ObjectFileMachO.cpp revision 0bf22380daa9ddbd8c0e8da81d3bc82c2601e5e6
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/FileSpecList.h" 19#include "lldb/Core/Log.h" 20#include "lldb/Core/Module.h" 21#include "lldb/Core/PluginManager.h" 22#include "lldb/Core/RangeMap.h" 23#include "lldb/Core/Section.h" 24#include "lldb/Core/StreamFile.h" 25#include "lldb/Core/StreamString.h" 26#include "lldb/Core/Timer.h" 27#include "lldb/Core/UUID.h" 28#include "lldb/Host/Host.h" 29#include "lldb/Host/FileSpec.h" 30#include "lldb/Symbol/ClangNamespaceDecl.h" 31#include "lldb/Symbol/ObjectFile.h" 32#include "lldb/Target/Platform.h" 33#include "lldb/Target/Process.h" 34#include "lldb/Target/Target.h" 35#include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 36#include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 37#include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 38 39using namespace lldb; 40using namespace lldb_private; 41using namespace llvm::MachO; 42 43class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 44{ 45public: 46 RegisterContextDarwin_x86_64_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 47 RegisterContextDarwin_x86_64 (thread, 0) 48 { 49 SetRegisterDataFrom_LC_THREAD (data); 50 } 51 52 virtual void 53 InvalidateAllRegisters () 54 { 55 // Do nothing... registers are always valid... 56 } 57 58 void 59 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 60 { 61 lldb::offset_t offset = 0; 62 SetError (GPRRegSet, Read, -1); 63 SetError (FPURegSet, Read, -1); 64 SetError (EXCRegSet, Read, -1); 65 bool done = false; 66 67 while (!done) 68 { 69 int flavor = data.GetU32 (&offset); 70 if (flavor == 0) 71 done = true; 72 else 73 { 74 uint32_t i; 75 uint32_t count = data.GetU32 (&offset); 76 switch (flavor) 77 { 78 case GPRRegSet: 79 for (i=0; i<count; ++i) 80 (&gpr.rax)[i] = data.GetU64(&offset); 81 SetError (GPRRegSet, Read, 0); 82 done = true; 83 84 break; 85 case FPURegSet: 86 // TODO: fill in FPU regs.... 87 //SetError (FPURegSet, Read, -1); 88 done = true; 89 90 break; 91 case EXCRegSet: 92 exc.trapno = data.GetU32(&offset); 93 exc.err = data.GetU32(&offset); 94 exc.faultvaddr = data.GetU64(&offset); 95 SetError (EXCRegSet, Read, 0); 96 done = true; 97 break; 98 case 7: 99 case 8: 100 case 9: 101 // fancy flavors that encapsulate of the the above 102 // falvors... 103 break; 104 105 default: 106 done = true; 107 break; 108 } 109 } 110 } 111 } 112protected: 113 virtual int 114 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 115 { 116 return 0; 117 } 118 119 virtual int 120 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 121 { 122 return 0; 123 } 124 125 virtual int 126 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 127 { 128 return 0; 129 } 130 131 virtual int 132 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 133 { 134 return 0; 135 } 136 137 virtual int 138 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 139 { 140 return 0; 141 } 142 143 virtual int 144 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 145 { 146 return 0; 147 } 148}; 149 150 151class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 152{ 153public: 154 RegisterContextDarwin_i386_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 155 RegisterContextDarwin_i386 (thread, 0) 156 { 157 SetRegisterDataFrom_LC_THREAD (data); 158 } 159 160 virtual void 161 InvalidateAllRegisters () 162 { 163 // Do nothing... registers are always valid... 