConstString g_sect_name_dwarf_debug_macinfo ("__debug_macinfo"); 1251 static ConstString g_sect_name_dwarf_debug_pubnames ("__debug_pubnames"); 1252 static ConstString g_sect_name_dwarf_debug_pubtypes ("__debug_pubtypes"); 1253 static ConstString g_sect_name_dwarf_debug_ranges ("__debug_ranges"); 1254 static ConstString g_sect_name_dwarf_debug_str ("__debug_str"); 1255 static ConstString g_sect_name_dwarf_apple_names ("__apple_names"); 1256 static ConstString g_sect_name_dwarf_apple_types ("__apple_types"); 1257 static ConstString g_sect_name_dwarf_apple_namespaces ("__apple_namespac"); 1258 static ConstString g_sect_name_dwarf_apple_objc ("__apple_objc"); 1259 static ConstString g_sect_name_eh_frame ("__eh_frame"); 1260 static ConstString g_sect_name_DATA ("__DATA"); 1261 static ConstString g_sect_name_TEXT ("__TEXT"); 1262 1263 lldb::SectionType sect_type = eSectionTypeOther; 1264 1265 if (section_name == g_sect_name_dwarf_debug_abbrev) 1266 sect_type = eSectionTypeDWARFDebugAbbrev; 1267 else if (section_name == g_sect_name_dwarf_debug_aranges) 1268 sect_type = eSectionTypeDWARFDebugAranges; 1269 else if (section_name == g_sect_name_dwarf_debug_frame) 1270 sect_type = eSectionTypeDWARFDebugFrame; 1271 else if (section_name == g_sect_name_dwarf_debug_info) 1272 sect_type = eSectionTypeDWARFDebugInfo; 1273 else if (section_name == g_sect_name_dwarf_debug_line) 1274 sect_type = eSectionTypeDWARFDebugLine; 1275 else if (section_name == g_sect_name_dwarf_debug_loc) 1276 sect_type = eSectionTypeDWARFDebugLoc; 1277 else if (section_name == g_sect_name_dwarf_debug_macinfo) 1278 sect_type = eSectionTypeDWARFDebugMacInfo; 1279 else if (section_name == g_sect_name_dwarf_debug_pubnames) 1280 sect_type = eSectionTypeDWARFDebugPubNames; 1281 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 1282 sect_type = eSectionTypeDWARFDebugPubTypes; 1283 else if (section_name == g_sect_name_dwarf_debug_ranges) 1284 sect_type = eSectionTypeDWARFDebugRanges; 1285 else if (section_name == g_sect_name_dwarf_debug_str) 1286 sect_type = eSectionTypeDWARFDebugStr; 1287 else if (section_name == g_sect_name_dwarf_apple_names) 1288 sect_type = eSectionTypeDWARFAppleNames; 1289 else if (section_name == g_sect_name_dwarf_apple_types) 1290 sect_type = eSectionTypeDWARFAppleTypes; 1291 else if (section_name == g_sect_name_dwarf_apple_namespaces) 1292 sect_type = eSectionTypeDWARFAppleNamespaces; 1293 else if (section_name == g_sect_name_dwarf_apple_objc) 1294 sect_type = eSectionTypeDWARFAppleObjC; 1295 else if (section_name == g_sect_name_objc_selrefs) 1296 sect_type = eSectionTypeDataCStringPointers; 1297 else if (section_name == g_sect_name_objc_msgrefs) 1298 sect_type = eSectionTypeDataObjCMessageRefs; 1299 else if (section_name == g_sect_name_eh_frame) 1300 sect_type = eSectionTypeEHFrame; 1301 else if (section_name == g_sect_name_cfstring) 1302 sect_type = eSectionTypeDataObjCCFStrings; 1303 else if (section_name == g_sect_name_objc_data || 1304 section_name == g_sect_name_objc_classrefs || 1305 section_name == g_sect_name_objc_superrefs || 1306 section_name == g_sect_name_objc_const || 1307 section_name == g_sect_name_objc_classlist) 1308 { 1309 sect_type = eSectionTypeDataPointers; 1310 } 1311 1312 if (sect_type == eSectionTypeOther) 1313 { 1314 switch (mach_sect_type) 1315 { 1316 // TODO: categorize sections by other flags for regular sections 1317 case S_REGULAR: 1318 if (segment_sp->GetName() == g_sect_name_TEXT) 1319 sect_type = eSectionTypeCode; 1320 else if (segment_sp->GetName() == g_sect_name_DATA) 1321 sect_type = eSectionTypeData; 1322 else 1323 sect_type = eSectionTypeOther; 1324 break; 1325 case S_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1326 case S_CSTRING_LITERALS: sect_type = eSectionTypeDataCString; break; // section with only literal C strings 1327 case S_4BYTE_LITERALS: sect_type = eSectionTypeData4; break; // section with only 4 byte literals 1328 case S_8BYTE_LITERALS: sect_type = eSectionTypeData8; break; // section with only 8 byte literals 1329 case S_LITERAL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only pointers to literals 1330 case S_NON_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only non-lazy symbol pointers 1331 case S_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only lazy symbol pointers 1332 case S_SYMBOL_STUBS: sect_type = eSectionTypeCode; break; // section with only symbol stubs, byte size of stub in the reserved2 field 1333 case S_MOD_INIT_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for initialization 1334 case S_MOD_TERM_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for termination 1335 case S_COALESCED: sect_type = eSectionTypeOther; break; 1336 case S_GB_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1337 case S_INTERPOSING: sect_type = eSectionTypeCode; break; // section with only pairs of function pointers for interposing 1338 case S_16BYTE_LITERALS: sect_type = eSectionTypeData16; break; // section with only 16 byte literals 1339 case S_DTRACE_DOF: sect_type = eSectionTypeDebug; break; 1340 case S_LAZY_DYLIB_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; 1341 default: break; 1342 } 1343 } 1344 1345 SectionSP section_sp(new Section (segment_sp, 1346 module_sp, 1347 this, 1348 ++sectID, 1349 section_name, 1350 sect_type, 1351 sect64.addr - segment_sp->GetFileAddress(), 1352 sect64.size, 1353 sect64.offset, 1354 sect64.offset == 0 ? 0 : sect64.size, 1355 sect64.flags)); 1356 // Set the section to be encrypted to match the segment 1357 1358 bool section_is_encrypted = false; 1359 if (!segment_is_encrypted && load_cmd.filesize != 0) 1360 section_is_encrypted = encrypted_file_ranges.FindEntryThatContains(sect64.offset) != NULL; 1361 1362 section_sp->SetIsEncrypted (segment_is_encrypted || section_is_encrypted); 1363 segment_sp->GetChildren().AddSection(section_sp); 1364 1365 if (segment_sp->IsFake()) 1366 { 1367 segment_sp.reset(); 1368 const_segname.Clear(); 1369 } 1370 } 1371 } 1372 if (segment_sp && is_dsym) 1373 { 1374 if (first_segment_sectID <= sectID) 1375 { 1376 lldb::user_id_t sect_uid; 1377 for (sect_uid = first_segment_sectID; sect_uid <= sectID; ++sect_uid) 1378 { 1379 SectionSP curr_section_sp(segment_sp->GetChildren().FindSectionByID (sect_uid)); 1380 SectionSP next_section_sp; 1381 if (sect_uid + 1 <= sectID) 1382 next_section_sp = segment_sp->GetChildren().FindSectionByID (sect_uid+1); 1383 1384 if (curr_section_sp.get()) 1385 { 1386 if (curr_section_sp->GetByteSize() == 0) 1387 { 1388 if (next_section_sp.get() != NULL) 1389 curr_section_sp->SetByteSize ( next_section_sp->GetFileAddress() - curr_section_sp->GetFileAddress() ); 1390 else 1391 curr_section_sp->SetByteSize ( load_cmd.vmsize ); 1392 } 1393 } 1394 } 1395 } 1396 } 1397 } 1398 } 1399 } 1400 else if (load_cmd.cmd == LC_DYSYMTAB) 1401 { 1402 m_dysymtab.cmd = load_cmd.cmd; 1403 m_dysymtab.cmdsize = load_cmd.cmdsize; 1404 m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1405 } 1406 1407 offset = load_cmd_offset + load_cmd.cmdsize; 1408 } 1409 1410 // StreamFile s(stdout, false); // REMOVE THIS LINE 1411 // s.Printf ("Sections for %s:\n", m_file.GetPath().c_str());// REMOVE THIS LINE 1412 // m_sections_ap->Dump(&s, NULL, true, UINT32_MAX);// REMOVE THIS LINE 1413 } 1414 } 1415 1416 class MachSymtabSectionInfo 1417 { 1418 public: 1419 1420 MachSymtabSectionInfo (SectionList *section_list) : 1421 m_section_list (section_list), 1422 m_section_infos() 1423 { 1424 // Get the number of sections down to a depth of 1 to include 1425 // all segments and their sections, but no other sections that 1426 // may be added for debug map or 1427 m_section_infos.resize(section_list->GetNumSections(1)); 1428 } 1429 1430 1431 SectionSP 1432 GetSection (uint8_t n_sect, addr_t file_addr) 1433 { 1434 if (n_sect == 0) 1435 return SectionSP(); 1436 if (n_sect < m_section_infos.size()) 1437 { 1438 if (!m_section_infos[n_sect].section_sp) 1439 { 1440 SectionSP section_sp (m_section_list->FindSectionByID (n_sect)); 1441 m_section_infos[n_sect].section_sp = section_sp; 1442 if (section_sp) 1443 { 1444 m_section_infos[n_sect].vm_range.SetBaseAddress (section_sp->GetFileAddress()); 1445 m_section_infos[n_sect].vm_range.SetByteSize (section_sp->GetByteSize()); 1446 } 1447 else 1448 { 1449 Host::SystemLog (Host::eSystemLogError, "error: unable to find section for section %u\n", n_sect); 1450 } 1451 } 1452 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) 1453 { 1454 // Symbol is in section. 1455 return m_section_infos[n_sect].section_sp; 1456 } 1457 else if (m_section_infos[n_sect].vm_range.GetByteSize () == 0 && 1458 m_section_infos[n_sect].vm_range.GetBaseAddress() == file_addr) 1459 { 1460 // Symbol is in section with zero size, but has the same start 1461 // address as the section. This can happen with linker symbols 1462 // (symbols that start with the letter 'l' or 'L'. 1463 return m_section_infos[n_sect].section_sp; 1464 } 1465 } 1466 return m_section_list->FindSectionContainingFileAddress(file_addr); 1467 } 1468 1469 protected: 1470 struct SectionInfo 1471 { 1472 SectionInfo () : 1473 vm_range(), 1474 section_sp () 1475 { 1476 } 1477 1478 VMRange vm_range; 1479 SectionSP section_sp; 1480 }; 1481 SectionList *m_section_list; 1482 std::vector<SectionInfo> m_section_infos; 1483 }; 1484 1485 struct TrieEntry 1486 { 1487 TrieEntry () : 1488 name(), 1489 address(LLDB_INVALID_ADDRESS), 1490 flags (0), 1491 other(0), 1492 import_name() 1493 { 1494 } 1495 1496 void 1497 Clear () 1498 { 1499 name.Clear(); 1500 address = LLDB_INVALID_ADDRESS; 1501 flags = 0; 1502 other = 0; 1503 import_name.Clear(); 1504 } 1505 1506 void 1507 Dump () const 1508 { 1509 printf ("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", address, flags, other, name.GetCString()); 1510 if (import_name) 1511 printf (" -> \"%s\"\n", import_name.GetCString()); 1512 else 1513 printf ("\n"); 1514 } 1515 ConstString name; 1516 uint64_t address; 1517 uint64_t flags; 1518 uint64_t other; 1519 ConstString import_name; 1520 }; 1521 1522 struct TrieEntryWithOffset 1523 { 1524 lldb::offset_t nodeOffset; 1525 TrieEntry entry; 1526 1527 TrieEntryWithOffset (lldb::offset_t offset) : 1528 nodeOffset (offset), 1529 entry() 1530 { 1531 } 1532 1533 void 1534 Dump (uint32_t idx) const 1535 { 1536 printf ("[%3u] 0x%16.16llx: ", idx, nodeOffset); 1537 entry.Dump(); 1538 } 1539 1540 bool 1541 operator<(const TrieEntryWithOffset& other) const 1542 { 1543 return ( nodeOffset < other.nodeOffset ); 1544 } 1545 }; 1546 1547 static void 1548 ParseTrieEntries (DataExtractor &data, 1549 lldb::offset_t offset, 1550 std::vector<llvm::StringRef> &nameSlices, 1551 std::set<lldb::addr_t> &resolver_addresses, 1552 std::vector<TrieEntryWithOffset>& output) 1553 { 1554 if (!data.ValidOffset(offset)) 1555 return; 1556 1557 const uint64_t terminalSize = data.GetULEB128(&offset); 1558 lldb::offset_t children_offset = offset + terminalSize; 1559 if ( terminalSize != 0 ) { 1560 TrieEntryWithOffset e (offset); 1561 e.entry.flags = data.GetULEB128(&offset); 1562 const char *import_name = NULL; 1563 if ( e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT ) { 1564 e.entry.address = 0; 1565 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 1566 import_name = data.GetCStr(&offset); 1567 } 1568 else { 1569 e.entry.address = data.GetULEB128(&offset); 1570 if ( e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER ) 1571 { 1572 resolver_addresses.insert(e.entry.address); 1573 e.entry.other = data.GetULEB128(&offset); 1574 } 1575 else 1576 e.entry.other = 0; 1577 } 1578 // Only add symbols that are reexport symbols with a valid import name 1579 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && import_name[0]) 1580 { 1581 std::string name; 1582 if (!nameSlices.empty()) 1583 { 1584 for (auto name_slice: nameSlices) 1585 name.append(name_slice.data(), name_slice.size()); 1586 } 1587 if (name.size() > 1) 1588 { 1589 // Skip the leading '_' 1590 e.entry.name.SetCStringWithLength(name.c_str() + 1,name.size() - 1); 1591 } 1592 if (import_name) 1593 { 1594 // Skip the leading '_' 1595 e.entry.import_name.SetCString(import_name+1); 1596 } 1597 output.push_back(e); 1598 } 1599 } 1600 1601 const uint8_t childrenCount = data.GetU8(&children_offset); 1602 for (uint8_t i=0; i < childrenCount; ++i) { 1603 nameSlices.push_back(data.GetCStr(&children_offset)); 1604 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 1605 if (childNodeOffset) 1606 { 1607 ParseTrieEntries(data, 1608 childNodeOffset, 1609 nameSlices, 1610 resolver_addresses, 1611 output); 1612 } 1613 nameSlices.pop_back(); 1614 } 1615 } 1616 1617 size_t 1618 ObjectFileMachO::ParseSymtab () 1619 { 1620 Timer scoped_timer(__PRETTY_FUNCTION__, 1621 "ObjectFileMachO::ParseSymtab () module = %s", 1622 m_file.GetFilename().AsCString("")); 1623 ModuleSP module_sp (GetModule()); 1624 if (!module_sp) 1625 return 0; 1626 1627 struct symtab_command symtab_load_command = { 0, 0, 0, 0, 0, 0 }; 1628 struct linkedit_data_command function_starts_load_command = { 0, 0, 0, 0 }; 1629 struct dyld_info_command dyld_info = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }; 1630 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 1631 FunctionStarts function_starts; 1632 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1633 uint32_t i; 1634 FileSpecList dylib_files; 1635 Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_SYMBOLS)); 1636 1637 for (i=0; i<m_header.ncmds; ++i) 1638 { 1639 const lldb::offset_t cmd_offset = offset; 1640 // Read in the load command and load command size 1641 struct load_command lc; 1642 if (m_data.GetU32(&offset, &lc, 2) == NULL) 1643 break; 1644 // Watch for the symbol table load command 1645 switch (lc.cmd) 1646 { 1647 case LC_SYMTAB: 1648 symtab_load_command.cmd = lc.cmd; 1649 symtab_load_command.cmdsize = lc.cmdsize; 1650 // Read in the rest of the symtab load command 1651 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 0) // fill in symoff, nsyms, stroff, strsize fields 1652 return 0; 1653 if (symtab_load_command.symoff == 0) 1654 { 1655 if (log) 1656 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 