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 HeaderMagic32: 406 case HeaderMagic32Swapped: 407 return sizeof(struct mach_header); 408 409 case HeaderMagic64: 410 case HeaderMagic64Swapped: 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 (spec.GetArchitecture().IsValid()) 527 { 528 GetUUID (header, data, data_offset, spec.GetUUID()); 529 specs.Append(spec); 530 } 531 } 532 } 533 } 534 return specs.GetSize() - initial_count; 535 } 536 537 538 539 const ConstString & 540 ObjectFileMachO::GetSegmentNameTEXT() 541 { 542 static ConstString g_segment_name_TEXT ("__TEXT"); 543 return g_segment_name_TEXT; 544 } 545 546 const ConstString & 547 ObjectFileMachO::GetSegmentNameDATA() 548 { 549 static ConstString g_segment_name_DATA ("__DATA"); 550 return g_segment_name_DATA; 551 } 552 553 const ConstString & 554 ObjectFileMachO::GetSegmentNameOBJC() 555 { 556 static ConstString g_segment_name_OBJC ("__OBJC"); 557 return g_segment_name_OBJC; 558 } 559 560 const ConstString & 561 ObjectFileMachO::GetSegmentNameLINKEDIT() 562 { 563 static ConstString g_section_name_LINKEDIT ("__LINKEDIT"); 564 return g_section_name_LINKEDIT; 565 } 566 567 const ConstString & 568 ObjectFileMachO::GetSectionNameEHFrame() 569 { 570 static ConstString g_section_name_eh_frame ("__eh_frame"); 571 return g_section_name_eh_frame; 572 } 573 574 bool 575 ObjectFileMachO::MagicBytesMatch (DataBufferSP& data_sp, 576 lldb::addr_t data_offset, 577 lldb::addr_t data_length) 578 { 579 DataExtractor data; 580 data.SetData (data_sp, data_offset, data_length); 581 lldb::offset_t offset = 0; 582 uint32_t magic = data.GetU32(&offset); 583 return MachHeaderSizeFromMagic(magic) != 0; 584 } 585 586 587 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 588 DataBufferSP& data_sp, 589 lldb::offset_t data_offset, 590 const FileSpec* file, 591 lldb::offset_t file_offset, 592 lldb::offset_t length) : 593 ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 594 m_mach_segments(), 595 m_mach_sections(), 596 m_entry_point_address(), 597 m_thread_context_offsets(), 598 m_thread_context_offsets_valid(false) 599 { 600 ::memset (&m_header, 0, sizeof(m_header)); 601 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 602 } 603 604 ObjectFileMachO::ObjectFileMachO (const lldb::ModuleSP &module_sp, 605 lldb::DataBufferSP& header_data_sp, 606 const lldb::ProcessSP &process_sp, 607 lldb::addr_t header_addr) : 608 ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 609 m_mach_segments(), 610 m_mach_sections(), 611 m_entry_point_address(), 612 m_thread_context_offsets(), 613 m_thread_context_offsets_valid(false) 614 { 615 ::memset (&m_header, 0, sizeof(m_header)); 616 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 617 } 618 619 ObjectFileMachO::~ObjectFileMachO() 620 { 621 } 622 623 bool 624 ObjectFileMachO::ParseHeader (DataExtractor &data, 625 lldb::offset_t *data_offset_ptr, 626 llvm::MachO::mach_header &header) 627 { 628 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 629 // Leave magic in the original byte order 630 header.magic = data.GetU32(data_offset_ptr); 631 bool can_parse = false; 632 bool is_64_bit = false; 633 switch (header.magic) 634 { 635 case HeaderMagic32: 636 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 637 data.SetAddressByteSize(4); 638 can_parse = true; 639 break; 640 641 case HeaderMagic64: 642 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 643 data.SetAddressByteSize(8); 644 can_parse = true; 645 is_64_bit = true; 646 break; 647 648 case HeaderMagic32Swapped: 649 data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 650 data.SetAddressByteSize(4); 651 can_parse = true; 652 break; 653 654 case HeaderMagic64Swapped: 655 data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 656 data.SetAddressByteSize(8); 657 is_64_bit = true; 658 can_parse = true; 659 break; 660 661 default: 662 break; 663 } 664 665 if (can_parse) 666 { 667 data.GetU32(data_offset_ptr, &header.cputype, 6); 668 if (is_64_bit) 669 *data_offset_ptr += 4; 670 return true; 671 } 672 else 673 { 674 memset(&header, 0, sizeof(header)); 675 } 676 return false; 677 } 678 679 bool 680 ObjectFileMachO::ParseHeader () 681 { 682 ModuleSP module_sp(GetModule()); 683 if (module_sp) 684 { 685 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 686 bool can_parse = false; 687 lldb::offset_t offset = 0; 688 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 689 // Leave magic in the original byte order 690 m_header.magic = m_data.GetU32(&offset); 691 switch (m_header.magic) 692 { 693 case HeaderMagic32: 694 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 695 m_data.SetAddressByteSize(4); 696 can_parse = true; 697 break; 698 699 case HeaderMagic64: 700 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 701 m_data.SetAddressByteSize(8); 702 can_parse = true; 703 break; 704 705 case HeaderMagic32Swapped: 706 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 707 m_data.SetAddressByteSize(4); 708 can_parse = true; 709 break; 710 711 case HeaderMagic64Swapped: 712 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 713 m_data.SetAddressByteSize(8); 714 can_parse = true; 715 break; 716 717 default: 718 break; 719 } 720 721 if (can_parse) 722 { 723 m_data.GetU32(&offset, &m_header.cputype, 6); 724 725 ArchSpec mach_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 726 727 // Check if the module has a required architecture 728 const ArchSpec &module_arch = module_sp->GetArchitecture(); 729 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 730 return false; 731 732 if (SetModulesArchitecture (mach_arch)) 733 { 734 const size_t header_and_lc_size = m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 735 if (m_data.GetByteSize() < header_and_lc_size) 736 { 737 DataBufferSP data_sp; 738 ProcessSP process_sp (m_process_wp.lock()); 739 if (process_sp) 740 { 741 data_sp = ReadMemory (process_sp, m_memory_addr, header_and_lc_size); 742 } 743 else 744 { 745 // Read in all only the load command data from the file on disk 746 data_sp = m_file.ReadFileContents(m_file_offset, header_and_lc_size); 747 if (data_sp->GetByteSize() != header_and_lc_size) 748 return false; 749 } 750 if (data_sp) 751 m_data.SetData (data_sp); 752 } 753 } 754 return true; 755 } 756 else 757 { 758 memset(&m_header, 0, sizeof(struct mach_header)); 759 } 760 } 761 return false; 762 } 763 764 765 ByteOrder 766 ObjectFileMachO::GetByteOrder () const 767 { 768 return m_data.GetByteOrder (); 769 } 770 771 bool 772 ObjectFileMachO::IsExecutable() const 773 { 774 return m_header.filetype == HeaderFileTypeExecutable; 775 } 776 777 uint32_t 778 ObjectFileMachO::GetAddressByteSize () const 779 { 780 return m_data.GetAddressByteSize (); 781 } 782 783 AddressClass 784 ObjectFileMachO::GetAddressClass (lldb::addr_t file_addr) 785 { 786 Symtab *symtab = GetSymtab(); 787 if (symtab) 788 { 789 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 790 if (symbol) 791 { 792 if (symbol->ValueIsAddress()) 793 { 794 SectionSP section_sp (symbol->GetAddress().GetSection()); 795 if (section_sp) 796 { 797 const SectionType section_type = section_sp->GetType(); 798 switch (section_type) 799 { 800 case eSectionTypeInvalid: return eAddressClassUnknown; 801 case eSectionTypeCode: 802 if (m_header.cputype == llvm::MachO::CPUTypeARM) 803 { 804 // For ARM we have a bit in the n_desc field of the symbol 805 // that tells us ARM/Thumb which is bit 0x0008. 806 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 807 return eAddressClassCodeAlternateISA; 808 } 809 return eAddressClassCode; 810 811 case eSectionTypeContainer: return eAddressClassUnknown; 812 case eSectionTypeData: 813 case eSectionTypeDataCString: 814 case eSectionTypeDataCStringPointers: 815 case eSectionTypeDataSymbolAddress: 816 case eSectionTypeData4: 817 case eSectionTypeData8: 818 case eSectionTypeData16: 819 case eSectionTypeDataPointers: 820 case eSectionTypeZeroFill: 821 case eSectionTypeDataObjCMessageRefs: 822 case eSectionTypeDataObjCCFStrings: 823 return eAddressClassData; 824 case eSectionTypeDebug: 825 case eSectionTypeDWARFDebugAbbrev: 826 case eSectionTypeDWARFDebugAranges: 827 case eSectionTypeDWARFDebugFrame: 828 case eSectionTypeDWARFDebugInfo: 829 case eSectionTypeDWARFDebugLine: 830 case eSectionTypeDWARFDebugLoc: 831 case eSectionTypeDWARFDebugMacInfo: 832 case eSectionTypeDWARFDebugPubNames: 833 case eSectionTypeDWARFDebugPubTypes: 834 case eSectionTypeDWARFDebugRanges: 835 case eSectionTypeDWARFDebugStr: 836 case eSectionTypeDWARFAppleNames: 837 case eSectionTypeDWARFAppleTypes: 838 case eSectionTypeDWARFAppleNamespaces: 839 case eSectionTypeDWARFAppleObjC: 840 return eAddressClassDebug; 841 case eSectionTypeEHFrame: return eAddressClassRuntime; 842 case eSectionTypeELFSymbolTable: 843 case eSectionTypeELFDynamicSymbols: 844 case eSectionTypeELFRelocationEntries: 845 case eSectionTypeELFDynamicLinkInfo: 846 case eSectionTypeOther: return eAddressClassUnknown; 847 } 848 } 849 } 850 851 const SymbolType symbol_type = symbol->GetType(); 852 switch (symbol_type) 853 { 854 case eSymbolTypeAny: return eAddressClassUnknown; 855 case eSymbolTypeAbsolute: return eAddressClassUnknown; 856 857 case eSymbolTypeCode: 858 case eSymbolTypeTrampoline: 859 case eSymbolTypeResolver: 860 if (m_header.cputype == llvm::MachO::CPUTypeARM) 861 { 862 // For ARM we have a bit in the n_desc field of the symbol 863 // that tells us ARM/Thumb which is bit 0x0008. 864 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 865 return eAddressClassCodeAlternateISA; 866 } 867 return eAddressClassCode; 868 869 case eSymbolTypeData: return eAddressClassData; 870 case eSymbolTypeRuntime: return eAddressClassRuntime; 871 case eSymbolTypeException: return eAddressClassRuntime; 872 case eSymbolTypeSourceFile: return eAddressClassDebug; 873 case eSymbolTypeHeaderFile: return eAddressClassDebug; 874 case eSymbolTypeObjectFile: return eAddressClassDebug; 875 case eSymbolTypeCommonBlock: return eAddressClassDebug; 876 case eSymbolTypeBlock: return eAddressClassDebug; 877 case eSymbolTypeLocal: return eAddressClassData; 878 case eSymbolTypeParam: return eAddressClassData; 879 case eSymbolTypeVariable: return eAddressClassData; 880 case eSymbolTypeVariableType: return eAddressClassDebug; 881 case eSymbolTypeLineEntry: return eAddressClassDebug; 882 case eSymbolTypeLineHeader: return eAddressClassDebug; 883 case eSymbolTypeScopeBegin: return eAddressClassDebug; 884 case eSymbolTypeScopeEnd: return eAddressClassDebug; 885 case eSymbolTypeAdditional: return eAddressClassUnknown; 886 case eSymbolTypeCompiler: return eAddressClassDebug; 887 case eSymbolTypeInstrumentation:return eAddressClassDebug; 888 case eSymbolTypeUndefined: return eAddressClassUnknown; 889 case eSymbolTypeObjCClass: return eAddressClassRuntime; 890 case eSymbolTypeObjCMetaClass: return eAddressClassRuntime; 891 case eSymbolTypeObjCIVar: return eAddressClassRuntime; 892 } 893 } 894 } 895 return eAddressClassUnknown; 896 } 897 898 Symtab * 899 ObjectFileMachO::GetSymtab() 900 { 901 ModuleSP module_sp(GetModule()); 902 if (module_sp) 903 { 904 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 905 if (m_symtab_ap.get() == NULL) 906 { 907 m_symtab_ap.reset(new Symtab(this)); 908 Mutex::Locker symtab_locker (m_symtab_ap->GetMutex()); 909 ParseSymtab (); 910 m_symtab_ap->Finalize (); 911 } 912 } 913 return m_symtab_ap.get(); 914 } 915 916 bool 917 ObjectFileMachO::IsStripped () 918 { 919 if (m_dysymtab.cmd == 0) 920 { 921 ModuleSP module_sp(GetModule()); 922 if (module_sp) 923 { 924 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 925 for (uint32_t i=0; i<m_header.ncmds; ++i) 926 { 927 const lldb::offset_t load_cmd_offset = offset; 928 929 load_command lc; 930 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 931 break; 932 if (lc.cmd == LoadCommandDynamicSymtabInfo) 933 { 934 m_dysymtab.cmd = lc.cmd; 935 m_dysymtab.cmdsize = lc.cmdsize; 936 if (m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == NULL) 937 { 938 // Clear m_dysymtab if we were unable to read all items from the load command 939 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 940 } 941 } 942 offset = load_cmd_offset + lc.cmdsize; 943 } 944 } 945 } 946 if (m_dysymtab.cmd) 947 return m_dysymtab.nlocalsym == 0; 948 return false; 949 } 950 951 void 952 ObjectFileMachO::CreateSections (SectionList &unified_section_list) 953 { 954 if (!m_sections_ap.get()) 955 { 956 m_sections_ap.reset(new SectionList()); 957 958 const bool is_dsym = (m_header.filetype == HeaderFileTypeDSYM); 959 lldb::user_id_t segID = 0; 960 lldb::user_id_t sectID = 0; 961 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 962 uint32_t i; 963 const bool is_core = GetType() == eTypeCoreFile; 964 //bool dump_sections = false; 965 ModuleSP module_sp (GetModule()); 966 // First look up any LC_ENCRYPTION_INFO load commands 967 typedef RangeArray<uint32_t, uint32_t, 8> EncryptedFileRanges; 968 EncryptedFileRanges encrypted_file_ranges; 969 encryption_info_command encryption_cmd; 970 for (i=0; i<m_header.ncmds; ++i) 971 { 972 const lldb::offset_t load_cmd_offset = offset; 973 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 974 break; 975 976 if (encryption_cmd.cmd == LoadCommandEncryptionInfo) 977 { 978 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) 979 { 980 if (encryption_cmd.cryptid != 0) 981 { 982 EncryptedFileRanges::Entry entry; 983 entry.SetRangeBase(encryption_cmd.cryptoff); 984 entry.SetByteSize(encryption_cmd.cryptsize); 985 encrypted_file_ranges.Append(entry); 986 } 987 } 988 } 989 offset = load_cmd_offset + encryption_cmd.cmdsize; 990 } 991 992 offset = MachHeaderSizeFromMagic(m_header.magic); 993 994 struct segment_command_64 load_cmd; 995 for (i=0; i<m_header.ncmds; ++i) 996 { 997 const lldb::offset_t load_cmd_offset = offset; 998 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 999 break; 1000 1001 if (load_cmd.cmd == LoadCommandSegment32 || load_cmd.cmd == LoadCommandSegment64) 1002 { 1003 if (m_data.GetU8(&offset, (uint8_t*)load_cmd.segname, 16)) 1004 { 1005 bool add_section = true; 1006 bool add_to_unified = true; 1007 ConstString const_segname (load_cmd.segname, std::min<size_t>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 1008 1009 SectionSP unified_section_sp(unified_section_list.FindSectionByName(const_segname)); 1010 if (is_dsym && unified_section_sp) 1011 { 1012 if (const_segname == GetSegmentNameLINKEDIT()) 1013 { 1014 // We need to keep the __LINKEDIT segment private to this object file only 1015 add_to_unified = false; 1016 } 1017 else 1018 { 1019 // This is the dSYM file and this section has already been created by 1020 // the object file, no need to create it. 1021 add_section = false; 1022 } 1023 } 1024 load_cmd.vmaddr = m_data.GetAddress(&offset); 1025 load_cmd.vmsize = m_data.GetAddress(&offset); 1026 load_cmd.fileoff = m_data.GetAddress(&offset); 1027 load_cmd.filesize = m_data.GetAddress(&offset); 1028 if (m_length != 0 && load_cmd.filesize != 0) 1029 { 1030 if (load_cmd.fileoff > m_length) 1031 { 1032 // We have a load command that says it extends past the end of hte file. This is likely 1033 // a corrupt file. We don't have any way to return an error condition here (this method 1034 // was likely invokved from something like ObjectFile::GetSectionList()) -- all we can do 1035 // is null out the SectionList vector and if a process has been set up, dump a message 1036 // to stdout. The most common case here is core file debugging with a truncated file. 1037 const char *lc_segment_name = load_cmd.cmd == LoadCommandSegment64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1038 module_sp->ReportError("is a corrupt mach-o file: load command %u %s has a fileoff (0x%" PRIx64 ") that extends beyond the end of the file (0x%" PRIx64 ")", 1039 i, 1040 lc_segment_name, 1041 load_cmd.fileoff, 1042 m_length); 1043 1044 load_cmd.fileoff = 0; 1045 load_cmd.filesize = 0; 1046 } 1047 1048 if (load_cmd.fileoff + load_cmd.filesize > m_length) 1049 { 1050 // We have a load command that says it extends past the end of hte file. This is likely 1051 // a corrupt file. We don't have any way to return an error condition here (this method 1052 // was likely invokved from something like ObjectFile::GetSectionList()) -- all we can do 1053 // is null out the SectionList vector and if a process has been set up, dump a message 1054 // to stdout. The most common case here is core file debugging with a truncated file. 1055 const char *lc_segment_name = load_cmd.cmd == LoadCommandSegment64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1056 module_sp->ReportError("is a corrupt mach-o file: 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", 1057 i, 1058 lc_segment_name, 1059 load_cmd.fileoff + load_cmd.filesize, 1060 m_length); 1061 1062 // Tuncase the length 1063 load_cmd.filesize = m_length - load_cmd.fileoff; 1064 } 1065 } 1066 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1067 { 1068 1069 const bool segment_is_encrypted = (load_cmd.flags & SegmentCommandFlagBitProtectedVersion1) != 0; 1070 1071 // Keep a list of mach segments around in case we need to 1072 // get at data that isn't stored in the abstracted Sections. 