164 } 165 166 void 167 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 168 { 169 lldb::offset_t offset = 0; 170 SetError (GPRRegSet, Read, -1); 171 SetError (FPURegSet, Read, -1); 172 SetError (EXCRegSet, Read, -1); 173 bool done = false; 174 175 while (!done) 176 { 177 int flavor = data.GetU32 (&offset); 178 if (flavor == 0) 179 done = true; 180 else 181 { 182 uint32_t i; 183 uint32_t count = data.GetU32 (&offset); 184 switch (flavor) 185 { 186 case GPRRegSet: 187 for (i=0; i<count; ++i) 188 (&gpr.eax)[i] = data.GetU32(&offset); 189 SetError (GPRRegSet, Read, 0); 190 done = true; 191 192 break; 193 case FPURegSet: 194 // TODO: fill in FPU regs.... 195 //SetError (FPURegSet, Read, -1); 196 done = true; 197 198 break; 199 case EXCRegSet: 200 exc.trapno = data.GetU32(&offset); 201 exc.err = data.GetU32(&offset); 202 exc.faultvaddr = data.GetU32(&offset); 203 SetError (EXCRegSet, Read, 0); 204 done = true; 205 break; 206 case 7: 207 case 8: 208 case 9: 209 // fancy flavors that encapsulate of the the above 210 // falvors... 211 break; 212 213 default: 214 done = true; 215 break; 216 } 217 } 218 } 219 } 220protected: 221 virtual int 222 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 223 { 224 return 0; 225 } 226 227 virtual int 228 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 229 { 230 return 0; 231 } 232 233 virtual int 234 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 235 { 236 return 0; 237 } 238 239 virtual int 240 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 241 { 242 return 0; 243 } 244 245 virtual int 246 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 247 { 248 return 0; 249 } 250 251 virtual int 252 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 253 { 254 return 0; 255 } 256}; 257 258class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm 259{ 260public: 261 RegisterContextDarwin_arm_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 262 RegisterContextDarwin_arm (thread, 0) 263 { 264 SetRegisterDataFrom_LC_THREAD (data); 265 } 266 267 virtual void 268 InvalidateAllRegisters () 269 { 270 // Do nothing... registers are always valid... 271 } 272 273 void 274 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 275 { 276 lldb::offset_t offset = 0; 277 SetError (GPRRegSet, Read, -1); 278 SetError (FPURegSet, Read, -1); 279 SetError (EXCRegSet, Read, -1); 280 int flavor = data.GetU32 (&offset); 281 uint32_t count = data.GetU32 (&offset); 282 switch (flavor) 283 { 284 case GPRRegSet: 285 for (uint32_t i=0; i<count; ++i) 286 gpr.r[i] = data.GetU32(&offset); 287 SetError (GPRRegSet, Read, 0); 288 break; 289 case FPURegSet: 290 // TODO: fill in FPU regs.... 291 //SetError (FPURegSet, Read, -1); 292 break; 293 case EXCRegSet: 294 exc.exception = data.GetU32(&offset); 295 exc.fsr = data.GetU32(&offset); 296 exc.far = data.GetU32(&offset); 297 SetError (EXCRegSet, Read, 0); 298 break; 299 } 300 } 301protected: 302 virtual int 303 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 304 { 305 return 0; 306 } 307 308 virtual int 309 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 310 { 311 return 0; 312 } 313 314 virtual int 315 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 316 { 317 return 0; 318 } 319 320 virtual int 321 DoReadDBG (lldb::tid_t tid, int flavor, DBG &dbg) 322 { 323 return -1; 324 } 325 326 virtual int 327 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 328 { 329 return 0; 330 } 331 332 virtual int 333 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 334 { 335 return 0; 336 } 337 338 virtual int 339 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 340 { 341 return 0; 342 } 343 344 virtual int 345 DoWriteDBG (lldb::tid_t tid, int flavor, const DBG &dbg) 346 { 347 return -1; 348 } 349}; 350 351#define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 