1657 return 0; 1658 } 1659 1660 if (symtab_load_command.stroff == 0) 1661 { 1662 if (log) 1663 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 1664 return 0; 1665 } 1666 1667 if (symtab_load_command.nsyms == 0) 1668 { 1669 if (log) 1670 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 1671 return 0; 1672 } 1673 1674 if (symtab_load_command.strsize == 0) 1675 { 1676 if (log) 1677 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 1678 return 0; 1679 } 1680 break; 1681 1682 case LC_DYLD_INFO: 1683 case LC_DYLD_INFO_ONLY: 1684 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) 1685 { 1686 dyld_info.cmd = lc.cmd; 1687 dyld_info.cmdsize = lc.cmdsize; 1688 } 1689 else 1690 { 1691 memset (&dyld_info, 0, sizeof(dyld_info)); 1692 } 1693 break; 1694 1695 case LC_LOAD_DYLIB: 1696 case LC_LOAD_WEAK_DYLIB: 1697 case LC_REEXPORT_DYLIB: 1698 case LC_LOADFVMLIB: 1699 case LC_LOAD_UPWARD_DYLIB: 1700 { 1701 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 1702 const char *path = m_data.PeekCStr(name_offset); 1703 if (path) 1704 { 1705 FileSpec file_spec(path, false); 1706 // Strip the path if there is @rpath, @executanble, etc so we just use the basename 1707 if (path[0] == '@') 1708 file_spec.GetDirectory().Clear(); 1709 1710 dylib_files.Append(file_spec); 1711 } 1712 } 1713 break; 1714 1715 case LC_FUNCTION_STARTS: 1716 function_starts_load_command.cmd = lc.cmd; 1717 function_starts_load_command.cmdsize = lc.cmdsize; 1718 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == NULL) // fill in symoff, nsyms, stroff, strsize fields 1719 memset (&function_starts_load_command, 0, sizeof(function_starts_load_command)); 1720 break; 1721 1722 default: 1723 break; 1724 } 1725 offset = cmd_offset + lc.cmdsize; 1726 } 1727 1728 if (symtab_load_command.cmd) 1729 { 1730 Symtab *symtab = m_symtab_ap.get(); 1731 SectionList *section_list = GetSectionList(); 1732 if (section_list == NULL) 1733 return 0; 1734 1735 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 1736 const ByteOrder byte_order = m_data.GetByteOrder(); 1737 bool bit_width_32 = addr_byte_size == 4; 1738 const size_t nlist_byte_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 1739 1740 DataExtractor nlist_data (NULL, 0, byte_order, addr_byte_size); 1741 DataExtractor strtab_data (NULL, 0, byte_order, addr_byte_size); 1742 DataExtractor function_starts_data (NULL, 0, byte_order, addr_byte_size); 1743 DataExtractor indirect_symbol_index_data (NULL, 0, byte_order, addr_byte_size); 1744 DataExtractor dyld_trie_data (NULL, 0, byte_order, addr_byte_size); 1745 1746 const addr_t nlist_data_byte_size = symtab_load_command.nsyms * nlist_byte_size; 1747 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 1748 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 1749 1750 ProcessSP process_sp (m_process_wp.lock()); 1751 Process *process = process_sp.get(); 1752 1753 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 1754 1755 if (process) 1756 { 1757 Target &target = process->GetTarget(); 1758 1759 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 1760 1761 SectionSP linkedit_section_sp(section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 1762 // Reading mach file from memory in a process or core file... 1763 1764 if (linkedit_section_sp) 1765 { 1766 const addr_t linkedit_load_addr = linkedit_section_sp->GetLoadBaseAddress(&target); 1767 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 1768 const addr_t symoff_addr = linkedit_load_addr + symtab_load_command.symoff - linkedit_file_offset; 1769 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - linkedit_file_offset; 1770 1771 bool data_was_read = false; 1772 1773 #if defined (__APPLE__) && defined (__arm__) 1774 if (m_header.flags & 0x80000000u) 1775 { 1776 // This mach-o memory file is in the dyld shared cache. If this 1777 // program is not remote and this is iOS, then this process will 1778 // share the same shared cache as the process we are debugging and 1779 // we can read the entire __LINKEDIT from the address space in this 1780 // process. This is a needed optimization that is used for local iOS 1781 // debugging only since all shared libraries in the shared cache do 1782 // not have corresponding files that exist in the file system of the 1783 // device. They have been combined into a single file. This means we 1784 // always have to load these files from memory. All of the symbol and 1785 // string tables from all of the __LINKEDIT sections from the shared 1786 // libraries in the shared cache have been merged into a single large 1787 // symbol and string table. Reading all of this symbol and string table 1788 // data across can slow down debug launch times, so we optimize this by 1789 // reading the memory for the __LINKEDIT section from this process. 1790 1791 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 1792 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 1793 bool use_lldb_cache = true; 1794 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && lldb_shared_cache != process_shared_cache) 1795 { 1796 use_lldb_cache = false; 1797 ModuleSP module_sp (GetModule()); 1798 if (module_sp) 1799 module_sp->ReportWarning ("shared cache in process does not match lldb's own shared cache, startup will be slow."); 1800 1801 } 1802 1803 PlatformSP platform_sp (target.GetPlatform()); 1804 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) 1805 { 1806 data_was_read = true; 1807 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, eByteOrderLittle); 1808 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, eByteOrderLittle); 1809 if (function_starts_load_command.cmd) 1810 { 1811 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 1812 function_starts_data.SetData ((void *)func_start_addr, function_starts_load_command.datasize, eByteOrderLittle); 1813 } 1814 } 1815 } 1816 #endif 1817 1818 if (!data_was_read) 1819 { 1820 if (memory_module_load_level == eMemoryModuleLoadLevelComplete) 1821 { 1822 DataBufferSP nlist_data_sp (ReadMemory (process_sp, symoff_addr, nlist_data_byte_size)); 1823 if (nlist_data_sp) 1824 nlist_data.SetData (nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 1825 // Load strings individually from memory when loading from memory since shared cache 1826 // string tables contain strings for all symbols from all shared cached libraries 1827 //DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, strtab_data_byte_size)); 1828 //if (strtab_data_sp) 1829 // strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 1830 if (m_dysymtab.nindirectsyms != 0) 1831 { 1832 const addr_t indirect_syms_addr = linkedit_load_addr + m_dysymtab.indirectsymoff - linkedit_file_offset; 1833 DataBufferSP indirect_syms_data_sp (ReadMemory (process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 1834 if (indirect_syms_data_sp) 1835 indirect_symbol_index_data.SetData (indirect_syms_data_sp, 0, indirect_syms_data_sp->GetByteSize()); 1836 } 1837 } 1838 1839 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) 1840 { 1841 if (function_starts_load_command.cmd) 1842 { 1843 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 1844 DataBufferSP func_start_data_sp (ReadMemory (process_sp, func_start_addr, function_starts_load_command.datasize)); 1845 if (func_start_data_sp) 1846 function_starts_data.SetData (func_start_data_sp, 0, func_start_data_sp->GetByteSize()); 1847 } 1848 } 1849 } 1850 } 1851 } 1852 else 1853 { 1854 nlist_data.SetData (m_data, 1855 symtab_load_command.symoff, 1856 nlist_data_byte_size); 1857 strtab_data.SetData (m_data, 1858 symtab_load_command.stroff, 1859 strtab_data_byte_size); 1860 1861 if (dyld_info.export_size > 0) 1862 { 1863 dyld_trie_data.SetData (m_data, 1864 dyld_info.export_off, 1865 dyld_info.export_size); 1866 } 1867 1868 if (m_dysymtab.nindirectsyms != 0) 1869 { 1870 indirect_symbol_index_data.SetData (m_data, 1871 m_dysymtab.indirectsymoff, 1872 m_dysymtab.nindirectsyms * 4); 1873 } 1874 if (function_starts_load_command.cmd) 1875 { 1876 function_starts_data.SetData (m_data, 1877 function_starts_load_command.dataoff, 1878 function_starts_load_command.datasize); 1879 } 1880 } 1881 1882 if (nlist_data.GetByteSize() == 0 && memory_module_load_level == eMemoryModuleLoadLevelComplete) 1883 { 1884 if (log) 1885 module_sp->LogMessage(log, "failed to read nlist data"); 1886 return 0; 1887 } 1888 1889 1890 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 1891 if (!have_strtab_data) 1892 { 1893 if (process) 1894 { 1895 if (strtab_addr == LLDB_INVALID_ADDRESS) 1896 { 1897 if (log) 1898 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 1899 return 0; 1900 } 1901 } 1902 else 1903 { 1904 if (log) 1905 module_sp->LogMessage(log, "failed to read strtab data"); 1906 return 0; 1907 } 1908 } 1909 1910 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 1911 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 1912 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 1913 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 1914 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT)); 1915 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA)); 1916 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC)); 1917 SectionSP eh_frame_section_sp; 1918 if (text_section_sp.get()) 1919 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame); 1920 else 1921 eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame); 1922 1923 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 1924 1925 // lldb works best if it knows the start addresss of all functions in a module. 1926 // Linker symbols or debug info are normally the best source of information for start addr / size but 1927 // they may be stripped in a released binary. 1928 // Two additional sources of information exist in Mach-O binaries: 1929 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each function's start address in the 1930 // binary, relative to the text section. 1931 // eh_frame - the eh_frame FDEs have the start addr & size of each function 1932 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on all modern binaries. 1933 // Binaries built to run on older releases may need to use eh_frame information. 1934 1935 if (text_section_sp && function_starts_data.GetByteSize()) 1936 { 1937 FunctionStarts::Entry function_start_entry; 1938 function_start_entry.data = false; 1939 lldb::offset_t function_start_offset = 0; 1940 function_start_entry.addr = text_section_sp->GetFileAddress(); 1941 uint64_t delta; 1942 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 0) 1943 { 1944 // Now append the current entry 1945 function_start_entry.addr += delta; 1946 function_starts.Append(function_start_entry); 1947 } 1948 } 1949 else 1950 { 1951 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the load command claiming an eh_frame 1952 // but it doesn't actually have the eh_frame content. And if we have a dSYM, we don't need to do any 1953 // of this fill-in-the-missing-symbols works anyway - the debug info should give us all the functions in 1954 // the module. 1955 if (text_section_sp.get() && eh_frame_section_sp.get() && m_type != eTypeDebugInfo) 1956 { 1957 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, eRegisterKindGCC, true); 1958 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 1959 eh_frame.GetFunctionAddressAndSizeVector (functions); 1960 addr_t text_base_addr = text_section_sp->GetFileAddress(); 1961 size_t count = functions.GetSize(); 1962 for (size_t i = 0; i < count; ++i) 1963 { 1964 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = functions.GetEntryAtIndex (i); 1965 if (func) 1966 { 1967 FunctionStarts::Entry function_start_entry; 1968 function_start_entry.addr = func->base - text_base_addr; 1969 function_starts.Append(function_start_entry); 1970 } 1971 } 1972 } 1973 } 1974 1975 const size_t function_starts_count = function_starts.GetSize(); 1976 1977 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() : NO_SECT; 1978 1979 lldb::offset_t nlist_data_offset = 0; 1980 1981 uint32_t N_SO_index = UINT32_MAX; 1982 1983 MachSymtabSectionInfo section_info (section_list); 1984 std::vector<uint32_t> N_FUN_indexes; 1985 std::vector<uint32_t> N_NSYM_indexes; 1986 std::vector<uint32_t> N_INCL_indexes; 1987 std::vector<uint32_t> N_BRAC_indexes; 1988 std::vector<uint32_t> N_COMM_indexes; 1989 typedef std::map <uint64_t, uint32_t> ValueToSymbolIndexMap; 1990 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap; 1991 typedef std::map <const char *, uint32_t> ConstNameToSymbolIndexMap; 1992 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 1993 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 1994 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 1995 // Any symbols that get merged into another will get an entry 1996 // in this map so we know 1997 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 1998 uint32_t nlist_idx = 0; 1999 Symbol *symbol_ptr = NULL; 2000 2001 uint32_t sym_idx = 0; 2002 Symbol *sym = NULL; 2003 size_t num_syms = 0; 2004 std::string memory_symbol_name; 2005 uint32_t unmapped_local_symbols_found = 0; 2006 2007 #if defined (__APPLE__) && defined (__arm__) 2008 2009 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been optimized by moving LOCAL 2010 // symbols out of the memory mapped portion of the DSC. The symbol information has all been retained, 2011 // but it isn't available in the normal nlist data. However, there *are* duplicate entries of *some* 2012 // LOCAL symbols in the normal nlist data. To handle this situation correctly, we must first attempt 2013 // to parse any DSC unmapped symbol information. If we find any, we set a flag that tells the normal 2014 // nlist parser to ignore all LOCAL symbols. 2015 2016 if (m_header.flags & 0x80000000u) 2017 { 2018 // Before we can start mapping the DSC, we need to make certain the target process is actually 2019 // using the cache we can find. 2020 2021 // Next we need to determine the correct path for the dyld shared cache. 