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 resolver_addresses.insert(e.entry.other); 1575 } 1576 else 1577 e.entry.other = 0; 1578 } 1579 // Only add symbols that are reexport symbols with a valid import name 1580 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && import_name[0]) 1581 { 1582 std::string name; 1583 if (!nameSlices.empty()) 1584 { 1585 for (auto name_slice: nameSlices) 1586 name.append(name_slice.data(), name_slice.size()); 1587 } 1588 if (name.size() > 1) 1589 { 1590 // Skip the leading '_' 1591 e.entry.name.SetCStringWithLength(name.c_str() + 1,name.size() - 1); 1592 } 1593 if (import_name) 1594 { 1595 // Skip the leading '_' 1596 e.entry.import_name.SetCString(import_name+1); 1597 } 1598 output.push_back(e); 1599 } 1600 } 1601 1602 const uint8_t childrenCount = data.GetU8(&children_offset); 1603 for (uint8_t i=0; i < childrenCount; ++i) { 1604 nameSlices.push_back(data.GetCStr(&children_offset)); 1605 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 1606 if (childNodeOffset) 1607 { 1608 ParseTrieEntries(data, 1609 childNodeOffset, 1610 nameSlices, 1611 resolver_addresses, 1612 output); 1613 } 1614 nameSlices.pop_back(); 1615 } 1616 } 1617 1618 size_t 1619 ObjectFileMachO::ParseSymtab () 1620 { 1621 Timer scoped_timer(__PRETTY_FUNCTION__, 1622 "ObjectFileMachO::ParseSymtab () module = %s", 1623 m_file.GetFilename().AsCString("")); 1624 ModuleSP module_sp (GetModule()); 1625 if (!module_sp) 1626 return 0; 1627 1628 struct symtab_command symtab_load_command = { 0, 0, 0, 0, 0, 0 }; 1629 struct linkedit_data_command function_starts_load_command = { 0, 0, 0, 0 }; 1630 struct dyld_info_command dyld_info = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }; 1631 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 1632 FunctionStarts function_starts; 1633 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1634 uint32_t i; 1635 FileSpecList dylib_files; 1636 Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_SYMBOLS)); 1637 1638 for (i=0; i<m_header.ncmds; ++i) 1639 { 1640 const lldb::offset_t cmd_offset = offset; 1641 // Read in the load command and load command size 1642 struct load_command lc; 1643 if (m_data.GetU32(&offset, &lc, 2) == NULL) 1644 break; 1645 // Watch for the symbol table load command 1646 switch (lc.cmd) 1647 { 1648 case LC_SYMTAB: 1649 symtab_load_command.cmd = lc.cmd; 1650 symtab_load_command.cmdsize = lc.cmdsize; 1651 // Read in the rest of the symtab load command 1652 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 0) // fill in symoff, nsyms, stroff, strsize fields 1653 return 0; 1654 if (symtab_load_command.symoff == 0) 1655 { 1656 if (log) 1657 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 1658 return 0; 1659 } 1660 1661 if (symtab_load_command.stroff == 0) 1662 { 1663 if (log) 1664 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 1665 return 0; 1666 } 1667 1668 if (symtab_load_command.nsyms == 0) 1669 { 1670 if (log) 1671 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 1672 return 0; 1673 } 1674 1675 if (symtab_load_command.strsize == 0) 1676 { 1677 if (log) 1678 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 1679 return 0; 1680 } 1681 break; 1682 1683 case LC_DYLD_INFO: 1684 case LC_DYLD_INFO_ONLY: 1685 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) 1686 { 1687 dyld_info.cmd = lc.cmd; 1688 dyld_info.cmdsize = lc.cmdsize; 1689 } 1690 else 1691 { 1692 memset (&dyld_info, 0, sizeof(dyld_info)); 1693 } 1694 break; 1695 1696 case LC_LOAD_DYLIB: 1697 case LC_LOAD_WEAK_DYLIB: 1698 case LC_REEXPORT_DYLIB: 1699 case LC_LOADFVMLIB: 1700 case LC_LOAD_UPWARD_DYLIB: 1701 { 1702 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 1703 const char *path = m_data.PeekCStr(name_offset); 1704 if (path) 1705 { 1706 FileSpec file_spec(path, false); 1707 // Strip the path if there is @rpath, @executanble, etc so we just use the basename 1708 if (path[0] == '@') 1709 file_spec.GetDirectory().Clear(); 1710 1711 dylib_files.Append(file_spec); 1712 } 1713 } 1714 break; 1715 1716 case LC_FUNCTION_STARTS: 1717 function_starts_load_command.cmd = lc.cmd; 1718 function_starts_load_command.cmdsize = lc.cmdsize; 1719 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == NULL) // fill in symoff, nsyms, stroff, strsize fields 1720 memset (&function_starts_load_command, 0, sizeof(function_starts_load_command)); 1721 break; 1722 1723 default: 1724 break; 1725 } 1726 offset = cmd_offset + lc.cmdsize; 1727 } 1728 1729 if (symtab_load_command.cmd) 1730 { 1731 Symtab *symtab = m_symtab_ap.get(); 1732 SectionList *section_list = GetSectionList(); 1733 if (section_list == NULL) 1734 return 0; 1735 1736 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 1737 const ByteOrder byte_order = m_data.GetByteOrder(); 1738 bool bit_width_32 = addr_byte_size == 4; 1739 const size_t nlist_byte_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 1740 1741 DataExtractor nlist_data (NULL, 0, byte_order, addr_byte_size); 1742 DataExtractor strtab_data (NULL, 0, byte_order, addr_byte_size); 1743 DataExtractor function_starts_data (NULL, 0, byte_order, addr_byte_size); 1744 DataExtractor indirect_symbol_index_data (NULL, 0, byte_order, addr_byte_size); 1745 DataExtractor dyld_trie_data (NULL, 0, byte_order, addr_byte_size); 1746 1747 const addr_t nlist_data_byte_size = symtab_load_command.nsyms * nlist_byte_size; 1748 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 1749 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 1750 1751 ProcessSP process_sp (m_process_wp.lock()); 1752 Process *process = process_sp.get(); 1753 1754 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 1755 1756 if (process) 1757 { 1758 Target &target = process->GetTarget(); 1759 1760 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 1761 1762 SectionSP linkedit_section_sp(section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 1763 // Reading mach file from memory in a process or core file... 1764 1765 if (linkedit_section_sp) 1766 { 1767 const addr_t linkedit_load_addr = linkedit_section_sp->GetLoadBaseAddress(&target); 1768 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 1769 const addr_t symoff_addr = linkedit_load_addr + symtab_load_command.symoff - linkedit_file_offset; 1770 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - linkedit_file_offset; 1771 1772 bool data_was_read = false; 1773 1774 #if defined (__APPLE__) && defined (__arm__) 1775 if (m_header.flags & 0x80000000u) 1776 { 1777 // This mach-o memory file is in the dyld shared cache. If this 1778 // program is not remote and this is iOS, then this process will 1779 // share the same shared cache as the process we are debugging and 1780 // we can read the entire __LINKEDIT from the address space in this 1781 // process. This is a needed optimization that is used for local iOS 1782 // debugging only since all shared libraries in the shared cache do 1783 // not have corresponding files that exist in the file system of the 1784 // device. They have been combined into a single file. This means we 1785 // always have to load these files from memory. All of the symbol and 1786 // string tables from all of the __LINKEDIT sections from the shared 1787 // libraries in the shared cache have been merged into a single large 1788 // symbol and string table. Reading all of this symbol and string table 1789 // data across can slow down debug launch times, so we optimize this by 1790 // reading the memory for the __LINKEDIT section from this process. 1791 1792 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 1793 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 1794 bool use_lldb_cache = true; 1795 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && lldb_shared_cache != process_shared_cache) 1796 { 1797 use_lldb_cache = false; 1798 ModuleSP module_sp (GetModule()); 1799 if (module_sp) 1800 module_sp->ReportWarning ("shared cache in process does not match lldb's own shared cache, startup will be slow."); 1801 1802 } 1803 1804 PlatformSP platform_sp (target.GetPlatform()); 1805 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) 1806 { 1807 data_was_read = true; 1808 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, eByteOrderLittle); 1809 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, eByteOrderLittle); 1810 if (function_starts_load_command.cmd) 1811 { 1812 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 1813 function_starts_data.SetData ((void *)func_start_addr, function_starts_load_command.datasize, eByteOrderLittle); 1814 } 1815 } 1816 } 1817 #endif 1818 1819 if (!data_was_read) 1820 { 1821 if (memory_module_load_level == eMemoryModuleLoadLevelComplete) 1822 { 1823 DataBufferSP nlist_data_sp (ReadMemory (process_sp, symoff_addr, nlist_data_byte_size)); 1824 if (nlist_data_sp) 1825 nlist_data.SetData (nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 1826 // Load strings individually from memory when loading from memory since shared cache 1827 // string tables contain strings for all symbols from all shared cached libraries 1828 //DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, strtab_data_byte_size)); 1829 //if (strtab_data_sp) 1830 // strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 1831 if (m_dysymtab.nindirectsyms != 0) 1832 { 1833 const addr_t indirect_syms_addr = linkedit_load_addr + m_dysymtab.indirectsymoff - linkedit_file_offset; 1834 DataBufferSP indirect_syms_data_sp (ReadMemory (process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 1835 if (indirect_syms_data_sp) 1836 indirect_symbol_index_data.SetData (indirect_syms_data_sp, 0, indirect_syms_data_sp->GetByteSize()); 1837 } 1838 } 1839 1840 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) 1841 { 1842 if (function_starts_load_command.cmd) 1843 { 1844 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 1845 DataBufferSP func_start_data_sp (ReadMemory (process_sp, func_start_addr, function_starts_load_command.datasize)); 1846 if (func_start_data_sp) 1847 function_starts_data.SetData (func_start_data_sp, 0, func_start_data_sp->GetByteSize()); 1848 } 1849 } 1850 } 1851 } 1852 } 1853 else 1854 { 1855 nlist_data.SetData (m_data, 1856 symtab_load_command.symoff, 1857 nlist_data_byte_size); 1858 strtab_data.SetData (m_data, 1859 symtab_load_command.stroff, 1860 strtab_data_byte_size); 1861 1862 if (dyld_info.export_size > 0) 1863 { 1864 dyld_trie_data.SetData (m_data, 1865 dyld_info.export_off, 1866 dyld_info.export_size); 1867 } 1868 1869 if (m_dysymtab.nindirectsyms != 0) 1870 { 1871 indirect_symbol_index_data.SetData (m_data, 1872 m_dysymtab.indirectsymoff, 1873 m_dysymtab.nindirectsyms * 4); 1874 } 1875 if (function_starts_load_command.cmd) 1876 { 1877 function_starts_data.SetData (m_data, 1878 function_starts_load_command.dataoff, 1879 function_starts_load_command.datasize); 1880 } 1881 } 1882 1883 if (nlist_data.GetByteSize() == 0 && memory_module_load_level == eMemoryModuleLoadLevelComplete) 1884 { 1885 if (log) 1886 module_sp->LogMessage(log, "failed to read nlist data"); 1887 return 0; 1888 } 1889 1890 1891 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 1892 if (!have_strtab_data) 1893 { 1894 if (process) 1895 { 1896 if (strtab_addr == LLDB_INVALID_ADDRESS) 1897 { 1898 if (log) 1899 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 1900 return 0; 1901 } 1902 } 1903 else 1904 { 1905 if (log) 1906 module_sp->LogMessage(log, "failed to read strtab data"); 1907 return 0; 1908 } 1909 } 1910 1911 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 1912 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 1913 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 1914 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 1915 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT)); 1916 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA)); 1917 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC)); 1918 SectionSP eh_frame_section_sp; 1919 if (text_section_sp.get()) 1920 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame); 1921 else 1922 eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame); 1923 1924 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 1925 1926 // lldb works best if it knows the start addresss of all functions in a module. 1927 // Linker symbols or debug info are normally the best source of information for start addr / size but 1928 // they may be stripped in a released binary. 1929 // Two additional sources of information exist in Mach-O binaries: 1930 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each function's start address in the 1931 // binary, relative to the text section. 1932 // eh_frame - the eh_frame FDEs have the start addr & size of each function 1933 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on all modern binaries. 1934 // Binaries built to run on older releases may need to use eh_frame information. 