352 353void 354ObjectFileMachO::Initialize() 355{ 356 PluginManager::RegisterPlugin (GetPluginNameStatic(), 357 GetPluginDescriptionStatic(), 358 CreateInstance, 359 CreateMemoryInstance); 360} 361 362void 363ObjectFileMachO::Terminate() 364{ 365 PluginManager::UnregisterPlugin (CreateInstance); 366} 367 368 369const char * 370ObjectFileMachO::GetPluginNameStatic() 371{ 372 return "object-file.mach-o"; 373} 374 375const char * 376ObjectFileMachO::GetPluginDescriptionStatic() 377{ 378 return "Mach-o object file reader (32 and 64 bit)"; 379} 380 381 382ObjectFile * 383ObjectFileMachO::CreateInstance (const lldb::ModuleSP &module_sp, DataBufferSP& data_sp, const FileSpec* file, addr_t offset, addr_t length) 384{ 385 if (ObjectFileMachO::MagicBytesMatch(data_sp, offset, length)) 386 { 387 std::auto_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module_sp, data_sp, file, offset, length)); 388 if (objfile_ap.get() && objfile_ap->ParseHeader()) 389 return objfile_ap.release(); 390 } 391 return NULL; 392} 393 394ObjectFile * 395ObjectFileMachO::CreateMemoryInstance (const lldb::ModuleSP &module_sp, 396 DataBufferSP& data_sp, 397 const ProcessSP &process_sp, 398 lldb::addr_t header_addr) 399{ 400 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) 401 { 402 std::auto_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module_sp, data_sp, process_sp, header_addr)); 403 if (objfile_ap.get() && objfile_ap->ParseHeader()) 404 return objfile_ap.release(); 405 } 406 return NULL; 407} 408 409 410const ConstString & 411ObjectFileMachO::GetSegmentNameTEXT() 412{ 413 static ConstString g_segment_name_TEXT ("__TEXT"); 414 return g_segment_name_TEXT; 415} 416 417const ConstString & 418ObjectFileMachO::GetSegmentNameDATA() 419{ 420 static ConstString g_segment_name_DATA ("__DATA"); 421 return g_segment_name_DATA; 422} 423 424const ConstString & 425ObjectFileMachO::GetSegmentNameOBJC() 426{ 427 static ConstString g_segment_name_OBJC ("__OBJC"); 428 return g_segment_name_OBJC; 429} 430 431const ConstString & 432ObjectFileMachO::GetSegmentNameLINKEDIT() 433{ 434 static ConstString g_section_name_LINKEDIT ("__LINKEDIT"); 435 return g_section_name_LINKEDIT; 436} 437 438const ConstString & 439ObjectFileMachO::GetSectionNameEHFrame() 440{ 441 static ConstString g_section_name_eh_frame ("__eh_frame"); 442 return g_section_name_eh_frame; 443} 444 445 446 447static uint32_t 448MachHeaderSizeFromMagic(uint32_t magic) 449{ 450 switch (magic) 451 { 452 case HeaderMagic32: 453 case HeaderMagic32Swapped: 454 return sizeof(struct mach_header); 455 456 case HeaderMagic64: 457 case HeaderMagic64Swapped: 458 return sizeof(struct mach_header_64); 459 break; 460 461 default: 462 break; 463 } 464 return 0; 465} 466 467 468bool 469ObjectFileMachO::MagicBytesMatch (DataBufferSP& data_sp, 470 lldb::addr_t data_offset, 471 lldb::addr_t data_length) 472{ 473 DataExtractor data; 474 data.SetData (data_sp, data_offset, data_length); 475 lldb::offset_t offset = 0; 476 uint32_t magic = data.GetU32(&offset); 477 return MachHeaderSizeFromMagic(magic) != 0; 478} 479 480 481ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, DataBufferSP& data_sp, const FileSpec* file, addr_t offset, addr_t length) : 482 ObjectFile(module_sp, file, offset, length, data_sp), 483 m_sections_ap(), 484 m_symtab_ap(), 485 m_mach_segments(), 486 m_mach_sections(), 487 m_entry_point_address(), 488 m_thread_context_offsets(), 489 m_thread_context_offsets_valid(false) 490{ 491 ::memset (&m_header, 0, sizeof(m_header)); 492 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 493} 494 495ObjectFileMachO::ObjectFileMachO (const lldb::ModuleSP &module_sp, 496 lldb::DataBufferSP& header_data_sp, 497 const lldb::ProcessSP &process_sp, 498 lldb::addr_t header_addr) : 499 ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 500 m_sections_ap(), 501 m_symtab_ap(), 502 m_mach_segments(), 503 m_mach_sections(), 504 m_entry_point_address(), 505 m_thread_context_offsets(), 506 m_thread_context_offsets_valid(false) 507{ 508 ::memset (&m_header, 0, sizeof(m_header)); 509 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 510} 511 512ObjectFileMachO::~ObjectFileMachO() 513{ 514} 515 516 517bool 518ObjectFileMachO::ParseHeader () 519{ 520 ModuleSP module_sp(GetModule()); 521 if (module_sp) 522 { 523 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 524 bool can_parse = false; 525 lldb::offset_t offset = 0; 526 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 527 // Leave magic in the original byte order 528 m_header.magic = m_data.GetU32(&offset); 529 switch (m_header.magic) 530 { 531 case HeaderMagic32: 532 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 533 m_data.SetAddressByteSize(4); 534 can_parse = true; 535 break; 536 537 case HeaderMagic64: 538 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 539 m_data.SetAddressByteSize(8); 540 can_parse = true; 541 break; 542 543 case HeaderMagic32Swapped: 544 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 545 m_data.SetAddressByteSize(4); 546 can_parse = true; 547 break; 548 549 case HeaderMagic64Swapped: 550 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 551 m_data.SetAddressByteSize(8); 552 can_parse = true; 553 break; 554 555 default: 556 break; 557 } 558 559 if (can_parse) 560 { 561 m_data.GetU32(&offset, &m_header.cputype, 6); 562 563 ArchSpec mach_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 564 565 // Check if the module has a required architecture 566 const ArchSpec &module_arch = module_sp->GetArchitecture(); 567 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 568 return false; 569 570 if (SetModulesArchitecture (mach_arch)) 571 { 572 const size_t header_and_lc_size = m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 573 if (m_data.GetByteSize() < header_and_lc_size) 574 { 575 DataBufferSP data_sp; 576 ProcessSP process_sp (m_process_wp.lock()); 577 if (process_sp) 578 { 579 data_sp = ReadMemory (process_sp, m_offset, header_and_lc_size); 580 } 581 else 582 { 583 // Read in all only the load command data from the file on disk 584 data_sp = m_file.ReadFileContents(m_offset, header_and_lc_size); 585 if (data_sp->GetByteSize() != header_and_lc_size) 586 return false; 587 } 588 if (data_sp) 589 m_data.SetData (data_sp); 590 } 591 } 592 return true; 593 } 594 else 595 { 596 memset(&m_header, 0, sizeof(struct mach_header)); 597 } 598 } 599 return false; 600} 601 602 603ByteOrder 604ObjectFileMachO::GetByteOrder () const 605{ 606 return m_data.GetByteOrder (); 607} 608 609bool 610ObjectFileMachO::IsExecutable() const 611{ 612 return m_header.filetype == HeaderFileTypeExecutable; 613} 614 615uint32_t 616ObjectFileMachO::GetAddressByteSize () const 617{ 618 return m_data.GetAddressByteSize (); 619} 620 621AddressClass 622ObjectFileMachO::GetAddressClass (lldb::addr_t file_addr) 623{ 624 Symtab *symtab = GetSymtab(); 625 if (symtab) 626 { 627 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 628 if (symbol) 629 { 630 if (symbol->ValueIsAddress()) 631 { 632 SectionSP section_sp (symbol->GetAddress().GetSection()); 633 if (section_sp) 634 { 635 const SectionType section_type = section_sp->GetType(); 636 switch (section_type) 637 { 638 case eSectionTypeInvalid: return eAddressClassUnknown; 639 case eSectionTypeCode: 640 if (m_header.cputype == llvm::MachO::CPUTypeARM) 641 { 642 // For ARM we have a bit in the n_desc field of the symbol 643 // that tells us ARM/Thumb which is bit 0x0008. 