1073 m_mach_segments.push_back (load_cmd); 1074 1075 // Use a segment ID of the segment index shifted left by 8 so they 1076 // never conflict with any of the sections. 1077 SectionSP segment_sp; 1078 if (add_section && (const_segname || is_core)) 1079 { 1080 segment_sp.reset(new Section (module_sp, // Module to which this section belongs 1081 this, // Object file to which this sections belongs 1082 ++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 1083 const_segname, // Name of this section 1084 eSectionTypeContainer, // This section is a container of other sections. 1085 load_cmd.vmaddr, // File VM address == addresses as they are found in the object file 1086 load_cmd.vmsize, // VM size in bytes of this section 1087 load_cmd.fileoff, // Offset to the data for this section in the file 1088 load_cmd.filesize, // Size in bytes of this section as found in the the file 1089 load_cmd.flags)); // Flags for this section 1090 1091 segment_sp->SetIsEncrypted (segment_is_encrypted); 1092 m_sections_ap->AddSection(segment_sp); 1093 if (add_to_unified) 1094 unified_section_list.AddSection(segment_sp); 1095 } 1096 else if (unified_section_sp) 1097 { 1098 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) 1099 { 1100 // Check to see if the module was read from memory? 1101 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) 1102 { 1103 // We have a module that is in memory and needs to have its 1104 // file address adjusted. We need to do this because when we 1105 // load a file from memory, its addresses will be slid already, 1106 // yet the addresses in the new symbol file will still be unslid. 1107 // Since everything is stored as section offset, this shouldn't 1108 // cause any problems. 1109 1110 // Make sure we've parsed the symbol table from the 1111 // ObjectFile before we go around changing its Sections. 1112 module_sp->GetObjectFile()->GetSymtab(); 1113 // eh_frame would present the same problems but we parse that on 1114 // a per-function basis as-needed so it's more difficult to 1115 // remove its use of the Sections. Realistically, the environments 1116 // where this code path will be taken will not have eh_frame sections. 1117 1118 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1119 } 1120 } 1121 m_sections_ap->AddSection(unified_section_sp); 1122 } 1123 1124 struct section_64 sect64; 1125 ::memset (§64, 0, sizeof(sect64)); 1126 // Push a section into our mach sections for the section at 1127 // index zero (NListSectionNoSection) if we don't have any 1128 // mach sections yet... 1129 if (m_mach_sections.empty()) 1130 m_mach_sections.push_back(sect64); 1131 uint32_t segment_sect_idx; 1132 const lldb::user_id_t first_segment_sectID = sectID + 1; 1133 1134 1135 const uint32_t num_u32s = load_cmd.cmd == LoadCommandSegment32 ? 7 : 8; 1136 for (segment_sect_idx=0; segment_sect_idx<load_cmd.nsects; ++segment_sect_idx) 1137 { 1138 if (m_data.GetU8(&offset, (uint8_t*)sect64.sectname, sizeof(sect64.sectname)) == NULL) 1139 break; 1140 if (m_data.GetU8(&offset, (uint8_t*)sect64.segname, sizeof(sect64.segname)) == NULL) 1141 break; 1142 sect64.addr = m_data.GetAddress(&offset); 1143 sect64.size = m_data.GetAddress(&offset); 1144 1145 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1146 break; 1147 1148 // Keep a list of mach sections around in case we need to 1149 // get at data that isn't stored in the abstracted Sections. 1150 m_mach_sections.push_back (sect64); 1151 1152 if (add_section) 1153 { 1154 ConstString section_name (sect64.sectname, std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 1155 if (!const_segname) 1156 { 1157 // We have a segment with no name so we need to conjure up 1158 // segments that correspond to the section's segname if there 1159 // isn't already such a section. If there is such a section, 1160 // we resize the section so that it spans all sections. 1161 // We also mark these sections as fake so address matches don't 1162 // hit if they land in the gaps between the child sections. 1163 const_segname.SetTrimmedCStringWithLength(sect64.segname, sizeof(sect64.segname)); 1164 segment_sp = unified_section_list.FindSectionByName (const_segname); 1165 if (segment_sp.get()) 1166 { 1167 Section *segment = segment_sp.get(); 1168 // Grow the section size as needed. 1169 const lldb::addr_t sect64_min_addr = sect64.addr; 1170 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1171 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1172 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1173 const lldb::addr_t curr_seg_max_addr = curr_seg_min_addr + curr_seg_byte_size; 1174 if (sect64_min_addr >= curr_seg_min_addr) 1175 { 1176 const lldb::addr_t new_seg_byte_size = sect64_max_addr - curr_seg_min_addr; 1177 // Only grow the section size if needed 1178 if (new_seg_byte_size > curr_seg_byte_size) 1179 segment->SetByteSize (new_seg_byte_size); 1180 } 1181 else 1182 { 1183 // We need to change the base address of the segment and 1184 // adjust the child section offsets for all existing children. 1185 const lldb::addr_t slide_amount = sect64_min_addr - curr_seg_min_addr; 1186 segment->Slide(slide_amount, false); 1187 segment->GetChildren().Slide(-slide_amount, false); 1188 segment->SetByteSize (curr_seg_max_addr - sect64_min_addr); 1189 } 1190 1191 // Grow the section size as needed. 1192 if (sect64.offset) 1193 { 1194 const lldb::addr_t segment_min_file_offset = segment->GetFileOffset(); 1195 const lldb::addr_t segment_max_file_offset = segment_min_file_offset + segment->GetFileSize(); 1196 1197 const lldb::addr_t section_min_file_offset = sect64.offset; 1198 const lldb::addr_t section_max_file_offset = section_min_file_offset + sect64.size; 1199 const lldb::addr_t new_file_offset = std::min (section_min_file_offset, segment_min_file_offset); 1200 const lldb::addr_t new_file_size = std::max (section_max_file_offset, segment_max_file_offset) - new_file_offset; 1201 segment->SetFileOffset (new_file_offset); 1202 segment->SetFileSize (new_file_size); 1203 } 1204 } 1205 else 1206 { 1207 // Create a fake section for the section's named segment 1208 segment_sp.reset(new Section (segment_sp, // Parent section 1209 module_sp, // Module to which this section belongs 1210 this, // Object file to which this section belongs 1211 ++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 1212 const_segname, // Name of this section 1213 eSectionTypeContainer, // This section is a container of other sections. 1214 sect64.addr, // File VM address == addresses as they are found in the object file 1215 sect64.size, // VM size in bytes of this section 1216 sect64.offset, // Offset to the data for this section in the file 1217 sect64.offset ? sect64.size : 0, // Size in bytes of this section as found in the the file 1218 load_cmd.flags)); // Flags for this section 1219 segment_sp->SetIsFake(true); 1220 1221 m_sections_ap->AddSection(segment_sp); 1222 if (add_to_unified) 1223 unified_section_list.AddSection(segment_sp); 1224 segment_sp->SetIsEncrypted (segment_is_encrypted); 1225 } 1226 } 1227 assert (segment_sp.get()); 1228 1229 uint32_t mach_sect_type = sect64.flags & SectionFlagMaskSectionType; 1230 static ConstString g_sect_name_objc_data ("__objc_data"); 1231 static ConstString g_sect_name_objc_msgrefs ("__objc_msgrefs"); 1232 static ConstString g_sect_name_objc_selrefs ("__objc_selrefs"); 1233 static ConstString g_sect_name_objc_classrefs ("__objc_classrefs"); 1234 static ConstString g_sect_name_objc_superrefs ("__objc_superrefs"); 1235 static ConstString g_sect_name_objc_const ("__objc_const"); 1236 static ConstString g_sect_name_objc_classlist ("__objc_classlist"); 1237 static ConstString g_sect_name_cfstring ("__cfstring"); 1238 1239 static ConstString g_sect_name_dwarf_debug_abbrev ("__debug_abbrev"); 1240 static ConstString g_sect_name_dwarf_debug_aranges ("__debug_aranges"); 1241 static ConstString g_sect_name_dwarf_debug_frame ("__debug_frame"); 1242 static ConstString g_sect_name_dwarf_debug_info ("__debug_info"); 1243 static ConstString g_sect_name_dwarf_debug_line ("__debug_line"); 1244 static ConstString g_sect_name_dwarf_debug_loc ("__debug_loc"); 1245 static ConstString g_sect_name_dwarf_debug_macinfo ("__debug_macinfo"); 1246 static ConstString g_sect_name_dwarf_debug_pubnames ("__debug_pubnames"); 1247 static ConstString g_sect_name_dwarf_debug_pubtypes ("__debug_pubtypes"); 1248 static ConstString g_sect_name_dwarf_debug_ranges ("__debug_ranges"); 1249 static ConstString g_sect_name_dwarf_debug_str ("__debug_str"); 1250 static ConstString g_sect_name_dwarf_apple_names ("__apple_names"); 1251 static ConstString g_sect_name_dwarf_apple_types ("__apple_types"); 1252 static ConstString g_sect_name_dwarf_apple_namespaces ("__apple_namespac"); 1253 static ConstString g_sect_name_dwarf_apple_objc ("__apple_objc"); 1254 static ConstString g_sect_name_eh_frame ("__eh_frame"); 1255 static ConstString g_sect_name_DATA ("__DATA"); 1256 static ConstString g_sect_name_TEXT ("__TEXT"); 1257 1258 SectionType sect_type = eSectionTypeOther; 1259 1260 if (section_name == g_sect_name_dwarf_debug_abbrev) 1261 sect_type = eSectionTypeDWARFDebugAbbrev; 1262 else if (section_name == g_sect_name_dwarf_debug_aranges) 1263 sect_type = eSectionTypeDWARFDebugAranges; 1264 else if (section_name == g_sect_name_dwarf_debug_frame) 1265 sect_type = eSectionTypeDWARFDebugFrame; 1266 else if (section_name == g_sect_name_dwarf_debug_info) 1267 sect_type = eSectionTypeDWARFDebugInfo; 1268 else if (section_name == g_sect_name_dwarf_debug_line) 1269 sect_type = eSectionTypeDWARFDebugLine; 1270 else if (section_name == g_sect_name_dwarf_debug_loc) 1271 sect_type = eSectionTypeDWARFDebugLoc; 1272 else if (section_name == g_sect_name_dwarf_debug_macinfo) 1273 sect_type = eSectionTypeDWARFDebugMacInfo; 1274 else if (section_name == g_sect_name_dwarf_debug_pubnames) 1275 sect_type = eSectionTypeDWARFDebugPubNames; 1276 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 1277 sect_type = eSectionTypeDWARFDebugPubTypes; 1278 else if (section_name == g_sect_name_dwarf_debug_ranges) 1279 sect_type = eSectionTypeDWARFDebugRanges; 1280 else if (section_name == g_sect_name_dwarf_debug_str) 1281 sect_type = eSectionTypeDWARFDebugStr; 1282 else if (section_name == g_sect_name_dwarf_apple_names) 1283 sect_type = eSectionTypeDWARFAppleNames; 1284 else if (section_name == g_sect_name_dwarf_apple_types) 1285 sect_type = eSectionTypeDWARFAppleTypes; 1286 else if (section_name == g_sect_name_dwarf_apple_namespaces) 1287 sect_type = eSectionTypeDWARFAppleNamespaces; 1288 else if (section_name == g_sect_name_dwarf_apple_objc) 1289 sect_type = eSectionTypeDWARFAppleObjC; 1290 else if (section_name == g_sect_name_objc_selrefs) 1291 sect_type = eSectionTypeDataCStringPointers; 1292 else if (section_name == g_sect_name_objc_msgrefs) 1293 sect_type = eSectionTypeDataObjCMessageRefs; 1294 else if (section_name == g_sect_name_eh_frame) 1295 sect_type = eSectionTypeEHFrame; 1296 else if (section_name == g_sect_name_cfstring) 1297 sect_type = eSectionTypeDataObjCCFStrings; 1298 else if (section_name == g_sect_name_objc_data || 1299 section_name == g_sect_name_objc_classrefs || 1300 section_name == g_sect_name_objc_superrefs || 1301 section_name == g_sect_name_objc_const || 1302 section_name == g_sect_name_objc_classlist) 1303 { 1304 sect_type = eSectionTypeDataPointers; 1305 } 1306 1307 if (sect_type == eSectionTypeOther) 1308 { 1309 switch (mach_sect_type) 1310 { 1311 // TODO: categorize sections by other flags for regular sections 1312 case SectionTypeRegular: 1313 if (segment_sp->GetName() == g_sect_name_TEXT) 1314 sect_type = eSectionTypeCode; 1315 else if (segment_sp->GetName() == g_sect_name_DATA) 1316 sect_type = eSectionTypeData; 1317 else 1318 sect_type = eSectionTypeOther; 1319 break; 1320 case SectionTypeZeroFill: sect_type = eSectionTypeZeroFill; break; 1321 case SectionTypeCStringLiterals: sect_type = eSectionTypeDataCString; break; // section with only literal C strings 1322 case SectionType4ByteLiterals: sect_type = eSectionTypeData4; break; // section with only 4 byte literals 1323 case SectionType8ByteLiterals: sect_type = eSectionTypeData8; break; // section with only 8 byte literals 1324 case SectionTypeLiteralPointers: sect_type = eSectionTypeDataPointers; break; // section with only pointers to literals 1325 case SectionTypeNonLazySymbolPointers: sect_type = eSectionTypeDataPointers; break; // section with only non-lazy symbol pointers 1326 case SectionTypeLazySymbolPointers: sect_type = eSectionTypeDataPointers; break; // section with only lazy symbol pointers 1327 case SectionTypeSymbolStubs: sect_type = eSectionTypeCode; break; // section with only symbol stubs, byte size of stub in the reserved2 field 1328 case SectionTypeModuleInitFunctionPointers: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for initialization 1329 case SectionTypeModuleTermFunctionPointers: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for termination 1330 case SectionTypeCoalesced: sect_type = eSectionTypeOther; break; 1331 case SectionTypeZeroFillLarge: sect_type = eSectionTypeZeroFill; break; 1332 case SectionTypeInterposing: sect_type = eSectionTypeCode; break; // section with only pairs of function pointers for interposing 1333 case SectionType16ByteLiterals: sect_type = eSectionTypeData16; break; // section with only 16 byte literals 1334 case SectionTypeDTraceObjectFormat: sect_type = eSectionTypeDebug; break; 1335 case SectionTypeLazyDylibSymbolPointers: sect_type = eSectionTypeDataPointers; break; 1336 default: break; 1337 } 1338 } 1339 1340 SectionSP section_sp(new Section (segment_sp, 1341 module_sp, 1342 this, 1343 ++sectID, 1344 section_name, 1345 sect_type, 1346 sect64.addr - segment_sp->GetFileAddress(), 1347 sect64.size, 1348 sect64.offset, 1349 sect64.offset == 0 ? 0 : sect64.size, 1350 sect64.flags)); 1351 // Set the section to be encrypted to match the segment 1352 1353 bool section_is_encrypted = false; 1354 if (!segment_is_encrypted && load_cmd.filesize != 0) 1355 section_is_encrypted = encrypted_file_ranges.FindEntryThatContains(sect64.offset) != NULL; 1356 1357 section_sp->SetIsEncrypted (segment_is_encrypted || section_is_encrypted); 1358 segment_sp->GetChildren().AddSection(section_sp); 1359 1360 if (segment_sp->IsFake()) 1361 { 1362 segment_sp.reset(); 1363 const_segname.Clear(); 1364 } 1365 } 1366 } 1367 if (segment_sp && is_dsym) 1368 { 1369 if (first_segment_sectID <= sectID) 1370 { 1371 lldb::user_id_t sect_uid; 1372 for (sect_uid = first_segment_sectID; sect_uid <= sectID; ++sect_uid) 1373 { 1374 SectionSP curr_section_sp(segment_sp->GetChildren().FindSectionByID (sect_uid)); 1375 SectionSP next_section_sp; 1376 if (sect_uid + 1 <= sectID) 1377 next_section_sp = segment_sp->GetChildren().FindSectionByID (sect_uid+1); 1378 1379 if (curr_section_sp.get()) 1380 { 1381 if (curr_section_sp->GetByteSize() == 0) 1382 { 1383 if (next_section_sp.get() != NULL) 1384 curr_section_sp->SetByteSize ( next_section_sp->GetFileAddress() - curr_section_sp->GetFileAddress() ); 1385 else 1386 curr_section_sp->SetByteSize ( load_cmd.vmsize ); 1387 } 1388 } 1389 } 1390 } 1391 } 1392 } 1393 } 1394 } 1395 else if (load_cmd.cmd == LoadCommandDynamicSymtabInfo) 1396 { 1397 m_dysymtab.cmd = load_cmd.cmd; 1398 m_dysymtab.cmdsize = load_cmd.cmdsize; 1399 m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1400 } 1401 1402 offset = load_cmd_offset + load_cmd.cmdsize; 1403 } 1404 1405 // StreamFile s(stdout, false); // REMOVE THIS LINE 1406 // s.Printf ("Sections for %s:\n", m_file.GetPath().c_str());// REMOVE THIS LINE 1407 // m_sections_ap->Dump(&s, NULL, true, UINT32_MAX);// REMOVE THIS LINE 1408 } 1409 } 1410 1411 class MachSymtabSectionInfo 1412 { 1413 public: 1414 1415 MachSymtabSectionInfo (SectionList *section_list) : 1416 m_section_list (section_list), 1417 m_section_infos() 1418 { 1419 // Get the number of sections down to a depth of 1 to include 1420 // all segments and their sections, but no other sections that 1421 // may be added for debug map or 1422 m_section_infos.resize(section_list->GetNumSections(1)); 1423 } 1424 1425 1426 SectionSP 1427 GetSection (uint8_t n_sect, addr_t file_addr) 1428 { 1429 if (n_sect == 0) 1430 return SectionSP(); 1431 if (n_sect < m_section_infos.size()) 1432 { 1433 if (!m_section_infos[n_sect].section_sp) 1434 { 1435 SectionSP section_sp (m_section_list->FindSectionByID (n_sect)); 1436 m_section_infos[n_sect].section_sp = section_sp; 1437 if (section_sp) 1438 { 1439 m_section_infos[n_sect].vm_range.SetBaseAddress (section_sp->GetFileAddress()); 1440 m_section_infos[n_sect].vm_range.SetByteSize (section_sp->GetByteSize()); 1441 } 1442 else 1443 { 1444 Host::SystemLog (Host::eSystemLogError, "error: unable to find section for section %u\n", n_sect); 1445 } 1446 } 1447 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) 1448 { 1449 // Symbol is in section. 