1935 1936 if (text_section_sp && function_starts_data.GetByteSize()) 1937 { 1938 FunctionStarts::Entry function_start_entry; 1939 function_start_entry.data = false; 1940 lldb::offset_t function_start_offset = 0; 1941 function_start_entry.addr = text_section_sp->GetFileAddress(); 1942 uint64_t delta; 1943 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 0) 1944 { 1945 // Now append the current entry 1946 function_start_entry.addr += delta; 1947 function_starts.Append(function_start_entry); 1948 } 1949 } 1950 else 1951 { 1952 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the load command claiming an eh_frame 1953 // but it doesn't actually have the eh_frame content. And if we have a dSYM, we don't need to do any 1954 // of this fill-in-the-missing-symbols works anyway - the debug info should give us all the functions in 1955 // the module. 1956 if (text_section_sp.get() && eh_frame_section_sp.get() && m_type != eTypeDebugInfo) 1957 { 1958 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, eRegisterKindGCC, true); 1959 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 1960 eh_frame.GetFunctionAddressAndSizeVector (functions); 1961 addr_t text_base_addr = text_section_sp->GetFileAddress(); 1962 size_t count = functions.GetSize(); 1963 for (size_t i = 0; i < count; ++i) 1964 { 1965 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = functions.GetEntryAtIndex (i); 1966 if (func) 1967 { 1968 FunctionStarts::Entry function_start_entry; 1969 function_start_entry.addr = func->base - text_base_addr; 1970 function_starts.Append(function_start_entry); 1971 } 1972 } 1973 } 1974 } 1975 1976 const size_t function_starts_count = function_starts.GetSize(); 1977 1978 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() : NO_SECT; 1979 1980 lldb::offset_t nlist_data_offset = 0; 1981 1982 uint32_t N_SO_index = UINT32_MAX; 1983 1984 MachSymtabSectionInfo section_info (section_list); 1985 std::vector<uint32_t> N_FUN_indexes; 1986 std::vector<uint32_t> N_NSYM_indexes; 1987 std::vector<uint32_t> N_INCL_indexes; 1988 std::vector<uint32_t> N_BRAC_indexes; 1989 std::vector<uint32_t> N_COMM_indexes; 1990 typedef std::multimap <uint64_t, uint32_t> ValueToSymbolIndexMap; 1991 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap; 1992 typedef std::map <const char *, uint32_t> ConstNameToSymbolIndexMap; 1993 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 1994 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 1995 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 1996 // Any symbols that get merged into another will get an entry 1997 // in this map so we know 1998 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 1999 uint32_t nlist_idx = 0; 2000 Symbol *symbol_ptr = NULL; 2001 2002 uint32_t sym_idx = 0; 2003 Symbol *sym = NULL; 2004 size_t num_syms = 0; 2005 std::string memory_symbol_name; 2006 uint32_t unmapped_local_symbols_found = 0; 2007 2008 std::vector<TrieEntryWithOffset> trie_entries; 2009 std::set<lldb::addr_t> resolver_addresses; 2010 2011 if (dyld_trie_data.GetByteSize() > 0) 2012 { 2013 std::vector<llvm::StringRef> nameSlices; 2014 ParseTrieEntries (dyld_trie_data, 2015 0, 2016 nameSlices, 2017 resolver_addresses, 2018 trie_entries); 2019 2020 ConstString text_segment_name ("__TEXT"); 2021 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 2022 if (text_segment_sp) 2023 { 2024 const lldb::addr_t text_segment_file_addr = text_segment_sp->GetFileAddress(); 2025 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) 2026 { 2027 for (auto &e : trie_entries) 2028 e.entry.address += text_segment_file_addr; 2029 } 2030 } 2031 } 2032 2033 #if defined (__APPLE__) && defined (__arm__) 2034 2035 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been optimized by moving LOCAL 2036 // symbols out of the memory mapped portion of the DSC. The symbol information has all been retained, 2037 // but it isn't available in the normal nlist data. However, there *are* duplicate entries of *some* 2038 // LOCAL symbols in the normal nlist data. To handle this situation correctly, we must first attempt 2039 // to parse any DSC unmapped symbol information. If we find any, we set a flag that tells the normal 2040 // nlist parser to ignore all LOCAL symbols. 2041 2042 if (m_header.flags & 0x80000000u) 2043 { 2044 // Before we can start mapping the DSC, we need to make certain the target process is actually 2045 // using the cache we can find. 2046 2047 // Next we need to determine the correct path for the dyld shared cache. 2048 2049 ArchSpec header_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 2050 char dsc_path[PATH_MAX]; 2051 2052 snprintf(dsc_path, sizeof(dsc_path), "%s%s%s", 2053 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR */ 2054 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2055 header_arch.GetArchitectureName()); 2056 2057 FileSpec dsc_filespec(dsc_path, false); 2058 2059 // We need definitions of two structures in the on-disk DSC, copy them here manually 2060 struct lldb_copy_dyld_cache_header_v0 2061 { 2062 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 2063 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 2064 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 2065 uint32_t imagesOffset; 2066 uint32_t imagesCount; 2067 uint64_t dyldBaseAddress; 2068 uint64_t codeSignatureOffset; 2069 uint64_t codeSignatureSize; 2070 uint64_t slideInfoOffset; 2071 uint64_t slideInfoSize; 2072 uint64_t localSymbolsOffset; // file offset of where local symbols are stored 2073 uint64_t localSymbolsSize; // size of local symbols information 2074 }; 2075 struct lldb_copy_dyld_cache_header_v1 2076 { 2077 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 2078 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 2079 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 2080 uint32_t imagesOffset; 2081 uint32_t imagesCount; 2082 uint64_t dyldBaseAddress; 2083 uint64_t codeSignatureOffset; 2084 uint64_t codeSignatureSize; 2085 uint64_t slideInfoOffset; 2086 uint64_t slideInfoSize; 2087 uint64_t localSymbolsOffset; 2088 uint64_t localSymbolsSize; 2089 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 and later 2090 }; 2091 2092 struct lldb_copy_dyld_cache_mapping_info 2093 { 2094 uint64_t address; 2095 uint64_t size; 2096 uint64_t fileOffset; 2097 uint32_t maxProt; 2098 uint32_t initProt; 2099 }; 2100 2101 struct lldb_copy_dyld_cache_local_symbols_info 2102 { 2103 uint32_t nlistOffset; 2104 uint32_t nlistCount; 2105 uint32_t stringsOffset; 2106 uint32_t stringsSize; 2107 uint32_t entriesOffset; 2108 uint32_t entriesCount; 2109 }; 2110 struct lldb_copy_dyld_cache_local_symbols_entry 2111 { 2112 uint32_t dylibOffset; 2113 uint32_t nlistStartIndex; 2114 uint32_t nlistCount; 2115 }; 2116 2117 /* The dyld_cache_header has a pointer to the dyld_cache_local_symbols_info structure (localSymbolsOffset). 2118 The dyld_cache_local_symbols_info structure gives us three things: 2119 1. The start and count of the nlist records in the dyld_shared_cache file 2120 2. The start and size of the strings for these nlist records 2121 3. The start and count of dyld_cache_local_symbols_entry entries 2122 2123 There is one dyld_cache_local_symbols_entry per dylib/framework in the dyld shared cache. 2124 The "dylibOffset" field is the Mach-O header of this dylib/framework in the dyld shared cache. 2125 The dyld_cache_local_symbols_entry also lists the start of this dylib/framework's nlist records 2126 and the count of how many nlist records there are for this dylib/framework. 2127 */ 2128 2129 // Process the dsc header to find the unmapped symbols 2130 // 2131 // Save some VM space, do not map the entire cache in one shot. 2132 2133 DataBufferSP dsc_data_sp; 2134 dsc_data_sp = dsc_filespec.MemoryMapFileContents(0, sizeof(struct lldb_copy_dyld_cache_header_v1)); 2135 2136 if (dsc_data_sp) 2137 { 2138 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2139 2140 char version_str[17]; 2141 int version = -1; 2142 lldb::offset_t offset = 0; 2143 memcpy (version_str, dsc_header_data.GetData (&offset, 16), 16); 2144 version_str[16] = '\0'; 2145 if (strncmp (version_str, "dyld_v", 6) == 0 && isdigit (version_str[6])) 2146 { 2147 int v; 2148 if (::sscanf (version_str + 6, "%d", &v) == 1) 2149 { 2150 version = v; 2151 } 2152 } 2153 2154 UUID dsc_uuid; 2155 if (version >= 1) 2156 { 2157 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, uuid); 2158 uint8_t uuid_bytes[sizeof (uuid_t)]; 2159 memcpy (uuid_bytes, dsc_header_data.GetData (&offset, sizeof (uuid_t)), sizeof (uuid_t)); 2160 dsc_uuid.SetBytes (uuid_bytes); 2161 } 2162 2163 bool uuid_match = true; 2164 if (dsc_uuid.IsValid() && process) 2165 { 2166 UUID shared_cache_uuid(GetProcessSharedCacheUUID(process)); 2167 2168 if (shared_cache_uuid.IsValid() && dsc_uuid != shared_cache_uuid) 2169 { 2170 // The on-disk dyld_shared_cache file is not the same as the one in this 2171 // process' memory, don't use it. 2172 uuid_match = false; 2173 ModuleSP module_sp (GetModule()); 2174 if (module_sp) 2175 module_sp->ReportWarning ("process shared cache does not match on-disk dyld_shared_cache file, some symbol names will be missing."); 2176 } 2177 } 2178 2179 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2180 2181 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2182 2183 // If the mappingOffset points to a location inside the header, we've 2184 // opened an old dyld shared cache, and should not proceed further. 2185 if (uuid_match && mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v0)) 2186 { 2187 2188 DataBufferSP dsc_mapping_info_data_sp = dsc_filespec.MemoryMapFileContents(mappingOffset, sizeof (struct lldb_copy_dyld_cache_mapping_info)); 2189 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, byte_order, addr_byte_size); 2190 offset = 0; 2191 2192 // The File addresses (from the in-memory Mach-O load commands) for the shared libraries 2193 // in the shared library cache need to be adjusted by an offset to match up with the 2194 // dylibOffset identifying field in the dyld_cache_local_symbol_entry's. This offset is 2195 // recorded in mapping_offset_value. 2196 const uint64_t mapping_offset_value = dsc_mapping_info_data.GetU64(&offset); 2197 2198 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2199 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2200 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2201 2202 if (localSymbolsOffset && localSymbolsSize) 2203 { 2204 // Map the local symbols 2205 if (DataBufferSP dsc_local_symbols_data_sp = dsc_filespec.MemoryMapFileContents(localSymbolsOffset, localSymbolsSize)) 2206 { 2207 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, byte_order, addr_byte_size); 2208 2209 offset = 0; 2210 2211 // Read the local_symbols_infos struct in one shot 2212 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2213 dsc_local_symbols_data.GetU32(&offset, &local_symbols_info.nlistOffset, 6); 2214 2215 SectionSP text_section_sp(section_list->FindSectionByName(GetSegmentNameTEXT())); 2216 2217 uint32_t header_file_offset = (text_section_sp->GetFileAddress() - mapping_offset_value); 2218 2219 offset = local_symbols_info.entriesOffset; 2220 for (uint32_t entry_index = 0; entry_index < local_symbols_info.entriesCount; entry_index++) 2221 { 2222 struct lldb_copy_dyld_cache_local_symbols_entry local_symbols_entry; 2223 local_symbols_entry.dylibOffset = dsc_local_symbols_data.GetU32(&offset); 2224 local_symbols_entry.nlistStartIndex = dsc_local_symbols_data.GetU32(&offset); 2225 local_symbols_entry.nlistCount = dsc_local_symbols_data.GetU32(&offset); 2226 2227 if (header_file_offset == local_symbols_entry.dylibOffset) 2228 { 2229 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2230 2231 // The normal nlist code cannot correctly size the Symbols array, we need to allocate it here. 2232 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms + unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2233 num_syms = symtab->GetNumSymbols(); 2234 2235 nlist_data_offset = local_symbols_info.nlistOffset + (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2236 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2237 2238 for (uint32_t nlist_index = 0; nlist_index < local_symbols_entry.nlistCount; nlist_index++) 2239 { 2240 ///////////////////////////// 2241 { 2242 struct nlist_64 nlist; 2243 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2244 break; 2245 2246 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked(&nlist_data_offset); 2247 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2248 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2249 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked (&nlist_data_offset); 2250 nlist.n_value = dsc_local_symbols_data.GetAddress_unchecked (&nlist_data_offset); 2251 2252 SymbolType type = eSymbolTypeInvalid; 2253 const char *symbol_name = dsc_local_symbols_data.PeekCStr(string_table_offset + nlist.n_strx); 2254 2255 if (symbol_name == NULL) 2256 { 2257 // No symbol should be NULL, even the symbols with no 2258 // string values should have an offset zero which points 2259 // to an empty C-string 2260 Host::SystemLog (Host::eSystemLogError, 2261 "error: DSC unmapped local symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 2262 entry_index, 2263 nlist.n_strx, 2264 module_sp->GetFileSpec().GetPath().c_str()); 2265 continue; 2266 } 2267 if (symbol_name[0] == '\0') 2268 symbol_name = NULL; 2269 2270 const char *symbol_name_non_abi_mangled = NULL; 2271 2272 SectionSP symbol_section; 2273 uint32_t symbol_byte_size = 0; 2274 bool add_nlist = true; 2275 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2276 bool demangled_is_synthesized = false; 2277 bool is_gsym = false; 2278 2279 assert (sym_idx < num_syms); 2280 2281 sym[sym_idx].SetDebug (is_debug); 2282 2283 if (is_debug) 2284 { 2285 switch (nlist.n_type) 2286 { 2287 case N_GSYM: 2288 // global symbol: name,,NO_SECT,type,0 2289 // Sometimes the N_GSYM value contains the address. 