2022 2023 ArchSpec header_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 2024 char dsc_path[PATH_MAX]; 2025 2026 snprintf(dsc_path, sizeof(dsc_path), "%s%s%s", 2027 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR */ 2028 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2029 header_arch.GetArchitectureName()); 2030 2031 FileSpec dsc_filespec(dsc_path, false); 2032 2033 // We need definitions of two structures in the on-disk DSC, copy them here manually 2034 struct lldb_copy_dyld_cache_header_v0 2035 { 2036 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 2037 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 2038 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 2039 uint32_t imagesOffset; 2040 uint32_t imagesCount; 2041 uint64_t dyldBaseAddress; 2042 uint64_t codeSignatureOffset; 2043 uint64_t codeSignatureSize; 2044 uint64_t slideInfoOffset; 2045 uint64_t slideInfoSize; 2046 uint64_t localSymbolsOffset; // file offset of where local symbols are stored 2047 uint64_t localSymbolsSize; // size of local symbols information 2048 }; 2049 struct lldb_copy_dyld_cache_header_v1 2050 { 2051 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 2052 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 2053 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 2054 uint32_t imagesOffset; 2055 uint32_t imagesCount; 2056 uint64_t dyldBaseAddress; 2057 uint64_t codeSignatureOffset; 2058 uint64_t codeSignatureSize; 2059 uint64_t slideInfoOffset; 2060 uint64_t slideInfoSize; 2061 uint64_t localSymbolsOffset; 2062 uint64_t localSymbolsSize; 2063 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 and later 2064 }; 2065 2066 struct lldb_copy_dyld_cache_mapping_info 2067 { 2068 uint64_t address; 2069 uint64_t size; 2070 uint64_t fileOffset; 2071 uint32_t maxProt; 2072 uint32_t initProt; 2073 }; 2074 2075 struct lldb_copy_dyld_cache_local_symbols_info 2076 { 2077 uint32_t nlistOffset; 2078 uint32_t nlistCount; 2079 uint32_t stringsOffset; 2080 uint32_t stringsSize; 2081 uint32_t entriesOffset; 2082 uint32_t entriesCount; 2083 }; 2084 struct lldb_copy_dyld_cache_local_symbols_entry 2085 { 2086 uint32_t dylibOffset; 2087 uint32_t nlistStartIndex; 2088 uint32_t nlistCount; 2089 }; 2090 2091 /* The dyld_cache_header has a pointer to the dyld_cache_local_symbols_info structure (localSymbolsOffset). 2092 The dyld_cache_local_symbols_info structure gives us three things: 2093 1. The start and count of the nlist records in the dyld_shared_cache file 2094 2. The start and size of the strings for these nlist records 2095 3. The start and count of dyld_cache_local_symbols_entry entries 2096 2097 There is one dyld_cache_local_symbols_entry per dylib/framework in the dyld shared cache. 2098 The "dylibOffset" field is the Mach-O header of this dylib/framework in the dyld shared cache. 2099 The dyld_cache_local_symbols_entry also lists the start of this dylib/framework's nlist records 2100 and the count of how many nlist records there are for this dylib/framework. 2101 */ 2102 2103 // Process the dsc header to find the unmapped symbols 2104 // 2105 // Save some VM space, do not map the entire cache in one shot. 2106 2107 DataBufferSP dsc_data_sp; 2108 dsc_data_sp = dsc_filespec.MemoryMapFileContents(0, sizeof(struct lldb_copy_dyld_cache_header_v1)); 2109 2110 if (dsc_data_sp) 2111 { 2112 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2113 2114 char version_str[17]; 2115 int version = -1; 2116 lldb::offset_t offset = 0; 2117 memcpy (version_str, dsc_header_data.GetData (&offset, 16), 16); 2118 version_str[16] = '\0'; 2119 if (strncmp (version_str, "dyld_v", 6) == 0 && isdigit (version_str[6])) 2120 { 2121 int v; 2122 if (::sscanf (version_str + 6, "%d", &v) == 1) 2123 { 2124 version = v; 2125 } 2126 } 2127 2128 UUID dsc_uuid; 2129 if (version >= 1) 2130 { 2131 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, uuid); 2132 uint8_t uuid_bytes[sizeof (uuid_t)]; 2133 memcpy (uuid_bytes, dsc_header_data.GetData (&offset, sizeof (uuid_t)), sizeof (uuid_t)); 2134 dsc_uuid.SetBytes (uuid_bytes); 2135 } 2136 2137 bool uuid_match = true; 2138 if (dsc_uuid.IsValid() && process) 2139 { 2140 UUID shared_cache_uuid(GetProcessSharedCacheUUID(process)); 2141 2142 if (shared_cache_uuid.IsValid() && dsc_uuid != shared_cache_uuid) 2143 { 2144 // The on-disk dyld_shared_cache file is not the same as the one in this 2145 // process' memory, don't use it. 2146 uuid_match = false; 2147 ModuleSP module_sp (GetModule()); 2148 if (module_sp) 2149 module_sp->ReportWarning ("process shared cache does not match on-disk dyld_shared_cache file, some symbol names will be missing."); 2150 } 2151 } 2152 2153 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2154 2155 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2156 2157 // If the mappingOffset points to a location inside the header, we've 2158 // opened an old dyld shared cache, and should not proceed further. 2159 if (uuid_match && mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v0)) 2160 { 2161 2162 DataBufferSP dsc_mapping_info_data_sp = dsc_filespec.MemoryMapFileContents(mappingOffset, sizeof (struct lldb_copy_dyld_cache_mapping_info)); 2163 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, byte_order, addr_byte_size); 2164 offset = 0; 2165 2166 // The File addresses (from the in-memory Mach-O load commands) for the shared libraries 2167 // in the shared library cache need to be adjusted by an offset to match up with the 2168 // dylibOffset identifying field in the dyld_cache_local_symbol_entry's. This offset is 2169 // recorded in mapping_offset_value. 2170 const uint64_t mapping_offset_value = dsc_mapping_info_data.GetU64(&offset); 2171 2172 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2173 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2174 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2175 2176 if (localSymbolsOffset && localSymbolsSize) 2177 { 2178 // Map the local symbols 2179 if (DataBufferSP dsc_local_symbols_data_sp = dsc_filespec.MemoryMapFileContents(localSymbolsOffset, localSymbolsSize)) 2180 { 2181 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, byte_order, addr_byte_size); 2182 2183 offset = 0; 2184 2185 // Read the local_symbols_infos struct in one shot 2186 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2187 dsc_local_symbols_data.GetU32(&offset, &local_symbols_info.nlistOffset, 6); 2188 2189 SectionSP text_section_sp(section_list->FindSectionByName(GetSegmentNameTEXT())); 2190 2191 uint32_t header_file_offset = (text_section_sp->GetFileAddress() - mapping_offset_value); 2192 2193 offset = local_symbols_info.entriesOffset; 2194 for (uint32_t entry_index = 0; entry_index < local_symbols_info.entriesCount; entry_index++) 2195 { 2196 struct lldb_copy_dyld_cache_local_symbols_entry local_symbols_entry; 2197 local_symbols_entry.dylibOffset = dsc_local_symbols_data.GetU32(&offset); 2198 local_symbols_entry.nlistStartIndex = dsc_local_symbols_data.GetU32(&offset); 2199 local_symbols_entry.nlistCount = dsc_local_symbols_data.GetU32(&offset); 2200 2201 if (header_file_offset == local_symbols_entry.dylibOffset) 2202 { 2203 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2204 2205 // The normal nlist code cannot correctly size the Symbols array, we need to allocate it here. 2206 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms + unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2207 num_syms = symtab->GetNumSymbols(); 2208 2209 nlist_data_offset = local_symbols_info.nlistOffset + (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2210 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2211 2212 for (uint32_t nlist_index = 0; nlist_index < local_symbols_entry.nlistCount; nlist_index++) 2213 { 2214 ///////////////////////////// 2215 { 2216 struct nlist_64 nlist; 2217 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2218 break; 2219 2220 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked(&nlist_data_offset); 2221 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2222 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2223 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked (&nlist_data_offset); 2224 nlist.n_value = dsc_local_symbols_data.GetAddress_unchecked (&nlist_data_offset); 2225 2226 SymbolType type = eSymbolTypeInvalid; 2227 const char *symbol_name = dsc_local_symbols_data.PeekCStr(string_table_offset + nlist.n_strx); 2228 2229 if (symbol_name == NULL) 2230 { 2231 // No symbol should be NULL, even the symbols with no 2232 // string values should have an offset zero which points 2233 // to an empty C-string 2234 Host::SystemLog (Host::eSystemLogError, 2235 "error: DSC unmapped local symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 2236 entry_index, 2237 nlist.n_strx, 2238 module_sp->GetFileSpec().GetPath().c_str()); 2239 continue; 2240 } 2241 if (symbol_name[0] == '\0') 2242 symbol_name = NULL; 2243 2244 const char *symbol_name_non_abi_mangled = NULL; 2245 2246 SectionSP symbol_section; 2247 uint32_t symbol_byte_size = 0; 2248 bool add_nlist = true; 2249 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2250 bool demangled_is_synthesized = false; 2251 bool is_gsym = false; 2252 2253 assert (sym_idx < num_syms); 2254 2255 sym[sym_idx].SetDebug (is_debug); 2256 2257 if (is_debug) 2258 { 2259 switch (nlist.n_type) 2260 { 2261 case N_GSYM: 2262 // global symbol: name,,NO_SECT,type,0 2263 // Sometimes the N_GSYM value contains the address. 2264 2265 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 2266 // have the same address, but we want to ensure that we always find only the real symbol, 2267 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 2268 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 2269 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 2270 // same address. 2271 2272 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 2273 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 2274 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 2275 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 2276 add_nlist = false; 2277 else 2278 { 2279 is_gsym = true; 2280 sym[sym_idx].SetExternal(true); 2281 if (nlist.n_value != 0) 2282 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2283 type = eSymbolTypeData; 2284 } 2285 break; 2286 2287 case N_FNAME: 2288 // procedure name (f77 kludge): name,,NO_SECT,0,0 2289 type = eSymbolTypeCompiler; 2290 break; 2291 2292 case N_FUN: 2293 // procedure: name,,n_sect,linenumber,address 2294 if (symbol_name) 2295 { 2296 type = eSymbolTypeCode; 2297 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2298 2299 N_FUN_addr_to_sym_idx[nlist.n_value] = sym_idx; 2300 // We use the current number of symbols in the symbol table in lieu of 2301 // using nlist_idx in case we ever start trimming entries out 2302 N_FUN_indexes.push_back(sym_idx); 2303 } 2304 else 2305 { 2306 type = eSymbolTypeCompiler; 2307 2308 if ( !N_FUN_indexes.empty() ) 2309 { 2310 // Copy the size of the function into the original STAB entry so we don't have 2311 // to hunt for it later 2312 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 2313 N_FUN_indexes.pop_back(); 2314 // We don't really need the end function STAB as it contains the size which 2315 // we already placed with the original symbol, so don't add it if we want a 2316 // minimal symbol table 2317 add_nlist = false; 2318 } 2319 } 2320 break; 2321 2322 case N_STSYM: 2323 // static symbol: name,,n_sect,type,address 2324 N_STSYM_addr_to_sym_idx[nlist.n_value] = sym_idx; 2325 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2326 type = eSymbolTypeData; 2327 break; 2328 2329 case N_LCSYM: 2330 // .lcomm symbol: name,,n_sect,type,address 2331 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2332 type = eSymbolTypeCommonBlock; 2333 break; 2334 2335 case N_BNSYM: 2336 // We use the current number of symbols in the symbol table in lieu of 2337 // using nlist_idx in case we ever start trimming entries out 2338 // Skip these if we want minimal symbol tables 2339 add_nlist = false; 2340 break; 2341 2342 case N_ENSYM: 2343 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 2344 // so that we can always skip the entire symbol if we need to navigate 2345 // more quickly at the source level when parsing STABS 2346 // Skip these if we want minimal symbol tables 2347 add_nlist = false; 2348 break; 2349 2350 2351 case N_OPT: 2352 // emitted with gcc2_compiled and in gcc source 2353 type = eSymbolTypeCompiler; 2354 break; 2355 2356 case N_RSYM: 2357 // register sym: name,,NO_SECT,type,register 2358 type = eSymbolTypeVariable; 2359 break; 2360 2361 case N_SLINE: 2362 // src line: 0,,n_sect,linenumber,address 2363 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2364 type = eSymbolTypeLineEntry; 2365 break; 2366 2367 case N_SSYM: 2368 // structure elt: name,,NO_SECT,type,struct_offset 2369 type = eSymbolTypeVariableType; 2370 break; 2371 2372 case N_SO: 2373 // source file name 2374 type = eSymbolTypeSourceFile; 2375 if (symbol_name == NULL) 2376 { 2377 add_nlist = false; 2378 if (N_SO_index != UINT32_MAX) 2379 { 2380 // Set the size of the N_SO to the terminating index of this N_SO 2381 // so that we can always skip the entire N_SO if we need to navigate 2382 // more quickly at the source level when parsing STABS 2383 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2384 symbol_ptr->SetByteSize(sym_idx); 2385 symbol_ptr->SetSizeIsSibling(true); 2386 } 2387 N_NSYM_indexes.clear(); 2388 N_INCL_indexes.clear(); 2389 N_BRAC_indexes.clear(); 2390 N_COMM_indexes.clear(); 2391 N_FUN_indexes.clear(); 2392 N_SO_index = UINT32_MAX; 2393 } 2394 else 2395 { 2396 // We use the current number of symbols in the symbol table in lieu of 2397 // using nlist_idx in case we ever start trimming entries out 2398 const bool N_SO_has_full_path = symbol_name[0] == '/'; 2399 if (N_SO_has_full_path) 2400 { 2401 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2402 { 2403 // We have two consecutive N_SO entries where the first contains a directory 2404 // and the second contains a full path. 2405 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 2406 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2407 add_nlist = false; 2408 } 2409 else 2410 { 2411 // This is the first entry in a