1450 return m_section_infos[n_sect].section_sp; 1451 } 1452 else if (m_section_infos[n_sect].vm_range.GetByteSize () == 0 && 1453 m_section_infos[n_sect].vm_range.GetBaseAddress() == file_addr) 1454 { 1455 // Symbol is in section with zero size, but has the same start 1456 // address as the section. This can happen with linker symbols 1457 // (symbols that start with the letter 'l' or 'L'. 1458 return m_section_infos[n_sect].section_sp; 1459 } 1460 } 1461 return m_section_list->FindSectionContainingFileAddress(file_addr); 1462 } 1463 1464 protected: 1465 struct SectionInfo 1466 { 1467 SectionInfo () : 1468 vm_range(), 1469 section_sp () 1470 { 1471 } 1472 1473 VMRange vm_range; 1474 SectionSP section_sp; 1475 }; 1476 SectionList *m_section_list; 1477 std::vector<SectionInfo> m_section_infos; 1478 }; 1479 1480 size_t 1481 ObjectFileMachO::ParseSymtab () 1482 { 1483 Timer scoped_timer(__PRETTY_FUNCTION__, 1484 "ObjectFileMachO::ParseSymtab () module = %s", 1485 m_file.GetFilename().AsCString("")); 1486 ModuleSP module_sp (GetModule()); 1487 if (!module_sp) 1488 return 0; 1489 1490 struct symtab_command symtab_load_command = { 0, 0, 0, 0, 0, 0 }; 1491 struct linkedit_data_command function_starts_load_command = { 0, 0, 0, 0 }; 1492 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 1493 FunctionStarts function_starts; 1494 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1495 uint32_t i; 1496 1497 Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_SYMBOLS)); 1498 1499 for (i=0; i<m_header.ncmds; ++i) 1500 { 1501 const lldb::offset_t cmd_offset = offset; 1502 // Read in the load command and load command size 1503 struct load_command lc; 1504 if (m_data.GetU32(&offset, &lc, 2) == NULL) 1505 break; 1506 // Watch for the symbol table load command 1507 switch (lc.cmd) 1508 { 1509 case LoadCommandSymtab: 1510 symtab_load_command.cmd = lc.cmd; 1511 symtab_load_command.cmdsize = lc.cmdsize; 1512 // Read in the rest of the symtab load command 1513 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 0) // fill in symoff, nsyms, stroff, strsize fields 1514 return 0; 1515 if (symtab_load_command.symoff == 0) 1516 { 1517 if (log) 1518 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 1519 return 0; 1520 } 1521 1522 if (symtab_load_command.stroff == 0) 1523 { 1524 if (log) 1525 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 1526 return 0; 1527 } 1528 1529 if (symtab_load_command.nsyms == 0) 1530 { 1531 if (log) 1532 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 1533 return 0; 1534 } 1535 1536 if (symtab_load_command.strsize == 0) 1537 { 1538 if (log) 1539 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 1540 return 0; 1541 } 1542 break; 1543 1544 case LoadCommandFunctionStarts: 1545 function_starts_load_command.cmd = lc.cmd; 1546 function_starts_load_command.cmdsize = lc.cmdsize; 1547 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == NULL) // fill in symoff, nsyms, stroff, strsize fields 1548 bzero (&function_starts_load_command, sizeof(function_starts_load_command)); 1549 break; 1550 1551 default: 1552 break; 1553 } 1554 offset = cmd_offset + lc.cmdsize; 1555 } 1556 1557 if (symtab_load_command.cmd) 1558 { 1559 Symtab *symtab = m_symtab_ap.get(); 1560 SectionList *section_list = GetSectionList(); 1561 if (section_list == NULL) 1562 return 0; 1563 1564 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 1565 const ByteOrder byte_order = m_data.GetByteOrder(); 1566 bool bit_width_32 = addr_byte_size == 4; 1567 const size_t nlist_byte_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 1568 1569 DataExtractor nlist_data (NULL, 0, byte_order, addr_byte_size); 1570 DataExtractor strtab_data (NULL, 0, byte_order, addr_byte_size); 1571 DataExtractor function_starts_data (NULL, 0, byte_order, addr_byte_size); 1572 DataExtractor indirect_symbol_index_data (NULL, 0, byte_order, addr_byte_size); 1573 1574 const addr_t nlist_data_byte_size = symtab_load_command.nsyms * nlist_byte_size; 1575 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 1576 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 1577 1578 ProcessSP process_sp (m_process_wp.lock()); 1579 Process *process = process_sp.get(); 1580 1581 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 1582 1583 if (process) 1584 { 1585 Target &target = process->GetTarget(); 1586 1587 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 1588 1589 SectionSP linkedit_section_sp(section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 1590 // Reading mach file from memory in a process or core file... 1591 1592 if (linkedit_section_sp) 1593 { 1594 const addr_t linkedit_load_addr = linkedit_section_sp->GetLoadBaseAddress(&target); 1595 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 1596 const addr_t symoff_addr = linkedit_load_addr + symtab_load_command.symoff - linkedit_file_offset; 1597 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - linkedit_file_offset; 1598 1599 bool data_was_read = false; 1600 1601 #if defined (__APPLE__) && defined (__arm__) 1602 if (m_header.flags & 0x80000000u) 1603 { 1604 // This mach-o memory file is in the dyld shared cache. If this 1605 // program is not remote and this is iOS, then this process will 1606 // share the same shared cache as the process we are debugging and 1607 // we can read the entire __LINKEDIT from the address space in this 1608 // process. This is a needed optimization that is used for local iOS 1609 // debugging only since all shared libraries in the shared cache do 1610 // not have corresponding files that exist in the file system of the 1611 // device. They have been combined into a single file. This means we 1612 // always have to load these files from memory. All of the symbol and 1613 // string tables from all of the __LINKEDIT sections from the shared 1614 // libraries in the shared cache have been merged into a single large 1615 // symbol and string table. Reading all of this symbol and string table 1616 // data across can slow down debug launch times, so we optimize this by 1617 // reading the memory for the __LINKEDIT section from this process. 1618 1619 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 1620 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 1621 bool use_lldb_cache = true; 1622 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && lldb_shared_cache != process_shared_cache) 1623 { 1624 use_lldb_cache = false; 1625 ModuleSP module_sp (GetModule()); 1626 if (module_sp) 1627 module_sp->ReportWarning ("shared cache in process does not match lldb's own shared cache, startup will be slow."); 1628 1629 } 1630 1631 PlatformSP platform_sp (target.GetPlatform()); 1632 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) 1633 { 1634 data_was_read = true; 1635 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, eByteOrderLittle); 1636 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, eByteOrderLittle); 1637 if (function_starts_load_command.cmd) 1638 { 1639 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 1640 function_starts_data.SetData ((void *)func_start_addr, function_starts_load_command.datasize, eByteOrderLittle); 1641 } 1642 } 1643 } 1644 #endif 1645 1646 if (!data_was_read) 1647 { 1648 if (memory_module_load_level == eMemoryModuleLoadLevelComplete) 1649 { 1650 DataBufferSP nlist_data_sp (ReadMemory (process_sp, symoff_addr, nlist_data_byte_size)); 1651 if (nlist_data_sp) 1652 nlist_data.SetData (nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 1653 // Load strings individually from memory when loading from memory since shared cache 1654 // string tables contain strings for all symbols from all shared cached libraries 1655 //DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, strtab_data_byte_size)); 1656 //if (strtab_data_sp) 1657 // strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 1658 if (m_dysymtab.nindirectsyms != 0) 1659 { 1660 const addr_t indirect_syms_addr = linkedit_load_addr + m_dysymtab.indirectsymoff - linkedit_file_offset; 1661 DataBufferSP indirect_syms_data_sp (ReadMemory (process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 1662 if (indirect_syms_data_sp) 1663 indirect_symbol_index_data.SetData (indirect_syms_data_sp, 0, indirect_syms_data_sp->GetByteSize()); 1664 } 1665 } 1666 1667 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) 1668 { 1669 if (function_starts_load_command.cmd) 1670 { 1671 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 1672 DataBufferSP func_start_data_sp (ReadMemory (process_sp, func_start_addr, function_starts_load_command.datasize)); 1673 if (func_start_data_sp) 1674 function_starts_data.SetData (func_start_data_sp, 0, func_start_data_sp->GetByteSize()); 1675 } 1676 } 1677 } 1678 } 1679 } 1680 else 1681 { 1682 nlist_data.SetData (m_data, 1683 symtab_load_command.symoff, 1684 nlist_data_byte_size); 1685 strtab_data.SetData (m_data, 1686 symtab_load_command.stroff, 1687 strtab_data_byte_size); 1688 if (m_dysymtab.nindirectsyms != 0) 1689 { 1690 indirect_symbol_index_data.SetData (m_data, 1691 m_dysymtab.indirectsymoff, 1692 m_dysymtab.nindirectsyms * 4); 1693 } 1694 if (function_starts_load_command.cmd) 1695 { 1696 function_starts_data.SetData (m_data, 1697 function_starts_load_command.dataoff, 1698 function_starts_load_command.datasize); 1699 } 1700 } 1701 1702 if (nlist_data.GetByteSize() == 0 && memory_module_load_level == eMemoryModuleLoadLevelComplete) 1703 { 1704 if (log) 1705 module_sp->LogMessage(log, "failed to read nlist data"); 1706 return 0; 1707 } 1708 1709 1710 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 1711 if (!have_strtab_data) 1712 { 1713 if (process) 1714 { 1715 if (strtab_addr == LLDB_INVALID_ADDRESS) 1716 { 1717 if (log) 1718 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 1719 return 0; 1720 } 1721 } 1722 else 1723 { 1724 if (log) 1725 module_sp->LogMessage(log, "failed to read strtab data"); 1726 return 0; 1727 } 1728 } 1729 1730 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 1731 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 1732 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 1733 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 1734 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT)); 1735 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA)); 1736 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC)); 1737 SectionSP eh_frame_section_sp; 1738 if (text_section_sp.get()) 1739 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame); 1740 else 1741 eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame); 1742 1743 const bool is_arm = (m_header.cputype == llvm::MachO::CPUTypeARM); 1744 1745 // lldb works best if it knows the start addresss of all functions in a module. 1746 // Linker symbols or debug info are normally the best source of information for start addr / size but 1747 // they may be stripped in a released binary. 1748 // Two additional sources of information exist in Mach-O binaries: 1749 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each function's start address in the 1750 // binary, relative to the text section. 1751 // eh_frame - the eh_frame FDEs have the start addr & size of each function 1752 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on all modern binaries. 1753 // Binaries built to run on older releases may need to use eh_frame information. 1754 1755 if (text_section_sp && function_starts_data.GetByteSize()) 1756 { 1757 FunctionStarts::Entry function_start_entry; 1758 function_start_entry.data = false; 1759 lldb::offset_t function_start_offset = 0; 1760 function_start_entry.addr = text_section_sp->GetFileAddress(); 1761 uint64_t delta; 1762 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 0) 1763 { 1764 // Now append the current entry 1765 function_start_entry.addr += delta; 1766 function_starts.Append(function_start_entry); 1767 } 1768 } 1769 else 1770 { 1771 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the load command claiming an eh_frame 1772 // but it doesn't actually have the eh_frame content. And if we have a dSYM, we don't need to do any 1773 // of this fill-in-the-missing-symbols works anyway - the debug info should give us all the functions in 1774 // the module. 1775 if (text_section_sp.get() && eh_frame_section_sp.get() && m_type != eTypeDebugInfo) 1776 { 1777 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, eRegisterKindGCC, true); 1778 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 1779 eh_frame.GetFunctionAddressAndSizeVector (functions); 1780 addr_t text_base_addr = text_section_sp->GetFileAddress(); 1781 size_t count = functions.GetSize(); 1782 for (size_t i = 0; i < count; ++i) 1783 { 1784 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = functions.GetEntryAtIndex (i); 1785 if (func) 1786 { 1787 FunctionStarts::Entry function_start_entry; 1788 function_start_entry.addr = func->base - text_base_addr; 1789 function_starts.Append(function_start_entry); 1790 } 1791 } 1792 } 1793 } 1794 1795 const size_t function_starts_count = function_starts.GetSize(); 1796 1797 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() : NListSectionNoSection; 1798 1799 lldb::offset_t nlist_data_offset = 0; 1800 1801 uint32_t N_SO_index = UINT32_MAX; 1802 1803 MachSymtabSectionInfo section_info (section_list); 1804 std::vector<uint32_t> N_FUN_indexes; 1805 std::vector<uint32_t> N_NSYM_indexes; 1806 std::vector<uint32_t> N_INCL_indexes; 1807 std::vector<uint32_t> N_BRAC_indexes; 1808 std::vector<uint32_t> N_COMM_indexes; 1809 typedef std::map <uint64_t, uint32_t> ValueToSymbolIndexMap; 1810 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap; 1811 typedef std::map <const char *, uint32_t> ConstNameToSymbolIndexMap; 1812 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 1813 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 1814 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 1815 // Any symbols that get merged into another will get an entry 1816 // in this map so we know 1817 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 1818 uint32_t nlist_idx = 0; 1819 Symbol *symbol_ptr = NULL; 1820 1821 uint32_t sym_idx = 0; 1822 Symbol *sym = NULL; 1823 size_t num_syms = 0; 1824 std::string memory_symbol_name; 1825 uint32_t unmapped_local_symbols_found = 0; 1826 1827 #if defined (__APPLE__) && defined (__arm__) 1828 1829 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been optimized by moving LOCAL 1830 // symbols out of the memory mapped portion of the DSC. The symbol information has all been retained, 1831 // but it isn't available in the normal nlist data. However, there *are* duplicate entries of *some* 1832 // LOCAL symbols in the normal nlist data. To handle this situation correctly, we must first attempt 1833 // to parse any DSC unmapped symbol information. If we find any, we set a flag that tells the normal 1834 // nlist parser to ignore all LOCAL symbols. 1835 1836 if (m_header.flags & 0x80000000u) 1837 { 1838 // Before we can start mapping the DSC, we need to make certain the target process is actually 1839 // using the cache we can find. 1840 1841 // Next we need to determine the correct path for the dyld shared cache. 