2290 2291 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 2292 // have the same address, but we want to ensure that we always find only the real symbol, 2293 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 2294 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 2295 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 2296 // same address. 2297 2298 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 2299 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 2300 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 2301 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 2302 add_nlist = false; 2303 else 2304 { 2305 is_gsym = true; 2306 sym[sym_idx].SetExternal(true); 2307 if (nlist.n_value != 0) 2308 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2309 type = eSymbolTypeData; 2310 } 2311 break; 2312 2313 case N_FNAME: 2314 // procedure name (f77 kludge): name,,NO_SECT,0,0 2315 type = eSymbolTypeCompiler; 2316 break; 2317 2318 case N_FUN: 2319 // procedure: name,,n_sect,linenumber,address 2320 if (symbol_name) 2321 { 2322 type = eSymbolTypeCode; 2323 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2324 2325 N_FUN_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 2326 // We use the current number of symbols in the symbol table in lieu of 2327 // using nlist_idx in case we ever start trimming entries out 2328 N_FUN_indexes.push_back(sym_idx); 2329 } 2330 else 2331 { 2332 type = eSymbolTypeCompiler; 2333 2334 if ( !N_FUN_indexes.empty() ) 2335 { 2336 // Copy the size of the function into the original STAB entry so we don't have 2337 // to hunt for it later 2338 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 2339 N_FUN_indexes.pop_back(); 2340 // We don't really need the end function STAB as it contains the size which 2341 // we already placed with the original symbol, so don't add it if we want a 2342 // minimal symbol table 2343 add_nlist = false; 2344 } 2345 } 2346 break; 2347 2348 case N_STSYM: 2349 // static symbol: name,,n_sect,type,address 2350 N_STSYM_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 2351 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2352 type = eSymbolTypeData; 2353 break; 2354 2355 case N_LCSYM: 2356 // .lcomm symbol: name,,n_sect,type,address 2357 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2358 type = eSymbolTypeCommonBlock; 2359 break; 2360 2361 case N_BNSYM: 2362 // We use the current number of symbols in the symbol table in lieu of 2363 // using nlist_idx in case we ever start trimming entries out 2364 // Skip these if we want minimal symbol tables 2365 add_nlist = false; 2366 break; 2367 2368 case N_ENSYM: 2369 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 2370 // so that we can always skip the entire symbol if we need to navigate 2371 // more quickly at the source level when parsing STABS 2372 // Skip these if we want minimal symbol tables 2373 add_nlist = false; 2374 break; 2375 2376 2377 case N_OPT: 2378 // emitted with gcc2_compiled and in gcc source 2379 type = eSymbolTypeCompiler; 2380 break; 2381 2382 case N_RSYM: 2383 // register sym: name,,NO_SECT,type,register 2384 type = eSymbolTypeVariable; 2385 break; 2386 2387 case N_SLINE: 2388 // src line: 0,,n_sect,linenumber,address 2389 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2390 type = eSymbolTypeLineEntry; 2391 break; 2392 2393 case N_SSYM: 2394 // structure elt: name,,NO_SECT,type,struct_offset 2395 type = eSymbolTypeVariableType; 2396 break; 2397 2398 case N_SO: 2399 // source file name 2400 type = eSymbolTypeSourceFile; 2401 if (symbol_name == NULL) 2402 { 2403 add_nlist = false; 2404 if (N_SO_index != UINT32_MAX) 2405 { 2406 // Set the size of the N_SO to the terminating index of this N_SO 2407 // so that we can always skip the entire N_SO if we need to navigate 2408 // more quickly at the source level when parsing STABS 2409 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2410 symbol_ptr->SetByteSize(sym_idx); 2411 symbol_ptr->SetSizeIsSibling(true); 2412 } 2413 N_NSYM_indexes.clear(); 2414 N_INCL_indexes.clear(); 2415 N_BRAC_indexes.clear(); 2416 N_COMM_indexes.clear(); 2417 N_FUN_indexes.clear(); 2418 N_SO_index = UINT32_MAX; 2419 } 2420 else 2421 { 2422 // We use the current number of symbols in the symbol table in lieu of 2423 // using nlist_idx in case we ever start trimming entries out 2424 const bool N_SO_has_full_path = symbol_name[0] == '/'; 2425 if (N_SO_has_full_path) 2426 { 2427 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2428 { 2429 // We have two consecutive N_SO entries where the first contains a directory 2430 // and the second contains a full path. 2431 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 2432 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2433 add_nlist = false; 2434 } 2435 else 2436 { 2437 // This is the first entry in a N_SO that contains a directory or 2438 // a full path to the source file 2439 N_SO_index = sym_idx; 2440 } 2441 } 2442 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2443 { 2444 // This is usually the second N_SO entry that contains just the filename, 2445 // so here we combine it with the first one if we are minimizing the symbol table 2446 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 2447 if (so_path && so_path[0]) 2448 { 2449 std::string full_so_path (so_path); 2450 const size_t double_slash_pos = full_so_path.find("//"); 2451 if (double_slash_pos != std::string::npos) 2452 { 2453 // The linker has been generating bad N_SO entries with doubled up paths 2454 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 2455 // and the second is the directory for the source file so you end up with 2456 // a path that looks like "/tmp/src//tmp/src/" 2457 FileSpec so_dir(so_path, false); 2458 if (!so_dir.Exists()) 2459 { 2460 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 2461 if (so_dir.Exists()) 2462 { 2463 // Trim off the incorrect path 2464 full_so_path.erase(0, double_slash_pos + 1); 2465 } 2466 } 2467 } 2468 if (*full_so_path.rbegin() != '/') 2469 full_so_path += '/'; 2470 full_so_path += symbol_name; 2471 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 2472 add_nlist = false; 2473 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2474 } 2475 } 2476 else 2477 { 2478 // This could be a relative path to a N_SO 2479 N_SO_index = sym_idx; 2480 } 2481 } 2482 break; 2483 2484 case N_OSO: 2485 // object file name: name,,0,0,st_mtime 2486 type = eSymbolTypeObjectFile; 2487 break; 2488 2489 case N_LSYM: 2490 // local sym: name,,NO_SECT,type,offset 2491 type = eSymbolTypeLocal; 2492 break; 2493 2494 //---------------------------------------------------------------------- 2495 // INCL scopes 2496 //---------------------------------------------------------------------- 2497 case N_BINCL: 2498 // include file beginning: name,,NO_SECT,0,sum 2499 // We use the current number of symbols in the symbol table in lieu of 2500 // using nlist_idx in case we ever start trimming entries out 2501 N_INCL_indexes.push_back(sym_idx); 2502 type = eSymbolTypeScopeBegin; 2503 break; 2504 2505 case N_EINCL: 2506 // include file end: name,,NO_SECT,0,0 2507 // Set the size of the N_BINCL to the terminating index of this N_EINCL 2508 // so that we can always skip the entire symbol if we need to navigate 2509 // more quickly at the source level when parsing STABS 2510 if ( !N_INCL_indexes.empty() ) 2511 { 2512 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 2513 symbol_ptr->SetByteSize(sym_idx + 1); 2514 symbol_ptr->SetSizeIsSibling(true); 2515 N_INCL_indexes.pop_back(); 2516 } 2517 type = eSymbolTypeScopeEnd; 2518 break; 2519 2520 case N_SOL: 2521 // #included file name: name,,n_sect,0,address 2522 type = eSymbolTypeHeaderFile; 2523 2524 // We currently don't use the header files on darwin 2525 add_nlist = false; 2526 break; 2527 2528 case N_PARAMS: 2529 // compiler parameters: name,,NO_SECT,0,0 2530 type = eSymbolTypeCompiler; 2531 break; 2532 2533 case N_VERSION: 2534 // compiler version: name,,NO_SECT,0,0 2535 type = eSymbolTypeCompiler; 2536 break; 2537 2538 case N_OLEVEL: 2539 // compiler -O level: name,,NO_SECT,0,0 2540 type = eSymbolTypeCompiler; 2541 break; 2542 2543 case N_PSYM: 2544 // parameter: name,,NO_SECT,type,offset 2545 type = eSymbolTypeVariable; 2546 break; 2547 2548 case N_ENTRY: 2549 // alternate entry: name,,n_sect,linenumber,address 2550 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2551 type = eSymbolTypeLineEntry; 2552 break; 2553 2554 //---------------------------------------------------------------------- 2555 // Left and Right Braces 2556 //---------------------------------------------------------------------- 2557 case N_LBRAC: 2558 // left bracket: 0,,NO_SECT,nesting level,address 2559 // We use the current number of symbols in the symbol table in lieu of 2560 // using nlist_idx in case we ever start trimming entries out 2561 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2562 N_BRAC_indexes.push_back(sym_idx); 2563 type = eSymbolTypeScopeBegin; 2564 break; 2565 2566 case N_RBRAC: 2567 // right bracket: 0,,NO_SECT,nesting level,address 2568 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 2569 // so that we can always skip the entire symbol if we need to navigate 2570 // more quickly at the source level when parsing STABS 2571 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2572 if ( !N_BRAC_indexes.empty() ) 2573 { 2574 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 2575 symbol_ptr->SetByteSize(sym_idx + 1); 2576 symbol_ptr->SetSizeIsSibling(true); 2577 N_BRAC_indexes.pop_back(); 2578 } 2579 type = eSymbolTypeScopeEnd; 2580 break; 2581 2582 case N_EXCL: 2583 // deleted include file: name,,NO_SECT,0,sum 2584 type = eSymbolTypeHeaderFile; 2585 break; 2586 2587 //---------------------------------------------------------------------- 2588 // COMM scopes 2589 //---------------------------------------------------------------------- 2590 case N_BCOMM: 2591 // begin common: name,,NO_SECT,0,0 2592 // We use the current number of symbols in the symbol table in lieu of 2593 // using nlist_idx in case we ever start trimming entries out 2594 type = eSymbolTypeScopeBegin; 2595 N_COMM_indexes.push_back(sym_idx); 2596 break; 2597 2598 case N_ECOML: 2599 // end common (local name): 0,,n_sect,0,address 2600 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2601 // Fall through 2602 2603 case N_ECOMM: 2604 // end common: name,,n_sect,0,0 2605 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 2606 // so that we can always skip the entire symbol if we need to navigate 2607 // more quickly at the source level when parsing STABS 2608 if ( !N_COMM_indexes.empty() ) 2609 { 2610 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 2611 symbol_ptr->SetByteSize(sym_idx + 1); 2612 symbol_ptr->SetSizeIsSibling(true); 2613 N_COMM_indexes.pop_back(); 2614 } 2615 type = eSymbolTypeScopeEnd; 2616 break; 2617 2618 case N_LENG: 2619 // second stab entry with length information 2620 type = eSymbolTypeAdditional; 2621 break; 2622 2623 default: break; 2624 } 2625 } 2626 else 2627 { 2628 //uint8_t n_pext = N_PEXT & nlist.n_type; 2629 uint8_t n_type = N_TYPE & nlist.n_type; 2630 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 2631 2632 switch (n_type) 2633 { 2634 case N_INDR: // Fall through 2635 case N_PBUD: // Fall through 2636 case N_UNDF: 2637 type = eSymbolTypeUndefined; 2638 break; 2639 2640 case N_ABS: 2641 type = eSymbolTypeAbsolute; 2642 break; 2643 2644 case N_SECT: 2645 { 2646 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2647 2648 if (symbol_section == NULL) 2649 { 2650 // TODO: warn about this? 2651 add_nlist = false; 2652 break; 2653 } 2654 2655 if (TEXT_eh_frame_sectID == nlist.n_sect) 2656 { 2657 type = eSymbolTypeException; 2658 } 2659 else 2660 { 2661 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 2662 2663 switch (section_type) 2664 { 2665 case S_REGULAR: break; // regular section 2666 //case S_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section 2667 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 2668 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 2669 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 2670 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 2671 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 2672 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 2673 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 2674 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 2675 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 2676 //case S_COALESCED: type = eSymbolType; break; // section contains symbols that are to be coalesced 2677 //case S_GB_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 2678 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 2679 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 2680 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 2681 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 2682 default: break; 2683 } 2684 2685 if (type == eSymbolTypeInvalid) 2686 { 2687 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 2688 if (symbol_section->IsDescendant (text_section_sp.get())) 2689 { 2690 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 2691 S_ATTR_SELF_MODIFYING_CODE | 2692 S_ATTR_SOME_INSTRUCTIONS)) 2693 type = eSymbolTypeData; 2694 else 2695 type = eSymbolTypeCode; 2696 } 2697 else if (symbol_section->IsDescendant(data_section_sp.get())) 2698 { 2699 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 2700 { 2701 type = eSymbolTypeRuntime; 2702 2703 if (symbol_name && 2704 symbol_name[0] == '_' && 2705 symbol_name[1] == 'O' && 2706 symbol_name[2] == 'B') 2707 { 2708 llvm::StringRef symbol_name_ref(symbol_name); 2709 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 2710 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 2711 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 2712 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 2713 { 2714 symbol_name_non_abi_mangled = symbol_name + 1; 2715 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2716 type = eSymbolTypeObjCClass; 2717 demangled_is_synthesized = true; 2718 } 2719 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 2720 { 2721 symbol_name_non_abi_mangled = symbol_name + 1; 2722 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2723 type = eSymbolTypeObjCMetaClass; 2724 demangled_is_synthesized = true; 2725 } 2726 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 2727 { 2728 symbol_name_non_abi_mangled = symbol_name + 1; 2729 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2730 type = eSymbolTypeObjCIVar; 2731 demangled_is_synthesized = true; 2732 } 2733 } 2734 } 2735 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 2736 { 2737 type = eSymbolTypeException; 2738 } 2739 else 2740 { 2741 type = eSymbolTypeData; 2742 } 2743 } 2744 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 2745 { 2746 type = eSymbolTypeTrampoline; 2747 } 2748 else if (symbol_section->IsDescendant(objc_section_sp.get())) 2749 { 2750 type = eSymbolTypeRuntime; 2751 if (symbol_name && symbol_name[0] == '.') 2752 { 2753 llvm::StringRef symbol_name_ref(symbol_name); 2754 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 2755 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 2756 { 2757 symbol_name_non_abi_mangled = symbol_name; 2758 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 2759 type = eSymbolTypeObjCClass; 2760 demangled_is_synthesized = true; 2761 } 2762 } 2763 } 2764 } 2765 } 2766 } 2767 break; 2768 } 2769 } 2770 2771 if (add_nlist) 2772 { 2773 uint64_t symbol_value = nlist.n_value; 2774 if (symbol_name_non_abi_mangled) 2775 { 2776 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 2777 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 2778 } 2779 else 2780 { 2781 bool symbol_name_is_mangled = false; 2782 2783 if (symbol_name && symbol_name[0] == '_') 2784 { 2785 symbol_name_is_mangled = symbol_name[1] == '_'; 2786 symbol_name++; // Skip the leading underscore 2787 } 2788 2789 if (symbol_name) 2790 { 2791 ConstString const_symbol_name(symbol_name); 2792 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 2793 if (is_gsym && is_debug) 2794 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 2795 } 2796 } 2797 if (symbol_section) 2798 { 2799 const addr_t section_file_addr = symbol_section->GetFileAddress(); 2800 if (symbol_byte_size == 0 && function_starts_count > 0) 2801 { 2802 addr_t symbol_lookup_file_addr = nlist.n_value; 2803 // Do an exact address match for non-ARM addresses, else get the closest since 2804 // the symbol might be a thumb symbol which has an address with bit zero set 2805 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 2806 if (is_arm && func_start_entry) 2807 { 2808 // Verify that the function start address is the symbol address (ARM) 2809 // or the symbol address + 1 (thumb) 2810 if (func_start_entry->addr != symbol_lookup_file_addr && 2811 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 2812 { 2813 // Not the right entry, NULL it out... 2814 func_start_entry = NULL; 2815 } 2816 } 2817 if (func_start_entry) 2818 { 2819 func_start_entry->data = true; 2820 2821 addr_t symbol_file_addr = func_start_entry->addr; 2822 uint32_t symbol_flags = 0; 2823 if (is_arm) 2824 { 2825 if (symbol_file_addr & 1) 2826 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 2827 symbol_file_addr &= 0xfffffffffffffffeull; 2828 } 2829 2830 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 2831 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 2832 if (next_func_start_entry) 2833 { 2834 addr_t next_symbol_file_addr = next_func_start_entry->addr; 2835 // Be sure the clear the Thumb address bit when we calculate the size 2836 // from the current and next address 2837 if (is_arm) 2838 next_symbol_file_addr &= 0xfffffffffffffffeull; 2839 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 2840 } 2841 else 2842 { 2843 symbol_byte_size = section_end_file_addr - symbol_file_addr; 2844 } 2845 } 2846 } 2847 symbol_value -= section_file_addr; 2848 } 2849 2850 if (is_debug == false) 2851 { 2852 if (type == eSymbolTypeCode) 2853 { 2854 // See if we can find a N_FUN entry for any code symbols. 2855 // If we do find a match, and the name matches, then we 2856 // can merge the two into just the function symbol to avoid 2857 // duplicate entries in the symbol table 2858 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 2859 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 2860 if (range.first != range.second) 2861 { 2862 bool found_it = false; 2863 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 2864 { 2865 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 2866 { 2867 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 2868 // We just need the flags from the linker symbol, so put these flags 2869 // into the N_FUN flags to avoid duplicate symbols in the symbol table 2870 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 2871 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2872 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 2873 sym[pos->second].SetType (eSymbolTypeResolver); 2874 sym[sym_idx].Clear(); 2875 found_it = true; 2876 break; 2877 } 2878 } 2879 if (found_it) 2880 continue; 2881 } 2882 else 2883 { 2884 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 2885 type = eSymbolTypeResolver; 2886 } 2887 } 2888 else if (type == eSymbolTypeData) 2889 { 2890 // See if we can find a N_STSYM entry for any data symbols. 2891 // If we do find a match, and the name matches, then we 2892 // can merge the two into just the Static symbol to avoid 2893 // duplicate entries in the symbol table 2894 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 2895 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 2896 if (range.first != range.second) 2897 { 2898 bool found_it = false; 2899 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 2900 { 2901 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 2902 { 2903 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 2904 // We just need the flags from the linker symbol, so put these flags 2905 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 2906 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 2907 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2908 sym[sym_idx].Clear(); 2909 found_it = true; 2910 break; 2911 } 2912 } 2913 if (found_it) 2914 continue; 2915 } 2916 else 2917 { 2918 // Combine N_GSYM stab entries with the non stab symbol 2919 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 2920 if (pos != N_GSYM_name_to_sym_idx.end()) 2921 { 2922 const uint32_t GSYM_sym_idx = pos->second; 2923 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 2924 // Copy the address, because often the N_GSYM address has an invalid address of zero 2925 // when the global is a common symbol 2926 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 2927 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 2928 // We just need the flags from the linker symbol, so put these flags 2929 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 2930 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2931 sym[sym_idx].Clear(); 2932 continue; 2933 } 2934 } 2935 } 2936 } 2937 2938 sym[sym_idx].SetID (nlist_idx); 2939 sym[sym_idx].SetType (type); 2940 sym[sym_idx].GetAddress().SetSection (symbol_section); 2941 sym[sym_idx].GetAddress().SetOffset (symbol_value); 2942 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2943 2944 if (symbol_byte_size > 0) 2945 sym[sym_idx].SetByteSize(symbol_byte_size); 2946 2947 if (demangled_is_synthesized) 2948 sym[sym_idx].SetDemangledNameIsSynthesized(true); 2949 ++sym_idx; 2950 } 2951 else 2952 { 2953 sym[sym_idx].Clear(); 2954 } 2955 2956 } 2957 ///////////////////////////// 2958 } 2959 break; // No more entries to consider 2960 } 2961 } 2962 } 2963 } 2964 } 2965 } 2966 } 2967 2968 // Must reset this in case it was mutated above! 2969 nlist_data_offset = 0; 2970 #endif 2971 2972 if (nlist_data.GetByteSize() > 0) 2973 { 2974 2975 // If the sym array was not created while parsing the DSC unmapped 2976 // symbols, create it now. 2977 if (sym == NULL) 2978 { 2979 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 2980 num_syms = symtab->GetNumSymbols(); 2981 } 2982 2983 if (unmapped_local_symbols_found) 2984 { 2985 assert(m_dysymtab.ilocalsym == 0); 2986 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 2987 nlist_idx = m_dysymtab.nlocalsym; 2988 } 2989 else 2990 { 2991 nlist_idx = 0; 2992 } 2993 2994 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) 2995 { 2996 struct nlist_64 nlist; 2997 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2998 break; 2999 3000 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3001 nlist.n_type = nlist_data.GetU8_unchecked (&nlist_data_offset); 3002 nlist.n_sect = nlist_data.GetU8_unchecked (&nlist_data_offset); 3003 nlist.n_desc = nlist_data.GetU16_unchecked (&nlist_data_offset); 3004 nlist.n_value = nlist_data.GetAddress_unchecked (&nlist_data_offset); 3005 3006 SymbolType type = eSymbolTypeInvalid; 3007 const char *symbol_name = NULL; 3008 3009 if (have_strtab_data) 3010 { 3011 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3012 3013 if (symbol_name == NULL) 3014 { 3015 // No symbol should be NULL, even the symbols with no 3016 // string values should have an offset zero which points 3017 // to an empty C-string 3018 Host::SystemLog (Host::eSystemLogError, 3019 "error: symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 3020 nlist_idx, 3021 nlist.n_strx, 3022 module_sp->GetFileSpec().GetPath().c_str()); 3023 continue; 3024 } 3025 if (symbol_name[0] == '\0') 3026 symbol_name = NULL; 3027 } 3028 else 3029 { 3030 const addr_t str_addr = strtab_addr + nlist.n_strx; 3031 Error str_error; 3032 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, str_error)) 3033 symbol_name = memory_symbol_name.c_str(); 3034 } 3035 const char *symbol_name_non_abi_mangled = NULL; 3036 3037 SectionSP symbol_section; 3038 lldb::addr_t symbol_byte_size = 0; 3039 bool add_nlist = true; 3040 bool is_gsym = false; 3041 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3042 bool demangled_is_synthesized = false; 3043 3044 assert (sym_idx < num_syms); 3045 3046 sym[sym_idx].SetDebug (is_debug); 3047 3048 if (is_debug) 3049 { 3050 switch (nlist.n_type) 3051 { 3052 case N_GSYM: 3053 // global symbol: name,,NO_SECT,type,0 3054 // Sometimes the N_GSYM value contains the address. 