N_SO that contains a directory or 2412 // a full path to the source file 2413 N_SO_index = sym_idx; 2414 } 2415 } 2416 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2417 { 2418 // This is usually the second N_SO entry that contains just the filename, 2419 // so here we combine it with the first one if we are minimizing the symbol table 2420 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 2421 if (so_path && so_path[0]) 2422 { 2423 std::string full_so_path (so_path); 2424 const size_t double_slash_pos = full_so_path.find("//"); 2425 if (double_slash_pos != std::string::npos) 2426 { 2427 // The linker has been generating bad N_SO entries with doubled up paths 2428 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 2429 // and the second is the directory for the source file so you end up with 2430 // a path that looks like "/tmp/src//tmp/src/" 2431 FileSpec so_dir(so_path, false); 2432 if (!so_dir.Exists()) 2433 { 2434 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 2435 if (so_dir.Exists()) 2436 { 2437 // Trim off the incorrect path 2438 full_so_path.erase(0, double_slash_pos + 1); 2439 } 2440 } 2441 } 2442 if (*full_so_path.rbegin() != '/') 2443 full_so_path += '/'; 2444 full_so_path += symbol_name; 2445 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 2446 add_nlist = false; 2447 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2448 } 2449 } 2450 else 2451 { 2452 // This could be a relative path to a N_SO 2453 N_SO_index = sym_idx; 2454 } 2455 } 2456 break; 2457 2458 case N_OSO: 2459 // object file name: name,,0,0,st_mtime 2460 type = eSymbolTypeObjectFile; 2461 break; 2462 2463 case N_LSYM: 2464 // local sym: name,,NO_SECT,type,offset 2465 type = eSymbolTypeLocal; 2466 break; 2467 2468 //---------------------------------------------------------------------- 2469 // INCL scopes 2470 //---------------------------------------------------------------------- 2471 case N_BINCL: 2472 // include file beginning: name,,NO_SECT,0,sum 2473 // We use the current number of symbols in the symbol table in lieu of 2474 // using nlist_idx in case we ever start trimming entries out 2475 N_INCL_indexes.push_back(sym_idx); 2476 type = eSymbolTypeScopeBegin; 2477 break; 2478 2479 case N_EINCL: 2480 // include file end: name,,NO_SECT,0,0 2481 // Set the size of the N_BINCL to the terminating index of this N_EINCL 2482 // so that we can always skip the entire symbol if we need to navigate 2483 // more quickly at the source level when parsing STABS 2484 if ( !N_INCL_indexes.empty() ) 2485 { 2486 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 2487 symbol_ptr->SetByteSize(sym_idx + 1); 2488 symbol_ptr->SetSizeIsSibling(true); 2489 N_INCL_indexes.pop_back(); 2490 } 2491 type = eSymbolTypeScopeEnd; 2492 break; 2493 2494 case N_SOL: 2495 // #included file name: name,,n_sect,0,address 2496 type = eSymbolTypeHeaderFile; 2497 2498 // We currently don't use the header files on darwin 2499 add_nlist = false; 2500 break; 2501 2502 case N_PARAMS: 2503 // compiler parameters: name,,NO_SECT,0,0 2504 type = eSymbolTypeCompiler; 2505 break; 2506 2507 case N_VERSION: 2508 // compiler version: name,,NO_SECT,0,0 2509 type = eSymbolTypeCompiler; 2510 break; 2511 2512 case N_OLEVEL: 2513 // compiler -O level: name,,NO_SECT,0,0 2514 type = eSymbolTypeCompiler; 2515 break; 2516 2517 case N_PSYM: 2518 // parameter: name,,NO_SECT,type,offset 2519 type = eSymbolTypeVariable; 2520 break; 2521 2522 case N_ENTRY: 2523 // alternate entry: name,,n_sect,linenumber,address 2524 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2525 type = eSymbolTypeLineEntry; 2526 break; 2527 2528 //---------------------------------------------------------------------- 2529 // Left and Right Braces 2530 //---------------------------------------------------------------------- 2531 case N_LBRAC: 2532 // left bracket: 0,,NO_SECT,nesting level,address 2533 // We use the current number of symbols in the symbol table in lieu of 2534 // using nlist_idx in case we ever start trimming entries out 2535 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2536 N_BRAC_indexes.push_back(sym_idx); 2537 type = eSymbolTypeScopeBegin; 2538 break; 2539 2540 case N_RBRAC: 2541 // right bracket: 0,,NO_SECT,nesting level,address 2542 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 2543 // so that we can always skip the entire symbol if we need to navigate 2544 // more quickly at the source level when parsing STABS 2545 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2546 if ( !N_BRAC_indexes.empty() ) 2547 { 2548 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 2549 symbol_ptr->SetByteSize(sym_idx + 1); 2550 symbol_ptr->SetSizeIsSibling(true); 2551 N_BRAC_indexes.pop_back(); 2552 } 2553 type = eSymbolTypeScopeEnd; 2554 break; 2555 2556 case N_EXCL: 2557 // deleted include file: name,,NO_SECT,0,sum 2558 type = eSymbolTypeHeaderFile; 2559 break; 2560 2561 //---------------------------------------------------------------------- 2562 // COMM scopes 2563 //---------------------------------------------------------------------- 2564 case N_BCOMM: 2565 // begin common: name,,NO_SECT,0,0 2566 // We use the current number of symbols in the symbol table in lieu of 2567 // using nlist_idx in case we ever start trimming entries out 2568 type = eSymbolTypeScopeBegin; 2569 N_COMM_indexes.push_back(sym_idx); 2570 break; 2571 2572 case N_ECOML: 2573 // end common (local name): 0,,n_sect,0,address 2574 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2575 // Fall through 2576 2577 case N_ECOMM: 2578 // end common: name,,n_sect,0,0 2579 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 2580 // so that we can always skip the entire symbol if we need to navigate 2581 // more quickly at the source level when parsing STABS 2582 if ( !N_COMM_indexes.empty() ) 2583 { 2584 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 2585 symbol_ptr->SetByteSize(sym_idx + 1); 2586 symbol_ptr->SetSizeIsSibling(true); 2587 N_COMM_indexes.pop_back(); 2588 } 2589 type = eSymbolTypeScopeEnd; 2590 break; 2591 2592 case N_LENG: 2593 // second stab entry with length information 2594 type = eSymbolTypeAdditional; 2595 break; 2596 2597 default: break; 2598 } 2599 } 2600 else 2601 { 2602 //uint8_t n_pext = N_PEXT & nlist.n_type; 2603 uint8_t n_type = N_TYPE & nlist.n_type; 2604 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 2605 2606 switch (n_type) 2607 { 2608 case N_INDR: // Fall through 2609 case N_PBUD: // Fall through 2610 case N_UNDF: 2611 type = eSymbolTypeUndefined; 2612 break; 2613 2614 case N_ABS: 2615 type = eSymbolTypeAbsolute; 2616 break; 2617 2618 case N_SECT: 2619 { 2620 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2621 2622 if (symbol_section == NULL) 2623 { 2624 // TODO: warn about this? 2625 add_nlist = false; 2626 break; 2627 } 2628 2629 if (TEXT_eh_frame_sectID == nlist.n_sect) 2630 { 2631 type = eSymbolTypeException; 2632 } 2633 else 2634 { 2635 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 2636 2637 switch (section_type) 2638 { 2639 case S_REGULAR: break; // regular section 2640 //case S_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section 2641 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 2642 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 2643 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 2644 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 2645 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 2646 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 2647 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 2648 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 2649 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 2650 //case S_COALESCED: type = eSymbolType; break; // section contains symbols that are to be coalesced 2651 //case S_GB_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 2652 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 2653 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 2654 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 2655 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 2656 default: break; 2657 } 2658 2659 if (type == eSymbolTypeInvalid) 2660 { 2661 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 2662 if (symbol_section->IsDescendant (text_section_sp.get())) 2663 { 2664 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 2665 S_ATTR_SELF_MODIFYING_CODE | 2666 S_ATTR_SOME_INSTRUCTIONS)) 2667 type = eSymbolTypeData; 2668 else 2669 type = eSymbolTypeCode; 2670 } 2671 else if (symbol_section->IsDescendant(data_section_sp.get())) 2672 { 2673 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 2674 { 2675 type = eSymbolTypeRuntime; 2676 2677 if (symbol_name && 2678 symbol_name[0] == '_' && 2679 symbol_name[1] == 'O' && 2680 symbol_name[2] == 'B') 2681 { 2682 llvm::StringRef symbol_name_ref(symbol_name); 2683 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 2684 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 2685 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 2686 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 2687 { 2688 symbol_name_non_abi_mangled = symbol_name + 1; 2689 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2690 type = eSymbolTypeObjCClass; 2691 demangled_is_synthesized = true; 2692 } 2693 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 2694 { 2695 symbol_name_non_abi_mangled = symbol_name + 1; 2696 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2697 type = eSymbolTypeObjCMetaClass; 2698 demangled_is_synthesized = true; 2699 } 2700 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 2701 { 2702 symbol_name_non_abi_mangled = symbol_name + 1; 2703 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2704 type = eSymbolTypeObjCIVar; 2705 demangled_is_synthesized = true; 2706 } 2707 } 2708 } 2709 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 2710 { 2711 type = eSymbolTypeException; 2712 } 2713 else 2714 { 2715 type = eSymbolTypeData; 2716 } 2717 } 2718 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 2719 { 2720 type = eSymbolTypeTrampoline; 2721 } 2722 else if (symbol_section->IsDescendant(objc_section_sp.get())) 2723 { 2724 type = eSymbolTypeRuntime; 2725 if (symbol_name && symbol_name[0] == '.') 2726 { 2727 llvm::StringRef symbol_name_ref(symbol_name); 2728 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 2729 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 2730 { 2731 symbol_name_non_abi_mangled = symbol_name; 2732 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 2733 type = eSymbolTypeObjCClass; 2734 demangled_is_synthesized = true; 2735 } 2736 } 2737 } 2738 } 2739 } 2740 } 2741 break; 2742 } 2743 } 2744 2745 if (add_nlist) 2746 { 2747 uint64_t symbol_value = nlist.n_value; 2748 if (symbol_name_non_abi_mangled) 2749 { 2750 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 2751 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 2752 } 2753 else 2754 { 2755 bool symbol_name_is_mangled = false; 2756 2757 if (symbol_name && symbol_name[0] == '_') 2758 { 2759 symbol_name_is_mangled = symbol_name[1] == '_'; 2760 symbol_name++; // Skip the leading underscore 2761 } 2762 2763 if (symbol_name) 2764 { 2765 ConstString const_symbol_name(symbol_name); 2766 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 2767 if (is_gsym && is_debug) 2768 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 2769 } 2770 } 2771 if (symbol_section) 2772 { 2773 const addr_t section_file_addr = symbol_section->GetFileAddress(); 2774 if (symbol_byte_size == 0 && function_starts_count > 0) 2775 { 2776 addr_t symbol_lookup_file_addr = nlist.n_value; 2777 // Do an exact address match for non-ARM addresses, else get the closest since 2778 // the symbol might be a thumb symbol which has an address with bit zero set 2779 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 2780 if (is_arm && func_start_entry) 2781 { 2782 // Verify that the function start address is the symbol address (ARM) 2783 // or the symbol address + 1 (thumb) 2784 if (func_start_entry->addr != symbol_lookup_file_addr && 2785 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 2786 { 2787 // Not the right entry, NULL it out... 2788 func_start_entry = NULL; 2789 } 2790 } 2791 if (func_start_entry) 2792 { 2793 func_start_entry->data = true; 2794 2795 addr_t symbol_file_addr = func_start_entry->addr; 2796 uint32_t symbol_flags = 0; 2797 if (is_arm) 2798 { 2799 if (symbol_file_addr & 1) 2800 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 2801 symbol_file_addr &= 0xfffffffffffffffeull; 2802 } 2803 2804 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 2805 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 2806 if (next_func_start_entry) 2807 { 2808 addr_t next_symbol_file_addr = next_func_start_entry->addr; 2809 // Be sure the clear the Thumb address bit when we calculate the size 2810 // from the current and next address 2811 if (is_arm) 2812 next_symbol_file_addr &= 0xfffffffffffffffeull; 2813 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 2814 } 2815 else 2816 { 2817 symbol_byte_size = section_end_file_addr - symbol_file_addr; 2818 } 2819 } 2820 } 2821 symbol_value -= section_file_addr; 2822 } 2823 2824 if (is_debug == false) 2825 { 2826 if (type == eSymbolTypeCode) 2827 { 2828 // See if we can find a N_FUN entry for any code symbols. 2829 // If we do find a match, and the name matches, then we 2830 // can merge the two into just the function symbol to avoid 2831 // duplicate entries in the symbol table 2832 ValueToSymbolIndexMap::const_iterator pos = N_FUN_addr_to_sym_idx.find (nlist.n_value); 2833 if (pos != N_FUN_addr_to_sym_idx.end()) 2834 { 2835 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 2836 { 2837 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 2838 // We just need the flags from the linker symbol, so put these flags 2839 // into the N_FUN flags to avoid duplicate symbols in the symbol table 2840 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 2841 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2842 sym[sym_idx].Clear(); 2843 continue; 2844 } 2845 } 2846 } 2847 else if (type == eSymbolTypeData) 2848 { 2849 // See if we can find a N_STSYM entry for any data symbols. 