1842 1843 ArchSpec header_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 1844 char dsc_path[PATH_MAX]; 1845 1846 snprintf(dsc_path, sizeof(dsc_path), "%s%s%s", 1847 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR */ 1848 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 1849 header_arch.GetArchitectureName()); 1850 1851 FileSpec dsc_filespec(dsc_path, false); 1852 1853 // We need definitions of two structures in the on-disk DSC, copy them here manually 1854 struct lldb_copy_dyld_cache_header_v0 1855 { 1856 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 1857 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 1858 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 1859 uint32_t imagesOffset; 1860 uint32_t imagesCount; 1861 uint64_t dyldBaseAddress; 1862 uint64_t codeSignatureOffset; 1863 uint64_t codeSignatureSize; 1864 uint64_t slideInfoOffset; 1865 uint64_t slideInfoSize; 1866 uint64_t localSymbolsOffset; // file offset of where local symbols are stored 1867 uint64_t localSymbolsSize; // size of local symbols information 1868 }; 1869 struct lldb_copy_dyld_cache_header_v1 1870 { 1871 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 1872 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 1873 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 1874 uint32_t imagesOffset; 1875 uint32_t imagesCount; 1876 uint64_t dyldBaseAddress; 1877 uint64_t codeSignatureOffset; 1878 uint64_t codeSignatureSize; 1879 uint64_t slideInfoOffset; 1880 uint64_t slideInfoSize; 1881 uint64_t localSymbolsOffset; 1882 uint64_t localSymbolsSize; 1883 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 and later 1884 }; 1885 1886 struct lldb_copy_dyld_cache_mapping_info 1887 { 1888 uint64_t address; 1889 uint64_t size; 1890 uint64_t fileOffset; 1891 uint32_t maxProt; 1892 uint32_t initProt; 1893 }; 1894 1895 struct lldb_copy_dyld_cache_local_symbols_info 1896 { 1897 uint32_t nlistOffset; 1898 uint32_t nlistCount; 1899 uint32_t stringsOffset; 1900 uint32_t stringsSize; 1901 uint32_t entriesOffset; 1902 uint32_t entriesCount; 1903 }; 1904 struct lldb_copy_dyld_cache_local_symbols_entry 1905 { 1906 uint32_t dylibOffset; 1907 uint32_t nlistStartIndex; 1908 uint32_t nlistCount; 1909 }; 1910 1911 /* The dyld_cache_header has a pointer to the dyld_cache_local_symbols_info structure (localSymbolsOffset). 1912 The dyld_cache_local_symbols_info structure gives us three things: 1913 1. The start and count of the nlist records in the dyld_shared_cache file 1914 2. The start and size of the strings for these nlist records 1915 3. The start and count of dyld_cache_local_symbols_entry entries 1916 1917 There is one dyld_cache_local_symbols_entry per dylib/framework in the dyld shared cache. 1918 The "dylibOffset" field is the Mach-O header of this dylib/framework in the dyld shared cache. 1919 The dyld_cache_local_symbols_entry also lists the start of this dylib/framework's nlist records 1920 and the count of how many nlist records there are for this dylib/framework. 1921 */ 1922 1923 // Process the dsc header to find the unmapped symbols 1924 // 1925 // Save some VM space, do not map the entire cache in one shot. 1926 1927 DataBufferSP dsc_data_sp; 1928 dsc_data_sp = dsc_filespec.MemoryMapFileContents(0, sizeof(struct lldb_copy_dyld_cache_header_v1)); 1929 1930 if (dsc_data_sp) 1931 { 1932 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 1933 1934 char version_str[17]; 1935 int version = -1; 1936 lldb::offset_t offset = 0; 1937 memcpy (version_str, dsc_header_data.GetData (&offset, 16), 16); 1938 version_str[16] = '\0'; 1939 if (strncmp (version_str, "dyld_v", 6) == 0 && isdigit (version_str[6])) 1940 { 1941 int v; 1942 if (::sscanf (version_str + 6, "%d", &v) == 1) 1943 { 1944 version = v; 1945 } 1946 } 1947 1948 UUID dsc_uuid; 1949 if (version >= 1) 1950 { 1951 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, uuid); 1952 uint8_t uuid_bytes[sizeof (uuid_t)]; 1953 memcpy (uuid_bytes, dsc_header_data.GetData (&offset, sizeof (uuid_t)), sizeof (uuid_t)); 1954 dsc_uuid.SetBytes (uuid_bytes); 1955 } 1956 1957 bool uuid_match = true; 1958 if (dsc_uuid.IsValid() && process) 1959 { 1960 UUID shared_cache_uuid(GetProcessSharedCacheUUID(process)); 1961 1962 if (shared_cache_uuid.IsValid() && dsc_uuid != shared_cache_uuid) 1963 { 1964 // The on-disk dyld_shared_cache file is not the same as the one in this 1965 // process' memory, don't use it. 1966 uuid_match = false; 1967 ModuleSP module_sp (GetModule()); 1968 if (module_sp) 1969 module_sp->ReportWarning ("process shared cache does not match on-disk dyld_shared_cache file, some symbol names will be missing."); 1970 } 1971 } 1972 1973 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, mappingOffset); 1974 1975 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 1976 1977 // If the mappingOffset points to a location inside the header, we've 1978 // opened an old dyld shared cache, and should not proceed further. 1979 if (uuid_match && mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v0)) 1980 { 1981 1982 DataBufferSP dsc_mapping_info_data_sp = dsc_filespec.MemoryMapFileContents(mappingOffset, sizeof (struct lldb_copy_dyld_cache_mapping_info)); 1983 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, byte_order, addr_byte_size); 1984 offset = 0; 1985 1986 // The File addresses (from the in-memory Mach-O load commands) for the shared libraries 1987 // in the shared library cache need to be adjusted by an offset to match up with the 1988 // dylibOffset identifying field in the dyld_cache_local_symbol_entry's. This offset is 1989 // recorded in mapping_offset_value. 1990 const uint64_t mapping_offset_value = dsc_mapping_info_data.GetU64(&offset); 1991 1992 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 1993 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 1994 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 1995 1996 if (localSymbolsOffset && localSymbolsSize) 1997 { 1998 // Map the local symbols 1999 if (DataBufferSP dsc_local_symbols_data_sp = dsc_filespec.MemoryMapFileContents(localSymbolsOffset, localSymbolsSize)) 2000 { 2001 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, byte_order, addr_byte_size); 2002 2003 offset = 0; 2004 2005 // Read the local_symbols_infos struct in one shot 2006 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2007 dsc_local_symbols_data.GetU32(&offset, &local_symbols_info.nlistOffset, 6); 2008 2009 SectionSP text_section_sp(section_list->FindSectionByName(GetSegmentNameTEXT())); 2010 2011 uint32_t header_file_offset = (text_section_sp->GetFileAddress() - mapping_offset_value); 2012 2013 offset = local_symbols_info.entriesOffset; 2014 for (uint32_t entry_index = 0; entry_index < local_symbols_info.entriesCount; entry_index++) 2015 { 2016 struct lldb_copy_dyld_cache_local_symbols_entry local_symbols_entry; 2017 local_symbols_entry.dylibOffset = dsc_local_symbols_data.GetU32(&offset); 2018 local_symbols_entry.nlistStartIndex = dsc_local_symbols_data.GetU32(&offset); 2019 local_symbols_entry.nlistCount = dsc_local_symbols_data.GetU32(&offset); 2020 2021 if (header_file_offset == local_symbols_entry.dylibOffset) 2022 { 2023 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2024 2025 // The normal nlist code cannot correctly size the Symbols array, we need to allocate it here. 2026 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms + unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2027 num_syms = symtab->GetNumSymbols(); 2028 2029 nlist_data_offset = local_symbols_info.nlistOffset + (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2030 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2031 2032 for (uint32_t nlist_index = 0; nlist_index < local_symbols_entry.nlistCount; nlist_index++) 2033 { 2034 ///////////////////////////// 2035 { 2036 struct nlist_64 nlist; 2037 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2038 break; 2039 2040 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked(&nlist_data_offset); 2041 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2042 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2043 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked (&nlist_data_offset); 2044 nlist.n_value = dsc_local_symbols_data.GetAddress_unchecked (&nlist_data_offset); 2045 2046 SymbolType type = eSymbolTypeInvalid; 2047 const char *symbol_name = dsc_local_symbols_data.PeekCStr(string_table_offset + nlist.n_strx); 2048 2049 if (symbol_name == NULL) 2050 { 2051 // No symbol should be NULL, even the symbols with no 2052 // string values should have an offset zero which points 2053 // to an empty C-string 2054 Host::SystemLog (Host::eSystemLogError, 2055 "error: DSC unmapped local symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 2056 entry_index, 2057 nlist.n_strx, 2058 module_sp->GetFileSpec().GetPath().c_str()); 2059 continue; 2060 } 2061 if (symbol_name[0] == '\0') 2062 symbol_name = NULL; 2063 2064 const char *symbol_name_non_abi_mangled = NULL; 2065 2066 SectionSP symbol_section; 2067 uint32_t symbol_byte_size = 0; 2068 bool add_nlist = true; 2069 bool is_debug = ((nlist.n_type & NlistMaskStab) != 0); 2070 bool demangled_is_synthesized = false; 2071 bool is_gsym = false; 2072 2073 assert (sym_idx < num_syms); 2074 2075 sym[sym_idx].SetDebug (is_debug); 2076 2077 if (is_debug) 2078 { 2079 switch (nlist.n_type) 2080 { 2081 case StabGlobalSymbol: 2082 // N_GSYM -- global symbol: name,,NO_SECT,type,0 2083 // Sometimes the N_GSYM value contains the address. 2084 2085 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 2086 // have the same address, but we want to ensure that we always find only the real symbol, 2087 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 2088 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 2089 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 2090 // same address. 2091 2092 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 2093 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 2094 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 2095 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 2096 add_nlist = false; 2097 else 2098 { 2099 is_gsym = true; 2100 sym[sym_idx].SetExternal(true); 2101 if (nlist.n_value != 0) 2102 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2103 type = eSymbolTypeData; 2104 } 2105 break; 2106 2107 case StabFunctionName: 2108 // N_FNAME -- procedure name (f77 kludge): name,,NO_SECT,0,0 2109 type = eSymbolTypeCompiler; 2110 break; 2111 2112 case StabFunction: 2113 // N_FUN -- procedure: name,,n_sect,linenumber,address 2114 if (symbol_name) 2115 { 2116 type = eSymbolTypeCode; 2117 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2118 2119 N_FUN_addr_to_sym_idx[nlist.n_value] = sym_idx; 2120 // We use the current number of symbols in the symbol table in lieu of 2121 // using nlist_idx in case we ever start trimming entries out 2122 N_FUN_indexes.push_back(sym_idx); 2123 } 2124 else 2125 { 2126 type = eSymbolTypeCompiler; 2127 2128 if ( !N_FUN_indexes.empty() ) 2129 { 2130 // Copy the size of the function into the original STAB entry so we don't have 2131 // to hunt for it later 2132 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 2133 N_FUN_indexes.pop_back(); 2134 // We don't really need the end function STAB as it contains the size which 2135 // we already placed with the original symbol, so don't add it if we want a 2136 // minimal symbol table 2137 add_nlist = false; 2138 } 2139 } 2140 break; 2141 2142 case StabStaticSymbol: 2143 // N_STSYM -- static symbol: name,,n_sect,type,address 2144 N_STSYM_addr_to_sym_idx[nlist.n_value] = sym_idx; 2145 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2146 type = eSymbolTypeData; 2147 break; 2148 2149 case StabLocalCommon: 2150 // N_LCSYM -- .lcomm symbol: name,,n_sect,type,address 2151 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2152 type = eSymbolTypeCommonBlock; 2153 break; 2154 2155 case StabBeginSymbol: 2156 // N_BNSYM 2157 // We use the current number of symbols in the symbol table in lieu of 2158 // using nlist_idx in case we ever start trimming entries out 2159 // Skip these if we want minimal symbol tables 2160 add_nlist = false; 2161 break; 2162 2163 case StabEndSymbol: 2164 // N_ENSYM 2165 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 2166 // so that we can always skip the entire symbol if we need to navigate 2167 // more quickly at the source level when parsing STABS 2168 // Skip these if we want minimal symbol tables 2169 add_nlist = false; 2170 break; 2171 2172 2173 case StabSourceFileOptions: 2174 // N_OPT - emitted with gcc2_compiled and in gcc source 2175 type = eSymbolTypeCompiler; 2176 break; 2177 2178 case StabRegisterSymbol: 2179 // N_RSYM - register sym: name,,NO_SECT,type,register 2180 type = eSymbolTypeVariable; 2181 break; 2182 2183 case StabSourceLine: 2184 // N_SLINE - src line: 0,,n_sect,linenumber,address 2185 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2186 type = eSymbolTypeLineEntry; 2187 break; 2188 2189 case StabStructureType: 2190 // N_SSYM - structure elt: name,,NO_SECT,type,struct_offset 2191 type = eSymbolTypeVariableType; 2192 break; 2193 2194 case StabSourceFileName: 2195 // N_SO - source file name 2196 type = eSymbolTypeSourceFile; 2197 if (symbol_name == NULL) 2198 { 2199 add_nlist = false; 2200 if (N_SO_index != UINT32_MAX) 2201 { 2202 // Set the size of the N_SO to the terminating index of this N_SO 2203 // so that we can always skip the entire N_SO if we need to navigate 2204 // more quickly at the source level when parsing STABS 2205 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2206 symbol_ptr->SetByteSize(sym_idx); 2207 symbol_ptr->SetSizeIsSibling(true); 2208 } 2209 N_NSYM_indexes.clear(); 2210 N_INCL_indexes.clear(); 2211 N_BRAC_indexes.clear(); 2212 N_COMM_indexes.clear(); 2213 N_FUN_indexes.clear(); 2214 N_SO_index = UINT32_MAX; 2215 } 2216 else 2217 { 2218 // We use the current number of symbols in the symbol table in lieu of 2219 // using nlist_idx in case we ever start trimming entries out 2220 const bool N_SO_has_full_path = symbol_name[0] == '/'; 2221 if (N_SO_has_full_path) 2222 { 2223 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2224 { 2225 // We have two consecutive N_SO entries where the first contains a directory 2226 // and the second contains a full path. 2227 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 2228 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2229 add_nlist = false; 2230 } 2231 else 2232 { 2233 // This is the first entry in a N_SO that contains a directory or 2234 // a full path to the source file 2235 N_SO_index = sym_idx; 2236 } 2237 } 2238 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2239 { 2240 // This is usually the second N_SO entry that contains just the filename, 2241 // so here we combine it with the first one if we are minimizing the symbol table 2242 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 2243 if (so_path && so_path[0]) 2244 { 2245 std::string full_so_path (so_path); 2246 const size_t double_slash_pos = full_so_path.find("//"); 2247 if (double_slash_pos != std::string::npos) 2248 { 2249 // The linker has been generating bad N_SO entries with doubled up paths 2250 // in the format "%s%s" where the first stirng in the DW_AT_comp_dir, 2251 // and the second is the directory for the source file so you end up with 2252 // a path that looks like "/tmp/src//tmp/src/" 2253 FileSpec so_dir(so_path, false); 2254 if (!so_dir.Exists()) 2255 { 2256 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 2257 if (so_dir.Exists()) 2258 { 2259 // Trim off the incorrect path 2260 full_so_path.erase(0, double_slash_pos + 1); 2261 } 2262 } 2263 } 2264 if (*full_so_path.rbegin() != '/') 2265 full_so_path += '/'; 2266 full_so_path += symbol_name; 2267 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 2268 add_nlist = false; 2269 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2270 } 2271 } 2272 else 2273 { 2274 // This could be a relative path to a N_SO 2275 N_SO_index = sym_idx; 2276 } 2277 } 2278 break; 2279 2280 case StabObjectFileName: 2281 // N_OSO - object file name: name,,0,0,st_mtime 2282 type = eSymbolTypeObjectFile; 2283 break; 2284 2285 case StabLocalSymbol: 2286 // N_LSYM - local sym: name,,NO_SECT,type,offset 2287 type = eSymbolTypeLocal; 2288 break; 2289 2290 //---------------------------------------------------------------------- 2291 // INCL scopes 2292 //---------------------------------------------------------------------- 2293 case StabBeginIncludeFileName: 2294 // N_BINCL - include file beginning: name,,NO_SECT,0,sum 2295 // We use the current number of symbols in the symbol table in lieu of 2296 // using nlist_idx in case we ever start trimming entries out 2297 N_INCL_indexes.push_back(sym_idx); 2298 type = eSymbolTypeScopeBegin; 2299 break; 2300 2301 case StabEndIncludeFile: 2302 // N_EINCL - include file end: name,,NO_SECT,0,0 2303 // Set the size of the N_BINCL to the terminating index of this N_EINCL 2304 // so that we can always skip the entire symbol if we need to navigate 2305 // more quickly at the source level when parsing STABS 2306 if ( !N_INCL_indexes.empty() ) 2307 { 2308 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 2309 symbol_ptr->SetByteSize(sym_idx + 1); 2310 symbol_ptr->SetSizeIsSibling(true); 2311 N_INCL_indexes.pop_back(); 2312 } 2313 type = eSymbolTypeScopeEnd; 2314 break; 2315 2316 case StabIncludeFileName: 2317 // N_SOL - #included file name: name,,n_sect,0,address 2318 type = eSymbolTypeHeaderFile; 2319 2320 // We currently don't use the header files on darwin 2321 add_nlist = false; 2322 break; 2323 2324 case