3055 3056 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 3057 // have the same address, but we want to ensure that we always find only the real symbol, 3058 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 3059 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 3060 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 3061 // same address. 3062 3063 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 3064 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 3065 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 3066 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 3067 add_nlist = false; 3068 else 3069 { 3070 is_gsym = true; 3071 sym[sym_idx].SetExternal(true); 3072 if (nlist.n_value != 0) 3073 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3074 type = eSymbolTypeData; 3075 } 3076 break; 3077 3078 case N_FNAME: 3079 // procedure name (f77 kludge): name,,NO_SECT,0,0 3080 type = eSymbolTypeCompiler; 3081 break; 3082 3083 case N_FUN: 3084 // procedure: name,,n_sect,linenumber,address 3085 if (symbol_name) 3086 { 3087 type = eSymbolTypeCode; 3088 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3089 3090 N_FUN_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3091 // We use the current number of symbols in the symbol table in lieu of 3092 // using nlist_idx in case we ever start trimming entries out 3093 N_FUN_indexes.push_back(sym_idx); 3094 } 3095 else 3096 { 3097 type = eSymbolTypeCompiler; 3098 3099 if ( !N_FUN_indexes.empty() ) 3100 { 3101 // Copy the size of the function into the original STAB entry so we don't have 3102 // to hunt for it later 3103 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 3104 N_FUN_indexes.pop_back(); 3105 // We don't really need the end function STAB as it contains the size which 3106 // we already placed with the original symbol, so don't add it if we want a 3107 // minimal symbol table 3108 add_nlist = false; 3109 } 3110 } 3111 break; 3112 3113 case N_STSYM: 3114 // static symbol: name,,n_sect,type,address 3115 N_STSYM_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3116 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3117 type = eSymbolTypeData; 3118 break; 3119 3120 case N_LCSYM: 3121 // .lcomm symbol: name,,n_sect,type,address 3122 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3123 type = eSymbolTypeCommonBlock; 3124 break; 3125 3126 case N_BNSYM: 3127 // We use the current number of symbols in the symbol table in lieu of 3128 // using nlist_idx in case we ever start trimming entries out 3129 // Skip these if we want minimal symbol tables 3130 add_nlist = false; 3131 break; 3132 3133 case N_ENSYM: 3134 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 3135 // so that we can always skip the entire symbol if we need to navigate 3136 // more quickly at the source level when parsing STABS 3137 // Skip these if we want minimal symbol tables 3138 add_nlist = false; 3139 break; 3140 3141 3142 case N_OPT: 3143 // emitted with gcc2_compiled and in gcc source 3144 type = eSymbolTypeCompiler; 3145 break; 3146 3147 case N_RSYM: 3148 // register sym: name,,NO_SECT,type,register 3149 type = eSymbolTypeVariable; 3150 break; 3151 3152 case N_SLINE: 3153 // src line: 0,,n_sect,linenumber,address 3154 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3155 type = eSymbolTypeLineEntry; 3156 break; 3157 3158 case N_SSYM: 3159 // structure elt: name,,NO_SECT,type,struct_offset 3160 type = eSymbolTypeVariableType; 3161 break; 3162 3163 case N_SO: 3164 // source file name 3165 type = eSymbolTypeSourceFile; 3166 if (symbol_name == NULL) 3167 { 3168 add_nlist = false; 3169 if (N_SO_index != UINT32_MAX) 3170 { 3171 // Set the size of the N_SO to the terminating index of this N_SO 3172 // so that we can always skip the entire N_SO if we need to navigate 3173 // more quickly at the source level when parsing STABS 3174 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3175 symbol_ptr->SetByteSize(sym_idx); 3176 symbol_ptr->SetSizeIsSibling(true); 3177 } 3178 N_NSYM_indexes.clear(); 3179 N_INCL_indexes.clear(); 3180 N_BRAC_indexes.clear(); 3181 N_COMM_indexes.clear(); 3182 N_FUN_indexes.clear(); 3183 N_SO_index = UINT32_MAX; 3184 } 3185 else 3186 { 3187 // We use the current number of symbols in the symbol table in lieu of 3188 // using nlist_idx in case we ever start trimming entries out 3189 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3190 if (N_SO_has_full_path) 3191 { 3192 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3193 { 3194 // We have two consecutive N_SO entries where the first contains a directory 3195 // and the second contains a full path. 3196 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 3197 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3198 add_nlist = false; 3199 } 3200 else 3201 { 3202 // This is the first entry in a N_SO that contains a directory or 3203 // a full path to the source file 3204 N_SO_index = sym_idx; 3205 } 3206 } 3207 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3208 { 3209 // This is usually the second N_SO entry that contains just the filename, 3210 // so here we combine it with the first one if we are minimizing the symbol table 3211 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3212 if (so_path && so_path[0]) 3213 { 3214 std::string full_so_path (so_path); 3215 const size_t double_slash_pos = full_so_path.find("//"); 3216 if (double_slash_pos != std::string::npos) 3217 { 3218 // The linker has been generating bad N_SO entries with doubled up paths 3219 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 3220 // and the second is the directory for the source file so you end up with 3221 // a path that looks like "/tmp/src//tmp/src/" 3222 FileSpec so_dir(so_path, false); 3223 if (!so_dir.Exists()) 3224 { 3225 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 3226 if (so_dir.Exists()) 3227 { 3228 // Trim off the incorrect path 3229 full_so_path.erase(0, double_slash_pos + 1); 3230 } 3231 } 3232 } 3233 if (*full_so_path.rbegin() != '/') 3234 full_so_path += '/'; 3235 full_so_path += symbol_name; 3236 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 3237 add_nlist = false; 3238 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3239 } 3240 } 3241 else 3242 { 3243 // This could be a relative path to a N_SO 3244 N_SO_index = sym_idx; 3245 } 3246 } 3247 3248 break; 3249 3250 case N_OSO: 3251 // object file name: name,,0,0,st_mtime 3252 type = eSymbolTypeObjectFile; 3253 break; 3254 3255 case N_LSYM: 3256 // local sym: name,,NO_SECT,type,offset 3257 type = eSymbolTypeLocal; 3258 break; 3259 3260 //---------------------------------------------------------------------- 3261 // INCL scopes 3262 //---------------------------------------------------------------------- 3263 case N_BINCL: 3264 // include file beginning: name,,NO_SECT,0,sum 3265 // We use the current number of symbols in the symbol table in lieu of 3266 // using nlist_idx in case we ever start trimming entries out 3267 N_INCL_indexes.push_back(sym_idx); 3268 type = eSymbolTypeScopeBegin; 3269 break; 3270 3271 case N_EINCL: 3272 // include file end: name,,NO_SECT,0,0 3273 // Set the size of the N_BINCL to the terminating index of this N_EINCL 3274 // so that we can always skip the entire symbol if we need to navigate 3275 // more quickly at the source level when parsing STABS 3276 if ( !N_INCL_indexes.empty() ) 3277 { 3278 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3279 symbol_ptr->SetByteSize(sym_idx + 1); 3280 symbol_ptr->SetSizeIsSibling(true); 3281 N_INCL_indexes.pop_back(); 3282 } 3283 type = eSymbolTypeScopeEnd; 3284 break; 3285 3286 case N_SOL: 3287 // #included file name: name,,n_sect,0,address 3288 type = eSymbolTypeHeaderFile; 3289 3290 // We currently don't use the header files on darwin 3291 add_nlist = false; 3292 break; 3293 3294 case N_PARAMS: 3295 // compiler parameters: name,,NO_SECT,0,0 3296 type = eSymbolTypeCompiler; 3297 break; 3298 3299 case N_VERSION: 3300 // compiler version: name,,NO_SECT,0,0 3301 type = eSymbolTypeCompiler; 3302 break; 3303 3304 case N_OLEVEL: 3305 // compiler -O level: name,,NO_SECT,0,0 3306 type = eSymbolTypeCompiler; 3307 break; 3308 3309 case N_PSYM: 3310 // parameter: name,,NO_SECT,type,offset 3311 type = eSymbolTypeVariable; 3312 break; 3313 3314 case N_ENTRY: 3315 // alternate entry: name,,n_sect,linenumber,address 3316 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3317 type = eSymbolTypeLineEntry; 3318 break; 3319 3320 //---------------------------------------------------------------------- 3321 // Left and Right Braces 3322 //---------------------------------------------------------------------- 3323 case N_LBRAC: 3324 // left bracket: 0,,NO_SECT,nesting level,address 3325 // We use the current number of symbols in the symbol table in lieu of 3326 // using nlist_idx in case we ever start trimming entries out 3327 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3328 N_BRAC_indexes.push_back(sym_idx); 3329 type = eSymbolTypeScopeBegin; 3330 break; 3331 3332 case N_RBRAC: 3333 // right bracket: 0,,NO_SECT,nesting level,address 3334 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 3335 // so that we can always skip the entire symbol if we need to navigate 3336 // more quickly at the source level when parsing STABS 3337 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3338 if ( !N_BRAC_indexes.empty() ) 3339 { 3340 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3341 symbol_ptr->SetByteSize(sym_idx + 1); 3342 symbol_ptr->SetSizeIsSibling(true); 3343 N_BRAC_indexes.pop_back(); 3344 } 3345 type = eSymbolTypeScopeEnd; 3346 break; 3347 3348 case N_EXCL: 3349 // deleted include file: name,,NO_SECT,0,sum 3350 type = eSymbolTypeHeaderFile; 3351 break; 3352 3353 //---------------------------------------------------------------------- 3354 // COMM scopes 3355 //---------------------------------------------------------------------- 3356 case N_BCOMM: 3357 // begin common: name,,NO_SECT,0,0 3358 // We use the current number of symbols in the symbol table in lieu of 3359 // using nlist_idx in case we ever start trimming entries out 3360 type = eSymbolTypeScopeBegin; 3361 N_COMM_indexes.push_back(sym_idx); 3362 break; 3363 3364 case N_ECOML: 3365 // end common (local name): 0,,n_sect,0,address 3366 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3367 // Fall through 3368 3369 case N_ECOMM: 3370 // end common: name,,n_sect,0,0 3371 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 3372 // so that we can always skip the entire symbol if we need to navigate 3373 // more quickly at the source level when parsing STABS 3374 if ( !N_COMM_indexes.empty() ) 3375 { 3376 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 3377 symbol_ptr->SetByteSize(sym_idx + 1); 3378 symbol_ptr->SetSizeIsSibling(true); 3379 N_COMM_indexes.pop_back(); 3380 } 3381 type = eSymbolTypeScopeEnd; 3382 break; 3383 3384 case N_LENG: 3385 // second stab entry with length information 3386 type = eSymbolTypeAdditional; 3387 break; 3388 3389 default: break; 3390 } 3391 } 3392 else 3393 { 3394 //uint8_t n_pext = N_PEXT & nlist.n_type; 3395 uint8_t n_type = N_TYPE & nlist.n_type; 3396 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3397 3398 switch (n_type) 3399 { 3400 case N_INDR:// Fall through 3401 case N_PBUD:// Fall through 3402 case N_UNDF: 3403 type = eSymbolTypeUndefined; 3404 break; 3405 3406 case N_ABS: 3407 type = eSymbolTypeAbsolute; 3408 break; 3409 3410 case N_SECT: 3411 { 3412 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3413 3414 if (!symbol_section) 3415 { 3416 // TODO: warn about this? 3417 add_nlist = false; 3418 break; 3419 } 3420 3421 if (TEXT_eh_frame_sectID == nlist.n_sect) 3422 { 3423 type = eSymbolTypeException; 3424 } 3425 else 3426 { 3427 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 3428 3429 switch (section_type) 3430 { 3431 case S_REGULAR: break; // regular section 3432 //case S_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section 3433 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 3434 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 3435 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 3436 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 3437 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 3438 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 3439 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 3440 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 3441 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 3442 //case S_COALESCED: type = eSymbolType; break; // section contains symbols that are to be coalesced 3443 //case S_GB_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 3444 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 3445 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 3446 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 3447 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 