1 //===-- ObjectFileMachO.cpp -------------------------------------*- C++ -*-===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 10 #include "llvm/ADT/StringRef.h" 11 #include "llvm/Support/MachO.h" 12 13 #include "ObjectFileMachO.h" 14 15 #include "lldb/lldb-private-log.h" 16 #include "lldb/Core/ArchSpec.h" 17 #include "lldb/Core/DataBuffer.h" 18 #include "lldb/Core/Debugger.h" 19 #include "lldb/Core/FileSpecList.h" 20 #include "lldb/Core/Log.h" 21 #include "lldb/Core/Module.h" 22 #include "lldb/Core/ModuleSpec.h" 23 #include "lldb/Core/PluginManager.h" 24 #include "lldb/Core/RangeMap.h" 25 #include "lldb/Core/Section.h" 26 #include "lldb/Core/StreamFile.h" 27 #include "lldb/Core/StreamString.h" 28 #include "lldb/Core/Timer.h" 29 #include "lldb/Core/UUID.h" 30 #include "lldb/Host/Host.h" 31 #include "lldb/Host/FileSpec.h" 32 #include "lldb/Symbol/ClangNamespaceDecl.h" 33 #include "lldb/Symbol/DWARFCallFrameInfo.h" 34 #include "lldb/Symbol/ObjectFile.h" 35 #include "lldb/Target/Platform.h" 36 #include "lldb/Target/Process.h" 37 #include "lldb/Target/Target.h" 38 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 39 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 40 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 41 42 #if defined (__APPLE__) && defined (__arm__) 43 // GetLLDBSharedCacheUUID() needs to call dlsym() 44 #include <dlfcn.h> 45 #endif 46 47 #ifndef __APPLE__ 48 #include "Utility/UuidCompatibility.h" 49 #endif 50 51 using namespace lldb; 52 using namespace lldb_private; 53 using namespace llvm::MachO; 54 55 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 56 { 57 public: 58 RegisterContextDarwin_x86_64_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 59 RegisterContextDarwin_x86_64 (thread, 0) 60 { 61 SetRegisterDataFrom_LC_THREAD (data); 62 } 63 64 virtual void 65 InvalidateAllRegisters () 66 { 67 // Do nothing... registers are always valid... 68 } 69 70 void 71 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 72 { 73 lldb::offset_t offset = 0; 74 SetError (GPRRegSet, Read, -1); 75 SetError (FPURegSet, Read, -1); 76 SetError (EXCRegSet, Read, -1); 77 bool done = false; 78 79 while (!done) 80 { 81 int flavor = data.GetU32 (&offset); 82 if (flavor == 0) 83 done = true; 84 else 85 { 86 uint32_t i; 87 uint32_t count = data.GetU32 (&offset); 88 switch (flavor) 89 { 90 case GPRRegSet: 91 for (i=0; i<count; ++i) 92 (&gpr.rax)[i] = data.GetU64(&offset); 93 SetError (GPRRegSet, Read, 0); 94 done = true; 95 96 break; 97 case FPURegSet: 98 // TODO: fill in FPU regs.... 99 //SetError (FPURegSet, Read, -1); 100 done = true; 101 102 break; 103 case EXCRegSet: 104 exc.trapno = data.GetU32(&offset); 105 exc.err = data.GetU32(&offset); 106 exc.faultvaddr = data.GetU64(&offset); 107 SetError (EXCRegSet, Read, 0); 108 done = true; 109 break; 110 case 7: 111 case 8: 112 case 9: 113 // fancy flavors that encapsulate of the the above 114 // falvors... 115 break; 116 117 default: 118 done = true; 119 break; 120 } 121 } 122 } 123 } 124 protected: 125 virtual int 126 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 127 { 128 return 0; 129 } 130 131 virtual int 132 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 133 { 134 return 0; 135 } 136 137 virtual int 138 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 139 { 140 return 0; 141 } 142 143 virtual int 144 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 145 { 146 return 0; 147 } 148 149 virtual int 150 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 151 { 152 return 0; 153 } 154 155 virtual int 156 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 157 { 158 return 0; 159 } 160 }; 161 162 163 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 164 { 165 public: 166 RegisterContextDarwin_i386_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 167 RegisterContextDarwin_i386 (thread, 0) 168 { 169 SetRegisterDataFrom_LC_THREAD (data); 170 } 171 172 virtual void 173 InvalidateAllRegisters () 174 { 175 // Do nothing... registers are always valid... 176 } 177 178 void 179 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 180 { 181 lldb::offset_t offset = 0; 182 SetError (GPRRegSet, Read, -1); 183 SetError (FPURegSet, Read, -1); 184 SetError (EXCRegSet, Read, -1); 185 bool done = false; 186 187 while (!done) 188 { 189 int flavor = data.GetU32 (&offset); 190 if (flavor == 0) 191 done = true; 192 else 193 { 194 uint32_t i; 195 uint32_t count = data.GetU32 (&offset); 196 switch (flavor) 197 { 198 case GPRRegSet: 199 for (i=0; i<count; ++i) 200 (&gpr.eax)[i] = data.GetU32(&offset); 201 SetError (GPRRegSet, Read, 0); 202 done = true; 203 204 break; 205 case FPURegSet: 206 // TODO: fill in FPU regs.... 207 //SetError (FPURegSet, Read, -1); 208 done = true; 209 210 break; 211 case EXCRegSet: 212 exc.trapno = data.GetU32(&offset); 213 exc.err = data.GetU32(&offset); 214 exc.faultvaddr = data.GetU32(&offset); 215 SetError (EXCRegSet, Read, 0); 216 done = true; 217 break; 218 case 7: 219 case 8: 220 case 9: 221 // fancy flavors that encapsulate of the the above 222 // falvors... 223 break; 224 225 default: 226 done = true; 227 break; 228 } 229 } 230 } 231 } 232 protected: 233 virtual int 234 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 235 { 236 return 0; 237 } 238 239 virtual int 240 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 241 { 242 return 0; 243 } 244 245 virtual int 246 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 247 { 248 return 0; 249 } 250 251 virtual int 252 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 253 { 254 return 0; 255 } 256 257 virtual int 258 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 259 { 260 return 0; 261 } 262 263 virtual int 264 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 265 { 266 return 0; 267 } 268 }; 269 270 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm 271 { 272 public: 273 RegisterContextDarwin_arm_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 274 RegisterContextDarwin_arm (thread, 0) 275 { 276 SetRegisterDataFrom_LC_THREAD (data); 277 } 278 279 virtual void 280 InvalidateAllRegisters () 281 { 282 // Do nothing... registers are always valid... 283 } 284 285 void 286 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 287 { 288 lldb::offset_t offset = 0; 289 SetError (GPRRegSet, Read, -1); 290 SetError (FPURegSet, Read, -1); 291 SetError (EXCRegSet, Read, -1); 292 bool done = false; 293 294 while (!done) 295 { 296 int flavor = data.GetU32 (&offset); 297 uint32_t count = data.GetU32 (&offset); 298 lldb::offset_t next_thread_state = offset + (count * 4); 299 switch (flavor) 300 { 301 case GPRRegSet: 302 for (uint32_t i=0; i<count; ++i) 303 { 304 gpr.r[i] = data.GetU32(&offset); 305 } 306 307 // Note that gpr.cpsr is also copied by the above loop; this loop technically extends 308 // one element past the end of the gpr.r[] array. 309 310 SetError (GPRRegSet, Read, 0); 311 offset = next_thread_state; 312 break; 313 314 case FPURegSet: 315 { 316 uint8_t *fpu_reg_buf = (uint8_t*) &fpu.floats.s[0]; 317 const int fpu_reg_buf_size = sizeof (fpu.floats); 318 if (data.ExtractBytes (offset, fpu_reg_buf_size, eByteOrderLittle, fpu_reg_buf) == fpu_reg_buf_size) 319 { 320 offset += fpu_reg_buf_size; 321 fpu.fpscr = data.GetU32(&offset); 322 SetError (FPURegSet, Read, 0); 323 } 324 else 325 { 326 done = true; 327 } 328 } 329 offset = next_thread_state; 330 break; 331 332 case EXCRegSet: 333 if (count == 3) 334 { 335 exc.exception = data.GetU32(&offset); 336 exc.fsr = data.GetU32(&offset); 337 exc.far = data.GetU32(&offset); 338 SetError (EXCRegSet, Read, 0); 339 } 340 done = true; 341 offset = next_thread_state; 342 break; 343 344 // Unknown register set flavor, stop trying to parse. 345 default: 346 done = true; 347 } 348 } 349 } 350 protected: 351 virtual int 352 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 353 { 354 return -1; 355 } 356 357 virtual int 358 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 359 { 360 return -1; 361 } 362 363 virtual int 364 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 365 { 366 return -1; 367 } 368 369 virtual int 370 DoReadDBG (lldb::tid_t tid, int flavor, DBG &dbg) 371 { 372 return -1; 373 } 374 375 virtual int 376 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 377 { 378 return 0; 379 } 380 381 virtual int 382 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 383 { 384 return 0; 385 } 386 387 virtual int 388 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 389 { 390 return 0; 391 } 392 393 virtual int 394 DoWriteDBG (lldb::tid_t tid, int flavor, const DBG &dbg) 395 { 396 return -1; 397 } 398 }; 399 400 static uint32_t 401 MachHeaderSizeFromMagic(uint32_t magic) 402 { 403 switch (magic) 404 { 405 case MH_MAGIC: 406 case MH_CIGAM: 407 return sizeof(struct mach_header); 408 409 case MH_MAGIC_64: 410 case MH_CIGAM_64: 411 return sizeof(struct mach_header_64); 412 break; 413 414 default: 415 break; 416 } 417 return 0; 418 } 419 420 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 421 422 void 423 ObjectFileMachO::Initialize() 424 { 425 PluginManager::RegisterPlugin (GetPluginNameStatic(), 426 GetPluginDescriptionStatic(), 427 CreateInstance, 428 CreateMemoryInstance, 429 GetModuleSpecifications); 430 } 431 432 void 433 ObjectFileMachO::Terminate() 434 { 435 PluginManager::UnregisterPlugin (CreateInstance); 436 } 437 438 439 lldb_private::ConstString 440 ObjectFileMachO::GetPluginNameStatic() 441 { 442 static ConstString g_name("mach-o"); 443 return g_name; 444 } 445 446 const char * 447 ObjectFileMachO::GetPluginDescriptionStatic() 448 { 449 return "Mach-o object file reader (32 and 64 bit)"; 450 } 451 452 ObjectFile * 453 ObjectFileMachO::CreateInstance (const lldb::ModuleSP &module_sp, 454 DataBufferSP& data_sp, 455 lldb::offset_t data_offset, 456 const FileSpec* file, 457 lldb::offset_t file_offset, 458 lldb::offset_t length) 459 { 460 if (!data_sp) 461 { 462 data_sp = file->MemoryMapFileContents(file_offset, length); 463 data_offset = 0; 464 } 465 466 if (ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 467 { 468 // Update the data to contain the entire file if it doesn't already 469 if (data_sp->GetByteSize() < length) 470 { 471 data_sp = file->MemoryMapFileContents(file_offset, length); 472 data_offset = 0; 473 } 474 std::unique_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module_sp, data_sp, data_offset, file, file_offset, length)); 475 if (objfile_ap.get() && objfile_ap->ParseHeader()) 476 return objfile_ap.release(); 477 } 478 return NULL; 479 } 480 481 ObjectFile * 482 ObjectFileMachO::CreateMemoryInstance (const lldb::ModuleSP &module_sp, 483 DataBufferSP& data_sp, 484 const ProcessSP &process_sp, 485 lldb::addr_t header_addr) 486 { 487 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) 488 { 489 std::unique_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module_sp, data_sp, process_sp, header_addr)); 490 if (objfile_ap.get() && objfile_ap->ParseHeader()) 491 return objfile_ap.release(); 492 } 493 return NULL; 494 } 495 496 size_t 497 ObjectFileMachO::GetModuleSpecifications (const lldb_private::FileSpec& file, 498 lldb::DataBufferSP& data_sp, 499 lldb::offset_t data_offset, 500 lldb::offset_t file_offset, 501 lldb::offset_t length, 502 lldb_private::ModuleSpecList &specs) 503 { 504 const size_t initial_count = specs.GetSize(); 505 506 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) 507 { 508 DataExtractor data; 509 data.SetData(data_sp); 510 llvm::MachO::mach_header header; 511 if (ParseHeader (data, &data_offset, header)) 512 { 513 if (header.sizeofcmds >= data_sp->GetByteSize()) 514 { 515 data_sp = file.ReadFileContents(file_offset, header.sizeofcmds); 516 data.SetData(data_sp); 517 data_offset = MachHeaderSizeFromMagic(header.magic); 518 } 519 if (data_sp) 520 { 521 ModuleSpec spec; 522 spec.GetFileSpec() = file; 523 spec.GetArchitecture().SetArchitecture(eArchTypeMachO, 524 header.cputype, 525 header.cpusubtype); 526 if (header.filetype == MH_PRELOAD) // 0x5u 527 { 528 // Set OS to "unknown" - this is a standalone binary with no dyld et al 529 spec.GetArchitecture().GetTriple().setOS (llvm::Triple::UnknownOS); 530 } 531 if (spec.GetArchitecture().IsValid()) 532 { 533 GetUUID (header, data, data_offset, spec.GetUUID()); 534 specs.Append(spec); 535 } 536 } 537 } 538 } 539 return specs.GetSize() - initial_count; 540 } 541 542 543 544 const ConstString & 545 ObjectFileMachO::GetSegmentNameTEXT() 546 { 547 static ConstString g_segment_name_TEXT ("__TEXT"); 548 return g_segment_name_TEXT; 549 } 550 551 const ConstString & 552 ObjectFileMachO::GetSegmentNameDATA() 553 { 554 static ConstString g_segment_name_DATA ("__DATA"); 555 return g_segment_name_DATA; 556 } 557 558 const ConstString & 559 ObjectFileMachO::GetSegmentNameOBJC() 560 { 561 static ConstString g_segment_name_OBJC ("__OBJC"); 562 return g_segment_name_OBJC; 563 } 564 565 const ConstString & 566 ObjectFileMachO::GetSegmentNameLINKEDIT() 