2850 // If we do find a match, and the name matches, then we 2851 // can merge the two into just the Static symbol to avoid 2852 // duplicate entries in the symbol table 2853 ValueToSymbolIndexMap::const_iterator pos = N_STSYM_addr_to_sym_idx.find (nlist.n_value); 2854 if (pos != N_STSYM_addr_to_sym_idx.end()) 2855 { 2856 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 2857 { 2858 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 2859 // We just need the flags from the linker symbol, so put these flags 2860 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 2861 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 2862 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2863 sym[sym_idx].Clear(); 2864 continue; 2865 } 2866 } 2867 else 2868 { 2869 // Combine N_GSYM stab entries with the non stab symbol 2870 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 2871 if (pos != N_GSYM_name_to_sym_idx.end()) 2872 { 2873 const uint32_t GSYM_sym_idx = pos->second; 2874 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 2875 // Copy the address, because often the N_GSYM address has an invalid address of zero 2876 // when the global is a common symbol 2877 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 2878 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 2879 // We just need the flags from the linker symbol, so put these flags 2880 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 2881 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2882 sym[sym_idx].Clear(); 2883 continue; 2884 } 2885 } 2886 } 2887 } 2888 2889 sym[sym_idx].SetID (nlist_idx); 2890 sym[sym_idx].SetType (type); 2891 sym[sym_idx].GetAddress().SetSection (symbol_section); 2892 sym[sym_idx].GetAddress().SetOffset (symbol_value); 2893 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2894 2895 if (symbol_byte_size > 0) 2896 sym[sym_idx].SetByteSize(symbol_byte_size); 2897 2898 if (demangled_is_synthesized) 2899 sym[sym_idx].SetDemangledNameIsSynthesized(true); 2900 ++sym_idx; 2901 } 2902 else 2903 { 2904 sym[sym_idx].Clear(); 2905 } 2906 2907 } 2908 ///////////////////////////// 2909 } 2910 break; // No more entries to consider 2911 } 2912 } 2913 } 2914 } 2915 } 2916 } 2917 } 2918 2919 // Must reset this in case it was mutated above! 2920 nlist_data_offset = 0; 2921 #endif 2922 2923 if (nlist_data.GetByteSize() > 0) 2924 { 2925 2926 // If the sym array was not created while parsing the DSC unmapped 2927 // symbols, create it now. 2928 if (sym == NULL) 2929 { 2930 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 2931 num_syms = symtab->GetNumSymbols(); 2932 } 2933 2934 if (unmapped_local_symbols_found) 2935 { 2936 assert(m_dysymtab.ilocalsym == 0); 2937 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 2938 nlist_idx = m_dysymtab.nlocalsym; 2939 } 2940 else 2941 { 2942 nlist_idx = 0; 2943 } 2944 2945 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) 2946 { 2947 struct nlist_64 nlist; 2948 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2949 break; 2950 2951 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 2952 nlist.n_type = nlist_data.GetU8_unchecked (&nlist_data_offset); 2953 nlist.n_sect = nlist_data.GetU8_unchecked (&nlist_data_offset); 2954 nlist.n_desc = nlist_data.GetU16_unchecked (&nlist_data_offset); 2955 nlist.n_value = nlist_data.GetAddress_unchecked (&nlist_data_offset); 2956 2957 SymbolType type = eSymbolTypeInvalid; 2958 const char *symbol_name = NULL; 2959 2960 if (have_strtab_data) 2961 { 2962 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 2963 2964 if (symbol_name == NULL) 2965 { 2966 // No symbol should be NULL, even the symbols with no 2967 // string values should have an offset zero which points 2968 // to an empty C-string 2969 Host::SystemLog (Host::eSystemLogError, 2970 "error: symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 2971 nlist_idx, 2972 nlist.n_strx, 2973 module_sp->GetFileSpec().GetPath().c_str()); 2974 continue; 2975 } 2976 if (symbol_name[0] == '\0') 2977 symbol_name = NULL; 2978 } 2979 else 2980 { 2981 const addr_t str_addr = strtab_addr + nlist.n_strx; 2982 Error str_error; 2983 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, str_error)) 2984 symbol_name = memory_symbol_name.c_str(); 2985 } 2986 const char *symbol_name_non_abi_mangled = NULL; 2987 2988 SectionSP symbol_section; 2989 lldb::addr_t symbol_byte_size = 0; 2990 bool add_nlist = true; 2991 bool is_gsym = false; 2992 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2993 bool demangled_is_synthesized = false; 2994 2995 assert (sym_idx < num_syms); 2996 2997 sym[sym_idx].SetDebug (is_debug); 2998 2999 if (is_debug) 3000 { 3001 switch (nlist.n_type) 3002 { 3003 case N_GSYM: 3004 // global symbol: name,,NO_SECT,type,0 3005 // Sometimes the N_GSYM value contains the address. 3006 3007 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 3008 // have the same address, but we want to ensure that we always find only the real symbol, 3009 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 3010 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 3011 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 3012 // same address. 3013 3014 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 3015 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 3016 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 3017 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 3018 add_nlist = false; 3019 else 3020 { 3021 is_gsym = true; 3022 sym[sym_idx].SetExternal(true); 3023 if (nlist.n_value != 0) 3024 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3025 type = eSymbolTypeData; 3026 } 3027 break; 3028 3029 case N_FNAME: 3030 // procedure name (f77 kludge): name,,NO_SECT,0,0 3031 type = eSymbolTypeCompiler; 3032 break; 3033 3034 case N_FUN: 3035 // procedure: name,,n_sect,linenumber,address 3036 if (symbol_name) 3037 { 3038 type = eSymbolTypeCode; 3039 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3040 3041 N_FUN_addr_to_sym_idx[nlist.n_value] = sym_idx; 3042 // We use the current number of symbols in the symbol table in lieu of 3043 // using nlist_idx in case we ever start trimming entries out 3044 N_FUN_indexes.push_back(sym_idx); 3045 } 3046 else 3047 { 3048 type = eSymbolTypeCompiler; 3049 3050 if ( !N_FUN_indexes.empty() ) 3051 { 3052 // Copy the size of the function into the original STAB entry so we don't have 3053 // to hunt for it later 3054 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 3055 N_FUN_indexes.pop_back(); 3056 // We don't really need the end function STAB as it contains the size which 3057 // we already placed with the original symbol, so don't add it if we want a 3058 // minimal symbol table 3059 add_nlist = false; 3060 } 3061 } 3062 break; 3063 3064 case N_STSYM: 3065 // static symbol: name,,n_sect,type,address 3066 N_STSYM_addr_to_sym_idx[nlist.n_value] = sym_idx; 3067 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3068 type = eSymbolTypeData; 3069 break; 3070 3071 case N_LCSYM: 3072 // .lcomm symbol: name,,n_sect,type,address 3073 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3074 type = eSymbolTypeCommonBlock; 3075 break; 3076 3077 case N_BNSYM: 3078 // We use the current number of symbols in the symbol table in lieu of 3079 // using nlist_idx in case we ever start trimming entries out 3080 // Skip these if we want minimal symbol tables 3081 add_nlist = false; 3082 break; 3083 3084 case N_ENSYM: 3085 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 3086 // so that we can always skip the entire symbol if we need to navigate 3087 // more quickly at the source level when parsing STABS 3088 // Skip these if we want minimal symbol tables 3089 add_nlist = false; 3090 break; 3091 3092 3093 case N_OPT: 3094 // emitted with gcc2_compiled and in gcc source 3095 type = eSymbolTypeCompiler; 3096 break; 3097 3098 case N_RSYM: 3099 // register sym: name,,NO_SECT,type,register 3100 type = eSymbolTypeVariable; 3101 break; 3102 3103 case N_SLINE: 3104 // src line: 0,,n_sect,linenumber,address 3105 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3106 type = eSymbolTypeLineEntry; 3107 break; 3108 3109 case N_SSYM: 3110 // structure elt: name,,NO_SECT,type,struct_offset 3111 type = eSymbolTypeVariableType; 3112 break; 3113 3114 case N_SO: 3115 // source file name 3116 type = eSymbolTypeSourceFile; 3117 if (symbol_name == NULL) 3118 { 3119 add_nlist = false; 3120 if (N_SO_index != UINT32_MAX) 3121 { 3122 // Set the size of the N_SO to the terminating index of this N_SO 3123 // so that we can always skip the entire N_SO if we need to navigate 3124 // more quickly at the source level when parsing STABS 3125 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3126 symbol_ptr->SetByteSize(sym_idx); 3127 symbol_ptr->SetSizeIsSibling(true); 3128 } 3129 N_NSYM_indexes.clear(); 3130 N_INCL_indexes.clear(); 3131 N_BRAC_indexes.clear(); 3132 N_COMM_indexes.clear(); 3133 N_FUN_indexes.clear(); 3134 N_SO_index = UINT32_MAX; 3135 } 3136 else 3137 { 3138 // We use the current number of symbols in the symbol table in lieu of 3139 // using nlist_idx in case we ever start trimming entries out 3140 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3141 if (N_SO_has_full_path) 3142 { 3143 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3144 { 3145 // We have two consecutive N_SO entries where the first contains a directory 3146 // and the second contains a full path. 3147 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 3148 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3149 add_nlist = false; 3150 } 3151 else 3152 { 3153 // This is the first entry in a N_SO that contains a directory or 3154 // a full path to the source file 3155 N_SO_index = sym_idx; 3156 } 3157 } 3158 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3159 { 3160 // This is usually the second N_SO entry that contains just the filename, 3161 // so here we combine it with the first one if we are minimizing the symbol table 3162 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3163 if (so_path && so_path[0]) 3164 { 3165 std::string full_so_path (so_path); 3166 const size_t double_slash_pos = full_so_path.find("//"); 3167 if (double_slash_pos != std::string::npos) 3168 { 3169 // The linker has been generating bad N_SO entries with doubled up paths 3170 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 3171 // and the second is the directory for the source file so you end up with 3172 // a path that looks like "/tmp/src//tmp/src/" 3173 FileSpec so_dir(so_path, false); 3174 if (!so_dir.Exists()) 3175 { 3176 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 3177 if (so_dir.Exists()) 3178 { 3179 // Trim off the incorrect path 3180 full_so_path.erase(0, double_slash_pos + 1); 3181 } 3182 } 3183 } 3184 if (*full_so_path.rbegin() != '/') 3185 full_so_path += '/'; 3186 full_so_path += symbol_name; 3187 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 3188 add_nlist = false; 3189 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3190 } 3191 } 3192 else 3193 { 3194 // This could be a relative path to a N_SO 3195 N_SO_index = sym_idx; 3196 } 3197 } 3198 3199 break; 3200 3201 case N_OSO: 3202 // object file name: name,,0,0,st_mtime 3203 type = eSymbolTypeObjectFile; 3204 break; 3205 3206 case N_LSYM: 3207 // local sym: name,,NO_SECT,type,offset 3208 type = eSymbolTypeLocal; 3209 break; 3210 3211 //---------------------------------------------------------------------- 3212 // INCL scopes 3213 //---------------------------------------------------------------------- 3214 case N_BINCL: 3215 // include file beginning: name,,NO_SECT,0,sum 3216 // We use the current number of symbols in the symbol table in lieu of 3217 // using nlist_idx in case we ever start trimming entries out 3218 N_INCL_indexes.push_back(sym_idx); 3219 type = eSymbolTypeScopeBegin; 3220 break; 3221 3222 case N_EINCL: 3223 // include file end: name,,NO_SECT,0,0 3224 // Set the size of the N_BINCL to the terminating index of this N_EINCL 3225 // so that we can always skip the entire symbol if we need to navigate 3226 // more quickly at the source level when parsing STABS 3227 if ( !N_INCL_indexes.empty() ) 3228 { 3229 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3230 symbol_ptr->SetByteSize(sym_idx + 1); 3231 symbol_ptr->SetSizeIsSibling(true); 3232 N_INCL_indexes.pop_back(); 3233 } 3234 type = eSymbolTypeScopeEnd; 3235 break; 3236 3237 case N_SOL: 3238 // #included file name: name,,n_sect,0,address 3239 type = eSymbolTypeHeaderFile; 3240 3241 // We currently don't use the header files on darwin 3242 add_nlist = false; 3243 break; 3244 3245 case N_PARAMS: 3246 // compiler parameters: name,,NO_SECT,0,0 3247 type = eSymbolTypeCompiler; 3248 break; 3249 3250 case N_VERSION: 3251 // compiler version: name,,NO_SECT,0,0 3252 type = eSymbolTypeCompiler; 3253 break; 3254 3255 case N_OLEVEL: 3256 // compiler -O level: name,,NO_SECT,0,0 3257 type = eSymbolTypeCompiler; 3258 break; 3259 3260 case N_PSYM: 3261 // parameter: name,,NO_SECT,type,offset 3262 type = eSymbolTypeVariable; 3263 break; 3264 3265 case N_ENTRY: 3266 // alternate entry: name,,n_sect,linenumber,address 3267 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3268 type = eSymbolTypeLineEntry; 3269 break; 3270 3271 //---------------------------------------------------------------------- 3272 // Left and Right Braces 3273 //---------------------------------------------------------------------- 3274 case N_LBRAC: 3275 // left bracket: 0,,NO_SECT,nesting level,address 3276 // We use the current number of symbols in the symbol table in lieu of 3277 // using nlist_idx in case we ever start trimming entries out 3278 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3279 N_BRAC_indexes.push_back(sym_idx); 3280 type = eSymbolTypeScopeBegin; 3281 break; 3282 3283 case N_RBRAC: 3284 // right bracket: 0,,NO_SECT,nesting level,address 3285 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 3286 // so that we can always skip the entire symbol if we need to navigate 3287 // more