3448 default: break; 3449 } 3450 3451 if (type == eSymbolTypeInvalid) 3452 { 3453 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 3454 if (symbol_section->IsDescendant (text_section_sp.get())) 3455 { 3456 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 3457 S_ATTR_SELF_MODIFYING_CODE | 3458 S_ATTR_SOME_INSTRUCTIONS)) 3459 type = eSymbolTypeData; 3460 else 3461 type = eSymbolTypeCode; 3462 } 3463 else 3464 if (symbol_section->IsDescendant(data_section_sp.get())) 3465 { 3466 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 3467 { 3468 type = eSymbolTypeRuntime; 3469 3470 if (symbol_name && 3471 symbol_name[0] == '_' && 3472 symbol_name[1] == 'O' && 3473 symbol_name[2] == 'B') 3474 { 3475 llvm::StringRef symbol_name_ref(symbol_name); 3476 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 3477 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 3478 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 3479 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 3480 { 3481 symbol_name_non_abi_mangled = symbol_name + 1; 3482 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3483 type = eSymbolTypeObjCClass; 3484 demangled_is_synthesized = true; 3485 } 3486 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 3487 { 3488 symbol_name_non_abi_mangled = symbol_name + 1; 3489 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3490 type = eSymbolTypeObjCMetaClass; 3491 demangled_is_synthesized = true; 3492 } 3493 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 3494 { 3495 symbol_name_non_abi_mangled = symbol_name + 1; 3496 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3497 type = eSymbolTypeObjCIVar; 3498 demangled_is_synthesized = true; 3499 } 3500 } 3501 } 3502 else 3503 if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 3504 { 3505 type = eSymbolTypeException; 3506 } 3507 else 3508 { 3509 type = eSymbolTypeData; 3510 } 3511 } 3512 else 3513 if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 3514 { 3515 type = eSymbolTypeTrampoline; 3516 } 3517 else 3518 if (symbol_section->IsDescendant(objc_section_sp.get())) 3519 { 3520 type = eSymbolTypeRuntime; 3521 if (symbol_name && symbol_name[0] == '.') 3522 { 3523 llvm::StringRef symbol_name_ref(symbol_name); 3524 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 3525 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 3526 { 3527 symbol_name_non_abi_mangled = symbol_name; 3528 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 3529 type = eSymbolTypeObjCClass; 3530 demangled_is_synthesized = true; 3531 } 3532 } 3533 } 3534 } 3535 } 3536 } 3537 break; 3538 } 3539 } 3540 3541 if (add_nlist) 3542 { 3543 uint64_t symbol_value = nlist.n_value; 3544 3545 if (symbol_name_non_abi_mangled) 3546 { 3547 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 3548 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 3549 } 3550 else 3551 { 3552 bool symbol_name_is_mangled = false; 3553 3554 if (symbol_name && symbol_name[0] == '_') 3555 { 3556 symbol_name_is_mangled = symbol_name[1] == '_'; 3557 symbol_name++; // Skip the leading underscore 3558 } 3559 3560 if (symbol_name) 3561 { 3562 ConstString const_symbol_name(symbol_name); 3563 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 3564 if (is_gsym && is_debug) 3565 { 3566 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 3567 } 3568 } 3569 } 3570 if (symbol_section) 3571 { 3572 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3573 if (symbol_byte_size == 0 && function_starts_count > 0) 3574 { 3575 addr_t symbol_lookup_file_addr = nlist.n_value; 3576 // Do an exact address match for non-ARM addresses, else get the closest since 3577 // the symbol might be a thumb symbol which has an address with bit zero set 3578 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 3579 if (is_arm && func_start_entry) 3580 { 3581 // Verify that the function start address is the symbol address (ARM) 3582 // or the symbol address + 1 (thumb) 3583 if (func_start_entry->addr != symbol_lookup_file_addr && 3584 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 3585 { 3586 // Not the right entry, NULL it out... 3587 func_start_entry = NULL; 3588 } 3589 } 3590 if (func_start_entry) 3591 { 3592 func_start_entry->data = true; 3593 3594 addr_t symbol_file_addr = func_start_entry->addr; 3595 if (is_arm) 3596 symbol_file_addr &= 0xfffffffffffffffeull; 3597 3598 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3599 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3600 if (next_func_start_entry) 3601 { 3602 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3603 // Be sure the clear the Thumb address bit when we calculate the size 3604 // from the current and next address 3605 if (is_arm) 3606 next_symbol_file_addr &= 0xfffffffffffffffeull; 3607 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3608 } 3609 else 3610 { 3611 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3612 } 3613 } 3614 } 3615 symbol_value -= section_file_addr; 3616 } 3617 3618 if (is_debug == false) 3619 { 3620 if (type == eSymbolTypeCode) 3621 { 3622 // See if we can find a N_FUN entry for any code symbols. 3623 // If we do find a match, and the name matches, then we 3624 // can merge the two into just the function symbol to avoid 3625 // duplicate entries in the symbol table 3626 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3627 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3628 if (range.first != range.second) 3629 { 3630 bool found_it = false; 3631 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3632 { 3633 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3634 { 3635 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3636 // We just need the flags from the linker symbol, so put these flags 3637 // into the N_FUN flags to avoid duplicate symbols in the symbol table 3638 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3639 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3640 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3641 sym[pos->second].SetType (eSymbolTypeResolver); 3642 sym[sym_idx].Clear(); 3643 found_it = true; 3644 break; 3645 } 3646 } 3647 if (found_it) 3648 continue; 3649 } 3650 else 3651 { 3652 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3653 type = eSymbolTypeResolver; 3654 } 3655 } 3656 else if (type == eSymbolTypeData) 3657 { 3658 // See if we can find a N_STSYM entry for any data symbols. 3659 // If we do find a match, and the name matches, then we 3660 // can merge the two into just the Static symbol to avoid 3661 // duplicate entries in the symbol table 3662 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3663 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 3664 if (range.first != range.second) 3665 { 3666 bool found_it = false; 3667 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3668 { 3669 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3670 { 3671 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3672 // We just need the flags from the linker symbol, so put these flags 3673 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3674 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3675 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3676 sym[sym_idx].Clear(); 3677 found_it = true; 3678 break; 3679 } 3680 } 3681 if (found_it) 3682 continue; 3683 } 3684 else 3685 { 3686 // Combine N_GSYM stab entries with the non stab symbol 3687 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 3688 if (pos != N_GSYM_name_to_sym_idx.end()) 3689 { 3690 const uint32_t GSYM_sym_idx = pos->second; 3691 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 3692 // Copy the address, because often the N_GSYM address has an invalid address of zero 3693 // when the global is a common symbol 3694 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 3695 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 3696 // We just need the flags from the linker symbol, so put these flags 3697 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3698 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3699 sym[sym_idx].Clear(); 3700 continue; 3701 } 3702 } 3703 } 3704 } 3705 3706 sym[sym_idx].SetID (nlist_idx); 3707 sym[sym_idx].SetType (type); 3708 sym[sym_idx].GetAddress().SetSection (symbol_section); 3709 sym[sym_idx].GetAddress().SetOffset (symbol_value); 3710 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3711 3712 if (symbol_byte_size > 0) 3713 sym[sym_idx].SetByteSize(symbol_byte_size); 3714 3715 if (demangled_is_synthesized) 3716 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3717 3718 ++sym_idx; 3719 } 3720 else 3721 { 3722 sym[sym_idx].Clear(); 3723 } 3724 } 3725 } 3726 3727 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 3728 3729 if (function_starts_count > 0) 3730 { 3731 char synthetic_function_symbol[PATH_MAX]; 3732 uint32_t num_synthetic_function_symbols = 0; 3733 for (i=0; i<function_starts_count; ++i) 3734 { 3735 if (function_starts.GetEntryRef (i).data == false) 3736 ++num_synthetic_function_symbols; 3737 } 3738 3739 if (num_synthetic_function_symbols > 0) 3740 { 3741 if (num_syms < sym_idx + num_synthetic_function_symbols) 3742 { 3743 num_syms = sym_idx + num_synthetic_function_symbols; 3744 sym = symtab->Resize (num_syms); 3745 } 3746 uint32_t synthetic_function_symbol_idx = 0; 3747 for (i=0; i<function_starts_count; ++i) 3748 { 3749 const FunctionStarts::Entry *func_start_entry = function_starts.GetEntryAtIndex (i); 3750 if (func_start_entry->data == false) 3751 { 3752 addr_t symbol_file_addr = func_start_entry->addr; 3753 uint32_t symbol_flags = 0; 3754 if (is_arm) 3755 { 3756 if (symbol_file_addr & 1) 3757 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3758 symbol_file_addr &= 0xfffffffffffffffeull; 3759 } 3760 Address symbol_addr; 3761 if (module_sp->ResolveFileAddress (symbol_file_addr, symbol_addr)) 3762 { 3763 SectionSP symbol_section (symbol_addr.GetSection()); 3764 uint32_t symbol_byte_size = 0; 3765 if (symbol_section) 3766 { 3767 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3768 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3769 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3770 if (next_func_start_entry) 3771 { 3772 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3773 if (is_arm) 3774 next_symbol_file_addr &= 0xfffffffffffffffeull; 3775 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3776 } 3777 else 3778 { 3779 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3780 } 3781 snprintf (synthetic_function_symbol, 3782 sizeof(synthetic_function_symbol), 3783 "___lldb_unnamed_function%u$$%s", 3784 ++synthetic_function_symbol_idx, 3785 module_sp->GetFileSpec().GetFilename().GetCString()); 3786 sym[sym_idx].SetID (synthetic_sym_id++); 3787 sym[sym_idx].GetMangled().SetDemangledName(ConstString(synthetic_function_symbol)); 3788 sym[sym_idx].SetType (eSymbolTypeCode); 3789 sym[sym_idx].SetIsSynthetic (true); 3790 sym[sym_idx].GetAddress() = symbol_addr; 3791 if (symbol_flags) 3792 sym[sym_idx].SetFlags (symbol_flags); 3793 if (symbol_byte_size) 3794 sym[sym_idx].SetByteSize (symbol_byte_size); 3795 ++sym_idx; 3796 } 3797 } 3798 } 3799 } 3800 } 3801 } 3802 3803 // Trim our symbols down to just what we ended up with after 3804 // removing any symbols. 3805 if (sym_idx < num_syms) 3806 { 3807 num_syms = sym_idx; 3808 sym = symtab->Resize (num_syms); 3809 } 3810 3811 // Now synthesize indirect symbols 3812 if (m_dysymtab.nindirectsyms != 0) 3813 { 3814 if (indirect_symbol_index_data.GetByteSize()) 3815 { 3816 NListIndexToSymbolIndexMap::const_iterator end_index_pos = m_nlist_idx_to_sym_idx.end(); 3817 3818 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); ++sect_idx) 3819 { 3820 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == S_SYMBOL_STUBS) 3821 { 3822 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 3823 if (symbol_stub_byte_size == 0) 3824 continue; 3825 3826 const uint32_t num_symbol_stubs = m_mach_sections[sect_idx].size / symbol_stub_byte_size; 3827 3828 if (num_symbol_stubs == 0) 3829 continue; 3830 3831 const uint32_t symbol_stub_index_offset = m_mach_sections[sect_idx].reserved1; 3832 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) 3833 { 3834 const uint32_t symbol_stub_index = symbol_stub_index_offset + stub_idx; 3835 const lldb::addr_t symbol_stub_addr = m_mach_sections[sect_idx].addr + (stub_idx * symbol_stub_byte_size); 3836 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 3837 if (indirect_symbol_index_data.ValidOffsetForDataOfSize(symbol_stub_offset, 4)) 3838 { 3839 const uint32_t stub_sym_id = indirect_symbol_index_data.GetU32 (&symbol_stub_offset); 3840 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 3841 continue; 3842 3843 NListIndexToSymbolIndexMap::const_iterator index_pos = m_nlist_idx_to_sym_idx.find (stub_sym_id); 3844 Symbol *stub_symbol = NULL; 3845 if (index_pos != end_index_pos) 3846 { 3847 // We have a remapping from the original nlist index to 3848 // a current symbol index, so just look this up by index 3849 stub_symbol = symtab->SymbolAtIndex (index_pos->second); 3850 } 3851 else 3852 { 3853 // We need to lookup a symbol using the original nlist 3854 // symbol index since this index is coming from the 3855 // S_SYMBOL_STUBS 3856 stub_symbol = symtab->FindSymbolByID (stub_sym_id); 3857 } 3858 3859 if (stub_symbol) 3860 { 3861 Address so_addr(symbol_stub_addr, section_list); 3862 3863 if (stub_symbol->GetType() == eSymbolTypeUndefined) 3864 { 3865 // Change the external symbol into a trampoline that makes sense 3866 // These symbols were N_UNDF N_EXT, and are useless to us, so we 3867 // can re-use them so we don't have to make up a synthetic symbol 3868 // for no good reason. 