567 { 568 static ConstString g_section_name_LINKEDIT ("__LINKEDIT"); 569 return g_section_name_LINKEDIT; 570 } 571 572 const ConstString & 573 ObjectFileMachO::GetSectionNameEHFrame() 574 { 575 static ConstString g_section_name_eh_frame ("__eh_frame"); 576 return g_section_name_eh_frame; 577 } 578 579 bool 580 ObjectFileMachO::MagicBytesMatch (DataBufferSP& data_sp, 581 lldb::addr_t data_offset, 582 lldb::addr_t data_length) 583 { 584 DataExtractor data; 585 data.SetData (data_sp, data_offset, data_length); 586 lldb::offset_t offset = 0; 587 uint32_t magic = data.GetU32(&offset); 588 return MachHeaderSizeFromMagic(magic) != 0; 589 } 590 591 592 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 593 DataBufferSP& data_sp, 594 lldb::offset_t data_offset, 595 const FileSpec* file, 596 lldb::offset_t file_offset, 597 lldb::offset_t length) : 598 ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 599 m_mach_segments(), 600 m_mach_sections(), 601 m_entry_point_address(), 602 m_thread_context_offsets(), 603 m_thread_context_offsets_valid(false) 604 { 605 ::memset (&m_header, 0, sizeof(m_header)); 606 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 607 } 608 609 ObjectFileMachO::ObjectFileMachO (const lldb::ModuleSP &module_sp, 610 lldb::DataBufferSP& header_data_sp, 611 const lldb::ProcessSP &process_sp, 612 lldb::addr_t header_addr) : 613 ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 614 m_mach_segments(), 615 m_mach_sections(), 616 m_entry_point_address(), 617 m_thread_context_offsets(), 618 m_thread_context_offsets_valid(false) 619 { 620 ::memset (&m_header, 0, sizeof(m_header)); 621 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 622 } 623 624 ObjectFileMachO::~ObjectFileMachO() 625 { 626 } 627 628 bool 629 ObjectFileMachO::ParseHeader (DataExtractor &data, 630 lldb::offset_t *data_offset_ptr, 631 llvm::MachO::mach_header &header) 632 { 633 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 634 // Leave magic in the original byte order 635 header.magic = data.GetU32(data_offset_ptr); 636 bool can_parse = false; 637 bool is_64_bit = false; 638 switch (header.magic) 639 { 640 case MH_MAGIC: 641 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 642 data.SetAddressByteSize(4); 643 can_parse = true; 644 break; 645 646 case MH_MAGIC_64: 647 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 648 data.SetAddressByteSize(8); 649 can_parse = true; 650 is_64_bit = true; 651 break; 652 653 case MH_CIGAM: 654 data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 655 data.SetAddressByteSize(4); 656 can_parse = true; 657 break; 658 659 case MH_CIGAM_64: 660 data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 661 data.SetAddressByteSize(8); 662 is_64_bit = true; 663 can_parse = true; 664 break; 665 666 default: 667 break; 668 } 669 670 if (can_parse) 671 { 672 data.GetU32(data_offset_ptr, &header.cputype, 6); 673 if (is_64_bit) 674 *data_offset_ptr += 4; 675 return true; 676 } 677 else 678 { 679 memset(&header, 0, sizeof(header)); 680 } 681 return false; 682 } 683 684 bool 685 ObjectFileMachO::ParseHeader () 686 { 687 ModuleSP module_sp(GetModule()); 688 if (module_sp) 689 { 690 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 691 bool can_parse = false; 692 lldb::offset_t offset = 0; 693 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 694 // Leave magic in the original byte order 695 m_header.magic = m_data.GetU32(&offset); 696 switch (m_header.magic) 697 { 698 case MH_MAGIC: 699 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 700 m_data.SetAddressByteSize(4); 701 can_parse = true; 702 break; 703 704 case MH_MAGIC_64: 705 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 706 m_data.SetAddressByteSize(8); 707 can_parse = true; 708 break; 709 710 case MH_CIGAM: 711 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 712 m_data.SetAddressByteSize(4); 713 can_parse = true; 714 break; 715 716 case MH_CIGAM_64: 717 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 718 m_data.SetAddressByteSize(8); 719 can_parse = true; 720 break; 721 722 default: 723 break; 724 } 725 726 if (can_parse) 727 { 728 m_data.GetU32(&offset, &m_header.cputype, 6); 729 730 ArchSpec mach_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 731 732 // Check if the module has a required architecture 733 const ArchSpec &module_arch = module_sp->GetArchitecture(); 734 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 735 return false; 736 737 if (SetModulesArchitecture (mach_arch)) 738 { 739 const size_t header_and_lc_size = m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 740 if (m_data.GetByteSize() < header_and_lc_size) 741 { 742 DataBufferSP data_sp; 743 ProcessSP process_sp (m_process_wp.lock()); 744 if (process_sp) 745 { 746 data_sp = ReadMemory (process_sp, m_memory_addr, header_and_lc_size); 747 } 748 else 749 { 750 // Read in all only the load command data from the file on disk 751 data_sp = m_file.ReadFileContents(m_file_offset, header_and_lc_size); 752 if (data_sp->GetByteSize() != header_and_lc_size) 753 return false; 754 } 755 if (data_sp) 756 m_data.SetData (data_sp); 757 } 758 } 759 return true; 760 } 761 else 762 { 763 memset(&m_header, 0, sizeof(struct mach_header)); 764 } 765 } 766 return false; 767 } 768 769 770 ByteOrder 771 ObjectFileMachO::GetByteOrder () const 772 { 773 return m_data.GetByteOrder (); 774 } 775 776 bool 777 ObjectFileMachO::IsExecutable() const 778 { 779 return m_header.filetype == MH_EXECUTE; 780 } 781 782 uint32_t 783 ObjectFileMachO::GetAddressByteSize () const 784 { 785 return m_data.GetAddressByteSize (); 786 } 787 788 AddressClass 789 ObjectFileMachO::GetAddressClass (lldb::addr_t file_addr) 790 { 791 Symtab *symtab = GetSymtab(); 792 if (symtab) 793 { 794 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 795 if (symbol) 796 { 797 if (symbol->ValueIsAddress()) 798 { 799 SectionSP section_sp (symbol->GetAddress().GetSection()); 800 if (section_sp) 801 { 802 const lldb::SectionType section_type = section_sp->GetType(); 803 switch (section_type) 804 { 805 case eSectionTypeInvalid: return eAddressClassUnknown; 806 case eSectionTypeCode: 807 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) 808 { 809 // For ARM we have a bit in the n_desc field of the symbol 810 // that tells us ARM/Thumb which is bit 0x0008. 811 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 812 return eAddressClassCodeAlternateISA; 813 } 814 return eAddressClassCode; 815 816 case eSectionTypeContainer: return eAddressClassUnknown; 817 case eSectionTypeData: 818 case eSectionTypeDataCString: 819 case eSectionTypeDataCStringPointers: 820 case eSectionTypeDataSymbolAddress: 821 case eSectionTypeData4: 822 case eSectionTypeData8: 823 case eSectionTypeData16: 824 case eSectionTypeDataPointers: 825 case eSectionTypeZeroFill: 826 case eSectionTypeDataObjCMessageRefs: 827 case eSectionTypeDataObjCCFStrings: 828 return eAddressClassData; 829 case eSectionTypeDebug: 830 case eSectionTypeDWARFDebugAbbrev: 831 case eSectionTypeDWARFDebugAranges: 832 case eSectionTypeDWARFDebugFrame: 833 case eSectionTypeDWARFDebugInfo: 834 case eSectionTypeDWARFDebugLine: 835 case eSectionTypeDWARFDebugLoc: 836 case eSectionTypeDWARFDebugMacInfo: 837 case eSectionTypeDWARFDebugPubNames: 838 case eSectionTypeDWARFDebugPubTypes: 839 case eSectionTypeDWARFDebugRanges: 840 case eSectionTypeDWARFDebugStr: 841 case eSectionTypeDWARFAppleNames: 842 case eSectionTypeDWARFAppleTypes: 843 case eSectionTypeDWARFAppleNamespaces: 844 case eSectionTypeDWARFAppleObjC: 845 return eAddressClassDebug; 846 case eSectionTypeEHFrame: return eAddressClassRuntime; 847 case eSectionTypeELFSymbolTable: 848 case eSectionTypeELFDynamicSymbols: 849 case eSectionTypeELFRelocationEntries: 850 case eSectionTypeELFDynamicLinkInfo: 851 case eSectionTypeOther: return eAddressClassUnknown; 852 } 853 } 854 } 855 856 const SymbolType symbol_type = symbol->GetType(); 857 switch (symbol_type) 858 { 859 case eSymbolTypeAny: return eAddressClassUnknown; 860 case eSymbolTypeAbsolute: return eAddressClassUnknown; 861 862 case eSymbolTypeCode: 863 case eSymbolTypeTrampoline: 864 case eSymbolTypeResolver: 865 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) 866 { 867 // For ARM we have a bit in the n_desc field of the symbol 868 // that tells us ARM/Thumb which is bit 0x0008. 869 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 870 return eAddressClassCodeAlternateISA; 871 } 872 return eAddressClassCode; 873 874 case eSymbolTypeData: return eAddressClassData; 875 case eSymbolTypeRuntime: return eAddressClassRuntime; 876 case eSymbolTypeException: return eAddressClassRuntime; 877 case eSymbolTypeSourceFile: return eAddressClassDebug; 878 case eSymbolTypeHeaderFile: return eAddressClassDebug; 879 case eSymbolTypeObjectFile: return eAddressClassDebug; 880 case eSymbolTypeCommonBlock: return eAddressClassDebug; 881 case eSymbolTypeBlock: return eAddressClassDebug; 882 case eSymbolTypeLocal: return eAddressClassData; 883 case eSymbolTypeParam: return eAddressClassData; 884 case eSymbolTypeVariable: return eAddressClassData; 885 case eSymbolTypeVariableType: return eAddressClassDebug; 886 case eSymbolTypeLineEntry: return eAddressClassDebug; 887 case eSymbolTypeLineHeader: return eAddressClassDebug; 888 case eSymbolTypeScopeBegin: return eAddressClassDebug; 889 case eSymbolTypeScopeEnd: return eAddressClassDebug; 890 case eSymbolTypeAdditional: return eAddressClassUnknown; 891 case eSymbolTypeCompiler: return eAddressClassDebug; 892 case eSymbolTypeInstrumentation:return eAddressClassDebug; 893 case eSymbolTypeUndefined: return eAddressClassUnknown; 894 case eSymbolTypeObjCClass: return eAddressClassRuntime; 895 case eSymbolTypeObjCMetaClass: return eAddressClassRuntime; 896 case eSymbolTypeObjCIVar: return eAddressClassRuntime; 897 } 898 } 899 } 900 return eAddressClassUnknown; 901 } 902 903 Symtab * 904 ObjectFileMachO::GetSymtab() 905 { 906 ModuleSP module_sp(GetModule()); 907 if (module_sp) 908 { 909 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 910 if (m_symtab_ap.get() == NULL) 911 { 912 m_symtab_ap.reset(new Symtab(this)); 913 Mutex::Locker symtab_locker (m_symtab_ap->GetMutex()); 914 ParseSymtab (); 915 m_symtab_ap->Finalize (); 916 } 917 } 918 return m_symtab_ap.get(); 919 } 920 921 bool 922 ObjectFileMachO::IsStripped () 923 { 924 if (m_dysymtab.cmd == 0) 925 { 926 ModuleSP module_sp(GetModule()); 927 if (module_sp) 928 { 929 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 930 for (uint32_t i=0; i<m_header.ncmds; ++i) 931 { 932 const lldb::offset_t load_cmd_offset = offset; 933 934 load_command lc; 935 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 936 break; 937 if (lc.cmd == LC_DYSYMTAB) 938 { 939 m_dysymtab.cmd = lc.cmd; 940 m_dysymtab.cmdsize = lc.cmdsize; 941 if (m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == NULL) 942 { 943 // Clear m_dysymtab if we were unable to read all items from the load command 944 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 945 } 946 } 947 offset = load_cmd_offset + lc.cmdsize; 948 } 949 } 950 } 951 if (m_dysymtab.cmd) 952 return m_dysymtab.nlocalsym == 0; 953 return false; 954 } 955 956 void 957 ObjectFileMachO::CreateSections (SectionList &unified_section_list) 958 { 959 if (!m_sections_ap.get()) 960 { 961 m_sections_ap.reset(new SectionList()); 962 963 const bool is_dsym = (m_header.filetype == MH_DSYM); 964 lldb::user_id_t segID = 0; 965 lldb::user_id_t sectID = 0; 966 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 967 uint32_t i; 968 const bool is_core = GetType() == eTypeCoreFile; 969 //bool dump_sections = false; 970 ModuleSP module_sp (GetModule()); 971 // First look up any LC_ENCRYPTION_INFO load commands 972 typedef RangeArray<uint32_t, uint32_t, 8> EncryptedFileRanges; 973 EncryptedFileRanges encrypted_file_ranges; 974 encryption_info_command encryption_cmd; 975 for (i=0; i<m_header.ncmds; ++i) 976 { 977 const lldb::offset_t load_cmd_offset = offset; 978 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 979 break; 980 981 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO) 982 { 983 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) 984 { 985 if (encryption_cmd.cryptid != 0) 986 { 987 EncryptedFileRanges::Entry entry; 988 entry.SetRangeBase(encryption_cmd.cryptoff); 989 entry.SetByteSize(encryption_cmd.cryptsize); 990 encrypted_file_ranges.Append(entry); 991 } 992 } 993 } 994 offset = load_cmd_offset + encryption_cmd.cmdsize; 995 } 996 997 offset = MachHeaderSizeFromMagic(m_header.magic); 998 999 struct segment_command_64 load_cmd; 1000 for (i=0; i<m_header.ncmds; ++i) 1001 { 1002 const lldb::offset_t load_cmd_offset = offset; 1003 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1004 break; 1005 1006 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1007 { 1008 if (m_data.GetU8(&offset, (uint8_t*)load_cmd.segname, 16)) 1009 { 1010 bool add_section = true; 1011 bool add_to_unified = true; 1012 ConstString const_segname (load_cmd.segname, std::min<size_t>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 1013 1014 SectionSP unified_section_sp(unified_section_list.FindSectionByName(const_segname)); 1015 if (is_dsym && unified_section_sp) 1016 { 1017 if (const_segname == GetSegmentNameLINKEDIT()) 1018 { 1019 // We need to keep the __LINKEDIT segment private to this object file only 1020 add_to_unified = false; 1021 } 1022 else 1023 { 1024 // This is the dSYM file and this section has already been created by 1025 // the object file, no need to create it. 1026 add_section = false; 1027 } 1028 } 1029 load_cmd.vmaddr = m_data.GetAddress(&offset); 1030 load_cmd.vmsize = m_data.GetAddress(&offset); 1031 load_cmd.fileoff = m_data.GetAddress(&offset); 1032 load_cmd.filesize = m_data.GetAddress(&offset); 1033 if (m_length != 0 && load_cmd.filesize != 0) 1034 { 1035 if (load_cmd.fileoff > m_length) 1036 { 1037 // We have a load command that says it extends past the end of hte file. This is likely 1038 // a corrupt file. We don't have any way to return an error condition here (this method 1039 // was likely invokved from something like ObjectFile::GetSectionList()) -- all we can do 1040 // is null out the SectionList vector and if a process has been set up, dump a message 1041 // to stdout. The most common case here is core file debugging with a truncated file. 1042 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1043 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 ")", 1044 i, 1045 lc_segment_name, 1046 load_cmd.fileoff, 1047 m_length); 1048 1049 load_cmd.fileoff = 0; 1050 load_cmd.filesize = 0; 1051 } 1052 1053 if (load_cmd.fileoff + load_cmd.filesize > m_length) 1054 { 1055 // We have a load command that says it extends past the end of hte file. This is likely 1056 // a corrupt file. We don't have any way to return an error condition here (this method 1057 // was likely invokved from something like ObjectFile::GetSectionList()) -- all we can do 1058 // is null out the SectionList vector and if a process has been set up, dump a message 1059 // to stdout. The most common case here is core file debugging with a truncated file. 1060 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1061 GetModule()->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", 1062 i, 1063 lc_segment_name, 1064 load_cmd.fileoff + load_cmd.filesize, 1065 m_length); 1066 1067 // Tuncase the length 1068 load_cmd.filesize = m_length - load_cmd.fileoff; 1069 } 1070 } 1071 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1072 { 1073 1074 const bool segment_is_encrypted = (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1075 1076 // Keep a list of mach segments around in case we need to 1077 // get at data that isn't stored in the abstracted Sections. 1078 m_mach_segments.push_back (load_cmd); 1079 1080 // Use a segment ID of the segment index shifted left by 8 so they 1081 // never conflict with any of the sections. 1082 SectionSP segment_sp; 1083 if (add_section && (const_segname || is_core)) 1084 { 1085 segment_sp.reset(new Section (module_sp, // Module to which this section belongs 1086 this, // Object file to which this sections belongs 1087 ++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 1088 const_segname, // Name of this section 1089 eSectionTypeContainer, // This section is a container of other sections. 1090 load_cmd.vmaddr, // File VM address == addresses as they are found in the object file 1091 load_cmd.vmsize, // VM size in bytes of this section 1092 load_cmd.fileoff, // Offset to the data for this section in the file 1093 load_cmd.filesize, // Size in bytes of this section as found in the the file 1094 load_cmd.flags)); // Flags for this section 1095 1096 segment_sp->SetIsEncrypted (segment_is_encrypted); 1097 m_sections_ap->AddSection(segment_sp); 1098 if (add_to_unified) 1099 unified_section_list.AddSection(segment_sp); 1100 } 1101 else if (unified_section_sp) 1102 { 1103 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) 1104 { 1105 // Check to see if the module was read from memory? 1106 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) 1107 { 1108 // We have a module that is in memory and needs to have its 1109 // file address adjusted. We need to do this because when we 1110 // load a file from memory, its addresses will be slid already, 1111 // yet the addresses in the new symbol file will still be unslid. 1112 // Since everything is stored as section offset, this shouldn't 1113 // cause any problems. 1114 1115 // Make sure we've parsed the symbol table from the 1116 // ObjectFile before we go around changing its Sections. 1117 module_sp->GetObjectFile()->GetSymtab(); 1118 // eh_frame would present the same problems but we parse that on 1119 // a per-function basis as-needed so it's more difficult to 1120 // remove its use of the Sections. Realistically, the environments 1121 // where this code path will be taken will not have eh_frame sections. 1122 1123 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1124 } 1125 } 1126 m_sections_ap->AddSection(unified_section_sp); 1127 } 1128 1129 struct section_64 sect64; 1130 ::memset (§64, 0, sizeof(sect64)); 1131 // Push a section into our mach sections for the section at 1132 // index zero (NO_SECT) if we don't have any mach sections yet... 1133 if (m_mach_sections.empty()) 1134 m_mach_sections.push_back(sect64); 1135 uint32_t segment_sect_idx; 1136 const lldb::user_id_t first_segment_sectID = sectID + 1; 1137 1138 1139 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1140 for (segment_sect_idx=0; segment_sect_idx<load_cmd.nsects; ++segment_sect_idx) 1141 { 1142 if (m_data.GetU8(&offset, (uint8_t*)sect64.sectname, sizeof(sect64.sectname)) == NULL) 1143 break; 1144 if (m_data.GetU8(&offset, (uint8_t*)sect64.segname, sizeof(sect64.segname)) == NULL) 1145 break; 1146 sect64.addr = m_data.GetAddress(&offset); 1147 sect64.size = m_data.GetAddress(&offset); 1148 1149 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1150 break; 1151 1152 // Keep a list of mach sections around in case we need to 1153 // get at data that isn't stored in the abstracted Sections. 1154 m_mach_sections.push_back (sect64); 1155 1156 if (add_section) 1157 { 1158 ConstString section_name (sect64.sectname, std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 1159 if (!const_segname) 1160 { 1161 // We have a segment with no name so we need to conjure up 1162 // segments that correspond to the section's segname if there 1163 // isn't already such a section. If there is such a section, 1164 // we resize the section so that it spans all sections. 1165 // We also mark these sections as fake so address matches don't 1166 // hit if they land in the gaps between the child sections. 1167 const_segname.SetTrimmedCStringWithLength(sect64.segname, sizeof(sect64.segname)); 1168 segment_sp = unified_section_list.FindSectionByName (const_segname); 1169 if (segment_sp.get()) 1170 { 1171 Section *segment = segment_sp.get(); 1172 // Grow the section size as needed. 1173 const lldb::addr_t sect64_min_addr = sect64.addr; 1174 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1175 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1176 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1177 const lldb::addr_t curr_seg_max_addr = curr_seg_min_addr + curr_seg_byte_size; 1178 if (sect64_min_addr >= curr_seg_min_addr) 1179 { 1180 const lldb::addr_t new_seg_byte_size = sect64_max_addr - curr_seg_min_addr; 1181 // Only grow the section size if needed 1182 if (new_seg_byte_size > curr_seg_byte_size) 1183 segment->SetByteSize (new_seg_byte_size); 1184 } 1185 else 1186 { 1187 // We need to change the base address of the segment and 1188 // adjust the child section offsets for all existing children. 1189 const lldb::addr_t slide_amount = sect64_min_addr - curr_seg_min_addr; 1190 segment->Slide(slide_amount, false); 1191 segment->GetChildren().Slide(-slide_amount, false); 1192 segment->SetByteSize (curr_seg_max_addr - sect64_min_addr); 1193 } 1194 1195 // Grow the section size as needed. 1196 if (sect64.offset) 1197 { 1198 const lldb::addr_t segment_min_file_offset = segment->GetFileOffset(); 1199 const lldb::addr_t segment_max_file_offset = segment_min_file_offset + segment->GetFileSize(); 1200 1201 const lldb::addr_t section_min_file_offset = sect64.offset; 1202 const lldb::addr_t section_max_file_offset = section_min_file_offset + sect64.size; 1203 const lldb::addr_t new_file_offset = std::min (section_min_file_offset, segment_min_file_offset); 1204 const lldb::addr_t new_file_size = std::max (section_max_file_offset, segment_max_file_offset) - new_file_offset; 1205 segment->SetFileOffset (new_file_offset); 1206 segment->SetFileSize (new_file_size); 1207 } 1208 } 1209 else 1210 { 1211 // Create a fake section for the section's named segment 1212 segment_sp.reset(new Section (segment_sp, // Parent section 1213 module_sp, // Module to which this section belongs 1214 this, // Object file to which this section belongs 1215 ++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 1216 const_segname, // Name of this section 1217 eSectionTypeContainer, // This section is a container of other sections. 1218 sect64.addr, // File VM address == addresses as they are found in the object file 1219 sect64.size, // VM size in bytes of this section 1220 sect64.offset, // Offset to the data for this section in the file 1221 sect64.offset ? sect64.size : 0, // Size in bytes of this section as found in the the file 1222 load_cmd.flags)); // Flags for this section 1223 segment_sp->SetIsFake(true); 1224 1225 m_sections_ap->AddSection(segment_sp); 1226 if (add_to_unified) 1227 unified_section_list.AddSection(segment_sp); 1228 segment_sp->SetIsEncrypted (segment_is_encrypted); 1229 } 1230 } 1231 assert (segment_sp.get()); 1232 1233 uint32_t mach_sect_type = sect64.flags & SECTION_TYPE; 1234 static ConstString g_sect_name_objc_data ("__objc_data"); 1235 static ConstString g_sect_name_objc_msgrefs ("__objc_msgrefs"); 1236 static ConstString g_sect_name_objc_selrefs ("__objc_selrefs"); 1237 static ConstString g_sect_name_objc_classrefs ("__objc_classrefs"); 1238 static ConstString g_sect_name_objc_superrefs ("__objc_superrefs"); 1239 static ConstString g_sect_name_objc_const ("__objc_const"); 1240 static ConstString g_sect_name_objc_classlist ("__objc_classlist"); 1241 static ConstString g_sect_name_cfstring ("__cfstring"); 1242 1243 static ConstString g_sect_name_dwarf_debug_abbrev ("__debug_abbrev"); 1244 static ConstString g_sect_name_dwarf_debug_aranges ("__debug_aranges"); 1245 static ConstString g_sect_name_dwarf_debug_frame ("__debug_frame"); 1246 static ConstString g_sect_name_dwarf_debug_info ("__debug_info"); 1247 static ConstString g_sect_name_dwarf_debug_line ("__debug_line"); 1248 static ConstString g_sect_name_dwarf_debug_loc ("__debug_loc"); 1249 static ConstString g_sect_name_dwarf_debug_macinfo ("__debug_macinfo"); 1250 static ConstString g_sect_name_dwarf_debug_pubnames ("__debug_pubnames"); 1251 static ConstString g_sect_name_dwarf_debug_pubtypes ("__debug_pubtypes"); 1252 static ConstString g_sect_name_dwarf_debug_ranges ("__debug_ranges"); 1253 static ConstString g_sect_name_dwarf_debug_str ("__debug_str"); 1254 static ConstString g_sect_name_dwarf_apple_names ("__apple_names"); 1255 static ConstString g_sect_name_dwarf_apple_types ("__apple_types"); 1256 static ConstString g_sect_name_dwarf_apple_namespaces ("__apple_namespac"); 1257 static ConstString g_sect_name_dwarf_apple_objc ("__apple_objc"); 1258 static ConstString g_sect_name_eh_frame ("__eh_frame"); 1259 static ConstString g_sect_name_DATA ("__DATA"); 1260 static ConstString g_sect_name_TEXT ("__TEXT"); 1261 1262 lldb::SectionType sect_type = eSectionTypeOther; 1263 1264 if (section_name == g_sect_name_dwarf_debug_abbrev) 1265 sect_type = eSectionTypeDWARFDebugAbbrev; 1266 else if (section_name == g_sect_name_dwarf_debug_aranges) 1267 sect_type = eSectionTypeDWARFDebugAranges; 1268 else if (section_name == g_sect_name_dwarf_debug_frame) 1269 sect_type = eSectionTypeDWARFDebugFrame; 1270 else if (section_name == g_sect_name_dwarf_debug_info) 1271 sect_type = eSectionTypeDWARFDebugInfo; 1272 else if (section_name == g_sect_name_dwarf_debug_line) 1273 sect_type = eSectionTypeDWARFDebugLine; 1274 else if (section_name == g_sect_name_dwarf_debug_loc) 1275 sect_type = eSectionTypeDWARFDebugLoc; 1276 else if (section_name == g_sect_name_dwarf_debug_macinfo) 1277 sect_type = eSectionTypeDWARFDebugMacInfo; 1278 else if (section_name == g_sect_name_dwarf_debug_pubnames) 1279 sect_type = eSectionTypeDWARFDebugPubNames; 1280 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 1281 sect_type = eSectionTypeDWARFDebugPubTypes; 1282 else if (section_name == g_sect_name_dwarf_debug_ranges) 1283 sect_type = eSectionTypeDWARFDebugRanges; 1284 else if (section_name == g_sect_name_dwarf_debug_str) 1285 sect_type = eSectionTypeDWARFDebugStr; 1286 else if (section_name == g_sect_name_dwarf_apple_names) 1287 sect_type = eSectionTypeDWARFAppleNames; 1288 else if (section_name == g_sect_name_dwarf_apple_types) 1289 sect_type = eSectionTypeDWARFAppleTypes; 1290 else if (section_name == g_sect_name_dwarf_apple_namespaces) 1291 sect_type = eSectionTypeDWARFAppleNamespaces; 1292 else if (section_name == g_sect_name_dwarf_apple_objc) 1293 sect_type = eSectionTypeDWARFAppleObjC; 1294 else if (section_name == g_sect_name_objc_selrefs) 1295 sect_type = eSectionTypeDataCStringPointers; 1296 else if (section_name == g_sect_name_objc_msgrefs) 1297 sect_type = eSectionTypeDataObjCMessageRefs; 1298 else if (section_name == g_sect_name_eh_frame) 1299 sect_type = eSectionTypeEHFrame; 1300 else if (section_name == g_sect_name_cfstring) 1301 sect_type = eSectionTypeDataObjCCFStrings; 1302 else if (section_name == g_sect_name_objc_data || 1303 section_name == g_sect_name_objc_classrefs || 1304 section_name == g_sect_name_objc_superrefs || 1305 section_name == g_sect_name_objc_const || 1306 section_name == g_sect_name_objc_classlist) 1307 { 1308 sect_type = eSectionTypeDataPointers; 1309 } 1310 1311 if (sect_type == eSectionTypeOther) 1312 { 1313 switch (mach_sect_type) 1314 { 1315 // TODO: categorize sections by other flags for regular sections 1316 case S_REGULAR: 1317 if (segment_sp->GetName() == g_sect_name_TEXT) 1318 sect_type = eSectionTypeCode; 1319 else if (segment_sp->GetName() == g_sect_name_DATA) 1320 sect_type = eSectionTypeData; 1321 else 1322 sect_type = eSectionTypeOther; 1323 break; 1324 case S_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1325 case S_CSTRING_LITERALS: sect_type = eSectionTypeDataCString; break; // section with only literal C strings 1326 case S_4BYTE_LITERALS: sect_type = eSectionTypeData4; break; // section with only 4 byte literals 1327 case S_8BYTE_LITERALS: sect_type = eSectionTypeData8; break; // section with only 8 byte literals 1328 case S_LITERAL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only pointers to literals 1329 case S_NON_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only non-lazy symbol pointers 1330 case S_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only lazy symbol pointers 1331 case S_SYMBOL_STUBS: sect_type = eSectionTypeCode; break; // section with only symbol stubs, byte size of stub in the reserved2 field 1332 case S_MOD_INIT_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for initialization 1333 case S_MOD_TERM_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for termination 1334 case S_COALESCED: sect_type = eSectionTypeOther; break; 1335 case S_GB_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1336 case S_INTERPOSING: sect_type = eSectionTypeCode; break; // section with only pairs of function pointers for interposing 1337 case S_16BYTE_LITERALS: sect_type = eSectionTypeData16; break; // section with only 16 byte literals 1338 case S_DTRACE_DOF: sect_type = eSectionTypeDebug; break; 1339 case S_LAZY_DYLIB_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; 1340 default: break; 1341 } 1342 } 1343 1344 SectionSP section_sp(new Section (segment_sp, 1345 module_sp, 1346 this, 1347 ++sectID, 1348 section_name, 1349 sect_type, 1350 sect64.addr - segment_sp->GetFileAddress(), 1351 sect64.size, 1352 sect64.offset, 1353 sect64.offset == 0 ? 0 : sect64.size, 1354 sect64.flags)); 1355 // Set the section to be encrypted to match the segment 1356 1357 bool section_is_encrypted = false; 1358 if (!segment_is_encrypted && load_cmd.filesize != 0) 1359 section_is_encrypted = encrypted_file_ranges.FindEntryThatContains(sect64.offset) != NULL; 1360 1361 section_sp->SetIsEncrypted (segment_is_encrypted || section_is_encrypted); 1362 segment_sp->GetChildren().AddSection(section_sp); 1363 1364 if (segment_sp->IsFake()) 1365 { 1366 segment_sp.reset(); 1367 const_segname.Clear(); 1368 } 1369 } 1370 } 1371 if (segment_sp && is_dsym) 1372 { 1373 if (first_segment_sectID <= sectID) 1374 { 1375 lldb::user_id_t sect_uid; 1376 for (sect_uid = first_segment_sectID; sect_uid <= sectID; ++sect_uid) 1377 { 1378 SectionSP curr_section_sp(segment_sp->GetChildren().FindSectionByID (sect_uid)); 1379 SectionSP next_section_sp; 1380 if (sect_uid + 1 <= sectID) 1381 next_section_sp = segment_sp->GetChildren().FindSectionByID (sect_uid+1); 1382 1383 if (curr_section_sp.get()) 1384 { 1385 if (curr_section_sp->GetByteSize() == 0) 1386 { 1387 if (next_section_sp.get() != NULL) 1388 curr_section_sp->SetByteSize ( next_section_sp->GetFileAddress() - curr_section_sp->GetFileAddress() ); 1389 else 1390 curr_section_sp->SetByteSize ( load_cmd.vmsize ); 1391 } 1392 } 1393 } 1394 } 1395 } 1396 } 1397 } 1398 } 1399 else if (load_cmd.cmd == LC_DYSYMTAB) 1400 { 1401 m_dysymtab.cmd = load_cmd.cmd; 1402 m_dysymtab.cmdsize = load_cmd.cmdsize; 1403 m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1404 } 1405 1406 offset = load_cmd_offset + load_cmd.cmdsize; 1407 } 1408 1409 // StreamFile s(stdout, false); // REMOVE THIS LINE 1410 // s.Printf ("Sections for %s:\n", m_file.GetPath().c_str());// REMOVE THIS LINE 1411 // m_sections_ap->Dump(&s, NULL, true, UINT32_MAX);// REMOVE THIS LINE 1412 } 1413 } 1414 1415 class MachSymtabSectionInfo 1416 { 1417 public: 1418 1419 MachSymtabSectionInfo (SectionList *section_list) : 1420 m_section_list (section_list), 1421 m_section_infos() 1422 { 1423 // Get the number of sections down to a depth of 1 to include 1424 // all segments and their sections, but no other sections that 1425 // may be added for debug map or 1426 m_section_infos.resize(section_list->GetNumSections(1)); 1427 } 1428 1429 1430 SectionSP 1431 GetSection (uint8_t n_sect, addr_t file_addr) 1432 { 1433 if (n_sect == 0) 1434 return SectionSP(); 1435 if (n_sect < m_section_infos.size()) 1436 { 1437 if (!m_section_infos[n_sect].section_sp) 1438 { 1439 SectionSP section_sp (m_section_list->FindSectionByID (n_sect)); 1440 m_section_infos[n_sect].section_sp = section_sp; 1441 if (section_sp) 1442 { 1443 m_section_infos[n_sect].vm_range.SetBaseAddress (section_sp->GetFileAddress()); 1444 m_section_infos[n_sect].vm_range.SetByteSize (section_sp->GetByteSize()); 1445 } 1446 else 1447 { 1448 Host::SystemLog (Host::eSystemLogError, "error: unable to find section for section %u\n", n_sect); 1449 } 1450 } 1451 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) 1452 { 1453 // Symbol is in section. 1454 return m_section_infos[n_sect].section_sp; 1455 } 1456 else if (m_section_infos[n_sect].vm_range.GetByteSize () == 0 && 1457 m_section_infos[n_sect].vm_range.GetBaseAddress() == file_addr) 1458 { 1459 // Symbol is in section with zero size, but has the same start 1460 // address as the section. This can happen with linker symbols 1461 // (symbols that start with the letter 'l' or 'L'. 1462 return m_section_infos[n_sect].section_sp; 1463 } 1464 } 1465 return m_section_list->FindSectionContainingFileAddress(file_addr); 1466 } 1467 1468 protected: 1469 struct SectionInfo 1470 { 1471 SectionInfo () : 1472 vm_range(), 1473 section_sp () 1474 { 1475 } 1476 1477 VMRange vm_range; 1478 SectionSP section_sp; 1479 }; 1480 SectionList *m_section_list; 1481 std::vector<SectionInfo> m_section_infos; 1482 }; 1483 1484 size_t 1485 ObjectFileMachO::ParseSymtab () 1486 { 1487 Timer scoped_timer(__PRETTY_FUNCTION__, 1488 "ObjectFileMachO::ParseSymtab () module = %s", 1489 m_file.GetFilename().AsCString("")); 1490 ModuleSP module_sp (GetModule()); 1491 if (!module_sp) 1492 return 0; 1493 1494 struct symtab_command symtab_load_command = { 0, 0, 0, 0, 0, 0 }; 1495 struct linkedit_data_command function_starts_load_command = { 0, 0, 0, 0 }; 1496 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 1497 FunctionStarts function_starts; 1498 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1499 uint32_t i; 1500 1501 Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_SYMBOLS)); 1502 1503 for (i=0; i<m_header.ncmds; ++i) 1504 { 1505 const lldb::offset_t cmd_offset = offset; 1506 // Read in the load command and load command size 1507 struct load_command lc; 1508 if (m_data.GetU32(&offset, &lc, 2) == NULL) 1509 break; 1510 // Watch for the symbol table load command 1511 switch (lc.cmd) 1512 { 1513 case LC_SYMTAB: 1514 symtab_load_command.cmd = lc.cmd; 1515 symtab_load_command.cmdsize = lc.cmdsize; 1516 // Read in the rest of the symtab load command 1517 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 0) // fill in symoff, nsyms, stroff, strsize fields 1518 return 0; 1519 if (symtab_load_command.symoff == 0) 1520 { 1521 if (log) 1522 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 1523 return 0; 1524 } 1525 1526 if (symtab_load_command.stroff == 0) 1527 { 1528 if (log) 1529 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 1530 return 0; 1531 } 1532 1533 if (symtab_load_command.nsyms == 0) 1534 { 1535 if (log) 1536 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 1537 return 0; 1538 } 1539 1540 if (symtab_load_command.strsize == 0) 1541 { 1542 if (log) 1543 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 1544 return 0; 1545 } 1546 break; 1547 1548 case LC_FUNCTION_STARTS: 1549 function_starts_load_command.cmd = lc.cmd; 1550 function_starts_load_command.cmdsize = lc.cmdsize; 1551 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == NULL) // fill in symoff, nsyms, stroff, strsize fields 1552 memset (&function_starts_load_command, 0, sizeof(function_starts_load_command)); 1553 break; 1554 1555 default: 1556 break; 1557 } 1558 offset = cmd_offset + lc.cmdsize; 1559 } 1560 1561 if (symtab_load_command.cmd) 1562 { 1563 Symtab *symtab = m_symtab_ap.get(); 1564 SectionList *section_list = GetSectionList(); 1565 if (section_list == NULL) 1566 return 0; 1567 1568 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 1569 const ByteOrder byte_order = m_data.GetByteOrder(); 1570 bool bit_width_32 = addr_byte_size == 4; 1571 const size_t nlist_byte_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 1572 1573 DataExtractor nlist_data (NULL, 0, byte_order, addr_byte_size); 1574 DataExtractor strtab_data (NULL, 0, byte_order, addr_byte_size); 1575 DataExtractor function_starts_data (NULL, 0, byte_order, addr_byte_size); 1576 DataExtractor indirect_symbol_index_data (NULL, 0, byte_order, addr_byte_size); 1577 1578 const addr_t nlist_data_byte_size = symtab_load_command.nsyms * nlist_byte_size; 1579 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 1580 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 1581 1582 ProcessSP process_sp (m_process_wp.lock()); 1583 Process *process = process_sp.get(); 1584 1585 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 1586 1587 if (process) 1588 { 1589 Target &target = process->GetTarget(); 1590 1591 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 1592 1593 SectionSP linkedit_section_sp(section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 1594 // Reading mach file from memory in a process or core file... 1595 1596 if (linkedit_section_sp) 1597 { 1598 const addr_t linkedit_load_addr = linkedit_section_sp->GetLoadBaseAddress(&target); 1599 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 1600 const addr_t symoff_addr = linkedit_load_addr + symtab_load_command.symoff - linkedit_file_offset; 1601 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - linkedit_file_offset; 1602 1603 bool data_was_read = false; 1604 1605 #if defined (__APPLE__) && defined (__arm__) 1606 if (m_header.flags & 0x80000000u) 1607 { 1608 // This mach-o memory file is in the dyld shared cache. If this 1609 // program is not remote and this is iOS, then this process will 1610 // share the same shared cache as the process we are debugging and 1611 // we can read the entire __LINKEDIT from the address space in this 1612 // process. This is a needed optimization that is used for local iOS 1613 // debugging only since all shared libraries in the shared cache do 1614 // not have corresponding files that exist in the file system of the 1615 // device. They have been combined into a single file. This means we 1616 // always have to load these files from memory. All of the symbol and 1617 // string tables from all of the __LINKEDIT sections from the shared 1618 // libraries in the shared cache have been merged into a single large 1619 // symbol and string table. Reading all of this symbol and string table 1620 // data across can slow down debug launch times, so we optimize this by 1621 // reading the memory for the __LINKEDIT section from this process. 