quickly at the source level when parsing STABS 3288 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3289 if ( !N_BRAC_indexes.empty() ) 3290 { 3291 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3292 symbol_ptr->SetByteSize(sym_idx + 1); 3293 symbol_ptr->SetSizeIsSibling(true); 3294 N_BRAC_indexes.pop_back(); 3295 } 3296 type = eSymbolTypeScopeEnd; 3297 break; 3298 3299 case N_EXCL: 3300 // deleted include file: name,,NO_SECT,0,sum 3301 type = eSymbolTypeHeaderFile; 3302 break; 3303 3304 //---------------------------------------------------------------------- 3305 // COMM scopes 3306 //---------------------------------------------------------------------- 3307 case N_BCOMM: 3308 // begin common: name,,NO_SECT,0,0 3309 // We use the current number of symbols in the symbol table in lieu of 3310 // using nlist_idx in case we ever start trimming entries out 3311 type = eSymbolTypeScopeBegin; 3312 N_COMM_indexes.push_back(sym_idx); 3313 break; 3314 3315 case N_ECOML: 3316 // end common (local name): 0,,n_sect,0,address 3317 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3318 // Fall through 3319 3320 case N_ECOMM: 3321 // end common: name,,n_sect,0,0 3322 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 3323 // so that we can always skip the entire symbol if we need to navigate 3324 // more quickly at the source level when parsing STABS 3325 if ( !N_COMM_indexes.empty() ) 3326 { 3327 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 3328 symbol_ptr->SetByteSize(sym_idx + 1); 3329 symbol_ptr->SetSizeIsSibling(true); 3330 N_COMM_indexes.pop_back(); 3331 } 3332 type = eSymbolTypeScopeEnd; 3333 break; 3334 3335 case N_LENG: 3336 // second stab entry with length information 3337 type = eSymbolTypeAdditional; 3338 break; 3339 3340 default: break; 3341 } 3342 } 3343 else 3344 { 3345 //uint8_t n_pext = N_PEXT & nlist.n_type; 3346 uint8_t n_type = N_TYPE & nlist.n_type; 3347 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3348 3349 switch (n_type) 3350 { 3351 case N_INDR:// Fall through 3352 case N_PBUD:// Fall through 3353 case N_UNDF: 3354 type = eSymbolTypeUndefined; 3355 break; 3356 3357 case N_ABS: 3358 type = eSymbolTypeAbsolute; 3359 break; 3360 3361 case N_SECT: 3362 { 3363 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3364 3365 if (!symbol_section) 3366 { 3367 // TODO: warn about this? 3368 add_nlist = false; 3369 break; 3370 } 3371 3372 if (TEXT_eh_frame_sectID == nlist.n_sect) 3373 { 3374 type = eSymbolTypeException; 3375 } 3376 else 3377 { 3378 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 3379 3380 switch (section_type) 3381 { 3382 case S_REGULAR: break; // regular section 3383 //case S_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section 3384 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 3385 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 3386 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 3387 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 3388 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 3389 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 3390 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 3391 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 3392 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 3393 //case S_COALESCED: type = eSymbolType; break; // section contains symbols that are to be coalesced 3394 //case S_GB_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 3395 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 3396 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 3397 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 3398 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 3399 default: break; 3400 } 3401 3402 if (type == eSymbolTypeInvalid) 3403 { 3404 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 3405 if (symbol_section->IsDescendant (text_section_sp.get())) 3406 { 3407 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 3408 S_ATTR_SELF_MODIFYING_CODE | 3409 S_ATTR_SOME_INSTRUCTIONS)) 3410 type = eSymbolTypeData; 3411 else 3412 type = eSymbolTypeCode; 3413 } 3414 else 3415 if (symbol_section->IsDescendant(data_section_sp.get())) 3416 { 3417 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 3418 { 3419 type = eSymbolTypeRuntime; 3420 3421 if (symbol_name && 3422 symbol_name[0] == '_' && 3423 symbol_name[1] == 'O' && 3424 symbol_name[2] == 'B') 3425 { 3426 llvm::StringRef symbol_name_ref(symbol_name); 3427 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 3428 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 3429 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 3430 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 3431 { 3432 symbol_name_non_abi_mangled = symbol_name + 1; 3433 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3434 type = eSymbolTypeObjCClass; 3435 demangled_is_synthesized = true; 3436 } 3437 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 3438 { 3439 symbol_name_non_abi_mangled = symbol_name + 1; 3440 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3441 type = eSymbolTypeObjCMetaClass; 3442 demangled_is_synthesized = true; 3443 } 3444 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 3445 { 3446 symbol_name_non_abi_mangled = symbol_name + 1; 3447 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3448 type = eSymbolTypeObjCIVar; 3449 demangled_is_synthesized = true; 3450 } 3451 } 3452 } 3453 else 3454 if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 3455 { 3456 type = eSymbolTypeException; 3457 } 3458 else 3459 { 3460 type = eSymbolTypeData; 3461 } 3462 } 3463 else 3464 if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 3465 { 3466 type = eSymbolTypeTrampoline; 3467 } 3468 else 3469 if (symbol_section->IsDescendant(objc_section_sp.get())) 3470 { 3471 type = eSymbolTypeRuntime; 3472 if (symbol_name && symbol_name[0] == '.') 3473 { 3474 llvm::StringRef symbol_name_ref(symbol_name); 3475 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 3476 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 3477 { 3478 symbol_name_non_abi_mangled = symbol_name; 3479 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 3480 type = eSymbolTypeObjCClass; 3481 demangled_is_synthesized = true; 3482 } 3483 } 3484 } 3485 } 3486 } 3487 } 3488 break; 3489 } 3490 } 3491 3492 if (add_nlist) 3493 { 3494 uint64_t symbol_value = nlist.n_value; 3495 3496 if (symbol_name_non_abi_mangled) 3497 { 3498 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 3499 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 3500 } 3501 else 3502 { 3503 bool symbol_name_is_mangled = false; 3504 3505 if (symbol_name && symbol_name[0] == '_') 3506 { 3507 symbol_name_is_mangled = symbol_name[1] == '_'; 3508 symbol_name++; // Skip the leading underscore 3509 } 3510 3511 if (symbol_name) 3512 { 3513 ConstString const_symbol_name(symbol_name); 3514 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 3515 if (is_gsym && is_debug) 3516 { 3517 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 3518 } 3519 } 3520 } 3521 if (symbol_section) 3522 { 3523 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3524 if (symbol_byte_size == 0 && function_starts_count > 0) 3525 { 3526 addr_t symbol_lookup_file_addr = nlist.n_value; 3527 // Do an exact address match for non-ARM addresses, else get the closest since 3528 // the symbol might be a thumb symbol which has an address with bit zero set 3529 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 3530 if (is_arm && func_start_entry) 3531 { 3532 // Verify that the function start address is the symbol address (ARM) 3533 // or the symbol address + 1 (thumb) 3534 if (func_start_entry->addr != symbol_lookup_file_addr && 3535 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 3536 { 3537 // Not the right entry, NULL it out... 3538 func_start_entry = NULL; 3539 } 3540 } 3541 if (func_start_entry) 3542 { 3543 func_start_entry->data = true; 3544 3545 addr_t symbol_file_addr = func_start_entry->addr; 3546 if (is_arm) 3547 symbol_file_addr &= 0xfffffffffffffffeull; 3548 3549 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3550 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3551 if (next_func_start_entry) 3552 { 3553 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3554 // Be sure the clear the Thumb address bit when we calculate the size 3555 // from the current and next address 3556 if (is_arm) 3557 next_symbol_file_addr &= 0xfffffffffffffffeull; 3558 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3559 } 3560 else 3561 { 3562 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3563 } 3564 } 3565 } 3566 symbol_value -= section_file_addr; 3567 } 3568 3569 if (is_debug == false) 3570 { 3571 if (type == eSymbolTypeCode) 3572 { 3573 // See if we can find a N_FUN entry for any code symbols. 3574 // If we do find a match, and the name matches, then we 3575 // can merge the two into just the function symbol to avoid 3576 // duplicate entries in the symbol table 3577 ValueToSymbolIndexMap::const_iterator pos = N_FUN_addr_to_sym_idx.find (nlist.n_value); 3578 if (pos != N_FUN_addr_to_sym_idx.end()) 3579 { 3580 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3581 { 3582 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3583 // We just need the flags from the linker symbol, so put these flags 3584 // into the N_FUN flags to avoid duplicate symbols in the symbol table 3585 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3586 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3587 sym[sym_idx].Clear(); 3588 continue; 3589 } 3590 } 3591 } 3592 else if (type == eSymbolTypeData) 3593 { 3594 // See if we can find a N_STSYM entry for any data symbols. 3595 // If we do find a match, and the name matches, then we 3596 // can merge the two into just the Static symbol to avoid 3597 // duplicate entries in the symbol table 3598 ValueToSymbolIndexMap::const_iterator pos = N_STSYM_addr_to_sym_idx.find (nlist.n_value); 3599 if (pos != N_STSYM_addr_to_sym_idx.end()) 3600 { 3601 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3602 { 3603 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3604 // We just need the flags from the linker symbol, so put these flags 3605 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3606 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3607 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3608 sym[sym_idx].Clear(); 3609 continue; 3610 } 3611 } 3612 else 3613 { 3614 // Combine N_GSYM stab entries with the non stab symbol 3615 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 3616 if (pos != N_GSYM_name_to_sym_idx.end()) 3617 { 3618 const uint32_t GSYM_sym_idx = pos->second; 3619 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 3620 // Copy the address, because often the N_GSYM address has an invalid address of zero 3621 // when the global is a common symbol 3622 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 3623 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 3624 // We just need the flags from the linker symbol, so put these flags 3625 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3626 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3627 sym[sym_idx].Clear(); 3628 continue; 3629 } 3630 } 3631 } 3632 } 3633 3634 sym[sym_idx].SetID (nlist_idx); 3635 sym[sym_idx].SetType (type); 3636 sym[sym_idx].GetAddress().SetSection (symbol_section); 3637 sym[sym_idx].GetAddress().SetOffset (symbol_value); 3638 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3639 3640 if (symbol_byte_size > 0) 3641 sym[sym_idx].SetByteSize(symbol_byte_size); 3642 3643 if (demangled_is_synthesized) 3644 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3645 3646 ++sym_idx; 3647 } 3648 else 3649 { 3650 sym[sym_idx].Clear(); 3651 } 3652 } 3653 } 3654 3655 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 3656 3657 if (function_starts_count > 0) 3658 { 3659 char synthetic_function_symbol[PATH_MAX]; 3660 uint32_t num_synthetic_function_symbols = 0; 3661 for (i=0; i<function_starts_count; ++i) 3662 { 3663 if (function_starts.GetEntryRef (i).data == false) 3664 ++num_synthetic_function_symbols; 3665 } 3666 3667 if (num_synthetic_function_symbols > 0) 3668 { 3669 if (num_syms < sym_idx + num_synthetic_function_symbols) 3670 { 3671 num_syms = sym_idx + num_synthetic_function_symbols; 3672 sym = symtab->Resize (num_syms); 3673 } 3674 uint32_t synthetic_function_symbol_idx = 0; 3675 for (i=0; i<function_starts_count; ++i) 3676 { 3677 const FunctionStarts::Entry *func_start_entry = function_starts.GetEntryAtIndex (i); 3678 if (func_start_entry->data == false) 3679 { 3680 addr_t symbol_file_addr = func_start_entry->addr; 3681 uint32_t symbol_flags = 0; 3682 if (is_arm) 3683 { 3684 if (symbol_file_addr & 1) 3685 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3686 symbol_file_addr &= 0xfffffffffffffffeull; 3687 } 3688 Address symbol_addr; 3689 if (module_sp->ResolveFileAddress (symbol_file_addr, symbol_addr)) 3690 { 3691 SectionSP symbol_section (symbol_addr.GetSection()); 3692 uint32_t symbol_byte_size = 0; 3693 if (symbol_section) 3694 { 3695 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3696 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3697 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3698 if (next_func_start_entry) 3699 { 3700 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3701 if (is_arm) 3702 next_symbol_file_addr &= 0xfffffffffffffffeull; 3703 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3704 } 3705 else 3706 { 3707 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3708 } 3709 snprintf (synthetic_function_symbol, 3710 sizeof(synthetic_function_symbol), 3711 "___lldb_unnamed_function%u$$%s", 3712 ++synthetic_function_symbol_idx, 3713 module_sp->GetFileSpec().GetFilename().GetCString()); 3714 sym[sym_idx].SetID (synthetic_sym_id++); 3715 sym[sym_idx].GetMangled().SetDemangledName(ConstString(synthetic_function_symbol)); 3716 sym[sym_idx].SetType (eSymbolTypeCode); 3717 sym[sym_idx].SetIsSynthetic (true); 3718 sym[sym_idx].GetAddress() = symbol_addr; 3719 if (symbol_flags) 3720 sym[sym_idx].SetFlags (symbol_flags); 3721 if (symbol_byte_size) 3722 sym[sym_idx].SetByteSize (symbol_byte_size); 3723 ++sym_idx; 3724 } 3725 } 3726 } 3727 } 3728 } 3729 } 3730 3731 // Trim our symbols down to just what we ended up with after 3732 // removing any symbols. 