3869 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 3870 stub_symbol->SetType (eSymbolTypeTrampoline); 3871 else 3872 stub_symbol->SetType (eSymbolTypeResolver); 3873 stub_symbol->SetExternal (false); 3874 stub_symbol->GetAddress() = so_addr; 3875 stub_symbol->SetByteSize (symbol_stub_byte_size); 3876 } 3877 else 3878 { 3879 // Make a synthetic symbol to describe the trampoline stub 3880 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 3881 if (sym_idx >= num_syms) 3882 { 3883 sym = symtab->Resize (++num_syms); 3884 stub_symbol = NULL; // this pointer no longer valid 3885 } 3886 sym[sym_idx].SetID (synthetic_sym_id++); 3887 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 3888 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 3889 sym[sym_idx].SetType (eSymbolTypeTrampoline); 3890 else 3891 sym[sym_idx].SetType (eSymbolTypeResolver); 3892 sym[sym_idx].SetIsSynthetic (true); 3893 sym[sym_idx].GetAddress() = so_addr; 3894 sym[sym_idx].SetByteSize (symbol_stub_byte_size); 3895 ++sym_idx; 3896 } 3897 } 3898 else 3899 { 3900 if (log) 3901 log->Warning ("symbol stub referencing symbol table symbol %u that isn't in our minimal symbol table, fix this!!!", stub_sym_id); 3902 } 3903 } 3904 } 3905 } 3906 } 3907 } 3908 } 3909 3910 3911 if (!trie_entries.empty()) 3912 { 3913 for (const auto &e : trie_entries) 3914 { 3915 if (e.entry.import_name) 3916 { 3917 // Make a synthetic symbol to describe re-exported symbol. 3918 if (sym_idx >= num_syms) 3919 sym = symtab->Resize (++num_syms); 3920 sym[sym_idx].SetID (synthetic_sym_id++); 3921 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 3922 sym[sym_idx].SetType (eSymbolTypeReExported); 3923 sym[sym_idx].SetIsSynthetic (true); 3924 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 3925 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) 3926 { 3927 sym[sym_idx].SetReExportedSymbolSharedLibrary(dylib_files.GetFileSpecAtIndex(e.entry.other-1)); 3928 } 3929 ++sym_idx; 3930 } 3931 } 3932 } 3933 3934 3935 3936 // StreamFile s(stdout, false); 3937 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 3938 // symtab->Dump(&s, NULL, eSortOrderNone); 3939 // Set symbol byte sizes correctly since mach-o nlist entries don't have sizes 3940 symtab->CalculateSymbolSizes(); 3941 3942 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 3943 // symtab->Dump(&s, NULL, eSortOrderNone); 3944 3945 return symtab->GetNumSymbols(); 3946 } 3947 return 0; 3948 } 3949 3950 3951 void 3952 ObjectFileMachO::Dump (Stream *s) 3953 { 3954 ModuleSP module_sp(GetModule()); 3955 if (module_sp) 3956 { 3957 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 3958 s->Printf("%p: ", this); 3959 s->Indent(); 3960 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 3961 s->PutCString("ObjectFileMachO64"); 3962 else 3963 s->PutCString("ObjectFileMachO32"); 3964 3965 ArchSpec header_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 3966 3967 *s << ", file = '" << m_file << "', arch = " << header_arch.GetArchitectureName() << "\n"; 3968 3969 SectionList *sections = GetSectionList(); 3970 if (sections) 3971 sections->Dump(s, NULL, true, UINT32_MAX); 3972 3973 if (m_symtab_ap.get()) 3974 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 3975 } 3976 } 3977 3978 bool 3979 ObjectFileMachO::GetUUID (const llvm::MachO::mach_header &header, 3980 const lldb_private::DataExtractor &data, 3981 lldb::offset_t lc_offset, 3982 lldb_private::UUID& uuid) 3983 { 3984 uint32_t i; 3985 struct uuid_command load_cmd; 3986 3987 lldb::offset_t offset = lc_offset; 3988 for (i=0; i<header.ncmds; ++i) 3989 { 3990 const lldb::offset_t cmd_offset = offset; 3991 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 3992 break; 3993 3994 if (load_cmd.cmd == LC_UUID) 3995 { 3996 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 3997 3998 if (uuid_bytes) 3999 { 4000 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4001 // We pretend these object files have no UUID to prevent crashing. 4002 4003 const uint8_t opencl_uuid[] = { 0x8c, 0x8e, 0xb3, 0x9b, 4004 0x3b, 0xa8, 4005 0x4b, 0x16, 4006 0xb6, 0xa4, 4007 0x27, 0x63, 0xbb, 0x14, 0xf0, 0x0d }; 4008 4009 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4010 return false; 4011 4012 uuid.SetBytes (uuid_bytes); 4013 return true; 4014 } 4015 return false; 4016 } 4017 offset = cmd_offset + load_cmd.cmdsize; 4018 } 4019 return false; 4020 } 4021 4022 bool 4023 ObjectFileMachO::GetUUID (lldb_private::UUID* uuid) 4024 { 4025 ModuleSP module_sp(GetModule()); 4026 if (module_sp) 4027 { 4028 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4029 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4030 return GetUUID (m_header, m_data, offset, *uuid); 4031 } 4032 return false; 4033 } 4034 4035 4036 uint32_t 4037 ObjectFileMachO::GetDependentModules (FileSpecList& files) 4038 { 4039 uint32_t count = 0; 4040 ModuleSP module_sp(GetModule()); 4041 if (module_sp) 4042 { 4043 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4044 struct load_command load_cmd; 4045 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4046 const bool resolve_path = false; // Don't resolve the dependend file paths since they may not reside on this system 4047 uint32_t i; 4048 for (i=0; i<m_header.ncmds; ++i) 4049 { 4050 const uint32_t cmd_offset = offset; 4051 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4052 break; 4053 4054 switch (load_cmd.cmd) 4055 { 4056 case LC_LOAD_DYLIB: 4057 case LC_LOAD_WEAK_DYLIB: 4058 case LC_REEXPORT_DYLIB: 4059 case LC_LOAD_DYLINKER: 4060 case LC_LOADFVMLIB: 4061 case LC_LOAD_UPWARD_DYLIB: 4062 { 4063 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 4064 const char *path = m_data.PeekCStr(name_offset); 4065 // Skip any path that starts with '@' since these are usually: 4066 // @executable_path/.../file 4067 // @rpath/.../file 4068 if (path && path[0] != '@') 4069 { 4070 FileSpec file_spec(path, resolve_path); 4071 if (files.AppendIfUnique(file_spec)) 4072 count++; 4073 } 4074 } 4075 break; 4076 4077 default: 4078 break; 4079 } 4080 offset = cmd_offset + load_cmd.cmdsize; 4081 } 4082 } 4083 return count; 4084 } 4085 4086 lldb_private::Address 4087 ObjectFileMachO::GetEntryPointAddress () 4088 { 4089 // If the object file is not an executable it can't hold the entry point. m_entry_point_address 4090 // is initialized to an invalid address, so we can just return that. 4091 // If m_entry_point_address is valid it means we've found it already, so return the cached value. 4092 4093 if (!IsExecutable() || m_entry_point_address.IsValid()) 4094 return m_entry_point_address; 4095 4096 // Otherwise, look for the UnixThread or Thread command. The data for the Thread command is given in 4097 // /usr/include/mach-o.h, but it is basically: 4098 // 4099 // uint32_t flavor - this is the flavor argument you would pass to thread_get_state 4100 // uint32_t count - this is the count of longs in the thread state data 4101 // struct XXX_thread_state state - this is the structure from <machine/thread_status.h> corresponding to the flavor. 4102 // <repeat this trio> 4103 // 4104 // So we just keep reading the various register flavors till we find the GPR one, then read the PC out of there. 4105 // FIXME: We will need to have a "RegisterContext data provider" class at some point that can get all the registers 4106 // out of data in this form & attach them to a given thread. That should underlie the MacOS X User process plugin, 4107 // and we'll also need it for the MacOS X Core File process plugin. When we have that we can also use it here. 4108 // 4109 // For now we hard-code the offsets and flavors we need: 4110 // 4111 // 4112 4113 ModuleSP module_sp(GetModule()); 4114 if (module_sp) 4115 { 4116 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4117 struct load_command load_cmd; 4118 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4119 uint32_t i; 4120 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 4121 bool done = false; 4122 4123 for (i=0; i<m_header.ncmds; ++i) 4124 { 4125 const lldb::offset_t cmd_offset = offset; 4126 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4127 break; 4128 4129 switch (load_cmd.cmd) 4130 { 4131 case LC_UNIXTHREAD: 4132 case LC_THREAD: 4133 { 4134 while (offset < cmd_offset + load_cmd.cmdsize) 4135 { 4136 uint32_t flavor = m_data.GetU32(&offset); 4137 uint32_t count = m_data.GetU32(&offset); 4138 if (count == 0) 4139 { 4140 // We've gotten off somehow, log and exit; 4141 return m_entry_point_address; 4142 } 4143 4144 switch (m_header.cputype) 4145 { 4146 case llvm::MachO::CPU_TYPE_ARM: 4147 if (flavor == 1) // ARM_THREAD_STATE from mach/arm/thread_status.h 4148 { 4149 offset += 60; // This is the offset of pc in the GPR thread state data structure. 4150 start_address = m_data.GetU32(&offset); 4151 done = true; 4152 } 4153 break; 4154 case llvm::MachO::CPU_TYPE_I386: 4155 if (flavor == 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 4156 { 4157 offset += 40; // This is the offset of eip in the GPR thread state data structure. 4158 start_address = m_data.GetU32(&offset); 4159 done = true; 4160 } 4161 break; 4162 case llvm::MachO::CPU_TYPE_X86_64: 4163 if (flavor == 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 4164 { 4165 offset += 16 * 8; // This is the offset of rip in the GPR thread state data structure. 4166 start_address = m_data.GetU64(&offset); 4167 done = true; 4168 } 4169 break; 4170 default: 4171 return m_entry_point_address; 4172 } 4173 // Haven't found the GPR flavor yet, skip over the data for this flavor: 4174 if (done) 4175 break; 4176 offset += count * 4; 4177 } 4178 } 4179 break; 4180 case LC_MAIN: 4181 { 4182 ConstString text_segment_name ("__TEXT"); 4183 uint64_t entryoffset = m_data.GetU64(&offset); 4184 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 4185 if (text_segment_sp) 4186 { 4187 done = true; 4188 start_address = text_segment_sp->GetFileAddress() + entryoffset; 4189 } 4190 } 4191 4192 default: 4193 break; 4194 } 4195 if (done) 4196 break; 4197 4198 // Go to the next load command: 4199 offset = cmd_offset + load_cmd.cmdsize; 4200 } 4201 4202 if (start_address != LLDB_INVALID_ADDRESS) 4203 { 4204 // We got the start address from the load commands, so now resolve that address in the sections 4205 // of this ObjectFile: 4206 if (!m_entry_point_address.ResolveAddressUsingFileSections (start_address, GetSectionList())) 4207 { 4208 m_entry_point_address.Clear(); 4209 } 4210 } 4211 else 4212 { 4213 // We couldn't read the UnixThread load command - maybe it wasn't there. As a fallback look for the 4214 // "start" symbol in the main executable. 