1622 1623 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 1624 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 1625 bool use_lldb_cache = true; 1626 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && lldb_shared_cache != process_shared_cache) 1627 { 1628 use_lldb_cache = false; 1629 ModuleSP module_sp (GetModule()); 1630 if (module_sp) 1631 module_sp->ReportWarning ("shared cache in process does not match lldb's own shared cache, startup will be slow."); 1632 1633 } 1634 1635 PlatformSP platform_sp (target.GetPlatform()); 1636 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) 1637 { 1638 data_was_read = true; 1639 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, eByteOrderLittle); 1640 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, eByteOrderLittle); 1641 if (function_starts_load_command.cmd) 1642 { 1643 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 1644 function_starts_data.SetData ((void *)func_start_addr, function_starts_load_command.datasize, eByteOrderLittle); 1645 } 1646 } 1647 } 1648 #endif 1649 1650 if (!data_was_read) 1651 { 1652 if (memory_module_load_level == eMemoryModuleLoadLevelComplete) 1653 { 1654 DataBufferSP nlist_data_sp (ReadMemory (process_sp, symoff_addr, nlist_data_byte_size)); 1655 if (nlist_data_sp) 1656 nlist_data.SetData (nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 1657 // Load strings individually from memory when loading from memory since shared cache 1658 // string tables contain strings for all symbols from all shared cached libraries 1659 //DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, strtab_data_byte_size)); 1660 //if (strtab_data_sp) 1661 // strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 1662 if (m_dysymtab.nindirectsyms != 0) 1663 { 1664 const addr_t indirect_syms_addr = linkedit_load_addr + m_dysymtab.indirectsymoff - linkedit_file_offset; 1665 DataBufferSP indirect_syms_data_sp (ReadMemory (process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 1666 if (indirect_syms_data_sp) 1667 indirect_symbol_index_data.SetData (indirect_syms_data_sp, 0, indirect_syms_data_sp->GetByteSize()); 1668 } 1669 } 1670 1671 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) 1672 { 1673 if (function_starts_load_command.cmd) 1674 { 1675 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 1676 DataBufferSP func_start_data_sp (ReadMemory (process_sp, func_start_addr, function_starts_load_command.datasize)); 1677 if (func_start_data_sp) 1678 function_starts_data.SetData (func_start_data_sp, 0, func_start_data_sp->GetByteSize()); 1679 } 1680 } 1681 } 1682 } 1683 } 1684 else 1685 { 1686 nlist_data.SetData (m_data, 1687 symtab_load_command.symoff, 1688 nlist_data_byte_size); 1689 strtab_data.SetData (m_data, 1690 symtab_load_command.stroff, 1691 strtab_data_byte_size); 1692 if (m_dysymtab.nindirectsyms != 0) 1693 { 1694 indirect_symbol_index_data.SetData (m_data, 1695 m_dysymtab.indirectsymoff, 1696 m_dysymtab.nindirectsyms * 4); 1697 } 1698 if (function_starts_load_command.cmd) 1699 { 1700 function_starts_data.SetData (m_data, 1701 function_starts_load_command.dataoff, 1702 function_starts_load_command.datasize); 1703 } 1704 } 1705 1706 if (nlist_data.GetByteSize() == 0 && memory_module_load_level == eMemoryModuleLoadLevelComplete) 1707 { 1708 if (log) 1709 module_sp->LogMessage(log, "failed to read nlist data"); 1710 return 0; 1711 } 1712 1713 1714 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 1715 if (!have_strtab_data) 1716 { 1717 if (process) 1718 { 1719 if (strtab_addr == LLDB_INVALID_ADDRESS) 1720 { 1721 if (log) 1722 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 1723 return 0; 1724 } 1725 } 1726 else 1727 { 1728 if (log) 1729 module_sp->LogMessage(log, "failed to read strtab data"); 1730 return 0; 1731 } 1732 } 1733 1734 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 1735 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 1736 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 1737 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 1738 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT)); 1739 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA)); 1740 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC)); 1741 SectionSP eh_frame_section_sp; 1742 if (text_section_sp.get()) 1743 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame); 1744 else 1745 eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame); 1746 1747 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 1748 1749 // lldb works best if it knows the start addresss of all functions in a module. 1750 // Linker symbols or debug info are normally the best source of information for start addr / size but 1751 // they may be stripped in a released binary. 1752 // Two additional sources of information exist in Mach-O binaries: 1753 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each function's start address in the 1754 // binary, relative to the text section. 1755 // eh_frame - the eh_frame FDEs have the start addr & size of each function 1756 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on all modern binaries. 1757 // Binaries built to run on older releases may need to use eh_frame information. 1758 1759 if (text_section_sp && function_starts_data.GetByteSize()) 1760 { 1761 FunctionStarts::Entry function_start_entry; 1762 function_start_entry.data = false; 1763 lldb::offset_t function_start_offset = 0; 1764 function_start_entry.addr = text_section_sp->GetFileAddress(); 1765 uint64_t delta; 1766 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 0) 1767 { 1768 // Now append the current entry 1769 function_start_entry.addr += delta; 1770 function_starts.Append(function_start_entry); 1771 } 1772 } 1773 else 1774 { 1775 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the load command claiming an eh_frame 1776 // but it doesn't actually have the eh_frame content. And if we have a dSYM, we don't need to do any 1777 // of this fill-in-the-missing-symbols works anyway - the debug info should give us all the functions in 1778 // the module. 1779 if (text_section_sp.get() && eh_frame_section_sp.get() && m_type != eTypeDebugInfo) 1780 { 1781 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, eRegisterKindGCC, true); 1782 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 1783 eh_frame.GetFunctionAddressAndSizeVector (functions); 1784 addr_t text_base_addr = text_section_sp->GetFileAddress(); 1785 size_t count = functions.GetSize(); 1786 for (size_t i = 0; i < count; ++i) 1787 { 1788 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = functions.GetEntryAtIndex (i); 1789 if (func) 1790 { 1791 FunctionStarts::Entry function_start_entry; 1792 function_start_entry.addr = func->base - text_base_addr; 1793 function_starts.Append(function_start_entry); 1794 } 1795 } 1796 } 1797 } 1798 1799 const size_t function_starts_count = function_starts.GetSize(); 1800 1801 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() : NO_SECT; 1802 1803 lldb::offset_t nlist_data_offset = 0; 1804 1805 uint32_t N_SO_index = UINT32_MAX; 1806 1807 MachSymtabSectionInfo section_info (section_list); 1808 std::vector<uint32_t> N_FUN_indexes; 1809 std::vector<uint32_t> N_NSYM_indexes; 1810 std::vector<uint32_t> N_INCL_indexes; 1811 std::vector<uint32_t> N_BRAC_indexes; 1812 std::vector<uint32_t> N_COMM_indexes; 1813 typedef std::map <uint64_t, uint32_t> ValueToSymbolIndexMap; 1814 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap; 1815 typedef std::map <const char *, uint32_t> ConstNameToSymbolIndexMap; 1816 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 1817 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 1818 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 1819 // Any symbols that get merged into another will get an entry 1820 // in this map so we know 1821 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 1822 uint32_t nlist_idx = 0; 1823 Symbol *symbol_ptr = NULL; 1824 1825 uint32_t sym_idx = 0; 1826 Symbol *sym = NULL; 1827 size_t num_syms = 0; 1828 std::string memory_symbol_name; 1829 uint32_t unmapped_local_symbols_found = 0; 1830 1831 #if defined (__APPLE__) && defined (__arm__) 1832 1833 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been optimized by moving LOCAL 1834 // symbols out of the memory mapped portion of the DSC. The symbol information has all been retained, 1835 // but it isn't available in the normal nlist data. However, there *are* duplicate entries of *some* 1836 // LOCAL symbols in the normal nlist data. To handle this situation correctly, we must first attempt 1837 // to parse any DSC unmapped symbol information. If we find any, we set a flag that tells the normal 1838 // nlist parser to ignore all LOCAL symbols. 1839 1840 if (m_header.flags & 0x80000000u) 1841 { 1842 // Before we can start mapping the DSC, we need to make certain the target process is actually 1843 // using the cache we can find. 1844 1845 // Next we need to determine the correct path for the dyld shared cache. 1846 1847 ArchSpec header_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 1848 char dsc_path[PATH_MAX]; 1849 1850 snprintf(dsc_path, sizeof(dsc_path), "%s%s%s", 1851 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR */ 1852 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 1853 header_arch.GetArchitectureName()); 1854 1855 FileSpec dsc_filespec(dsc_path, false); 1856 1857 // We need definitions of two structures in the on-disk DSC, copy them here manually 1858 struct lldb_copy_dyld_cache_header_v0 1859 { 1860 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 1861 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 1862 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 1863 uint32_t imagesOffset; 1864 uint32_t imagesCount; 1865 uint64_t dyldBaseAddress; 1866 uint64_t codeSignatureOffset; 1867 uint64_t codeSignatureSize; 1868 uint64_t slideInfoOffset; 1869 uint64_t slideInfoSize; 1870 uint64_t localSymbolsOffset; // file offset of where local symbols are stored 1871 uint64_t localSymbolsSize; // size of local symbols information 1872 }; 1873 struct lldb_copy_dyld_cache_header_v1 1874 { 1875 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 1876 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 1877 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 1878 uint32_t imagesOffset; 1879 uint32_t imagesCount; 1880 uint64_t dyldBaseAddress; 1881 uint64_t codeSignatureOffset; 1882 uint64_t codeSignatureSize; 1883 uint64_t slideInfoOffset; 1884 uint64_t slideInfoSize; 1885 uint64_t localSymbolsOffset; 1886 uint64_t localSymbolsSize; 1887 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 and later 1888 }; 1889 1890 struct lldb_copy_dyld_cache_mapping_info 1891 { 1892 uint64_t address; 1893 uint64_t size; 1894 uint64_t fileOffset; 1895 uint32_t maxProt; 1896 uint32_t initProt; 1897 }; 1898 1899 struct lldb_copy_dyld_cache_local_symbols_info 1900 { 1901 uint32_t nlistOffset; 1902 uint32_t nlistCount; 1903 uint32_t stringsOffset; 1904 uint32_t stringsSize; 1905 uint32_t entriesOffset; 1906 uint32_t entriesCount; 1907 }; 1908 struct lldb_copy_dyld_cache_local_symbols_entry 1909 { 1910 uint32_t dylibOffset; 1911 uint32_t nlistStartIndex; 1912 uint32_t nlistCount; 1913 }; 1914 1915 /* The dyld_cache_header has a pointer to the dyld_cache_local_symbols_info structure (localSymbolsOffset). 1916 The dyld_cache_local_symbols_info structure gives us three things: 1917 1. The start and count of the nlist records in the dyld_shared_cache file 1918 2. The start and size of the strings for these nlist records 1919 3. The start and count of dyld_cache_local_symbols_entry entries 1920 1921 There is one dyld_cache_local_symbols_entry per dylib/framework in the dyld shared cache. 1922 The "dylibOffset" field is the Mach-O header of this dylib/framework in the dyld shared cache. 1923 The dyld_cache_local_symbols_entry also lists the start of this dylib/framework's nlist records 1924 and the count of how many nlist records there are for this dylib/framework. 1925 */ 1926 1927 // Process the dsc header to find the unmapped symbols 1928 // 1929 // Save some VM space, do not map the entire cache in one shot. 1930 1931 DataBufferSP dsc_data_sp; 1932 dsc_data_sp = dsc_filespec.MemoryMapFileContents(0, sizeof(struct lldb_copy_dyld_cache_header_v1)); 1933 1934 if (dsc_data_sp) 1935 { 1936 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 1937 1938 char version_str[17]; 1939 int version = -1; 1940 lldb::offset_t offset = 0; 1941 memcpy (version_str, dsc_header_data.GetData (&offset, 16), 16); 1942 version_str[16] = '\0'; 1943 if (strncmp (version_str, "dyld_v", 6) == 0 && isdigit (version_str[6])) 1944 { 1945 int v; 1946 if (::sscanf (version_str + 6, "%d", &v) == 1) 1947 { 1948 version = v; 1949 } 1950 } 1951 1952 UUID dsc_uuid; 1953 if (version >= 1) 1954 { 1955 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, uuid); 1956 uint8_t uuid_bytes[sizeof (uuid_t)]; 1957 memcpy (uuid_bytes, dsc_header_data.GetData (&offset, sizeof (uuid_t)), sizeof (uuid_t)); 1958 dsc_uuid.SetBytes (uuid_bytes); 1959 } 1960 1961 bool uuid_match = true; 1962 if (dsc_uuid.IsValid() && process) 1963 { 1964 UUID shared_cache_uuid(GetProcessSharedCacheUUID(process)); 1965 1966 if (shared_cache_uuid.IsValid() && dsc_uuid != shared_cache_uuid) 1967 { 1968 // The on-disk dyld_shared_cache file is not the same as the one in this 1969 // process' memory, don't use it. 1970 uuid_match = false; 1971 ModuleSP module_sp (GetModule()); 1972 if (module_sp) 1973 module_sp->ReportWarning ("process shared cache does not match on-disk dyld_shared_cache file, some symbol names will be missing."); 1974 } 1975 } 1976 1977 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, mappingOffset); 1978 1979 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 1980 1981 // If the mappingOffset points to a location inside the header, we've 1982 // opened an old dyld shared cache, and should not proceed further. 