StabCompilerParameters: 2325 // N_PARAMS - compiler parameters: name,,NO_SECT,0,0 2326 type = eSymbolTypeCompiler; 2327 break; 2328 2329 case StabCompilerVersion: 2330 // N_VERSION - compiler version: name,,NO_SECT,0,0 2331 type = eSymbolTypeCompiler; 2332 break; 2333 2334 case StabCompilerOptLevel: 2335 // N_OLEVEL - compiler -O level: name,,NO_SECT,0,0 2336 type = eSymbolTypeCompiler; 2337 break; 2338 2339 case StabParameter: 2340 // N_PSYM - parameter: name,,NO_SECT,type,offset 2341 type = eSymbolTypeVariable; 2342 break; 2343 2344 case StabAlternateEntry: 2345 // N_ENTRY - alternate entry: name,,n_sect,linenumber,address 2346 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2347 type = eSymbolTypeLineEntry; 2348 break; 2349 2350 //---------------------------------------------------------------------- 2351 // Left and Right Braces 2352 //---------------------------------------------------------------------- 2353 case StabLeftBracket: 2354 // N_LBRAC - left bracket: 0,,NO_SECT,nesting level,address 2355 // We use the current number of symbols in the symbol table in lieu of 2356 // using nlist_idx in case we ever start trimming entries out 2357 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2358 N_BRAC_indexes.push_back(sym_idx); 2359 type = eSymbolTypeScopeBegin; 2360 break; 2361 2362 case StabRightBracket: 2363 // N_RBRAC - right bracket: 0,,NO_SECT,nesting level,address 2364 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 2365 // so that we can always skip the entire symbol if we need to navigate 2366 // more quickly at the source level when parsing STABS 2367 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2368 if ( !N_BRAC_indexes.empty() ) 2369 { 2370 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 2371 symbol_ptr->SetByteSize(sym_idx + 1); 2372 symbol_ptr->SetSizeIsSibling(true); 2373 N_BRAC_indexes.pop_back(); 2374 } 2375 type = eSymbolTypeScopeEnd; 2376 break; 2377 2378 case StabDeletedIncludeFile: 2379 // N_EXCL - deleted include file: name,,NO_SECT,0,sum 2380 type = eSymbolTypeHeaderFile; 2381 break; 2382 2383 //---------------------------------------------------------------------- 2384 // COMM scopes 2385 //---------------------------------------------------------------------- 2386 case StabBeginCommon: 2387 // N_BCOMM - begin common: name,,NO_SECT,0,0 2388 // We use the current number of symbols in the symbol table in lieu of 2389 // using nlist_idx in case we ever start trimming entries out 2390 type = eSymbolTypeScopeBegin; 2391 N_COMM_indexes.push_back(sym_idx); 2392 break; 2393 2394 case StabEndCommonLocal: 2395 // N_ECOML - end common (local name): 0,,n_sect,0,address 2396 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2397 // Fall through 2398 2399 case StabEndCommon: 2400 // N_ECOMM - end common: name,,n_sect,0,0 2401 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 2402 // so that we can always skip the entire symbol if we need to navigate 2403 // more quickly at the source level when parsing STABS 2404 if ( !N_COMM_indexes.empty() ) 2405 { 2406 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 2407 symbol_ptr->SetByteSize(sym_idx + 1); 2408 symbol_ptr->SetSizeIsSibling(true); 2409 N_COMM_indexes.pop_back(); 2410 } 2411 type = eSymbolTypeScopeEnd; 2412 break; 2413 2414 case StabLength: 2415 // N_LENG - second stab entry with length information 2416 type = eSymbolTypeAdditional; 2417 break; 2418 2419 default: break; 2420 } 2421 } 2422 else 2423 { 2424 //uint8_t n_pext = NlistMaskPrivateExternal & nlist.n_type; 2425 uint8_t n_type = NlistMaskType & nlist.n_type; 2426 sym[sym_idx].SetExternal((NlistMaskExternal & nlist.n_type) != 0); 2427 2428 switch (n_type) 2429 { 2430 case NListTypeIndirect: // N_INDR - Fall through 2431 case NListTypePreboundUndefined:// N_PBUD - Fall through 2432 case NListTypeUndefined: // N_UNDF 2433 type = eSymbolTypeUndefined; 2434 break; 2435 2436 case NListTypeAbsolute: // N_ABS 2437 type = eSymbolTypeAbsolute; 2438 break; 2439 2440 case NListTypeSection: // N_SECT 2441 { 2442 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2443 2444 if (symbol_section == NULL) 2445 { 2446 // TODO: warn about this? 2447 add_nlist = false; 2448 break; 2449 } 2450 2451 if (TEXT_eh_frame_sectID == nlist.n_sect) 2452 { 2453 type = eSymbolTypeException; 2454 } 2455 else 2456 { 2457 uint32_t section_type = symbol_section->Get() & SectionFlagMaskSectionType; 2458 2459 switch (section_type) 2460 { 2461 case SectionTypeRegular: break; // regular section 2462 //case SectionTypeZeroFill: type = eSymbolTypeData; break; // zero fill on demand section 2463 case SectionTypeCStringLiterals: type = eSymbolTypeData; break; // section with only literal C strings 2464 case SectionType4ByteLiterals: type = eSymbolTypeData; break; // section with only 4 byte literals 2465 case SectionType8ByteLiterals: type = eSymbolTypeData; break; // section with only 8 byte literals 2466 case SectionTypeLiteralPointers: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 2467 case SectionTypeNonLazySymbolPointers: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 2468 case SectionTypeLazySymbolPointers: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 2469 case SectionTypeSymbolStubs: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 2470 case SectionTypeModuleInitFunctionPointers: type = eSymbolTypeCode; break; // section with only function pointers for initialization 2471 case SectionTypeModuleTermFunctionPointers: type = eSymbolTypeCode; break; // section with only function pointers for termination 2472 //case SectionTypeCoalesced: type = eSymbolType; break; // section contains symbols that are to be coalesced 2473 //case SectionTypeZeroFillLarge: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 2474 case SectionTypeInterposing: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 2475 case SectionType16ByteLiterals: type = eSymbolTypeData; break; // section with only 16 byte literals 2476 case SectionTypeDTraceObjectFormat: type = eSymbolTypeInstrumentation; break; 2477 case SectionTypeLazyDylibSymbolPointers: type = eSymbolTypeTrampoline; break; 2478 default: break; 2479 } 2480 2481 if (type == eSymbolTypeInvalid) 2482 { 2483 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 2484 if (symbol_section->IsDescendant (text_section_sp.get())) 2485 { 2486 if (symbol_section->IsClear(SectionAttrUserPureInstructions | 2487 SectionAttrUserSelfModifyingCode | 2488 SectionAttrSytemSomeInstructions)) 2489 type = eSymbolTypeData; 2490 else 2491 type = eSymbolTypeCode; 2492 } 2493 else if (symbol_section->IsDescendant(data_section_sp.get())) 2494 { 2495 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 2496 { 2497 type = eSymbolTypeRuntime; 2498 2499 if (symbol_name && 2500 symbol_name[0] == '_' && 2501 symbol_name[1] == 'O' && 2502 symbol_name[2] == 'B') 2503 { 2504 llvm::StringRef symbol_name_ref(symbol_name); 2505 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 2506 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 2507 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 2508 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 2509 { 2510 symbol_name_non_abi_mangled = symbol_name + 1; 2511 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2512 type = eSymbolTypeObjCClass; 2513 demangled_is_synthesized = true; 2514 } 2515 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 2516 { 2517 symbol_name_non_abi_mangled = symbol_name + 1; 2518 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2519 type = eSymbolTypeObjCMetaClass; 2520 demangled_is_synthesized = true; 2521 } 2522 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 2523 { 2524 symbol_name_non_abi_mangled = symbol_name + 1; 2525 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2526 type = eSymbolTypeObjCIVar; 2527 demangled_is_synthesized = true; 2528 } 2529 } 2530 } 2531 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 2532 { 2533 type = eSymbolTypeException; 2534 } 2535 else 2536 { 2537 type = eSymbolTypeData; 2538 } 2539 } 2540 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 2541 { 2542 type = eSymbolTypeTrampoline; 2543 } 2544 else if (symbol_section->IsDescendant(objc_section_sp.get())) 2545 { 2546 type = eSymbolTypeRuntime; 2547 if (symbol_name && symbol_name[0] == '.') 2548 { 2549 llvm::StringRef symbol_name_ref(symbol_name); 2550 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 2551 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 2552 { 2553 symbol_name_non_abi_mangled = symbol_name; 2554 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 2555 type = eSymbolTypeObjCClass; 2556 demangled_is_synthesized = true; 2557 } 2558 } 2559 } 2560 } 2561 } 2562 } 2563 break; 2564 } 2565 } 2566 2567 if (add_nlist) 2568 { 2569 uint64_t symbol_value = nlist.n_value; 2570 if (symbol_name_non_abi_mangled) 2571 { 2572 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 2573 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 2574 } 2575 else 2576 { 2577 bool symbol_name_is_mangled = false; 2578 2579 if (symbol_name && symbol_name[0] == '_') 2580 { 2581 symbol_name_is_mangled = symbol_name[1] == '_'; 2582 symbol_name++; // Skip the leading underscore 2583 } 2584 2585 if (symbol_name) 2586 { 2587 ConstString const_symbol_name(symbol_name); 2588 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 2589 if (is_gsym && is_debug) 2590 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 2591 } 2592 } 2593 if (symbol_section) 2594 { 2595 const addr_t section_file_addr = symbol_section->GetFileAddress(); 2596 if (symbol_byte_size == 0 && function_starts_count > 0) 2597 { 2598 addr_t symbol_lookup_file_addr = nlist.n_value; 2599 // Do an exact address match for non-ARM addresses, else get the closest since 2600 // the symbol might be a thumb symbol which has an address with bit zero set 2601 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 2602 if (is_arm && func_start_entry) 2603 { 2604 // Verify that the function start address is the symbol address (ARM) 2605 // or the symbol address + 1 (thumb) 2606 if (func_start_entry->addr != symbol_lookup_file_addr && 2607 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 2608 { 2609 // Not the right entry, NULL it out... 2610 func_start_entry = NULL; 2611 } 2612 } 2613 if (func_start_entry) 2614 { 2615 func_start_entry->data = true; 2616 2617 addr_t symbol_file_addr = func_start_entry->addr; 2618 uint32_t symbol_flags = 0; 2619 if (is_arm) 2620 { 2621 if (symbol_file_addr & 1) 2622 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 2623 symbol_file_addr &= 0xfffffffffffffffeull; 2624 } 2625 2626 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 2627 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 2628 if (next_func_start_entry) 2629 { 2630 addr_t next_symbol_file_addr = next_func_start_entry->addr; 2631 // Be sure the clear the Thumb address bit when we calculate the size 2632 // from the current and next address 2633 if (is_arm) 2634 next_symbol_file_addr &= 0xfffffffffffffffeull; 2635 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 2636 } 2637 else 2638 { 2639 symbol_byte_size = section_end_file_addr - symbol_file_addr; 2640 } 2641 } 2642 } 2643 symbol_value -= section_file_addr; 2644 } 2645 2646 if (is_debug == false) 2647 { 2648 if (type == eSymbolTypeCode) 2649 { 2650 // See if we can find a N_FUN entry for any code symbols. 2651 // If we do find a match, and the name matches, then we 2652 // can merge the two into just the function symbol to avoid 2653 // duplicate entries in the symbol table 2654 ValueToSymbolIndexMap::const_iterator pos = N_FUN_addr_to_sym_idx.find (nlist.n_value); 2655 if (pos != N_FUN_addr_to_sym_idx.end()) 2656 { 2657 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 2658 { 2659 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 2660 // We just need the flags from the linker symbol, so put these flags 2661 // into the N_FUN flags to avoid duplicate symbols in the symbol table 2662 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2663 sym[sym_idx].Clear(); 2664 continue; 2665 } 2666 } 2667 } 2668 else if (type == eSymbolTypeData) 2669 { 2670 // See if we can find a N_STSYM entry for any data symbols. 2671 // If we do find a match, and the name matches, then we 2672 // can merge the two into just the Static symbol to avoid 2673 // duplicate entries in the symbol table 2674 ValueToSymbolIndexMap::const_iterator pos = N_STSYM_addr_to_sym_idx.find (nlist.n_value); 2675 if (pos != N_STSYM_addr_to_sym_idx.end()) 2676 { 2677 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 2678 { 2679 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 2680 // We just need the flags from the linker symbol, so put these flags 2681 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 2682 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2683 sym[sym_idx].Clear(); 2684 continue; 2685 } 2686 } 2687 else 2688 { 2689 // Combine N_GSYM stab entries with the non stab symbol 2690 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 2691 if (pos != N_GSYM_name_to_sym_idx.end()) 2692 { 2693 const uint32_t GSYM_sym_idx = pos->second; 2694 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 2695 // Copy the address, because often the N_GSYM address has an invalid address of zero 2696 // when the global is a common symbol 2697 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 2698 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 2699 // We just need the flags from the linker symbol, so put these flags 2700 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 2701 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2702 sym[sym_idx].Clear(); 2703 continue; 2704 } 2705 } 2706 } 2707 } 2708 2709 sym[sym_idx].SetID (nlist_idx); 2710 sym[sym_idx].SetType (type); 2711 sym[sym_idx].GetAddress().SetSection (symbol_section); 2712 sym[sym_idx].GetAddress().SetOffset (symbol_value); 2713 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2714 2715 if (symbol_byte_size > 0) 2716 sym[sym_idx].SetByteSize(symbol_byte_size); 2717 2718 if (demangled_is_synthesized) 2719 sym[sym_idx].SetDemangledNameIsSynthesized(true); 2720 ++sym_idx; 2721 } 2722 else 2723 { 2724 sym[sym_idx].Clear(); 2725 } 2726 2727 } 2728 ///////////////////////////// 2729 } 2730 break; // No more entries to consider 2731 } 2732 } 2733 } 2734 } 2735 } 2736 } 2737 } 2738 2739 // Must reset this in case it was mutated above! 2740 nlist_data_offset = 0; 2741 #endif 2742 2743 if (nlist_data.GetByteSize() > 0) 2744 { 2745 2746 // If the sym array was not created while parsing the DSC unmapped 2747 // symbols, create it now. 2748 if (sym == NULL) 2749 { 2750 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 2751 num_syms = symtab->GetNumSymbols(); 2752 } 2753 2754 if (unmapped_local_symbols_found) 2755 { 2756 assert(m_dysymtab.ilocalsym == 0); 2757 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 2758 nlist_idx = m_dysymtab.nlocalsym; 2759 } 2760 else 2761 { 2762 nlist_idx = 0; 2763 } 2764 2765 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) 2766 { 2767 struct nlist_64 nlist; 2768 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2769 break; 2770 2771 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 2772 nlist.n_type = nlist_data.GetU8_unchecked (&nlist_data_offset); 2773 nlist.n_sect = nlist_data.GetU8_unchecked (&nlist_data_offset); 2774 nlist.n_desc = nlist_data.GetU16_unchecked (&nlist_data_offset); 2775 nlist.n_value = nlist_data.GetAddress_unchecked (&nlist_data_offset); 2776 2777 SymbolType type = eSymbolTypeInvalid; 2778 const char *symbol_name = NULL; 2779 2780 if (have_strtab_data) 2781 { 2782 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 2783 2784 if (symbol_name == NULL) 2785 { 2786 // No symbol should be NULL, even the symbols with no 2787 // string values should have an offset zero which points 2788 // to an empty C-string 2789 Host::SystemLog (Host::eSystemLogError, 2790 "error: symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 2791 nlist_idx, 2792 nlist.n_strx, 2793 module_sp->GetFileSpec().GetPath().c_str()); 2794 continue; 2795 } 2796 if (symbol_name[0] == '\0') 2797 symbol_name = NULL; 2798 } 2799 else 2800 { 2801 const addr_t str_addr = strtab_addr + nlist.n_strx; 2802 Error str_error; 2803 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, str_error)) 2804 symbol_name = memory_symbol_name.c_str(); 2805 } 2806 const char *symbol_name_non_abi_mangled = NULL; 2807 2808 SectionSP symbol_section; 2809 lldb::addr_t symbol_byte_size = 0; 2810 bool add_nlist = true; 2811 bool is_gsym = false; 2812 bool is_debug = ((nlist.n_type & NlistMaskStab) != 0); 2813 bool demangled_is_synthesized = false; 2814 2815 assert (sym_idx < num_syms); 2816 2817 sym[sym_idx].SetDebug (is_debug); 2818 2819 if (is_debug) 2820 { 2821 switch (nlist.n_type) 2822 { 2823 case StabGlobalSymbol: 2824 // N_GSYM -- global symbol: name,,NO_SECT,type,0 2825 // Sometimes the N_GSYM value contains the address. 