3733 if (sym_idx < num_syms) 3734 { 3735 num_syms = sym_idx; 3736 sym = symtab->Resize (num_syms); 3737 } 3738 3739 std::vector<TrieEntryWithOffset> trie_entries; 3740 std::set<lldb::addr_t> resolver_addresses; 3741 3742 if (dyld_trie_data.GetByteSize() > 0) 3743 { 3744 std::vector<llvm::StringRef> nameSlices; 3745 ParseTrieEntries (dyld_trie_data, 3746 0, 3747 nameSlices, 3748 resolver_addresses, 3749 trie_entries); 3750 3751 ConstString text_segment_name ("__TEXT"); 3752 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 3753 if (text_segment_sp) 3754 { 3755 const lldb::addr_t text_segment_file_addr = text_segment_sp->GetFileAddress(); 3756 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) 3757 { 3758 for (auto &e : trie_entries) 3759 e.entry.address += text_segment_file_addr; 3760 } 3761 } 3762 } 3763 3764 // Now synthesize indirect symbols 3765 if (m_dysymtab.nindirectsyms != 0) 3766 { 3767 if (indirect_symbol_index_data.GetByteSize()) 3768 { 3769 NListIndexToSymbolIndexMap::const_iterator end_index_pos = m_nlist_idx_to_sym_idx.end(); 3770 3771 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); ++sect_idx) 3772 { 3773 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == S_SYMBOL_STUBS) 3774 { 3775 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 3776 if (symbol_stub_byte_size == 0) 3777 continue; 3778 3779 const uint32_t num_symbol_stubs = m_mach_sections[sect_idx].size / symbol_stub_byte_size; 3780 3781 if (num_symbol_stubs == 0) 3782 continue; 3783 3784 const uint32_t symbol_stub_index_offset = m_mach_sections[sect_idx].reserved1; 3785 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) 3786 { 3787 const uint32_t symbol_stub_index = symbol_stub_index_offset + stub_idx; 3788 const lldb::addr_t symbol_stub_addr = m_mach_sections[sect_idx].addr + (stub_idx * symbol_stub_byte_size); 3789 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 3790 if (indirect_symbol_index_data.ValidOffsetForDataOfSize(symbol_stub_offset, 4)) 3791 { 3792 const uint32_t stub_sym_id = indirect_symbol_index_data.GetU32 (&symbol_stub_offset); 3793 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 3794 continue; 3795 3796 NListIndexToSymbolIndexMap::const_iterator index_pos = m_nlist_idx_to_sym_idx.find (stub_sym_id); 3797 Symbol *stub_symbol = NULL; 3798 if (index_pos != end_index_pos) 3799 { 3800 // We have a remapping from the original nlist index to 3801 // a current symbol index, so just look this up by index 3802 stub_symbol = symtab->SymbolAtIndex (index_pos->second); 3803 } 3804 else 3805 { 3806 // We need to lookup a symbol using the original nlist 3807 // symbol index since this index is coming from the 3808 // S_SYMBOL_STUBS 3809 stub_symbol = symtab->FindSymbolByID (stub_sym_id); 3810 } 3811 3812 if (stub_symbol) 3813 { 3814 Address so_addr(symbol_stub_addr, section_list); 3815 3816 if (stub_symbol->GetType() == eSymbolTypeUndefined) 3817 { 3818 // Change the external symbol into a trampoline that makes sense 3819 // These symbols were N_UNDF N_EXT, and are useless to us, so we 3820 // can re-use them so we don't have to make up a synthetic symbol 3821 // for no good reason. 3822 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 3823 stub_symbol->SetType (eSymbolTypeTrampoline); 3824 else 3825 stub_symbol->SetType (eSymbolTypeResolver); 3826 stub_symbol->SetExternal (false); 3827 stub_symbol->GetAddress() = so_addr; 3828 stub_symbol->SetByteSize (symbol_stub_byte_size); 3829 } 3830 else 3831 { 3832 // Make a synthetic symbol to describe the trampoline stub 3833 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 3834 if (sym_idx >= num_syms) 3835 { 3836 sym = symtab->Resize (++num_syms); 3837 stub_symbol = NULL; // this pointer no longer valid 3838 } 3839 sym[sym_idx].SetID (synthetic_sym_id++); 3840 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 3841 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 3842 sym[sym_idx].SetType (eSymbolTypeTrampoline); 3843 else 3844 sym[sym_idx].SetType (eSymbolTypeResolver); 3845 sym[sym_idx].SetIsSynthetic (true); 3846 sym[sym_idx].GetAddress() = so_addr; 3847 sym[sym_idx].SetByteSize (symbol_stub_byte_size); 3848 ++sym_idx; 3849 } 3850 } 3851 else 3852 { 3853 if (log) 3854 log->Warning ("symbol stub referencing symbol table symbol %u that isn't in our minimal symbol table, fix this!!!", stub_sym_id); 3855 } 3856 } 3857 } 3858 } 3859 } 3860 } 3861 } 3862 3863 3864 if (!trie_entries.empty()) 3865 { 3866 for (const auto &e : trie_entries) 3867 { 3868 if (e.entry.import_name) 3869 { 3870 // Make a synthetic symbol to describe re-exported symbol. 3871 if (sym_idx >= num_syms) 3872 sym = symtab->Resize (++num_syms); 3873 sym[sym_idx].SetID (synthetic_sym_id++); 3874 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 3875 sym[sym_idx].SetType (eSymbolTypeReExported); 3876 sym[sym_idx].SetIsSynthetic (true); 3877 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 3878 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) 3879 { 3880 sym[sym_idx].SetReExportedSymbolSharedLibrary(dylib_files.GetFileSpecAtIndex(e.entry.other-1)); 3881 } 3882 ++sym_idx; 3883 } 3884 } 3885 } 3886 3887 3888 3889 // StreamFile s(stdout, false); 3890 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 3891 // symtab->Dump(&s, NULL, eSortOrderNone); 3892 // Set symbol byte sizes correctly since mach-o nlist entries don't have sizes 3893 symtab->CalculateSymbolSizes(); 3894 3895 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 3896 // symtab->Dump(&s, NULL, eSortOrderNone); 3897 3898 return symtab->GetNumSymbols(); 3899 } 3900 return 0; 3901 } 3902 3903 3904 void 3905 ObjectFileMachO::Dump (Stream *s) 3906 { 3907 ModuleSP module_sp(GetModule()); 3908 if (module_sp) 3909 { 3910 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 3911 s->Printf("%p: ", this); 3912 s->Indent(); 3913 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 3914 s->PutCString("ObjectFileMachO64"); 3915 else 3916 s->PutCString("ObjectFileMachO32"); 3917 3918 ArchSpec header_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 3919 3920 *s << ", file = '" << m_file << "', arch = " << header_arch.GetArchitectureName() << "\n"; 3921 3922 SectionList *sections = GetSectionList(); 3923 if (sections) 3924 sections->Dump(s, NULL, true, UINT32_MAX); 3925 3926 if (m_symtab_ap.get()) 3927 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 3928 } 3929 } 3930 3931 bool 3932 ObjectFileMachO::GetUUID (const llvm::MachO::mach_header &header, 3933 const lldb_private::DataExtractor &data, 3934 lldb::offset_t lc_offset, 3935 lldb_private::UUID& uuid) 3936 { 3937 uint32_t i; 3938 struct uuid_command load_cmd; 3939 3940 lldb::offset_t offset = lc_offset; 3941 for (i=0; i<header.ncmds; ++i) 3942 { 3943 const lldb::offset_t cmd_offset = offset; 3944 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 3945 break; 3946 3947 if (load_cmd.cmd == LC_UUID) 3948 { 3949 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 3950 3951 if (uuid_bytes) 3952 { 3953 // OpenCL on Mac OS X uses the same UUID for each of its object files. 3954 // We pretend these object files have no UUID to prevent crashing. 3955 3956 const uint8_t opencl_uuid[] = { 0x8c, 0x8e, 0xb3, 0x9b, 3957 0x3b, 0xa8, 3958 0x4b, 0x16, 3959 0xb6, 0xa4, 3960 0x27, 0x63, 0xbb, 0x14, 0xf0, 0x0d }; 3961 3962 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 3963 return false; 3964 3965 uuid.SetBytes (uuid_bytes); 3966 return true; 3967 } 3968 return false; 3969 } 3970 offset = cmd_offset + load_cmd.cmdsize; 3971 } 3972 return false; 3973 } 3974 3975 bool 3976 ObjectFileMachO::GetUUID (lldb_private::UUID* uuid) 3977 { 3978 ModuleSP module_sp(GetModule()); 3979 if (module_sp) 3980 { 3981 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 3982 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 3983 return GetUUID (m_header, m_data, offset, *uuid); 3984 } 3985 return false; 3986 } 3987 3988 3989 uint32_t 3990 ObjectFileMachO::GetDependentModules (FileSpecList& files) 3991 { 3992 uint32_t count = 0; 3993 ModuleSP module_sp(GetModule()); 3994 if (module_sp) 3995 { 3996 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 3997 struct load_command load_cmd; 3998 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 3999 const bool resolve_path = false; // Don't resolve the dependend file paths since they may not reside on this system 4000 uint32_t i; 4001 for (i=0; i<m_header.ncmds; ++i) 4002 { 4003 const uint32_t cmd_offset = offset; 4004 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4005 break; 4006 4007 switch (load_cmd.cmd) 4008 { 4009 case LC_LOAD_DYLIB: 4010 case LC_LOAD_WEAK_DYLIB: 4011 case LC_REEXPORT_DYLIB: 4012 case LC_LOAD_DYLINKER: 4013 case LC_LOADFVMLIB: 4014 case LC_LOAD_UPWARD_DYLIB: 4015 { 4016 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 4017 const char *path = m_data.PeekCStr(name_offset); 4018 // Skip any path that starts with '@' since these are usually: 4019 // @executable_path/.../file 4020 // @rpath/.../file 4021 if (path && path[0] != '@') 4022 { 4023 FileSpec file_spec(path, resolve_path); 4024 if (files.AppendIfUnique(file_spec)) 4025 count++; 4026 } 4027 } 4028 break; 4029 4030 default: 4031 break; 4032 } 4033 offset = cmd_offset + load_cmd.cmdsize; 4034 } 4035 } 4036 return count; 4037 } 4038 4039 lldb_private::Address 4040 ObjectFileMachO::GetEntryPointAddress () 4041 { 4042 // If the object file is not an executable it can't hold the entry point. m_entry_point_address 4043 // is initialized to an invalid address, so we can just return that. 4044 // If m_entry_point_address is valid it means we've found it already, so return the cached value. 4045 4046 if (!IsExecutable() || m_entry_point_address.IsValid()) 4047 return m_entry_point_address; 4048 4049 // Otherwise, look for the UnixThread or Thread command. The data for the Thread command is given in 4050 // /usr/include/mach-o.h, but it is basically: 4051 // 4052 // uint32_t flavor - this is the flavor argument you would pass to thread_get_state 4053 // uint32_t count - this is the count of longs in the thread state data 4054 // struct XXX_thread_state state - this is the structure from <machine/thread_status.h> corresponding to the flavor. 4055 // <repeat this trio> 4056 // 4057 // So we just keep reading the various register flavors till we find the GPR one, then read the PC out of there. 4058 // FIXME: We will need to have a "RegisterContext data provider" class at some point that can get all the registers 4059 // out of data in this form & attach them to a given thread. That should underlie the MacOS X User process plugin, 4060 // and we'll also need it for the MacOS X Core File process plugin. When we have that we can also use it here. 4061 // 4062 // For now we hard-code the offsets and flavors we need: 4063 // 4064 // 4065 4066 ModuleSP module_sp(GetModule()); 4067 if (module_sp) 4068 { 4069 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4070 struct load_command load_cmd; 4071 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4072 uint32_t i; 4073 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 4074 bool done = false; 4075 4076 for (i=0; i<m_header.ncmds; ++i) 4077 { 4078 const lldb::offset_t cmd_offset = offset; 4079 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4080 break; 4081 4082 switch (load_cmd.cmd) 4083 { 4084 case LC_UNIXTHREAD: 4085 case LC_THREAD: 4086 { 4087 while (offset < cmd_offset + load_cmd.cmdsize) 4088 { 4089 uint32_t flavor = m_data.GetU32(&offset); 4090 uint32_t count = m_data.GetU32(&offset); 4091 if (count == 0) 4092 { 4093 // We've gotten off somehow, log and exit; 4094 return m_entry_point_address; 4095 } 4096 4097 switch (m_header.cputype) 4098 { 4099 case llvm::MachO::CPU_TYPE_ARM: 4100 if (flavor == 1) // ARM_THREAD_STATE from mach/arm/thread_status.h 4101 { 4102 offset += 60; // This is the offset of pc in the GPR thread state data structure. 4103 start_address = m_data.GetU32(&offset); 4104 done = true; 4105 } 4106 break; 4107 case llvm::MachO::CPU_TYPE_I386: 4108 if (flavor == 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 4109 { 4110 offset += 40; // This is the offset of eip in the GPR thread state data structure. 4111 start_address = m_data.GetU32(&offset); 4112 done = true; 4113 } 4114 break; 4115 case llvm::MachO::CPU_TYPE_X86_64: 4116 if (flavor == 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 4117 { 4118 offset += 16 * 8; // This is the offset of rip in the GPR thread state data structure. 4119 start_address = m_data.GetU64(&offset); 4120 done = true; 4121 } 4122 break; 4123 default: 4124 return m_entry_point_address; 4125 } 4126 // Haven't found the GPR flavor yet, skip over the data for this flavor: 4127 if (done) 4128 break; 4129 offset += count * 4; 4130 } 4131 } 4132 break; 4133 case LC_MAIN: 4134 { 4135 ConstString text_segment_name ("__TEXT"); 4136 uint64_t entryoffset = m_data.GetU64(&offset); 4137 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 4138 if (text_segment_sp) 4139 { 4140 done = true; 4141 start_address = text_segment_sp->GetFileAddress() + entryoffset; 4142 } 4143 } 4144 4145 default: 4146 break; 4147 } 4148 if (done) 4149 break; 4150 4151 // Go to the next load command: 4152 offset = cmd_offset + load_cmd.cmdsize; 4153 } 4154 4155 if (start_address != LLDB_INVALID_ADDRESS) 4156 { 4157 // We got the start address from the load commands, so now resolve that address in the sections 4158 // of this ObjectFile: 4159 if (!m_entry_point_address.ResolveAddressUsingFileSections (start_address, GetSectionList())) 4160 { 4161 m_entry_point_address.Clear(); 4162 } 4163 } 4164 else 4165 { 4166 // We couldn't read the UnixThread load command - maybe it wasn't there. As a fallback look for the 4167 // "start" symbol in the main executable. 