1983 if (uuid_match && mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v0)) 1984 { 1985 1986 DataBufferSP dsc_mapping_info_data_sp = dsc_filespec.MemoryMapFileContents(mappingOffset, sizeof (struct lldb_copy_dyld_cache_mapping_info)); 1987 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, byte_order, addr_byte_size); 1988 offset = 0; 1989 1990 // The File addresses (from the in-memory Mach-O load commands) for the shared libraries 1991 // in the shared library cache need to be adjusted by an offset to match up with the 1992 // dylibOffset identifying field in the dyld_cache_local_symbol_entry's. This offset is 1993 // recorded in mapping_offset_value. 1994 const uint64_t mapping_offset_value = dsc_mapping_info_data.GetU64(&offset); 1995 1996 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 1997 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 1998 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 1999 2000 if (localSymbolsOffset && localSymbolsSize) 2001 { 2002 // Map the local symbols 2003 if (DataBufferSP dsc_local_symbols_data_sp = dsc_filespec.MemoryMapFileContents(localSymbolsOffset, localSymbolsSize)) 2004 { 2005 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, byte_order, addr_byte_size); 2006 2007 offset = 0; 2008 2009 // Read the local_symbols_infos struct in one shot 2010 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2011 dsc_local_symbols_data.GetU32(&offset, &local_symbols_info.nlistOffset, 6); 2012 2013 SectionSP text_section_sp(section_list->FindSectionByName(GetSegmentNameTEXT())); 2014 2015 uint32_t header_file_offset = (text_section_sp->GetFileAddress() - mapping_offset_value); 2016 2017 offset = local_symbols_info.entriesOffset; 2018 for (uint32_t entry_index = 0; entry_index < local_symbols_info.entriesCount; entry_index++) 2019 { 2020 struct lldb_copy_dyld_cache_local_symbols_entry local_symbols_entry; 2021 local_symbols_entry.dylibOffset = dsc_local_symbols_data.GetU32(&offset); 2022 local_symbols_entry.nlistStartIndex = dsc_local_symbols_data.GetU32(&offset); 2023 local_symbols_entry.nlistCount = dsc_local_symbols_data.GetU32(&offset); 2024 2025 if (header_file_offset == local_symbols_entry.dylibOffset) 2026 { 2027 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2028 2029 // The normal nlist code cannot correctly size the Symbols array, we need to allocate it here. 2030 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms + unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2031 num_syms = symtab->GetNumSymbols(); 2032 2033 nlist_data_offset = local_symbols_info.nlistOffset + (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2034 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2035 2036 for (uint32_t nlist_index = 0; nlist_index < local_symbols_entry.nlistCount; nlist_index++) 2037 { 2038 ///////////////////////////// 2039 { 2040 struct nlist_64 nlist; 2041 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2042 break; 2043 2044 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked(&nlist_data_offset); 2045 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2046 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2047 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked (&nlist_data_offset); 2048 nlist.n_value = dsc_local_symbols_data.GetAddress_unchecked (&nlist_data_offset); 2049 2050 SymbolType type = eSymbolTypeInvalid; 2051 const char *symbol_name = dsc_local_symbols_data.PeekCStr(string_table_offset + nlist.n_strx); 2052 2053 if (symbol_name == NULL) 2054 { 2055 // No symbol should be NULL, even the symbols with no 2056 // string values should have an offset zero which points 2057 // to an empty C-string 2058 Host::SystemLog (Host::eSystemLogError, 2059 "error: DSC unmapped local symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 2060 entry_index, 2061 nlist.n_strx, 2062 module_sp->GetFileSpec().GetPath().c_str()); 2063 continue; 2064 } 2065 if (symbol_name[0] == '\0') 2066 symbol_name = NULL; 2067 2068 const char *symbol_name_non_abi_mangled = NULL; 2069 2070 SectionSP symbol_section; 2071 uint32_t symbol_byte_size = 0; 2072 bool add_nlist = true; 2073 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2074 bool demangled_is_synthesized = false; 2075 bool is_gsym = false; 2076 2077 assert (sym_idx < num_syms); 2078 2079 sym[sym_idx].SetDebug (is_debug); 2080 2081 if (is_debug) 2082 { 2083 switch (nlist.n_type) 2084 { 2085 case N_GSYM: 2086 // global symbol: name,,NO_SECT,type,0 2087 // Sometimes the N_GSYM value contains the address. 2088 2089 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 2090 // have the same address, but we want to ensure that we always find only the real symbol, 2091 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 2092 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 2093 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 2094 // same address. 2095 2096 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 2097 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 2098 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 2099 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 2100 add_nlist = false; 2101 else 2102 { 2103 is_gsym = true; 2104 sym[sym_idx].SetExternal(true); 2105 if (nlist.n_value != 0) 2106 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2107 type = eSymbolTypeData; 2108 } 2109 break; 2110 2111 case N_FNAME: 2112 // procedure name (f77 kludge): name,,NO_SECT,0,0 2113 type = eSymbolTypeCompiler; 2114 break; 2115 2116 case N_FUN: 2117 // procedure: name,,n_sect,linenumber,address 2118 if (symbol_name) 2119 { 2120 type = eSymbolTypeCode; 2121 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2122 2123 N_FUN_addr_to_sym_idx[nlist.n_value] = sym_idx; 2124 // We use the current number of symbols in the symbol table in lieu of 2125 // using nlist_idx in case we ever start trimming entries out 2126 N_FUN_indexes.push_back(sym_idx); 2127 } 2128 else 2129 { 2130 type = eSymbolTypeCompiler; 2131 2132 if ( !N_FUN_indexes.empty() ) 2133 { 2134 // Copy the size of the function into the original STAB entry so we don't have 2135 // to hunt for it later 2136 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 2137 N_FUN_indexes.pop_back(); 2138 // We don't really need the end function STAB as it contains the size which 2139 // we already placed with the original symbol, so don't add it if we want a 2140 // minimal symbol table 2141 add_nlist = false; 2142 } 2143 } 2144 break; 2145 2146 case N_STSYM: 2147 // static symbol: name,,n_sect,type,address 2148 N_STSYM_addr_to_sym_idx[nlist.n_value] = sym_idx; 2149 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2150 type = eSymbolTypeData; 2151 break; 2152 2153 case N_LCSYM: 2154 // .lcomm symbol: name,,n_sect,type,address 2155 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2156 type = eSymbolTypeCommonBlock; 2157 break; 2158 2159 case N_BNSYM: 2160 // We use the current number of symbols in the symbol table in lieu of 2161 // using nlist_idx in case we ever start trimming entries out 2162 // Skip these if we want minimal symbol tables 2163 add_nlist = false; 2164 break; 2165 2166 case N_ENSYM: 2167 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 2168 // so that we can always skip the entire symbol if we need to navigate 2169 // more quickly at the source level when parsing STABS 2170 // Skip these if we want minimal symbol tables 2171 add_nlist = false; 2172 break; 2173 2174 2175 case N_OPT: 2176 // emitted with gcc2_compiled and in gcc source 2177 type = eSymbolTypeCompiler; 2178 break; 2179 2180 case N_RSYM: 2181 // register sym: name,,NO_SECT,type,register 2182 type = eSymbolTypeVariable; 2183 break; 2184 2185 case N_SLINE: 2186 // src line: 0,,n_sect,linenumber,address 2187 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2188 type = eSymbolTypeLineEntry; 2189 break; 2190 2191 case N_SSYM: 2192 // structure elt: name,,NO_SECT,type,struct_offset 2193 type = eSymbolTypeVariableType; 2194 break; 2195 2196 case N_SO: 2197 // source file name 2198 type = eSymbolTypeSourceFile; 2199 if (symbol_name == NULL) 2200 { 2201 add_nlist = false; 2202 if (N_SO_index != UINT32_MAX) 2203 { 2204 // Set the size of the N_SO to the terminating index of this N_SO 2205 // so that we can always skip the entire N_SO if we need to navigate 2206 // more quickly at the source level when parsing STABS 2207 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2208 symbol_ptr->SetByteSize(sym_idx); 2209 symbol_ptr->SetSizeIsSibling(true); 2210 } 2211 N_NSYM_indexes.clear(); 2212 N_INCL_indexes.clear(); 2213 N_BRAC_indexes.clear(); 2214 N_COMM_indexes.clear(); 2215 N_FUN_indexes.clear(); 2216 N_SO_index = UINT32_MAX; 2217 } 2218 else 2219 { 2220 // We use the current number of symbols in the symbol table in lieu of 2221 // using nlist_idx in case we ever start trimming entries out 2222 const bool N_SO_has_full_path = symbol_name[0] == '/'; 2223 if (N_SO_has_full_path) 2224 { 2225 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2226 { 2227 // We have two consecutive N_SO entries where the first contains a directory 2228 // and the second contains a full path. 2229 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 2230 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2231 add_nlist = false; 2232 } 2233 else 2234 { 2235 // This is the first entry in a N_SO that contains a directory or 2236 // a full path to the source file 2237 N_SO_index = sym_idx; 2238 } 2239 } 2240 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2241 { 2242 // This is usually the second N_SO entry that contains just the filename, 2243 // so here we combine it with the first one if we are minimizing the symbol table 2244 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 2245 if (so_path && so_path[0]) 2246 { 2247 std::string full_so_path (so_path); 2248 const size_t double_slash_pos = full_so_path.find("//"); 2249 if (double_slash_pos != std::string::npos) 2250 { 2251 // The linker has been generating bad N_SO entries with doubled up paths 2252 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 2253 // and the second is the directory for the source file so you end up with 2254 // a path that looks like "/tmp/src//tmp/src/" 2255 FileSpec so_dir(so_path, false); 2256 if (!so_dir.Exists()) 2257 { 2258 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 2259 if (so_dir.Exists()) 2260 { 2261 // Trim off the incorrect path 2262 full_so_path.erase(0, double_slash_pos + 1); 2263 } 2264 } 2265 } 2266 if (*full_so_path.rbegin() != '/') 2267 full_so_path += '/'; 2268 full_so_path += symbol_name; 2269 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 2270 add_nlist = false; 2271 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2272 } 2273 } 2274 else 2275 { 2276 // This could be a relative path to a N_SO 2277 N_SO_index = sym_idx; 2278 } 2279 } 2280 break; 2281 2282 case N_OSO: 2283 // object file name: name,,0,0,st_mtime 2284 type = eSymbolTypeObjectFile; 2285 break; 2286 2287 case N_LSYM: 2288 // local sym: name,,NO_SECT,type,offset 2289 type = eSymbolTypeLocal; 2290 break; 2291 2292 //---------------------------------------------------------------------- 2293 // INCL scopes 2294 //---------------------------------------------------------------------- 2295 case N_BINCL: 2296 // include file beginning: name,,NO_SECT,0,sum 2297 // We use the current number of symbols in the symbol table in lieu of 2298 // using nlist_idx in case we ever start trimming entries out 2299 N_INCL_indexes.push_back(sym_idx); 2300 type = eSymbolTypeScopeBegin; 2301 break; 2302 2303 case N_EINCL: 2304 // include file end: name,,NO_SECT,0,0 2305 // Set the size of the N_BINCL to the terminating index of this N_EINCL 2306 // so that we can always skip the entire symbol if we need to navigate 2307 // more quickly at the source level when parsing STABS 2308 if ( !N_INCL_indexes.empty() ) 2309 { 2310 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 2311 symbol_ptr->SetByteSize(sym_idx + 1); 2312 symbol_ptr->SetSizeIsSibling(true); 2313 N_INCL_indexes.pop_back(); 2314 } 2315 type = eSymbolTypeScopeEnd; 2316 break; 2317 2318 case N_SOL: 2319 // #included file name: name,,n_sect,0,address 2320 type = eSymbolTypeHeaderFile; 2321 2322 // We currently don't use the header files on darwin 2323 add_nlist = false; 2324 break; 2325 