2826 2827 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 2828 // have the same address, but we want to ensure that we always find only the real symbol, 2829 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 2830 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 2831 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 2832 // same address. 2833 2834 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 2835 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 2836 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 2837 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 2838 add_nlist = false; 2839 else 2840 { 2841 is_gsym = true; 2842 sym[sym_idx].SetExternal(true); 2843 if (nlist.n_value != 0) 2844 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2845 type = eSymbolTypeData; 2846 } 2847 break; 2848 2849 case StabFunctionName: 2850 // N_FNAME -- procedure name (f77 kludge): name,,NO_SECT,0,0 2851 type = eSymbolTypeCompiler; 2852 break; 2853 2854 case StabFunction: 2855 // N_FUN -- procedure: name,,n_sect,linenumber,address 2856 if (symbol_name) 2857 { 2858 type = eSymbolTypeCode; 2859 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2860 2861 N_FUN_addr_to_sym_idx[nlist.n_value] = sym_idx; 2862 // We use the current number of symbols in the symbol table in lieu of 2863 // using nlist_idx in case we ever start trimming entries out 2864 N_FUN_indexes.push_back(sym_idx); 2865 } 2866 else 2867 { 2868 type = eSymbolTypeCompiler; 2869 2870 if ( !N_FUN_indexes.empty() ) 2871 { 2872 // Copy the size of the function into the original STAB entry so we don't have 2873 // to hunt for it later 2874 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 2875 N_FUN_indexes.pop_back(); 2876 // We don't really need the end function STAB as it contains the size which 2877 // we already placed with the original symbol, so don't add it if we want a 2878 // minimal symbol table 2879 add_nlist = false; 2880 } 2881 } 2882 break; 2883 2884 case StabStaticSymbol: 2885 // N_STSYM -- static symbol: name,,n_sect,type,address 2886 N_STSYM_addr_to_sym_idx[nlist.n_value] = sym_idx; 2887 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2888 type = eSymbolTypeData; 2889 break; 2890 2891 case StabLocalCommon: 2892 // N_LCSYM -- .lcomm symbol: name,,n_sect,type,address 2893 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2894 type = eSymbolTypeCommonBlock; 2895 break; 2896 2897 case StabBeginSymbol: 2898 // N_BNSYM 2899 // We use the current number of symbols in the symbol table in lieu of 2900 // using nlist_idx in case we ever start trimming entries out 2901 // Skip these if we want minimal symbol tables 2902 add_nlist = false; 2903 break; 2904 2905 case StabEndSymbol: 2906 // N_ENSYM 2907 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 2908 // so that we can always skip the entire symbol if we need to navigate 2909 // more quickly at the source level when parsing STABS 2910 // Skip these if we want minimal symbol tables 2911 add_nlist = false; 2912 break; 2913 2914 2915 case StabSourceFileOptions: 2916 // N_OPT - emitted with gcc2_compiled and in gcc source 2917 type = eSymbolTypeCompiler; 2918 break; 2919 2920 case StabRegisterSymbol: 2921 // N_RSYM - register sym: name,,NO_SECT,type,register 2922 type = eSymbolTypeVariable; 2923 break; 2924 2925 case StabSourceLine: 2926 // N_SLINE - src line: 0,,n_sect,linenumber,address 2927 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2928 type = eSymbolTypeLineEntry; 2929 break; 2930 2931 case StabStructureType: 2932 // N_SSYM - structure elt: name,,NO_SECT,type,struct_offset 2933 type = eSymbolTypeVariableType; 2934 break; 2935 2936 case StabSourceFileName: 2937 // N_SO - source file name 2938 type = eSymbolTypeSourceFile; 2939 if (symbol_name == NULL) 2940 { 2941 add_nlist = false; 2942 if (N_SO_index != UINT32_MAX) 2943 { 2944 // Set the size of the N_SO to the terminating index of this N_SO 2945 // so that we can always skip the entire N_SO if we need to navigate 2946 // more quickly at the source level when parsing STABS 2947 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2948 symbol_ptr->SetByteSize(sym_idx); 2949 symbol_ptr->SetSizeIsSibling(true); 2950 } 2951 N_NSYM_indexes.clear(); 2952 N_INCL_indexes.clear(); 2953 N_BRAC_indexes.clear(); 2954 N_COMM_indexes.clear(); 2955 N_FUN_indexes.clear(); 2956 N_SO_index = UINT32_MAX; 2957 } 2958 else 2959 { 2960 // We use the current number of symbols in the symbol table in lieu of 2961 // using nlist_idx in case we ever start trimming entries out 2962 const bool N_SO_has_full_path = symbol_name[0] == '/'; 2963 if (N_SO_has_full_path) 2964 { 2965 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2966 { 2967 // We have two consecutive N_SO entries where the first contains a directory 2968 // and the second contains a full path. 2969 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 2970 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2971 add_nlist = false; 2972 } 2973 else 2974 { 2975 // This is the first entry in a N_SO that contains a directory or 2976 // a full path to the source file 2977 N_SO_index = sym_idx; 2978 } 2979 } 2980 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2981 { 2982 // This is usually the second N_SO entry that contains just the filename, 2983 // so here we combine it with the first one if we are minimizing the symbol table 2984 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 2985 if (so_path && so_path[0]) 2986 { 2987 std::string full_so_path (so_path); 2988 const size_t double_slash_pos = full_so_path.find("//"); 2989 if (double_slash_pos != std::string::npos) 2990 { 2991 // The linker has been generating bad N_SO entries with doubled up paths 2992 // in the format "%s%s" where the first stirng in the DW_AT_comp_dir, 2993 // and the second is the directory for the source file so you end up with 2994 // a path that looks like "/tmp/src//tmp/src/" 2995 FileSpec so_dir(so_path, false); 2996 if (!so_dir.Exists()) 2997 { 2998 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 2999 if (so_dir.Exists()) 3000 { 3001 // Trim off the incorrect path 3002 full_so_path.erase(0, double_slash_pos + 1); 3003 } 3004 } 3005 } 3006 if (*full_so_path.rbegin() != '/') 3007 full_so_path += '/'; 3008 full_so_path += symbol_name; 3009 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 3010 add_nlist = false; 3011 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3012 } 3013 } 3014 else 3015 { 3016 // This could be a relative path to a N_SO 3017 N_SO_index = sym_idx; 3018 } 3019 } 3020 3021 break; 3022 3023 case StabObjectFileName: 3024 // N_OSO - object file name: name,,0,0,st_mtime 3025 type = eSymbolTypeObjectFile; 3026 break; 3027 3028 case StabLocalSymbol: 3029 // N_LSYM - local sym: name,,NO_SECT,type,offset 3030 type = eSymbolTypeLocal; 3031 break; 3032 3033 //---------------------------------------------------------------------- 3034 // INCL scopes 3035 //---------------------------------------------------------------------- 3036 case StabBeginIncludeFileName: 3037 // N_BINCL - include file beginning: name,,NO_SECT,0,sum 3038 // We use the current number of symbols in the symbol table in lieu of 3039 // using nlist_idx in case we ever start trimming entries out 3040 N_INCL_indexes.push_back(sym_idx); 3041 type = eSymbolTypeScopeBegin; 3042 break; 3043 3044 case StabEndIncludeFile: 3045 // N_EINCL - include file end: name,,NO_SECT,0,0 3046 // Set the size of the N_BINCL to the terminating index of this N_EINCL 3047 // so that we can always skip the entire symbol if we need to navigate 3048 // more quickly at the source level when parsing STABS 3049 if ( !N_INCL_indexes.empty() ) 3050 { 3051 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3052 symbol_ptr->SetByteSize(sym_idx + 1); 3053 symbol_ptr->SetSizeIsSibling(true); 3054 N_INCL_indexes.pop_back(); 3055 } 3056 type = eSymbolTypeScopeEnd; 3057 break; 3058 3059 case StabIncludeFileName: 3060 // N_SOL - #included file name: name,,n_sect,0,address 3061 type = eSymbolTypeHeaderFile; 3062 3063 // We currently don't use the header files on darwin 3064 add_nlist = false; 3065 break; 3066 3067 case StabCompilerParameters: 3068 // N_PARAMS - compiler parameters: name,,NO_SECT,0,0 3069 type = eSymbolTypeCompiler; 3070 break; 3071 3072 case StabCompilerVersion: 3073 // N_VERSION - compiler version: name,,NO_SECT,0,0 3074 type = eSymbolTypeCompiler; 3075 break; 3076 3077 case StabCompilerOptLevel: 3078 // N_OLEVEL - compiler -O level: name,,NO_SECT,0,0 3079 type = eSymbolTypeCompiler; 3080 break; 3081 3082 case StabParameter: 3083 // N_PSYM - parameter: name,,NO_SECT,type,offset 3084 type = eSymbolTypeVariable; 3085 break; 3086 3087 case StabAlternateEntry: 3088 // N_ENTRY - alternate entry: name,,n_sect,linenumber,address 3089 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3090 type = eSymbolTypeLineEntry; 3091 break; 3092 3093 //---------------------------------------------------------------------- 3094 // Left and Right Braces 3095 //---------------------------------------------------------------------- 3096 case StabLeftBracket: 3097 // N_LBRAC - left bracket: 0,,NO_SECT,nesting level,address 3098 // We use the current number of symbols in the symbol table in lieu of 3099 // using nlist_idx in case we ever start trimming entries out 3100 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3101 N_BRAC_indexes.push_back(sym_idx); 3102 type = eSymbolTypeScopeBegin; 3103 break; 3104 3105 case StabRightBracket: 3106 // N_RBRAC - right bracket: 0,,NO_SECT,nesting level,address 3107 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 3108 // so that we can always skip the entire symbol if we need to navigate 3109 // more quickly at the source level when parsing STABS 3110 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3111 if ( !N_BRAC_indexes.empty() ) 3112 { 3113 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3114 symbol_ptr->SetByteSize(sym_idx + 1); 3115 symbol_ptr->SetSizeIsSibling(true); 3116 N_BRAC_indexes.pop_back(); 3117 } 3118 type = eSymbolTypeScopeEnd; 3119 break; 3120 3121 case StabDeletedIncludeFile: 3122 // N_EXCL - deleted include file: name,,NO_SECT,0,sum 3123 type = eSymbolTypeHeaderFile; 3124 break; 3125 3126 //---------------------------------------------------------------------- 3127 // COMM scopes 3128 //---------------------------------------------------------------------- 3129 case StabBeginCommon: 3130 // N_BCOMM - begin common: name,,NO_SECT,0,0 3131 // We use the current number of symbols in the symbol table in lieu of 3132 // using nlist_idx in case we ever start trimming entries out 3133 type = eSymbolTypeScopeBegin; 3134 N_COMM_indexes.push_back(sym_idx); 3135 break; 3136 3137 case StabEndCommonLocal: 3138 // N_ECOML - end common (local name): 0,,n_sect,0,address 3139 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3140 // Fall through 3141 3142 case StabEndCommon: 3143 // N_ECOMM - end common: name,,n_sect,0,0 3144 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 3145 // so that we can always skip the entire symbol if we need to navigate 3146 // more quickly at the source level when parsing STABS 3147 if ( !N_COMM_indexes.empty() ) 3148 { 3149 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 3150 symbol_ptr->SetByteSize(sym_idx + 1); 3151 symbol_ptr->SetSizeIsSibling(true); 3152 N_COMM_indexes.pop_back(); 3153 } 3154 type = eSymbolTypeScopeEnd; 3155 break; 3156 3157 case StabLength: 3158 // N_LENG - second stab entry with length information 3159 type = eSymbolTypeAdditional; 3160 break; 3161 3162 default: break; 3163 } 3164 } 3165 else 3166 { 3167 //uint8_t n_pext = NlistMaskPrivateExternal & nlist.n_type; 3168 uint8_t n_type = NlistMaskType & nlist.n_type; 3169 sym[sym_idx].SetExternal((NlistMaskExternal & nlist.n_type) != 0); 3170 3171 switch (n_type) 3172 { 3173 case NListTypeIndirect: // N_INDR - Fall through 3174 case NListTypePreboundUndefined:// N_PBUD - Fall through 3175 case NListTypeUndefined: // N_UNDF 3176 type = eSymbolTypeUndefined; 3177 break; 3178 3179 case NListTypeAbsolute: // N_ABS 3180 type = eSymbolTypeAbsolute; 3181 break; 3182 3183 case NListTypeSection: // N_SECT 3184 { 3185 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3186 3187 if (!symbol_section) 3188 { 3189 // TODO: warn about this? 3190 add_nlist = false; 3191 break; 3192 } 3193 3194 if (TEXT_eh_frame_sectID == nlist.n_sect) 3195 { 3196 type = eSymbolTypeException; 3197 } 3198 else 3199 { 3200 uint32_t section_type = symbol_section->Get() & SectionFlagMaskSectionType; 3201 3202 switch (section_type) 3203 { 3204 case SectionTypeRegular: break; // regular section 3205 //case SectionTypeZeroFill: type = eSymbolTypeData; break; // zero fill on demand section 3206 case SectionTypeCStringLiterals: type = eSymbolTypeData; break; // section with only literal C strings 3207 case SectionType4ByteLiterals: type = eSymbolTypeData; break; // section with only 4 byte literals 3208 case SectionType8ByteLiterals: type = eSymbolTypeData; break; // section with only 8 byte literals 3209 case SectionTypeLiteralPointers: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 3210 case SectionTypeNonLazySymbolPointers: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 3211 case SectionTypeLazySymbolPointers: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 3212 case SectionTypeSymbolStubs: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 3213 case SectionTypeModuleInitFunctionPointers: type = eSymbolTypeCode; break; // section with only function pointers for initialization 3214 case SectionTypeModuleTermFunctionPointers: type = eSymbolTypeCode; break; // section with only function pointers for termination 3215 //case SectionTypeCoalesced: type = eSymbolType; break; // section contains symbols that are to be coalesced 3216 //case SectionTypeZeroFillLarge: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 3217 case SectionTypeInterposing: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 3218 case SectionType16ByteLiterals: type = eSymbolTypeData; break; // section with only 16 byte literals 3219 case SectionTypeDTraceObjectFormat: type = eSymbolTypeInstrumentation; break; 3220 case SectionTypeLazyDylibSymbolPointers: type = eSymbolTypeTrampoline; break; 3221 default: break; 3222 } 3223 3224 if (type == eSymbolTypeInvalid) 3225 { 3226 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 3227 if (symbol_section->IsDescendant (text_section_sp.get())) 3228 { 3229 if (symbol_section->IsClear(SectionAttrUserPureInstructions | 3230 SectionAttrUserSelfModifyingCode | 3231 SectionAttrSytemSomeInstructions)) 3232 type = eSymbolTypeData; 3233 else 3234 type = eSymbolTypeCode; 3235 } 3236 else 3237 if (symbol_section->IsDescendant(data_section_sp.get())) 3238 { 3239 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 3240 { 3241 type = eSymbolTypeRuntime; 3242 3243 if (symbol_name && 3244 symbol_name[0] == '_' && 3245 symbol_name[1] == 'O' && 3246 symbol_name[2] == 'B') 3247 { 3248 llvm::StringRef symbol_name_ref(symbol_name); 3249 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 3250 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 3251 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 3252 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 3253 { 3254 symbol_name_non_abi_mangled = symbol_name + 1; 3255 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3256 type = eSymbolTypeObjCClass; 3257 demangled_is_synthesized = true; 3258 } 3259 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 3260 { 3261 symbol_name_non_abi_mangled = symbol_name + 1; 3262 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3263 type = eSymbolTypeObjCMetaClass; 3264 demangled_is_synthesized = true; 3265 } 3266 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 3267 { 3268 symbol_name_non_abi_mangled = symbol_name + 1; 3269 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3270 type = eSymbolTypeObjCIVar; 3271 demangled_is_synthesized = true; 3272 } 3273 } 3274 } 3275 else 3276 if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 3277 { 3278 type = eSymbolTypeException; 3279 } 3280 else 3281 { 3282 type = eSymbolTypeData; 3283 } 3284 } 3285 else 3286 if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 3287 { 3288 type = eSymbolTypeTrampoline; 3289 } 3290 else 3291 if (symbol_section->IsDescendant(objc_section_sp.get())) 3292 { 3293 type = eSymbolTypeRuntime; 3294 if (symbol_name && symbol_name[0] == '.') 