4168 4169 ModuleSP module_sp (GetModule()); 4170 4171 if (module_sp) 4172 { 4173 SymbolContextList contexts; 4174 SymbolContext context; 4175 if (module_sp->FindSymbolsWithNameAndType(ConstString ("start"), eSymbolTypeCode, contexts)) 4176 { 4177 if (contexts.GetContextAtIndex(0, context)) 4178 m_entry_point_address = context.symbol->GetAddress(); 4179 } 4180 } 4181 } 4182 } 4183 4184 return m_entry_point_address; 4185 4186 } 4187 4188 lldb_private::Address 4189 ObjectFileMachO::GetHeaderAddress () 4190 { 4191 lldb_private::Address header_addr; 4192 SectionList *section_list = GetSectionList(); 4193 if (section_list) 4194 { 4195 SectionSP text_segment_sp (section_list->FindSectionByName (GetSegmentNameTEXT())); 4196 if (text_segment_sp) 4197 { 4198 header_addr.SetSection (text_segment_sp); 4199 header_addr.SetOffset (0); 4200 } 4201 } 4202 return header_addr; 4203 } 4204 4205 uint32_t 4206 ObjectFileMachO::GetNumThreadContexts () 4207 { 4208 ModuleSP module_sp(GetModule()); 4209 if (module_sp) 4210 { 4211 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4212 if (!m_thread_context_offsets_valid) 4213 { 4214 m_thread_context_offsets_valid = true; 4215 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4216 FileRangeArray::Entry file_range; 4217 thread_command thread_cmd; 4218 for (uint32_t i=0; i<m_header.ncmds; ++i) 4219 { 4220 const uint32_t cmd_offset = offset; 4221 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 4222 break; 4223 4224 if (thread_cmd.cmd == LC_THREAD) 4225 { 4226 file_range.SetRangeBase (offset); 4227 file_range.SetByteSize (thread_cmd.cmdsize - 8); 4228 m_thread_context_offsets.Append (file_range); 4229 } 4230 offset = cmd_offset + thread_cmd.cmdsize; 4231 } 4232 } 4233 } 4234 return m_thread_context_offsets.GetSize(); 4235 } 4236 4237 lldb::RegisterContextSP 4238 ObjectFileMachO::GetThreadContextAtIndex (uint32_t idx, lldb_private::Thread &thread) 4239 { 4240 lldb::RegisterContextSP reg_ctx_sp; 4241 4242 ModuleSP module_sp(GetModule()); 4243 if (module_sp) 4244 { 4245 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4246 if (!m_thread_context_offsets_valid) 4247 GetNumThreadContexts (); 4248 4249 const FileRangeArray::Entry *thread_context_file_range = m_thread_context_offsets.GetEntryAtIndex (idx); 4250 if (thread_context_file_range) 4251 { 4252 4253 DataExtractor data (m_data, 4254 thread_context_file_range->GetRangeBase(), 4255 thread_context_file_range->GetByteSize()); 4256 4257 switch (m_header.cputype) 4258 { 4259 case llvm::MachO::CPU_TYPE_ARM: 4260 reg_ctx_sp.reset (new RegisterContextDarwin_arm_Mach (thread, data)); 4261 break; 4262 4263 case llvm::MachO::CPU_TYPE_I386: 4264 reg_ctx_sp.reset (new RegisterContextDarwin_i386_Mach (thread, data)); 4265 break; 4266 4267 case llvm::MachO::CPU_TYPE_X86_64: 4268 reg_ctx_sp.reset (new RegisterContextDarwin_x86_64_Mach (thread, data)); 4269 break; 4270 } 4271 } 4272 } 4273 return reg_ctx_sp; 4274 } 4275 4276 4277 ObjectFile::Type 4278 ObjectFileMachO::CalculateType() 4279 { 4280 switch (m_header.filetype) 4281 { 4282 case MH_OBJECT: // 0x1u 4283 if (GetAddressByteSize () == 4) 4284 { 4285 // 32 bit kexts are just object files, but they do have a valid 4286 // UUID load command. 4287 UUID uuid; 4288 if (GetUUID(&uuid)) 4289 { 4290 // this checking for the UUID load command is not enough 4291 // we could eventually look for the symbol named 4292 // "OSKextGetCurrentIdentifier" as this is required of kexts 4293 if (m_strata == eStrataInvalid) 4294 m_strata = eStrataKernel; 4295 return eTypeSharedLibrary; 4296 } 4297 } 4298 return eTypeObjectFile; 4299 4300 case MH_EXECUTE: return eTypeExecutable; // 0x2u 4301 case MH_FVMLIB: return eTypeSharedLibrary; // 0x3u 4302 case MH_CORE: return eTypeCoreFile; // 0x4u 4303 case MH_PRELOAD: return eTypeSharedLibrary; // 0x5u 4304 case MH_DYLIB: return eTypeSharedLibrary; // 0x6u 4305 case MH_DYLINKER: return eTypeDynamicLinker; // 0x7u 4306 case MH_BUNDLE: return eTypeSharedLibrary; // 0x8u 4307 case MH_DYLIB_STUB: return eTypeStubLibrary; // 0x9u 4308 case MH_DSYM: return eTypeDebugInfo; // 0xAu 4309 case MH_KEXT_BUNDLE: return eTypeSharedLibrary; // 0xBu 4310 default: 4311 break; 4312 } 4313 return eTypeUnknown; 4314 } 4315 4316 ObjectFile::Strata 4317 ObjectFileMachO::CalculateStrata() 4318 { 4319 switch (m_header.filetype) 4320 { 4321 case MH_OBJECT: // 0x1u 4322 { 4323 // 32 bit kexts are just object files, but they do have a valid 4324 // UUID load command. 4325 UUID uuid; 4326 if (GetUUID(&uuid)) 4327 { 4328 // this checking for the UUID load command is not enough 4329 // we could eventually look for the symbol named 4330 // "OSKextGetCurrentIdentifier" as this is required of kexts 4331 if (m_type == eTypeInvalid) 4332 m_type = eTypeSharedLibrary; 4333 4334 return eStrataKernel; 4335 } 4336 } 4337 return eStrataUnknown; 4338 4339 case MH_EXECUTE: // 0x2u 4340 // Check for the MH_DYLDLINK bit in the flags 4341 if (m_header.flags & MH_DYLDLINK) 4342 { 4343 return eStrataUser; 4344 } 4345 else 4346 { 4347 SectionList *section_list = GetSectionList(); 4348 if (section_list) 4349 { 4350 static ConstString g_kld_section_name ("__KLD"); 4351 if (section_list->FindSectionByName(g_kld_section_name)) 4352 return eStrataKernel; 4353 } 4354 } 4355 return eStrataRawImage; 4356 4357 case MH_FVMLIB: return eStrataUser; // 0x3u 4358 case MH_CORE: return eStrataUnknown; // 0x4u 4359 case MH_PRELOAD: return eStrataRawImage; // 0x5u 4360 case MH_DYLIB: return eStrataUser; // 0x6u 4361 case MH_DYLINKER: return eStrataUser; // 0x7u 4362 case MH_BUNDLE: return eStrataUser; // 0x8u 4363 case MH_DYLIB_STUB: return eStrataUser; // 0x9u 4364 case MH_DSYM: return eStrataUnknown; // 0xAu 4365 case MH_KEXT_BUNDLE: return eStrataKernel; // 0xBu 4366 default: 4367 break; 4368 } 4369 return eStrataUnknown; 4370 } 4371 4372 4373 uint32_t 4374 ObjectFileMachO::GetVersion (uint32_t *versions, uint32_t num_versions) 4375 { 4376 ModuleSP module_sp(GetModule()); 4377 if (module_sp) 4378 { 4379 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4380 struct dylib_command load_cmd; 4381 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4382 uint32_t version_cmd = 0; 4383 uint64_t version = 0; 4384 uint32_t i; 4385 for (i=0; i<m_header.ncmds; ++i) 4386 { 4387 const lldb::offset_t cmd_offset = offset; 4388 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4389 break; 4390 4391 if (load_cmd.cmd == LC_ID_DYLIB) 4392 { 4393 if (version_cmd == 0) 4394 { 4395 version_cmd = load_cmd.cmd; 4396 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 4397 break; 4398 version = load_cmd.dylib.current_version; 4399 } 4400 break; // Break for now unless there is another more complete version 4401 // number load command in the future. 4402 } 4403 offset = cmd_offset + load_cmd.cmdsize; 4404 } 4405 4406 if (version_cmd == LC_ID_DYLIB) 4407 { 4408 if (versions != NULL && num_versions > 0) 4409 { 4410 if (num_versions > 0) 4411 versions[0] = (version & 0xFFFF0000ull) >> 16; 4412 if (num_versions > 1) 4413 versions[1] = (version & 0x0000FF00ull) >> 8; 4414 if (num_versions > 2) 4415 versions[2] = (version & 0x000000FFull); 4416 // Fill in an remaining version numbers with invalid values 4417 for (i=3; i<num_versions; ++i) 4418 versions[i] = UINT32_MAX; 4419 } 4420 // The LC_ID_DYLIB load command has a version with 3 version numbers 4421 // in it, so always return 3 4422 return 3; 4423 } 4424 } 4425 return false; 4426 } 4427 4428 bool 4429 ObjectFileMachO::GetArchitecture (ArchSpec &arch) 4430 { 4431 ModuleSP module_sp(GetModule()); 4432 if (module_sp) 4433 { 4434 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4435 arch.SetArchitecture (eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 4436 4437 // Files with type MH_PRELOAD are currently used in cases where the image 4438 // debugs at the addresses in the file itself. Below we set the OS to 4439 // unknown to make sure we use the DynamicLoaderStatic()... 4440 if (m_header.filetype == MH_PRELOAD) 4441 { 4442 arch.GetTriple().setOS (llvm::Triple::UnknownOS); 4443 } 4444 return true; 4445 } 4446 return false; 4447 } 4448 4449 4450 UUID 4451 ObjectFileMachO::GetProcessSharedCacheUUID (Process *process) 4452 { 4453 UUID uuid; 4454 if (process) 4455 { 4456 addr_t all_image_infos = process->GetImageInfoAddress(); 4457 4458 // The address returned by GetImageInfoAddress may be the address of dyld (don't want) 4459 // or it may be the address of the dyld_all_image_infos structure (want). The first four 4460 // bytes will be either the version field (all_image_infos) or a Mach-O file magic constant. 4461 // Version 13 and higher of dyld_all_image_infos is required to get the sharedCacheUUID field. 4462 4463 Error err; 4464 uint32_t version_or_magic = process->ReadUnsignedIntegerFromMemory (all_image_infos, 4, -1, err); 4465 if (version_or_magic != -1 4466 && version_or_magic != MH_MAGIC 4467 && version_or_magic != MH_CIGAM 4468 && version_or_magic != MH_MAGIC_64 4469 && version_or_magic != MH_CIGAM_64 4470 && version_or_magic >= 13) 4471 { 4472 addr_t sharedCacheUUID_address = LLDB_INVALID_ADDRESS; 4473 int wordsize = process->GetAddressByteSize(); 4474 if (wordsize == 8) 4475 { 4476 sharedCacheUUID_address = all_image_infos + 160; // sharedCacheUUID <mach-o/dyld_images.h> 4477 } 4478 if (wordsize == 4) 4479 { 4480 sharedCacheUUID_address = all_image_infos + 84; // sharedCacheUUID <mach-o/dyld_images.h> 4481 } 4482 if (sharedCacheUUID_address != LLDB_INVALID_ADDRESS) 4483 { 4484 uuid_t shared_cache_uuid; 4485 if (process->ReadMemory (sharedCacheUUID_address, shared_cache_uuid, sizeof (uuid_t), err) == sizeof (uuid_t)) 4486 { 4487 uuid.SetBytes (shared_cache_uuid); 4488 } 4489 } 4490 } 4491 } 4492 return uuid; 4493 } 4494 4495 UUID 4496 ObjectFileMachO::GetLLDBSharedCacheUUID () 4497 { 4498 UUID uuid; 4499 #if defined (__APPLE__) && defined (__arm__) 4500 uint8_t *(*dyld_get_all_image_infos)(void); 4501 dyld_get_all_image_infos = (uint8_t*(*)()) dlsym (RTLD_DEFAULT, "_dyld_get_all_image_infos"); 4502 if (dyld_get_all_image_infos) 4503 { 4504 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 4505 if (dyld_all_image_infos_address) 4506 { 4507 uint32_t *version = (uint32_t*) dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 4508 if (*version >= 13) 4509 { 4510 uuid_t *sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 84); // sharedCacheUUID <mach-o/dyld_images.h> 4511 uuid.SetBytes (sharedCacheUUID_address); 4512 } 4513 } 4514 } 4515 #endif 4516 return uuid; 4517 } 4518 4519 uint32_t 4520 ObjectFileMachO::GetMinimumOSVersion (uint32_t *versions, uint32_t num_versions) 4521 { 4522 if (m_min_os_versions.empty()) 4523 { 4524 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4525 bool success = false; 4526 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 4527 { 4528 const lldb::offset_t load_cmd_offset = offset; 4529 4530 version_min_command lc; 4531 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 4532 break; 4533 if (lc.cmd == LC_VERSION_MIN_MACOSX || lc.cmd == LC_VERSION_MIN_IPHONEOS) 4534 { 4535 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 4536 { 4537 const uint32_t xxxx = lc.version >> 16; 4538 const uint32_t yy = (lc.version >> 8) & 0xffu; 4539 const uint32_t zz = lc.version & 0xffu; 4540 if (xxxx) 4541 { 4542 m_min_os_versions.push_back(xxxx); 4543 if (yy) 4544 { 4545 m_min_os_versions.push_back(yy); 4546 if (zz) 4547 m_min_os_versions.push_back(zz); 4548 } 4549 } 4550 success = true; 4551 } 4552 } 4553 offset = load_cmd_offset + lc.cmdsize; 4554 } 4555 4556 if (success == false) 4557 { 4558 // Push an invalid value so we don't keep trying to 4559 m_min_os_versions.push_back(UINT32_MAX); 4560 } 4561 } 4562 4563 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) 4564 { 4565 if (versions != NULL && num_versions > 0) 4566 { 4567 for (size_t i=0; i<num_versions; ++i) 4568 { 4569 if (i < m_min_os_versions.size()) 4570 versions[i] = m_min_os_versions[i]; 4571 else 4572 versions[i] = 0; 4573 } 4574 } 4575 return m_min_os_versions.size(); 4576 } 4577 // Call the superclasses version that will empty out the data 4578 return ObjectFile::GetMinimumOSVersion (versions, num_versions); 4579 } 4580 4581 uint32_t 4582 ObjectFileMachO::GetSDKVersion(uint32_t *versions, uint32_t num_versions) 4583 { 4584 if (m_sdk_versions.empty()) 4585 { 4586 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4587 bool success = false; 4588 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 4589 { 4590 const lldb::offset_t load_cmd_offset = offset; 4591 4592 version_min_command lc; 4593 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 4594 break; 4595 if (lc.cmd == LC_VERSION_MIN_MACOSX || lc.cmd == LC_VERSION_MIN_IPHONEOS) 4596 { 4597 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 4598 { 4599 const uint32_t xxxx = lc.reserved >> 16; 4600 const uint32_t yy = (lc.reserved >> 8) & 0xffu; 4601 const uint32_t zz = lc.reserved & 0xffu; 4602 if (xxxx) 4603 { 4604 m_sdk_versions.push_back(xxxx); 4605 if (yy) 4606 { 4607 m_sdk_versions.push_back(yy); 4608 if (zz) 4609 m_sdk_versions.push_back(zz); 4610 } 4611 } 4612 success = true; 4613 } 4614 } 4615 offset = load_cmd_offset + lc.cmdsize; 4616 } 4617 4618 if (success == false) 4619 { 4620 // Push an invalid value so we don't keep trying to 4621 m_sdk_versions.push_back(UINT32_MAX); 4622 } 4623 } 4624 4625 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) 4626 { 4627 if (versions != NULL && num_versions > 0) 4628 { 4629 for (size_t i=0; i<num_versions; ++i) 4630 { 4631 if (i < m_sdk_versions.size()) 4632 versions[i] = m_sdk_versions[i]; 4633 else 4634 versions[i] = 0; 4635 } 4636 } 4637 return m_sdk_versions.size(); 4638 } 4639 // Call the superclasses version that will empty out the data 4640 return ObjectFile::GetSDKVersion (versions, num_versions); 4641 } 4642 4643 4644 //------------------------------------------------------------------ 4645 // PluginInterface protocol 4646 //------------------------------------------------------------------ 4647 lldb_private::ConstString 4648 ObjectFileMachO::GetPluginName() 4649 { 4650 return GetPluginNameStatic(); 4651 } 4652 4653 uint32_t 4654 ObjectFileMachO::GetPluginVersion() 4655 { 4656 return 1; 4657 } 4658 4659