2326 case N_PARAMS: 2327 // compiler parameters: name,,NO_SECT,0,0 2328 type = eSymbolTypeCompiler; 2329 break; 2330 2331 case N_VERSION: 2332 // compiler version: name,,NO_SECT,0,0 2333 type = eSymbolTypeCompiler; 2334 break; 2335 2336 case N_OLEVEL: 2337 // compiler -O level: name,,NO_SECT,0,0 2338 type = eSymbolTypeCompiler; 2339 break; 2340 2341 case N_PSYM: 2342 // parameter: name,,NO_SECT,type,offset 2343 type = eSymbolTypeVariable; 2344 break; 2345 2346 case N_ENTRY: 2347 // alternate entry: name,,n_sect,linenumber,address 2348 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2349 type = eSymbolTypeLineEntry; 2350 break; 2351 2352 //---------------------------------------------------------------------- 2353 // Left and Right Braces 2354 //---------------------------------------------------------------------- 2355 case N_LBRAC: 2356 // left bracket: 0,,NO_SECT,nesting level,address 2357 // We use the current number of symbols in the symbol table in lieu of 2358 // using nlist_idx in case we ever start trimming entries out 2359 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2360 N_BRAC_indexes.push_back(sym_idx); 2361 type = eSymbolTypeScopeBegin; 2362 break; 2363 2364 case N_RBRAC: 2365 // right bracket: 0,,NO_SECT,nesting level,address 2366 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 2367 // so that we can always skip the entire symbol if we need to navigate 2368 // more quickly at the source level when parsing STABS 2369 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2370 if ( !N_BRAC_indexes.empty() ) 2371 { 2372 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 2373 symbol_ptr->SetByteSize(sym_idx + 1); 2374 symbol_ptr->SetSizeIsSibling(true); 2375 N_BRAC_indexes.pop_back(); 2376 } 2377 type = eSymbolTypeScopeEnd; 2378 break; 2379 2380 case N_EXCL: 2381 // deleted include file: name,,NO_SECT,0,sum 2382 type = eSymbolTypeHeaderFile; 2383 break; 2384 2385 //---------------------------------------------------------------------- 2386 // COMM scopes 2387 //---------------------------------------------------------------------- 2388 case N_BCOMM: 2389 // begin common: name,,NO_SECT,0,0 2390 // We use the current number of symbols in the symbol table in lieu of 2391 // using nlist_idx in case we ever start trimming entries out 2392 type = eSymbolTypeScopeBegin; 2393 N_COMM_indexes.push_back(sym_idx); 2394 break; 2395 2396 case N_ECOML: 2397 // end common (local name): 0,,n_sect,0,address 2398 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2399 // Fall through 2400 2401 case N_ECOMM: 2402 // end common: name,,n_sect,0,0 2403 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 2404 // so that we can always skip the entire symbol if we need to navigate 2405 // more quickly at the source level when parsing STABS 2406 if ( !N_COMM_indexes.empty() ) 2407 { 2408 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 2409 symbol_ptr->SetByteSize(sym_idx + 1); 2410 symbol_ptr->SetSizeIsSibling(true); 2411 N_COMM_indexes.pop_back(); 2412 } 2413 type = eSymbolTypeScopeEnd; 2414 break; 2415 2416 case N_LENG: 2417 // second stab entry with length information 2418 type = eSymbolTypeAdditional; 2419 break; 2420 2421 default: break; 2422 } 2423 } 2424 else 2425 { 2426 //uint8_t n_pext = N_PEXT & nlist.n_type; 2427 uint8_t n_type = N_TYPE & nlist.n_type; 2428 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 2429 2430 switch (n_type) 2431 { 2432 case N_INDR: // Fall through 2433 case N_PBUD: // Fall through 2434 case N_UNDF: 2435 type = eSymbolTypeUndefined; 2436 break; 2437 2438 case N_ABS: 2439 type = eSymbolTypeAbsolute; 2440 break; 2441 2442 case N_SECT: 2443 { 2444 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2445 2446 if (symbol_section == NULL) 2447 { 2448 // TODO: warn about this? 2449 add_nlist = false; 2450 break; 2451 } 2452 2453 if (TEXT_eh_frame_sectID == nlist.n_sect) 2454 { 2455 type = eSymbolTypeException; 2456 } 2457 else 2458 { 2459 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 2460 2461 switch (section_type) 2462 { 2463 case S_REGULAR: break; // regular section 2464 //case S_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section 2465 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 2466 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 2467 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 2468 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 2469 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 2470 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 2471 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 2472 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 2473 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 2474 //case S_COALESCED: type = eSymbolType; break; // section contains symbols that are to be coalesced 2475 //case S_GB_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 2476 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 2477 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 2478 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 2479 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 2480 default: break; 2481 } 2482 2483 if (type == eSymbolTypeInvalid) 2484 { 2485 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 2486 if (symbol_section->IsDescendant (text_section_sp.get())) 2487 { 2488 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 2489 S_ATTR_SELF_MODIFYING_CODE | 2490 S_ATTR_SOME_INSTRUCTIONS)) 2491 type = eSymbolTypeData; 2492 else 2493 type = eSymbolTypeCode; 2494 } 2495 else if (symbol_section->IsDescendant(data_section_sp.get())) 2496 { 2497 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 2498 { 2499 type = eSymbolTypeRuntime; 2500 2501 if (symbol_name && 2502 symbol_name[0] == '_' && 2503 symbol_name[1] == 'O' && 2504 symbol_name[2] == 'B') 2505 { 2506 llvm::StringRef symbol_name_ref(symbol_name); 2507 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 2508 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 2509 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 2510 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 2511 { 2512 symbol_name_non_abi_mangled = symbol_name + 1; 2513 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2514 type = eSymbolTypeObjCClass; 2515 demangled_is_synthesized = true; 2516 } 2517 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 2518 { 2519 symbol_name_non_abi_mangled = symbol_name + 1; 2520 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2521 type = eSymbolTypeObjCMetaClass; 2522 demangled_is_synthesized = true; 2523 } 2524 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 2525 { 2526 symbol_name_non_abi_mangled = symbol_name + 1; 2527 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2528 type = eSymbolTypeObjCIVar; 2529 demangled_is_synthesized = true; 2530 } 2531 } 2532 } 2533 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 2534 { 2535 type = eSymbolTypeException; 2536 } 2537 else 2538 { 2539 type = eSymbolTypeData; 2540 } 2541 } 2542 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 2543 { 2544 type = eSymbolTypeTrampoline; 2545 } 2546 else if (symbol_section->IsDescendant(objc_section_sp.get())) 2547 { 2548 type = eSymbolTypeRuntime; 2549 if (symbol_name && symbol_name[0] == '.') 2550 { 2551 llvm::StringRef symbol_name_ref(symbol_name); 2552 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 2553 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 2554 { 2555 symbol_name_non_abi_mangled = symbol_name; 2556 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 2557 type = eSymbolTypeObjCClass; 2558 demangled_is_synthesized = true; 2559 } 2560 } 2561 } 2562 } 2563 } 2564 } 2565 break; 2566 } 2567 } 2568 2569 if (add_nlist) 2570 { 2571 uint64_t symbol_value = nlist.n_value; 2572 if (symbol_name_non_abi_mangled) 2573 { 2574 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 2575 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 2576 } 2577 else 2578 { 2579 bool symbol_name_is_mangled = false; 2580 2581 if (symbol_name && symbol_name[0] == '_') 2582 { 2583 symbol_name_is_mangled = symbol_name[1] == '_'; 2584 symbol_name++; // Skip the leading underscore 2585 } 2586 2587 if (symbol_name) 2588 { 2589 ConstString const_symbol_name(symbol_name); 2590 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 2591 if (is_gsym && is_debug) 2592 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 2593 } 2594 } 2595 if (symbol_section) 2596 { 2597 const addr_t section_file_addr = symbol_section->GetFileAddress(); 2598 if (symbol_byte_size == 0 && function_starts_count > 0) 2599 { 2600 addr_t symbol_lookup_file_addr = nlist.n_value; 2601 // Do an exact address match for non-ARM addresses, else get the closest since 2602 // the symbol might be a thumb symbol which has an address with bit zero set 2603 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 2604 if (is_arm && func_start_entry) 2605 { 2606 // Verify that the function start address is the symbol address (ARM) 2607 // or the symbol address + 1 (thumb) 2608 if (func_start_entry->addr != symbol_lookup_file_addr && 2609 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 2610 { 2611 // Not the right entry, NULL it out... 2612 func_start_entry = NULL; 2613 } 2614 } 2615 if (func_start_entry) 2616 { 2617 func_start_entry->data = true; 2618 2619 addr_t symbol_file_addr = func_start_entry->addr; 2620 uint32_t symbol_flags = 0; 2621 if (is_arm) 2622 { 2623 if (symbol_file_addr & 1) 2624 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 2625 symbol_file_addr &= 0xfffffffffffffffeull; 2626 } 2627 2628 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 2629 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 2630 if (next_func_start_entry) 2631 { 2632 addr_t next_symbol_file_addr = next_func_start_entry->addr; 2633 // Be sure the clear the Thumb address bit when we calculate the size 2634 // from the current and next address 2635 if (is_arm) 2636 next_symbol_file_addr &= 0xfffffffffffffffeull; 2637 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 2638 } 2639 else 2640 { 2641 symbol_byte_size = section_end_file_addr - symbol_file_addr; 2642 } 2643 } 2644 } 2645 symbol_value -= section_file_addr; 2646 } 2647 2648 if (is_debug == false) 2649 { 2650 if (type == eSymbolTypeCode) 2651 { 2652 // See if we can find a N_FUN entry for any code symbols. 2653 // If we do find a match, and the name matches, then we 2654 // can merge the two into just the function symbol to avoid 2655 // duplicate entries in the symbol table 2656 ValueToSymbolIndexMap::const_iterator pos = N_FUN_addr_to_sym_idx.find (nlist.n_value); 2657 if (pos != N_FUN_addr_to_sym_idx.end()) 2658 { 2659 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 2660 { 2661 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 2662 // We just need the flags from the linker symbol, so put these flags 2663 // into the N_FUN flags to avoid duplicate symbols in the symbol table 2664 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2665 sym[sym_idx].Clear(); 2666 continue; 2667 } 2668 } 2669 } 2670 else if (type == eSymbolTypeData) 2671 { 2672 // See if we can find a N_STSYM entry for any data symbols. 2673 // If we do find a match, and the name matches, then we 2674 // can merge the two into just the Static symbol to avoid 2675 // duplicate entries in the symbol table 2676 ValueToSymbolIndexMap::const_iterator pos = N_STSYM_addr_to_sym_idx.find (nlist.n_value); 2677 if (pos != N_STSYM_addr_to_sym_idx.end()) 2678 { 2679 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 2680 { 2681 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 2682 // We just need the flags from the linker symbol, so put these flags 2683 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 2684 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2685 sym[sym_idx].Clear(); 2686 continue; 2687 } 2688 } 2689 else 2690 { 2691 // Combine N_GSYM stab entries with the non stab symbol 2692 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 2693 if (pos != N_GSYM_name_to_sym_idx.end()) 2694 { 2695 const uint32_t GSYM_sym_idx = pos->second; 2696 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 2697 // Copy the address, because often the N_GSYM address has an invalid address of zero 2698 // when the global is a common symbol 2699 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 2700 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 2701 // We just need the flags from the linker symbol, so put these flags 2702 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 2703 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2704 sym[sym_idx].Clear(); 2705 continue; 2706 } 2707 } 2708 } 2709 } 2710 2711 sym[sym_idx].SetID (nlist_idx); 2712 sym[sym_idx].SetType (type); 2713 sym[sym_idx].GetAddress().SetSection (symbol_section); 2714 sym[sym_idx].GetAddress().SetOffset (symbol_value); 2715 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2716 2717 if (symbol_byte_size > 0) 2718 sym[sym_idx].SetByteSize(symbol_byte_size); 2719 2720 if (demangled_is_synthesized) 2721 sym[sym_idx].SetDemangledNameIsSynthesized(true); 2722 ++sym_idx; 2723 } 2724 else 2725 { 2726 sym[sym_idx].Clear(); 2727 } 2728 2729 } 2730 ///////////////////////////// 2731 } 2732 break; // No more entries to consider 2733 } 2734 } 2735 } 2736 } 2737 } 2738 } 2739 } 2740 2741 // Must reset this in case it was mutated above! 