3295 { 3296 llvm::StringRef symbol_name_ref(symbol_name); 3297 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 3298 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 3299 { 3300 symbol_name_non_abi_mangled = symbol_name; 3301 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 3302 type = eSymbolTypeObjCClass; 3303 demangled_is_synthesized = true; 3304 } 3305 } 3306 } 3307 } 3308 } 3309 } 3310 break; 3311 } 3312 } 3313 3314 if (add_nlist) 3315 { 3316 uint64_t symbol_value = nlist.n_value; 3317 3318 if (symbol_name_non_abi_mangled) 3319 { 3320 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 3321 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 3322 } 3323 else 3324 { 3325 bool symbol_name_is_mangled = false; 3326 3327 if (symbol_name && symbol_name[0] == '_') 3328 { 3329 symbol_name_is_mangled = symbol_name[1] == '_'; 3330 symbol_name++; // Skip the leading underscore 3331 } 3332 3333 if (symbol_name) 3334 { 3335 ConstString const_symbol_name(symbol_name); 3336 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 3337 if (is_gsym && is_debug) 3338 { 3339 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 3340 } 3341 } 3342 } 3343 if (symbol_section) 3344 { 3345 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3346 if (symbol_byte_size == 0 && function_starts_count > 0) 3347 { 3348 addr_t symbol_lookup_file_addr = nlist.n_value; 3349 // Do an exact address match for non-ARM addresses, else get the closest since 3350 // the symbol might be a thumb symbol which has an address with bit zero set 3351 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 3352 if (is_arm && func_start_entry) 3353 { 3354 // Verify that the function start address is the symbol address (ARM) 3355 // or the symbol address + 1 (thumb) 3356 if (func_start_entry->addr != symbol_lookup_file_addr && 3357 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 3358 { 3359 // Not the right entry, NULL it out... 3360 func_start_entry = NULL; 3361 } 3362 } 3363 if (func_start_entry) 3364 { 3365 func_start_entry->data = true; 3366 3367 addr_t symbol_file_addr = func_start_entry->addr; 3368 if (is_arm) 3369 symbol_file_addr &= 0xfffffffffffffffeull; 3370 3371 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3372 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3373 if (next_func_start_entry) 3374 { 3375 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3376 // Be sure the clear the Thumb address bit when we calculate the size 3377 // from the current and next address 3378 if (is_arm) 3379 next_symbol_file_addr &= 0xfffffffffffffffeull; 3380 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3381 } 3382 else 3383 { 3384 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3385 } 3386 } 3387 } 3388 symbol_value -= section_file_addr; 3389 } 3390 3391 if (is_debug == false) 3392 { 3393 if (type == eSymbolTypeCode) 3394 { 3395 // See if we can find a N_FUN entry for any code symbols. 3396 // If we do find a match, and the name matches, then we 3397 // can merge the two into just the function symbol to avoid 3398 // duplicate entries in the symbol table 3399 ValueToSymbolIndexMap::const_iterator pos = N_FUN_addr_to_sym_idx.find (nlist.n_value); 3400 if (pos != N_FUN_addr_to_sym_idx.end()) 3401 { 3402 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3403 { 3404 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3405 // We just need the flags from the linker symbol, so put these flags 3406 // into the N_FUN flags to avoid duplicate symbols in the symbol table 3407 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3408 sym[sym_idx].Clear(); 3409 continue; 3410 } 3411 } 3412 } 3413 else if (type == eSymbolTypeData) 3414 { 3415 // See if we can find a N_STSYM entry for any data symbols. 3416 // If we do find a match, and the name matches, then we 3417 // can merge the two into just the Static symbol to avoid 3418 // duplicate entries in the symbol table 3419 ValueToSymbolIndexMap::const_iterator pos = N_STSYM_addr_to_sym_idx.find (nlist.n_value); 3420 if (pos != N_STSYM_addr_to_sym_idx.end()) 3421 { 3422 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3423 { 3424 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3425 // We just need the flags from the linker symbol, so put these flags 3426 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3427 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3428 sym[sym_idx].Clear(); 3429 continue; 3430 } 3431 } 3432 else 3433 { 3434 // Combine N_GSYM stab entries with the non stab symbol 3435 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 3436 if (pos != N_GSYM_name_to_sym_idx.end()) 3437 { 3438 const uint32_t GSYM_sym_idx = pos->second; 3439 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 3440 // Copy the address, because often the N_GSYM address has an invalid address of zero 3441 // when the global is a common symbol 3442 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 3443 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 3444 // We just need the flags from the linker symbol, so put these flags 3445 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3446 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3447 sym[sym_idx].Clear(); 3448 continue; 3449 } 3450 } 3451 } 3452 } 3453 3454 sym[sym_idx].SetID (nlist_idx); 3455 sym[sym_idx].SetType (type); 3456 sym[sym_idx].GetAddress().SetSection (symbol_section); 3457 sym[sym_idx].GetAddress().SetOffset (symbol_value); 3458 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3459 3460 if (symbol_byte_size > 0) 3461 sym[sym_idx].SetByteSize(symbol_byte_size); 3462 3463 if (demangled_is_synthesized) 3464 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3465 3466 ++sym_idx; 3467 } 3468 else 3469 { 3470 sym[sym_idx].Clear(); 3471 } 3472 3473 } 3474 3475 // STAB N_GSYM entries end up having a symbol type eSymbolTypeGlobal and when the symbol value 3476 // is zero, the address of the global ends up being in a non-STAB entry. Try and fix up all 3477 // such entries by figuring out what the address for the global is by looking up this non-STAB 3478 // entry and copying the value into the debug symbol's value to save us the hassle in the 3479 // debug symbol parser. 3480 3481 Symbol *global_symbol = NULL; 3482 for (nlist_idx = 0; 3483 nlist_idx < symtab_load_command.nsyms && (global_symbol = symtab->FindSymbolWithType (eSymbolTypeData, Symtab::eDebugYes, Symtab::eVisibilityAny, nlist_idx)) != NULL; 3484 nlist_idx++) 3485 { 3486 if (global_symbol->GetAddress().GetFileAddress() == 0) 3487 { 3488 std::vector<uint32_t> indexes; 3489 if (symtab->AppendSymbolIndexesWithName (global_symbol->GetMangled().GetName(), indexes) > 0) 3490 { 3491 std::vector<uint32_t>::const_iterator pos; 3492 std::vector<uint32_t>::const_iterator end = indexes.end(); 3493 for (pos = indexes.begin(); pos != end; ++pos) 3494 { 3495 symbol_ptr = symtab->SymbolAtIndex(*pos); 3496 if (symbol_ptr != global_symbol && symbol_ptr->IsDebug() == false) 3497 { 3498 global_symbol->GetAddress() = symbol_ptr->GetAddress(); 3499 break; 3500 } 3501 } 3502 } 3503 } 3504 } 3505 } 3506 3507 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 3508 3509 if (function_starts_count > 0) 3510 { 3511 char synthetic_function_symbol[PATH_MAX]; 3512 uint32_t num_synthetic_function_symbols = 0; 3513 for (i=0; i<function_starts_count; ++i) 3514 { 3515 if (function_starts.GetEntryRef (i).data == false) 3516 ++num_synthetic_function_symbols; 3517 } 3518 3519 if (num_synthetic_function_symbols > 0) 3520 { 3521 if (num_syms < sym_idx + num_synthetic_function_symbols) 3522 { 3523 num_syms = sym_idx + num_synthetic_function_symbols; 3524 sym = symtab->Resize (num_syms); 3525 } 3526 uint32_t synthetic_function_symbol_idx = 0; 3527 for (i=0; i<function_starts_count; ++i) 3528 { 3529 const FunctionStarts::Entry *func_start_entry = function_starts.GetEntryAtIndex (i); 3530 if (func_start_entry->data == false) 3531 { 3532 addr_t symbol_file_addr = func_start_entry->addr; 3533 uint32_t symbol_flags = 0; 3534 if (is_arm) 3535 { 3536 if (symbol_file_addr & 1) 3537 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3538 symbol_file_addr &= 0xfffffffffffffffeull; 3539 } 3540 Address symbol_addr; 3541 if (module_sp->ResolveFileAddress (symbol_file_addr, symbol_addr)) 3542 { 3543 SectionSP symbol_section (symbol_addr.GetSection()); 3544 uint32_t symbol_byte_size = 0; 3545 if (symbol_section) 3546 { 3547 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3548 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3549 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3550 if (next_func_start_entry) 3551 { 3552 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3553 if (is_arm) 3554 next_symbol_file_addr &= 0xfffffffffffffffeull; 3555 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3556 } 3557 else 3558 { 3559 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3560 } 3561 snprintf (synthetic_function_symbol, 3562 sizeof(synthetic_function_symbol), 3563 "___lldb_unnamed_function%u$$%s", 3564 ++synthetic_function_symbol_idx, 3565 module_sp->GetFileSpec().GetFilename().GetCString()); 3566 sym[sym_idx].SetID (synthetic_sym_id++); 3567 sym[sym_idx].GetMangled().SetDemangledName(ConstString(synthetic_function_symbol)); 3568 sym[sym_idx].SetType (eSymbolTypeCode); 3569 sym[sym_idx].SetIsSynthetic (true); 3570 sym[sym_idx].GetAddress() = symbol_addr; 3571 if (symbol_flags) 3572 sym[sym_idx].SetFlags (symbol_flags); 3573 if (symbol_byte_size) 3574 sym[sym_idx].SetByteSize (symbol_byte_size); 3575 ++sym_idx; 3576 } 3577 } 3578 } 3579 } 3580 } 3581 } 3582 3583 // Trim our symbols down to just what we ended up with after 3584 // removing any symbols. 3585 if (sym_idx < num_syms) 3586 { 3587 num_syms = sym_idx; 3588 sym = symtab->Resize (num_syms); 3589 } 3590 3591 // Now synthesize indirect symbols 3592 if (m_dysymtab.nindirectsyms != 0) 3593 { 3594 if (indirect_symbol_index_data.GetByteSize()) 3595 { 3596 NListIndexToSymbolIndexMap::const_iterator end_index_pos = m_nlist_idx_to_sym_idx.end(); 3597 3598 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); ++sect_idx) 3599 { 3600 if ((m_mach_sections[sect_idx].flags & SectionFlagMaskSectionType) == SectionTypeSymbolStubs) 3601 { 3602 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 3603 if (symbol_stub_byte_size == 0) 3604 continue; 3605 3606 const uint32_t num_symbol_stubs = m_mach_sections[sect_idx].size / symbol_stub_byte_size; 3607 3608 if (num_symbol_stubs == 0) 3609 continue; 3610 3611 const uint32_t symbol_stub_index_offset = m_mach_sections[sect_idx].reserved1; 3612 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) 3613 { 3614 const uint32_t symbol_stub_index = symbol_stub_index_offset + stub_idx; 3615 const lldb::addr_t symbol_stub_addr = m_mach_sections[sect_idx].addr + (stub_idx * symbol_stub_byte_size); 3616 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 3617 if (indirect_symbol_index_data.ValidOffsetForDataOfSize(symbol_stub_offset, 4)) 3618 { 3619 const uint32_t stub_sym_id = indirect_symbol_index_data.GetU32 (&symbol_stub_offset); 3620 if (stub_sym_id & (IndirectSymbolAbsolute | IndirectSymbolLocal)) 3621 continue; 3622 3623 NListIndexToSymbolIndexMap::const_iterator index_pos = m_nlist_idx_to_sym_idx.find (stub_sym_id); 3624 Symbol *stub_symbol = NULL; 3625 if (index_pos != end_index_pos) 3626 { 3627 // We have a remapping from the original nlist index to 3628 // a current symbol index, so just look this up by index 3629 stub_symbol = symtab->SymbolAtIndex (index_pos->second); 3630 } 3631 else 3632 { 3633 // We need to lookup a symbol using the original nlist 3634 // symbol index since this index is coming from the 3635 // S_SYMBOL_STUBS 3636 stub_symbol = symtab->FindSymbolByID (stub_sym_id); 3637 } 3638 3639 if (stub_symbol) 3640 { 3641 Address so_addr(symbol_stub_addr, section_list); 3642 3643 if (stub_symbol->GetType() == eSymbolTypeUndefined) 3644 { 3645 // Change the external symbol into a trampoline that makes sense 3646 // These symbols were N_UNDF N_EXT, and are useless to us, so we 3647 // can re-use them so we don't have to make up a synthetic symbol 3648 // for no good reason. 3649 stub_symbol->SetType (eSymbolTypeTrampoline); 3650 stub_symbol->SetExternal (false); 3651 stub_symbol->GetAddress() = so_addr; 3652 stub_symbol->SetByteSize (symbol_stub_byte_size); 3653 } 3654 else 3655 { 3656 // Make a synthetic symbol to describe the trampoline stub 3657 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 3658 if (sym_idx >= num_syms) 3659 { 3660 sym = symtab->Resize (++num_syms); 3661 stub_symbol = NULL; // this pointer no longer valid 3662 } 3663 sym[sym_idx].SetID (synthetic_sym_id++); 3664 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 3665 sym[sym_idx].SetType (eSymbolTypeTrampoline); 3666 sym[sym_idx].SetIsSynthetic (true); 3667 sym[sym_idx].GetAddress() = so_addr; 3668 sym[sym_idx].SetByteSize (symbol_stub_byte_size); 3669 ++sym_idx; 3670 } 3671 } 3672 else 3673 { 3674 if (log) 3675 log->Warning ("symbol stub referencing symbol table symbol %u that isn't in our minimal symbol table, fix this!!!", stub_sym_id); 3676 } 3677 } 3678 } 3679 } 3680 } 3681 } 3682 } 3683 3684 // StreamFile s(stdout, false); 3685 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 3686 // symtab->Dump(&s, NULL, eSortOrderNone); 3687 // Set symbol byte sizes correctly since mach-o nlist entries don't have sizes 3688 symtab->CalculateSymbolSizes(); 3689 3690 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 3691 // symtab->Dump(&s, NULL, eSortOrderNone); 3692 3693 return symtab->GetNumSymbols(); 3694 } 3695 return 0; 3696 } 3697 3698 3699 void 3700 ObjectFileMachO::Dump (Stream *s) 3701 { 3702 ModuleSP module_sp(GetModule()); 3703 if (module_sp) 3704 { 3705 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 3706 s->Printf("%p: ", this); 3707 s->Indent(); 3708 if (m_header.magic == HeaderMagic64 || m_header.magic == HeaderMagic64Swapped) 3709 s->PutCString("ObjectFileMachO64"); 3710 else 3711 s->PutCString("ObjectFileMachO32"); 3712 3713 ArchSpec header_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 3714 3715 *s << ", file = '" << m_file << "', arch = " << header_arch.GetArchitectureName() << "\n"; 3716 3717 SectionList *sections = GetSectionList(); 3718 if (sections) 3719 sections->Dump(s, NULL, true, UINT32_MAX); 3720 3721 if (m_symtab_ap.get()) 3722 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 3723 } 3724 } 3725 3726 bool 3727 ObjectFileMachO::GetUUID (const llvm::MachO::mach_header &header, 3728 const lldb_private::DataExtractor &data, 3729 lldb::offset_t lc_offset, 3730 lldb_private::UUID& uuid) 3731 { 3732 uint32_t i; 3733 struct uuid_command load_cmd; 3734 3735 lldb::offset_t offset = lc_offset; 3736 for (i=0; i<header.ncmds; ++i) 3737 { 3738 const lldb::offset_t cmd_offset = offset; 3739 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 3740 break; 3741 3742 if (load_cmd.cmd == LoadCommandUUID) 3743 { 3744 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 3745 3746 if (uuid_bytes) 3747 { 3748 // OpenCL on Mac OS X uses the same UUID for each of its object files. 3749 // We pretend these object files have no UUID to prevent crashing. 3750 3751 const uint8_t opencl_uuid[] = { 0x8c, 0x8e, 0xb3, 0x9b, 3752 0x3b, 0xa8, 3753 0x4b, 0x16, 3754 0xb6, 0xa4, 3755 0x27, 0x63, 0xbb, 0x14, 0xf0, 0x0d }; 3756 3757 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 3758 return false; 3759 3760 uuid.SetBytes (uuid_bytes); 3761 return true; 3762 } 3763 return false; 3764 } 3765 offset = cmd_offset + load_cmd.cmdsize; 3766 } 3767 return false; 3768 } 3769 3770 bool 3771 ObjectFileMachO::GetUUID (lldb_private::UUID* uuid) 3772 { 3773 ModuleSP module_sp(GetModule()); 3774 if (module_sp) 3775 { 3776 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 3777 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 3778 return GetUUID (m_header, m_data, offset, *uuid); 3779 } 3780 return false; 3781 } 3782 3783 3784 uint32_t 3785 ObjectFileMachO::GetDependentModules (FileSpecList& files) 3786 { 3787 uint32_t count = 0; 3788 ModuleSP module_sp(GetModule()); 3789 if (module_sp) 3790 { 3791 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 3792 struct load_command load_cmd; 3793 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 3794 const bool resolve_path = false; // Don't resolve the dependend file paths since they may not reside on this system 3795 uint32_t i; 3796 for (i=0; i<m_header.ncmds; ++i) 3797 { 3798 const uint32_t cmd_offset = offset; 3799 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 3800 break; 3801 3802 switch (load_cmd.cmd) 3803 { 3804 case LoadCommandDylibLoad: 3805 case LoadCommandDylibLoadWeak: 3806 case LoadCommandDylibReexport: 3807 case LoadCommandDynamicLinkerLoad: 3808 case LoadCommandFixedVMShlibLoad: 3809 case LoadCommandDylibLoadUpward: 3810 { 3811 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 3812 const char *path = m_data.PeekCStr(name_offset); 3813 // Skip any path that starts with '@' since these are usually: 3814 // @executable_path/.../file 3815 // @rpath/.../file 3816 if (path && path[0] != '@') 3817 { 3818 FileSpec file_spec(path, resolve_path); 3819 if (files.AppendIfUnique(file_spec)) 3820 count++; 3821 } 3822 } 3823 break; 3824 3825 default: 3826 break; 3827 } 3828 offset = cmd_offset + load_cmd.cmdsize; 3829 } 3830 } 3831 return count; 3832 } 3833 3834 lldb_private::Address 3835 ObjectFileMachO::GetEntryPointAddress () 3836 { 3837 // If the object file is not an executable it can't hold the entry point. m_entry_point_address 3838 // is initialized to an invalid address, so we can just return that. 3839 // If m_entry_point_address is valid it means we've found it already, so return the cached value. 