4215 4216 ModuleSP module_sp (GetModule()); 4217 4218 if (module_sp) 4219 { 4220 SymbolContextList contexts; 4221 SymbolContext context; 4222 if (module_sp->FindSymbolsWithNameAndType(ConstString ("start"), eSymbolTypeCode, contexts)) 4223 { 4224 if (contexts.GetContextAtIndex(0, context)) 4225 m_entry_point_address = context.symbol->GetAddress(); 4226 } 4227 } 4228 } 4229 } 4230 4231 return m_entry_point_address; 4232 4233 } 4234 4235 lldb_private::Address 4236 ObjectFileMachO::GetHeaderAddress () 4237 { 4238 lldb_private::Address header_addr; 4239 SectionList *section_list = GetSectionList(); 4240 if (section_list) 4241 { 4242 SectionSP text_segment_sp (section_list->FindSectionByName (GetSegmentNameTEXT())); 4243 if (text_segment_sp) 4244 { 4245 header_addr.SetSection (text_segment_sp); 4246 header_addr.SetOffset (0); 4247 } 4248 } 4249 return header_addr; 4250 } 4251 4252 uint32_t 4253 ObjectFileMachO::GetNumThreadContexts () 4254 { 4255 ModuleSP module_sp(GetModule()); 4256 if (module_sp) 4257 { 4258 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4259 if (!m_thread_context_offsets_valid) 4260 { 4261 m_thread_context_offsets_valid = true; 4262 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4263 FileRangeArray::Entry file_range; 4264 thread_command thread_cmd; 4265 for (uint32_t i=0; i<m_header.ncmds; ++i) 4266 { 4267 const uint32_t cmd_offset = offset; 4268 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 4269 break; 4270 4271 if (thread_cmd.cmd == LC_THREAD) 4272 { 4273 file_range.SetRangeBase (offset); 4274 file_range.SetByteSize (thread_cmd.cmdsize - 8); 4275 m_thread_context_offsets.Append (file_range); 4276 } 4277 offset = cmd_offset + thread_cmd.cmdsize; 4278 } 4279 } 4280 } 4281 return m_thread_context_offsets.GetSize(); 4282 } 4283 4284 lldb::RegisterContextSP 4285 ObjectFileMachO::GetThreadContextAtIndex (uint32_t idx, lldb_private::Thread &thread) 4286 { 4287 lldb::RegisterContextSP reg_ctx_sp; 4288 4289 ModuleSP module_sp(GetModule()); 4290 if (module_sp) 4291 { 4292 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4293 if (!m_thread_context_offsets_valid) 4294 GetNumThreadContexts (); 4295 4296 const FileRangeArray::Entry *thread_context_file_range = m_thread_context_offsets.GetEntryAtIndex (idx); 4297 if (thread_context_file_range) 4298 { 4299 4300 DataExtractor data (m_data, 4301 thread_context_file_range->GetRangeBase(), 4302 thread_context_file_range->GetByteSize()); 4303 4304 switch (m_header.cputype) 4305 { 4306 case llvm::MachO::CPU_TYPE_ARM: 4307 reg_ctx_sp.reset (new RegisterContextDarwin_arm_Mach (thread, data)); 4308 break; 4309 4310 case llvm::MachO::CPU_TYPE_I386: 4311 reg_ctx_sp.reset (new RegisterContextDarwin_i386_Mach (thread, data)); 4312 break; 4313 4314 case llvm::MachO::CPU_TYPE_X86_64: 4315 reg_ctx_sp.reset (new RegisterContextDarwin_x86_64_Mach (thread, data)); 4316 break; 4317 } 4318 } 4319 } 4320 return reg_ctx_sp; 4321 } 4322 4323 4324 ObjectFile::Type 4325 ObjectFileMachO::CalculateType() 4326 { 4327 switch (m_header.filetype) 4328 { 4329 case MH_OBJECT: // 0x1u 4330 if (GetAddressByteSize () == 4) 4331 { 4332 // 32 bit kexts are just object files, but they do have a valid 4333 // UUID load command. 4334 UUID uuid; 4335 if (GetUUID(&uuid)) 4336 { 4337 // this checking for the UUID load command is not enough 4338 // we could eventually look for the symbol named 4339 // "OSKextGetCurrentIdentifier" as this is required of kexts 4340 if (m_strata == eStrataInvalid) 4341 m_strata = eStrataKernel; 4342 return eTypeSharedLibrary; 4343 } 4344 } 4345 return eTypeObjectFile; 4346 4347 case MH_EXECUTE: return eTypeExecutable; // 0x2u 4348 case MH_FVMLIB: return eTypeSharedLibrary; // 0x3u 4349 case MH_CORE: return eTypeCoreFile; // 0x4u 4350 case MH_PRELOAD: return eTypeSharedLibrary; // 0x5u 4351 case MH_DYLIB: return eTypeSharedLibrary; // 0x6u 4352 case MH_DYLINKER: return eTypeDynamicLinker; // 0x7u 4353 case MH_BUNDLE: return eTypeSharedLibrary; // 0x8u 4354 case MH_DYLIB_STUB: return eTypeStubLibrary; // 0x9u 4355 case MH_DSYM: return eTypeDebugInfo; // 0xAu 4356 case MH_KEXT_BUNDLE: return eTypeSharedLibrary; // 0xBu 4357 default: 4358 break; 4359 } 4360 return eTypeUnknown; 4361 } 4362 4363 ObjectFile::Strata 4364 ObjectFileMachO::CalculateStrata() 4365 { 4366 switch (m_header.filetype) 4367 { 4368 case MH_OBJECT: // 0x1u 4369 { 4370 // 32 bit kexts are just object files, but they do have a valid 4371 // UUID load command. 4372 UUID uuid; 4373 if (GetUUID(&uuid)) 4374 { 4375 // this checking for the UUID load command is not enough 4376 // we could eventually look for the symbol named 4377 // "OSKextGetCurrentIdentifier" as this is required of kexts 4378 if (m_type == eTypeInvalid) 4379 m_type = eTypeSharedLibrary; 4380 4381 return eStrataKernel; 4382 } 4383 } 4384 return eStrataUnknown; 4385 4386 case MH_EXECUTE: // 0x2u 4387 // Check for the MH_DYLDLINK bit in the flags 4388 if (m_header.flags & MH_DYLDLINK) 4389 { 4390 return eStrataUser; 4391 } 4392 else 4393 { 4394 SectionList *section_list = GetSectionList(); 4395 if (section_list) 4396 { 4397 static ConstString g_kld_section_name ("__KLD"); 4398 if (section_list->FindSectionByName(g_kld_section_name)) 4399 return eStrataKernel; 4400 } 4401 } 4402 return eStrataRawImage; 4403 4404 case MH_FVMLIB: return eStrataUser; // 0x3u 4405 case MH_CORE: return eStrataUnknown; // 0x4u 4406 case MH_PRELOAD: return eStrataRawImage; // 0x5u 4407 case MH_DYLIB: return eStrataUser; // 0x6u 4408 case MH_DYLINKER: return eStrataUser; // 0x7u 4409 case MH_BUNDLE: return eStrataUser; // 0x8u 4410 case MH_DYLIB_STUB: return eStrataUser; // 0x9u 4411 case MH_DSYM: return eStrataUnknown; // 0xAu 4412 case MH_KEXT_BUNDLE: return eStrataKernel; // 0xBu 4413 default: 4414 break; 4415 } 4416 return eStrataUnknown; 4417 } 4418 4419 4420 uint32_t 4421 ObjectFileMachO::GetVersion (uint32_t *versions, uint32_t num_versions) 4422 { 4423 ModuleSP module_sp(GetModule()); 4424 if (module_sp) 4425 { 4426 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4427 struct dylib_command load_cmd; 4428 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4429 uint32_t version_cmd = 0; 4430 uint64_t version = 0; 4431 uint32_t i; 4432 for (i=0; i<m_header.ncmds; ++i) 4433 { 4434 const lldb::offset_t cmd_offset = offset; 4435 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4436 break; 4437 4438 if (load_cmd.cmd == LC_ID_DYLIB) 4439 { 4440 if (version_cmd == 0) 4441 { 4442 version_cmd = load_cmd.cmd; 4443 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 4444 break; 4445 version = load_cmd.dylib.current_version; 4446 } 4447 break; // Break for now unless there is another more complete version 4448 // number load command in the future. 4449 } 4450 offset = cmd_offset + load_cmd.cmdsize; 4451 } 4452 4453 if (version_cmd == LC_ID_DYLIB) 4454 { 4455 if (versions != NULL && num_versions > 0) 4456 { 4457 if (num_versions > 0) 4458 versions[0] = (version & 0xFFFF0000ull) >> 16; 4459 if (num_versions > 1) 4460 versions[1] = (version & 0x0000FF00ull) >> 8; 4461 if (num_versions > 2) 4462 versions[2] = (version & 0x000000FFull); 4463 // Fill in an remaining version numbers with invalid values 4464 for (i=3; i<num_versions; ++i) 4465 versions[i] = UINT32_MAX; 4466 } 4467 // The LC_ID_DYLIB load command has a version with 3 version numbers 4468 // in it, so always return 3 4469 return 3; 4470 } 4471 } 4472 return false; 4473 } 4474 4475 bool 4476 ObjectFileMachO::GetArchitecture (ArchSpec &arch) 4477 { 4478 ModuleSP module_sp(GetModule()); 4479 if (module_sp) 4480 { 4481 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4482 arch.SetArchitecture (eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 4483 4484 // Files with type MH_PRELOAD are currently used in cases where the image 4485 // debugs at the addresses in the file itself. Below we set the OS to 4486 // unknown to make sure we use the DynamicLoaderStatic()... 4487 if (m_header.filetype == MH_PRELOAD) 4488 { 4489 arch.GetTriple().setOS (llvm::Triple::UnknownOS); 4490 } 4491 return true; 4492 } 4493 return false; 4494 } 4495 4496 4497 UUID 4498 ObjectFileMachO::GetProcessSharedCacheUUID (Process *process) 4499 { 4500 UUID uuid; 4501 if (process) 4502 { 4503 addr_t all_image_infos = process->GetImageInfoAddress(); 4504 4505 // The address returned by GetImageInfoAddress may be the address of dyld (don't want) 4506 // or it may be the address of the dyld_all_image_infos structure (want). The first four 4507 // bytes will be either the version field (all_image_infos) or a Mach-O file magic constant. 4508 // Version 13 and higher of dyld_all_image_infos is required to get the sharedCacheUUID field. 4509 4510 Error err; 4511 uint32_t version_or_magic = process->ReadUnsignedIntegerFromMemory (all_image_infos, 4, -1, err); 4512 if (version_or_magic != -1 4513 && version_or_magic != MH_MAGIC 4514 && version_or_magic != MH_CIGAM 4515 && version_or_magic != MH_MAGIC_64 4516 && version_or_magic != MH_CIGAM_64 4517 && version_or_magic >= 13) 4518 { 4519 addr_t sharedCacheUUID_address = LLDB_INVALID_ADDRESS; 4520 int wordsize = process->GetAddressByteSize(); 4521 if (wordsize == 8) 4522 { 4523 sharedCacheUUID_address = all_image_infos + 160; // sharedCacheUUID <mach-o/dyld_images.h> 4524 } 4525 if (wordsize == 4) 4526 { 4527 sharedCacheUUID_address = all_image_infos + 84; // sharedCacheUUID <mach-o/dyld_images.h> 4528 } 4529 if (sharedCacheUUID_address != LLDB_INVALID_ADDRESS) 4530 { 4531 uuid_t shared_cache_uuid; 4532 if (process->ReadMemory (sharedCacheUUID_address, shared_cache_uuid, sizeof (uuid_t), err) == sizeof (uuid_t)) 4533 { 4534 uuid.SetBytes (shared_cache_uuid); 4535 } 4536 } 4537 } 4538 } 4539 return uuid; 4540 } 4541 4542 UUID 4543 ObjectFileMachO::GetLLDBSharedCacheUUID () 4544 { 4545 UUID uuid; 4546 #if defined (__APPLE__) && defined (__arm__) 4547 uint8_t *(*dyld_get_all_image_infos)(void); 4548 dyld_get_all_image_infos = (uint8_t*(*)()) dlsym (RTLD_DEFAULT, "_dyld_get_all_image_infos"); 4549 if (dyld_get_all_image_infos) 4550 { 4551 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 4552 if (dyld_all_image_infos_address) 4553 { 4554 uint32_t *version = (uint32_t*) dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 4555 if (*version >= 13) 4556 { 4557 uuid_t *sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 84); // sharedCacheUUID <mach-o/dyld_images.h> 4558 uuid.SetBytes (sharedCacheUUID_address); 4559 } 4560 } 4561 } 4562 #endif 4563 return uuid; 4564 } 4565 4566 uint32_t 4567 ObjectFileMachO::GetMinimumOSVersion (uint32_t *versions, uint32_t num_versions) 4568 { 4569 if (m_min_os_versions.empty()) 4570 { 4571 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4572 bool success = false; 4573 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 4574 { 4575 const lldb::offset_t load_cmd_offset = offset; 4576 4577 version_min_command lc; 4578 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 4579 break; 4580 if (lc.cmd == LC_VERSION_MIN_MACOSX || lc.cmd == LC_VERSION_MIN_IPHONEOS) 4581 { 4582 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 4583 { 4584 const uint32_t xxxx = lc.version >> 16; 4585 const uint32_t yy = (lc.version >> 8) & 0xffu; 4586 const uint32_t zz = lc.version & 0xffu; 4587 if (xxxx) 4588 { 4589 m_min_os_versions.push_back(xxxx); 4590 if (yy) 4591 { 4592 m_min_os_versions.push_back(yy); 4593 if (zz) 4594 m_min_os_versions.push_back(zz); 4595 } 4596 } 4597 success = true; 4598 } 4599 } 4600 offset = load_cmd_offset + lc.cmdsize; 4601 } 4602 4603 if (success == false) 4604 { 4605 // Push an invalid value so we don't keep trying to 4606 m_min_os_versions.push_back(UINT32_MAX); 4607 } 4608 } 4609 4610 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) 4611 { 4612 if (versions != NULL && num_versions > 0) 4613 { 4614 for (size_t i=0; i<num_versions; ++i) 4615 { 4616 if (i < m_min_os_versions.size()) 4617 versions[i] = m_min_os_versions[i]; 4618 else 4619 versions[i] = 0; 4620 } 4621 } 4622 return m_min_os_versions.size(); 4623 } 4624 // Call the superclasses version that will empty out the data 4625 return ObjectFile::GetMinimumOSVersion (versions, num_versions); 4626 } 4627 4628 uint32_t 4629 ObjectFileMachO::GetSDKVersion(uint32_t *versions, uint32_t num_versions) 4630 { 4631 if (m_sdk_versions.empty()) 4632 { 4633 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4634 bool success = false; 4635 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 4636 { 4637 const lldb::offset_t load_cmd_offset = offset; 4638 4639 version_min_command lc; 4640 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 4641 break; 4642 if (lc.cmd == LC_VERSION_MIN_MACOSX || lc.cmd == LC_VERSION_MIN_IPHONEOS) 4643 { 4644 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 4645 { 4646 const uint32_t xxxx = lc.reserved >> 16; 4647 const uint32_t yy = (lc.reserved >> 8) & 0xffu; 4648 const uint32_t zz = lc.reserved & 0xffu; 4649 if (xxxx) 4650 { 4651 m_sdk_versions.push_back(xxxx); 4652 if (yy) 4653 { 4654 m_sdk_versions.push_back(yy); 4655 if (zz) 4656 m_sdk_versions.push_back(zz); 4657 } 4658 } 4659 success = true; 4660 } 4661 } 4662 offset = load_cmd_offset + lc.cmdsize; 4663 } 4664 4665 if (success == false) 4666 { 4667 // Push an invalid value so we don't keep trying to 4668 m_sdk_versions.push_back(UINT32_MAX); 4669 } 4670 } 4671 4672 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) 4673 { 4674 if (versions != NULL && num_versions > 0) 4675 { 4676 for (size_t i=0; i<num_versions; ++i) 4677 { 4678 if (i < m_sdk_versions.size()) 4679 versions[i] = m_sdk_versions[i]; 4680 else 4681 versions[i] = 0; 4682 } 4683 } 4684 return m_sdk_versions.size(); 4685 } 4686 // Call the superclasses version that will empty out the data 4687 return ObjectFile::GetSDKVersion (versions, num_versions); 4688 } 4689 4690 4691 //------------------------------------------------------------------ 4692 // PluginInterface protocol 4693 //------------------------------------------------------------------ 4694 lldb_private::ConstString 4695 ObjectFileMachO::GetPluginName() 4696 { 4697 return GetPluginNameStatic(); 4698 } 4699 4700 uint32_t 4701 ObjectFileMachO::GetPluginVersion() 4702 { 4703 return 1; 4704 } 4705 4706