2742 nlist_data_offset = 0; 2743 #endif 2744 2745 if (nlist_data.GetByteSize() > 0) 2746 { 2747 2748 // If the sym array was not created while parsing the DSC unmapped 2749 // symbols, create it now. 2750 if (sym == NULL) 2751 { 2752 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 2753 num_syms = symtab->GetNumSymbols(); 2754 } 2755 2756 if (unmapped_local_symbols_found) 2757 { 2758 assert(m_dysymtab.ilocalsym == 0); 2759 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 2760 nlist_idx = m_dysymtab.nlocalsym; 2761 } 2762 else 2763 { 2764 nlist_idx = 0; 2765 } 2766 2767 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) 2768 { 2769 struct nlist_64 nlist; 2770 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2771 break; 2772 2773 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 2774 nlist.n_type = nlist_data.GetU8_unchecked (&nlist_data_offset); 2775 nlist.n_sect = nlist_data.GetU8_unchecked (&nlist_data_offset); 2776 nlist.n_desc = nlist_data.GetU16_unchecked (&nlist_data_offset); 2777 nlist.n_value = nlist_data.GetAddress_unchecked (&nlist_data_offset); 2778 2779 SymbolType type = eSymbolTypeInvalid; 2780 const char *symbol_name = NULL; 2781 2782 if (have_strtab_data) 2783 { 2784 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 2785 2786 if (symbol_name == NULL) 2787 { 2788 // No symbol should be NULL, even the symbols with no 2789 // string values should have an offset zero which points 2790 // to an empty C-string 2791 Host::SystemLog (Host::eSystemLogError, 2792 "error: symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 2793 nlist_idx, 2794 nlist.n_strx, 2795 module_sp->GetFileSpec().GetPath().c_str()); 2796 continue; 2797 } 2798 if (symbol_name[0] == '\0') 2799 symbol_name = NULL; 2800 } 2801 else 2802 { 2803 const addr_t str_addr = strtab_addr + nlist.n_strx; 2804 Error str_error; 2805 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, str_error)) 2806 symbol_name = memory_symbol_name.c_str(); 2807 } 2808 const char *symbol_name_non_abi_mangled = NULL; 2809 2810 SectionSP symbol_section; 2811 lldb::addr_t symbol_byte_size = 0; 2812 bool add_nlist = true; 2813 bool is_gsym = false; 2814 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2815 bool demangled_is_synthesized = false; 2816 2817 assert (sym_idx < num_syms); 2818 2819 sym[sym_idx].SetDebug (is_debug); 2820 2821 if (is_debug) 2822 { 2823 switch (nlist.n_type) 2824 { 2825 case N_GSYM: 2826 // global symbol: name,,NO_SECT,type,0 2827 // Sometimes the N_GSYM value contains the address. 2828 2829 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 2830 // have the same address, but we want to ensure that we always find only the real symbol, 2831 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 2832 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 2833 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 2834 // same address. 2835 2836 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 2837 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 2838 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 2839 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 2840 add_nlist = false; 2841 else 2842 { 2843 is_gsym = true; 2844 sym[sym_idx].SetExternal(true); 2845 if (nlist.n_value != 0) 2846 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2847 type = eSymbolTypeData; 2848 } 2849 break; 2850 2851 case N_FNAME: 2852 // procedure name (f77 kludge): name,,NO_SECT,0,0 2853 type = eSymbolTypeCompiler; 2854 break; 2855 2856 case N_FUN: 2857 // procedure: name,,n_sect,linenumber,address 2858 if (symbol_name) 2859 { 2860 type = eSymbolTypeCode; 2861 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2862 2863 N_FUN_addr_to_sym_idx[nlist.n_value] = sym_idx; 2864 // We use the current number of symbols in the symbol table in lieu of 2865 // using nlist_idx in case we ever start trimming entries out 2866 N_FUN_indexes.push_back(sym_idx); 2867 } 2868 else 2869 { 2870 type = eSymbolTypeCompiler; 2871 2872 if ( !N_FUN_indexes.empty() ) 2873 { 2874 // Copy the size of the function into the original STAB entry so we don't have 2875 // to hunt for it later 2876 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 2877 N_FUN_indexes.pop_back(); 2878 // We don't really need the end function STAB as it contains the size which 2879 // we already placed with the original symbol, so don't add it if we want a 2880 // minimal symbol table 2881 add_nlist = false; 2882 } 2883 } 2884 break; 2885 2886 case N_STSYM: 2887 // static symbol: name,,n_sect,type,address 2888 N_STSYM_addr_to_sym_idx[nlist.n_value] = sym_idx; 2889 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2890 type = eSymbolTypeData; 2891 break; 2892 2893 case N_LCSYM: 2894 // .lcomm symbol: name,,n_sect,type,address 2895 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2896 type = eSymbolTypeCommonBlock; 2897 break; 2898 2899 case N_BNSYM: 2900 // We use the current number of symbols in the symbol table in lieu of 2901 // using nlist_idx in case we ever start trimming entries out 2902 // Skip these if we want minimal symbol tables 2903 add_nlist = false; 2904 break; 2905 2906 case 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 N_OPT: 2916 // emitted with gcc2_compiled and in gcc source 2917 type = eSymbolTypeCompiler; 2918 break; 2919 2920 case N_RSYM: 2921 // register sym: name,,NO_SECT,type,register 2922 type = eSymbolTypeVariable; 2923 break; 2924 2925 case N_SLINE: 2926 // 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 N_SSYM: 2932 // structure elt: name,,NO_SECT,type,struct_offset 2933 type = eSymbolTypeVariableType; 2934 break; 2935 2936 case N_SO: 2937 // 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 string 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 N_OSO: 3024 // object file name: name,,0,0,st_mtime 3025 type = eSymbolTypeObjectFile; 3026 break; 3027 3028 case N_LSYM: 3029 // local sym: name,,NO_SECT,type,offset 3030 type = eSymbolTypeLocal; 3031 break; 3032 3033 //---------------------------------------------------------------------- 3034 // INCL scopes 3035 //---------------------------------------------------------------------- 3036 case N_BINCL: 3037 // 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 N_EINCL: 3045 // 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 N_SOL: 3060 // #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 N_PARAMS: 3068 // compiler parameters: name,,NO_SECT,0,0 3069 type = eSymbolTypeCompiler; 3070 break; 3071 3072 case N_VERSION: 3073 // compiler version: name,,NO_SECT,0,0 3074 type = eSymbolTypeCompiler; 3075 break; 3076 3077 case N_OLEVEL: 3078 // compiler -O level: name,,NO_SECT,0,0 3079 type = eSymbolTypeCompiler; 3080 break; 3081 3082 case N_PSYM: 3083 // parameter: name,,NO_SECT,type,offset 3084 type = eSymbolTypeVariable; 3085 break; 3086 3087 case N_ENTRY: 3088 // 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 N_LBRAC: 3097 // 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 N_RBRAC: 3106 // 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 N_EXCL: 3122 // deleted include file: name,,NO_SECT,0,sum 3123 type = eSymbolTypeHeaderFile; 3124 break; 3125 3126 //---------------------------------------------------------------------- 3127 // COMM scopes 3128 //---------------------------------------------------------------------- 3129 case N_BCOMM: 3130 // 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 N_ECOML: 3138 // 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 N_ECOMM: 3143 // 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 N_LENG: 3158 // 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 = N_PEXT & nlist.n_type; 3168 uint8_t n_type = N_TYPE & nlist.n_type; 3169 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3170 3171 switch (n_type) 3172 { 3173 case N_INDR:// Fall through 3174 case N_PBUD:// Fall through 3175 case N_UNDF: 3176 type = eSymbolTypeUndefined; 3177 break; 3178 3179 case N_ABS: 3180 type = eSymbolTypeAbsolute; 3181 break; 3182 3183 case 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() & SECTION_TYPE; 3201 3202 switch (section_type) 3203 { 3204 case S_REGULAR: break; // regular section 3205 //case S_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section 3206 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 3207 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 3208 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 3209 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 3210 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 3211 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 3212 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 3213 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 3214 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 3215 //case S_COALESCED: type = eSymbolType; break; // section contains symbols that are to be coalesced 3216 //case S_GB_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 3217 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 3218 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 3219 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 3220 case S_LAZY_DYLIB_SYMBOL_POINTERS: 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(S_ATTR_PURE_INSTRUCTIONS | 3230 S_ATTR_SELF_MODIFYING_CODE | 3231 S_ATTR_SOME_INSTRUCTIONS)) 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 & SECTION_TYPE) == S_SYMBOL_STUBS) 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 & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 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 == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 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 == LC_UUID) 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 LC_LOAD_DYLIB: 3805 case LC_LOAD_WEAK_DYLIB: 3806 case LC_REEXPORT_DYLIB: 3807 case LC_LOAD_DYLINKER: 3808 case LC_LOADFVMLIB: 3809 case LC_LOAD_UPWARD_DYLIB: 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 LC_UNIXTHREAD: 3880 case LC_THREAD: 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::CPU_TYPE_ARM: 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::CPU_TYPE_I386: 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::CPU_TYPE_X86_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 LC_MAIN: 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 == LC_THREAD) 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::CPU_TYPE_ARM: 4055 reg_ctx_sp.reset (new RegisterContextDarwin_arm_Mach (thread, data)); 4056 break; 4057 4058 case llvm::MachO::CPU_TYPE_I386: 4059 reg_ctx_sp.reset (new RegisterContextDarwin_i386_Mach (thread, data)); 4060 break; 4061 4062 case llvm::MachO::CPU_TYPE_X86_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 MH_OBJECT: // 0x1u 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 MH_EXECUTE: return eTypeExecutable; // 0x2u 4096 case MH_FVMLIB: return eTypeSharedLibrary; // 0x3u 4097 case MH_CORE: return eTypeCoreFile; // 0x4u 4098 case MH_PRELOAD: return eTypeSharedLibrary; // 0x5u 4099 case MH_DYLIB: return eTypeSharedLibrary; // 0x6u 4100 case MH_DYLINKER: return eTypeDynamicLinker; // 0x7u 4101 case MH_BUNDLE: return eTypeSharedLibrary; // 0x8u 4102 case MH_DYLIB_STUB: return eTypeStubLibrary; // 0x9u 4103 case MH_DSYM: return eTypeDebugInfo; // 0xAu 4104 case MH_KEXT_BUNDLE: return eTypeSharedLibrary; // 0xBu 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 MH_OBJECT: // 0x1u 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 MH_EXECUTE: // 0x2u 4135 // Check for the MH_DYLDLINK bit in the flags 4136 if (m_header.flags & MH_DYLDLINK) 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 MH_FVMLIB: return eStrataUser; // 0x3u 4153 case MH_CORE: return eStrataUnknown; // 0x4u 4154 case MH_PRELOAD: return eStrataRawImage; // 0x5u 4155 case MH_DYLIB: return eStrataUser; // 0x6u 4156 case MH_DYLINKER: return eStrataUser; // 0x7u 4157 case MH_BUNDLE: return eStrataUser; // 0x8u 4158 case MH_DYLIB_STUB: return eStrataUser; // 0x9u 4159 case MH_DSYM: return eStrataUnknown; // 0xAu 4160 case MH_KEXT_BUNDLE: return eStrataKernel; // 0xBu 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 == LC_ID_DYLIB) 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 == LC_ID_DYLIB) 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 == MH_PRELOAD) 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 != MH_MAGIC 4262 && version_or_magic != MH_CIGAM 4263 && version_or_magic != MH_MAGIC_64 4264 && version_or_magic != MH_CIGAM_64 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