3840 3841 if (!IsExecutable() || m_entry_point_address.IsValid()) 3842 return m_entry_point_address; 3843 3844 // Otherwise, look for the UnixThread or Thread command. The data for the Thread command is given in 3845 // /usr/include/mach-o.h, but it is basically: 3846 // 3847 // uint32_t flavor - this is the flavor argument you would pass to thread_get_state 3848 // uint32_t count - this is the count of longs in the thread state data 3849 // struct XXX_thread_state state - this is the structure from <machine/thread_status.h> corresponding to the flavor. 3850 // <repeat this trio> 3851 // 3852 // So we just keep reading the various register flavors till we find the GPR one, then read the PC out of there. 3853 // FIXME: We will need to have a "RegisterContext data provider" class at some point that can get all the registers 3854 // out of data in this form & attach them to a given thread. That should underlie the MacOS X User process plugin, 3855 // and we'll also need it for the MacOS X Core File process plugin. When we have that we can also use it here. 3856 // 3857 // For now we hard-code the offsets and flavors we need: 3858 // 3859 // 3860 3861 ModuleSP module_sp(GetModule()); 3862 if (module_sp) 3863 { 3864 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 3865 struct load_command load_cmd; 3866 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 3867 uint32_t i; 3868 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 3869 bool done = false; 3870 3871 for (i=0; i<m_header.ncmds; ++i) 3872 { 3873 const lldb::offset_t cmd_offset = offset; 3874 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 3875 break; 3876 3877 switch (load_cmd.cmd) 3878 { 3879 case LoadCommandUnixThread: 3880 case LoadCommandThread: 3881 { 3882 while (offset < cmd_offset + load_cmd.cmdsize) 3883 { 3884 uint32_t flavor = m_data.GetU32(&offset); 3885 uint32_t count = m_data.GetU32(&offset); 3886 if (count == 0) 3887 { 3888 // We've gotten off somehow, log and exit; 3889 return m_entry_point_address; 3890 } 3891 3892 switch (m_header.cputype) 3893 { 3894 case llvm::MachO::CPUTypeARM: 3895 if (flavor == 1) // ARM_THREAD_STATE from mach/arm/thread_status.h 3896 { 3897 offset += 60; // This is the offset of pc in the GPR thread state data structure. 3898 start_address = m_data.GetU32(&offset); 3899 done = true; 3900 } 3901 break; 3902 case llvm::MachO::CPUTypeI386: 3903 if (flavor == 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 3904 { 3905 offset += 40; // This is the offset of eip in the GPR thread state data structure. 3906 start_address = m_data.GetU32(&offset); 3907 done = true; 3908 } 3909 break; 3910 case llvm::MachO::CPUTypeX86_64: 3911 if (flavor == 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 3912 { 3913 offset += 16 * 8; // This is the offset of rip in the GPR thread state data structure. 3914 start_address = m_data.GetU64(&offset); 3915 done = true; 3916 } 3917 break; 3918 default: 3919 return m_entry_point_address; 3920 } 3921 // Haven't found the GPR flavor yet, skip over the data for this flavor: 3922 if (done) 3923 break; 3924 offset += count * 4; 3925 } 3926 } 3927 break; 3928 case LoadCommandMain: 3929 { 3930 ConstString text_segment_name ("__TEXT"); 3931 uint64_t entryoffset = m_data.GetU64(&offset); 3932 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 3933 if (text_segment_sp) 3934 { 3935 done = true; 3936 start_address = text_segment_sp->GetFileAddress() + entryoffset; 3937 } 3938 } 3939 3940 default: 3941 break; 3942 } 3943 if (done) 3944 break; 3945 3946 // Go to the next load command: 3947 offset = cmd_offset + load_cmd.cmdsize; 3948 } 3949 3950 if (start_address != LLDB_INVALID_ADDRESS) 3951 { 3952 // We got the start address from the load commands, so now resolve that address in the sections 3953 // of this ObjectFile: 3954 if (!m_entry_point_address.ResolveAddressUsingFileSections (start_address, GetSectionList())) 3955 { 3956 m_entry_point_address.Clear(); 3957 } 3958 } 3959 else 3960 { 3961 // We couldn't read the UnixThread load command - maybe it wasn't there. As a fallback look for the 3962 // "start" symbol in the main executable. 3963 3964 ModuleSP module_sp (GetModule()); 3965 3966 if (module_sp) 3967 { 3968 SymbolContextList contexts; 3969 SymbolContext context; 3970 if (module_sp->FindSymbolsWithNameAndType(ConstString ("start"), eSymbolTypeCode, contexts)) 3971 { 3972 if (contexts.GetContextAtIndex(0, context)) 3973 m_entry_point_address = context.symbol->GetAddress(); 3974 } 3975 } 3976 } 3977 } 3978 3979 return m_entry_point_address; 3980 3981 } 3982 3983 lldb_private::Address 3984 ObjectFileMachO::GetHeaderAddress () 3985 { 3986 lldb_private::Address header_addr; 3987 SectionList *section_list = GetSectionList(); 3988 if (section_list) 3989 { 3990 SectionSP text_segment_sp (section_list->FindSectionByName (GetSegmentNameTEXT())); 3991 if (text_segment_sp) 3992 { 3993 header_addr.SetSection (text_segment_sp); 3994 header_addr.SetOffset (0); 3995 } 3996 } 3997 return header_addr; 3998 } 3999 4000 uint32_t 4001 ObjectFileMachO::GetNumThreadContexts () 4002 { 4003 ModuleSP module_sp(GetModule()); 4004 if (module_sp) 4005 { 4006 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4007 if (!m_thread_context_offsets_valid) 4008 { 4009 m_thread_context_offsets_valid = true; 4010 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4011 FileRangeArray::Entry file_range; 4012 thread_command thread_cmd; 4013 for (uint32_t i=0; i<m_header.ncmds; ++i) 4014 { 4015 const uint32_t cmd_offset = offset; 4016 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 4017 break; 4018 4019 if (thread_cmd.cmd == LoadCommandThread) 4020 { 4021 file_range.SetRangeBase (offset); 4022 file_range.SetByteSize (thread_cmd.cmdsize - 8); 4023 m_thread_context_offsets.Append (file_range); 4024 } 4025 offset = cmd_offset + thread_cmd.cmdsize; 4026 } 4027 } 4028 } 4029 return m_thread_context_offsets.GetSize(); 4030 } 4031 4032 lldb::RegisterContextSP 4033 ObjectFileMachO::GetThreadContextAtIndex (uint32_t idx, lldb_private::Thread &thread) 4034 { 4035 lldb::RegisterContextSP reg_ctx_sp; 4036 4037 ModuleSP module_sp(GetModule()); 4038 if (module_sp) 4039 { 4040 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4041 if (!m_thread_context_offsets_valid) 4042 GetNumThreadContexts (); 4043 4044 const FileRangeArray::Entry *thread_context_file_range = m_thread_context_offsets.GetEntryAtIndex (idx); 4045 if (thread_context_file_range) 4046 { 4047 4048 DataExtractor data (m_data, 4049 thread_context_file_range->GetRangeBase(), 4050 thread_context_file_range->GetByteSize()); 4051 4052 switch (m_header.cputype) 4053 { 4054 case llvm::MachO::CPUTypeARM: 4055 reg_ctx_sp.reset (new RegisterContextDarwin_arm_Mach (thread, data)); 4056 break; 4057 4058 case llvm::MachO::CPUTypeI386: 4059 reg_ctx_sp.reset (new RegisterContextDarwin_i386_Mach (thread, data)); 4060 break; 4061 4062 case llvm::MachO::CPUTypeX86_64: 4063 reg_ctx_sp.reset (new RegisterContextDarwin_x86_64_Mach (thread, data)); 4064 break; 4065 } 4066 } 4067 } 4068 return reg_ctx_sp; 4069 } 4070 4071 4072 ObjectFile::Type 4073 ObjectFileMachO::CalculateType() 4074 { 4075 switch (m_header.filetype) 4076 { 4077 case HeaderFileTypeObject: // 0x1u MH_OBJECT 4078 if (GetAddressByteSize () == 4) 4079 { 4080 // 32 bit kexts are just object files, but they do have a valid 4081 // UUID load command. 4082 UUID uuid; 4083 if (GetUUID(&uuid)) 4084 { 4085 // this checking for the UUID load command is not enough 4086 // we could eventually look for the symbol named 4087 // "OSKextGetCurrentIdentifier" as this is required of kexts 4088 if (m_strata == eStrataInvalid) 4089 m_strata = eStrataKernel; 4090 return eTypeSharedLibrary; 4091 } 4092 } 4093 return eTypeObjectFile; 4094 4095 case HeaderFileTypeExecutable: return eTypeExecutable; // 0x2u MH_EXECUTE 4096 case HeaderFileTypeFixedVMShlib: return eTypeSharedLibrary; // 0x3u MH_FVMLIB 4097 case HeaderFileTypeCore: return eTypeCoreFile; // 0x4u MH_CORE 4098 case HeaderFileTypePreloadedExecutable: return eTypeSharedLibrary; // 0x5u MH_PRELOAD 4099 case HeaderFileTypeDynamicShlib: return eTypeSharedLibrary; // 0x6u MH_DYLIB 4100 case HeaderFileTypeDynamicLinkEditor: return eTypeDynamicLinker; // 0x7u MH_DYLINKER 4101 case HeaderFileTypeBundle: return eTypeSharedLibrary; // 0x8u MH_BUNDLE 4102 case HeaderFileTypeDynamicShlibStub: return eTypeStubLibrary; // 0x9u MH_DYLIB_STUB 4103 case HeaderFileTypeDSYM: return eTypeDebugInfo; // 0xAu MH_DSYM 4104 case HeaderFileTypeKextBundle: return eTypeSharedLibrary; // 0xBu MH_KEXT_BUNDLE 4105 default: 4106 break; 4107 } 4108 return eTypeUnknown; 4109 } 4110 4111 ObjectFile::Strata 4112 ObjectFileMachO::CalculateStrata() 4113 { 4114 switch (m_header.filetype) 4115 { 4116 case HeaderFileTypeObject: // 0x1u MH_OBJECT 4117 { 4118 // 32 bit kexts are just object files, but they do have a valid 4119 // UUID load command. 4120 UUID uuid; 4121 if (GetUUID(&uuid)) 4122 { 4123 // this checking for the UUID load command is not enough 4124 // we could eventually look for the symbol named 4125 // "OSKextGetCurrentIdentifier" as this is required of kexts 4126 if (m_type == eTypeInvalid) 4127 m_type = eTypeSharedLibrary; 4128 4129 return eStrataKernel; 4130 } 4131 } 4132 return eStrataUnknown; 4133 4134 case HeaderFileTypeExecutable: // 0x2u MH_EXECUTE 4135 // Check for the MH_DYLDLINK bit in the flags 4136 if (m_header.flags & HeaderFlagBitIsDynamicLinkObject) 4137 { 4138 return eStrataUser; 4139 } 4140 else 4141 { 4142 SectionList *section_list = GetSectionList(); 4143 if (section_list) 4144 { 4145 static ConstString g_kld_section_name ("__KLD"); 4146 if (section_list->FindSectionByName(g_kld_section_name)) 4147 return eStrataKernel; 4148 } 4149 } 4150 return eStrataRawImage; 4151 4152 case HeaderFileTypeFixedVMShlib: return eStrataUser; // 0x3u MH_FVMLIB 4153 case HeaderFileTypeCore: return eStrataUnknown; // 0x4u MH_CORE 4154 case HeaderFileTypePreloadedExecutable: return eStrataRawImage; // 0x5u MH_PRELOAD 4155 case HeaderFileTypeDynamicShlib: return eStrataUser; // 0x6u MH_DYLIB 4156 case HeaderFileTypeDynamicLinkEditor: return eStrataUser; // 0x7u MH_DYLINKER 4157 case HeaderFileTypeBundle: return eStrataUser; // 0x8u MH_BUNDLE 4158 case HeaderFileTypeDynamicShlibStub: return eStrataUser; // 0x9u MH_DYLIB_STUB 4159 case HeaderFileTypeDSYM: return eStrataUnknown; // 0xAu MH_DSYM 4160 case HeaderFileTypeKextBundle: return eStrataKernel; // 0xBu MH_KEXT_BUNDLE 4161 default: 4162 break; 4163 } 4164 return eStrataUnknown; 4165 } 4166 4167 4168 uint32_t 4169 ObjectFileMachO::GetVersion (uint32_t *versions, uint32_t num_versions) 4170 { 4171 ModuleSP module_sp(GetModule()); 4172 if (module_sp) 4173 { 4174 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4175 struct dylib_command load_cmd; 4176 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4177 uint32_t version_cmd = 0; 4178 uint64_t version = 0; 4179 uint32_t i; 4180 for (i=0; i<m_header.ncmds; ++i) 4181 { 4182 const lldb::offset_t cmd_offset = offset; 4183 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4184 break; 4185 4186 if (load_cmd.cmd == LoadCommandDylibIdent) 4187 { 4188 if (version_cmd == 0) 4189 { 4190 version_cmd = load_cmd.cmd; 4191 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 4192 break; 4193 version = load_cmd.dylib.current_version; 4194 } 4195 break; // Break for now unless there is another more complete version 4196 // number load command in the future. 4197 } 4198 offset = cmd_offset + load_cmd.cmdsize; 4199 } 4200 4201 if (version_cmd == LoadCommandDylibIdent) 4202 { 4203 if (versions != NULL && num_versions > 0) 4204 { 4205 if (num_versions > 0) 4206 versions[0] = (version & 0xFFFF0000ull) >> 16; 4207 if (num_versions > 1) 4208 versions[1] = (version & 0x0000FF00ull) >> 8; 4209 if (num_versions > 2) 4210 versions[2] = (version & 0x000000FFull); 4211 // Fill in an remaining version numbers with invalid values 4212 for (i=3; i<num_versions; ++i) 4213 versions[i] = UINT32_MAX; 4214 } 4215 // The LC_ID_DYLIB load command has a version with 3 version numbers 4216 // in it, so always return 3 4217 return 3; 4218 } 4219 } 4220 return false; 4221 } 4222 4223 bool 4224 ObjectFileMachO::GetArchitecture (ArchSpec &arch) 4225 { 4226 ModuleSP module_sp(GetModule()); 4227 if (module_sp) 4228 { 4229 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4230 arch.SetArchitecture (eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 4231 4232 // Files with type MH_PRELOAD are currently used in cases where the image 4233 // debugs at the addresses in the file itself. Below we set the OS to 4234 // unknown to make sure we use the DynamicLoaderStatic()... 4235 if (m_header.filetype == HeaderFileTypePreloadedExecutable) 4236 { 4237 arch.GetTriple().setOS (llvm::Triple::UnknownOS); 4238 } 4239 return true; 4240 } 4241 return false; 4242 } 4243 4244 4245 UUID 4246 ObjectFileMachO::GetProcessSharedCacheUUID (Process *process) 4247 { 4248 UUID uuid; 4249 if (process) 4250 { 4251 addr_t all_image_infos = process->GetImageInfoAddress(); 4252 4253 // The address returned by GetImageInfoAddress may be the address of dyld (don't want) 4254 // or it may be the address of the dyld_all_image_infos structure (want). The first four 4255 // bytes will be either the version field (all_image_infos) or a Mach-O file magic constant. 4256 // Version 13 and higher of dyld_all_image_infos is required to get the sharedCacheUUID field. 4257 4258 Error err; 4259 uint32_t version_or_magic = process->ReadUnsignedIntegerFromMemory (all_image_infos, 4, -1, err); 4260 if (version_or_magic != -1 4261 && version_or_magic != HeaderMagic32 4262 && version_or_magic != HeaderMagic32Swapped 4263 && version_or_magic != HeaderMagic64 4264 && version_or_magic != HeaderMagic64Swapped 4265 && version_or_magic >= 13) 4266 { 4267 addr_t sharedCacheUUID_address = LLDB_INVALID_ADDRESS; 4268 int wordsize = process->GetAddressByteSize(); 4269 if (wordsize == 8) 4270 { 4271 sharedCacheUUID_address = all_image_infos + 160; // sharedCacheUUID <mach-o/dyld_images.h> 4272 } 4273 if (wordsize == 4) 4274 { 4275 sharedCacheUUID_address = all_image_infos + 84; // sharedCacheUUID <mach-o/dyld_images.h> 4276 } 4277 if (sharedCacheUUID_address != LLDB_INVALID_ADDRESS) 4278 { 4279 uuid_t shared_cache_uuid; 4280 if (process->ReadMemory (sharedCacheUUID_address, shared_cache_uuid, sizeof (uuid_t), err) == sizeof (uuid_t)) 4281 { 4282 uuid.SetBytes (shared_cache_uuid); 4283 } 4284 } 4285 } 4286 } 4287 return uuid; 4288 } 4289 4290 UUID 4291 ObjectFileMachO::GetLLDBSharedCacheUUID () 4292 { 4293 UUID uuid; 4294 #if defined (__APPLE__) && defined (__arm__) 4295 uint8_t *(*dyld_get_all_image_infos)(void); 4296 dyld_get_all_image_infos = (uint8_t*(*)()) dlsym (RTLD_DEFAULT, "_dyld_get_all_image_infos"); 4297 if (dyld_get_all_image_infos) 4298 { 4299 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 4300 if (dyld_all_image_infos_address) 4301 { 4302 uint32_t *version = (uint32_t*) dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 4303 if (*version >= 13) 4304 { 4305 uuid_t *sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 84); // sharedCacheUUID <mach-o/dyld_images.h> 4306 uuid.SetBytes (sharedCacheUUID_address); 4307 } 4308 } 4309 } 4310 #endif 4311 return uuid; 4312 } 4313 4314 4315 //------------------------------------------------------------------ 4316 // PluginInterface protocol 4317 //------------------------------------------------------------------ 4318 lldb_private::ConstString 4319 ObjectFileMachO::GetPluginName() 4320 { 4321 return GetPluginNameStatic(); 4322 } 4323 4324 uint32_t 4325 ObjectFileMachO::GetPluginVersion() 4326 { 4327 return 1; 4328 } 4329 4330