644 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 645 return eAddressClassCodeAlternateISA; 646 } 647 return eAddressClassCode; 648 649 case eSectionTypeContainer: return eAddressClassUnknown; 650 case eSectionTypeData: 651 case eSectionTypeDataCString: 652 case eSectionTypeDataCStringPointers: 653 case eSectionTypeDataSymbolAddress: 654 case eSectionTypeData4: 655 case eSectionTypeData8: 656 case eSectionTypeData16: 657 case eSectionTypeDataPointers: 658 case eSectionTypeZeroFill: 659 case eSectionTypeDataObjCMessageRefs: 660 case eSectionTypeDataObjCCFStrings: 661 return eAddressClassData; 662 case eSectionTypeDebug: 663 case eSectionTypeDWARFDebugAbbrev: 664 case eSectionTypeDWARFDebugAranges: 665 case eSectionTypeDWARFDebugFrame: 666 case eSectionTypeDWARFDebugInfo: 667 case eSectionTypeDWARFDebugLine: 668 case eSectionTypeDWARFDebugLoc: 669 case eSectionTypeDWARFDebugMacInfo: 670 case eSectionTypeDWARFDebugPubNames: 671 case eSectionTypeDWARFDebugPubTypes: 672 case eSectionTypeDWARFDebugRanges: 673 case eSectionTypeDWARFDebugStr: 674 case eSectionTypeDWARFAppleNames: 675 case eSectionTypeDWARFAppleTypes: 676 case eSectionTypeDWARFAppleNamespaces: 677 case eSectionTypeDWARFAppleObjC: 678 return eAddressClassDebug; 679 case eSectionTypeEHFrame: return eAddressClassRuntime; 680 case eSectionTypeOther: return eAddressClassUnknown; 681 } 682 } 683 } 684 685 const SymbolType symbol_type = symbol->GetType(); 686 switch (symbol_type) 687 { 688 case eSymbolTypeAny: return eAddressClassUnknown; 689 case eSymbolTypeAbsolute: return eAddressClassUnknown; 690 691 case eSymbolTypeCode: 692 case eSymbolTypeTrampoline: 693 if (m_header.cputype == llvm::MachO::CPUTypeARM) 694 { 695 // For ARM we have a bit in the n_desc field of the symbol 696 // that tells us ARM/Thumb which is bit 0x0008. 697 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 698 return eAddressClassCodeAlternateISA; 699 } 700 return eAddressClassCode; 701 702 case eSymbolTypeData: return eAddressClassData; 703 case eSymbolTypeRuntime: return eAddressClassRuntime; 704 case eSymbolTypeException: return eAddressClassRuntime; 705 case eSymbolTypeSourceFile: return eAddressClassDebug; 706 case eSymbolTypeHeaderFile: return eAddressClassDebug; 707 case eSymbolTypeObjectFile: return eAddressClassDebug; 708 case eSymbolTypeCommonBlock: return eAddressClassDebug; 709 case eSymbolTypeBlock: return eAddressClassDebug; 710 case eSymbolTypeLocal: return eAddressClassData; 711 case eSymbolTypeParam: return eAddressClassData; 712 case eSymbolTypeVariable: return eAddressClassData; 713 case eSymbolTypeVariableType: return eAddressClassDebug; 714 case eSymbolTypeLineEntry: return eAddressClassDebug; 715 case eSymbolTypeLineHeader: return eAddressClassDebug; 716 case eSymbolTypeScopeBegin: return eAddressClassDebug; 717 case eSymbolTypeScopeEnd: return eAddressClassDebug; 718 case eSymbolTypeAdditional: return eAddressClassUnknown; 719 case eSymbolTypeCompiler: return eAddressClassDebug; 720 case eSymbolTypeInstrumentation:return eAddressClassDebug; 721 case eSymbolTypeUndefined: return eAddressClassUnknown; 722 case eSymbolTypeObjCClass: return eAddressClassRuntime; 723 case eSymbolTypeObjCMetaClass: return eAddressClassRuntime; 724 case eSymbolTypeObjCIVar: return eAddressClassRuntime; 725 } 726 } 727 } 728 return eAddressClassUnknown; 729} 730 731Symtab * 732ObjectFileMachO::GetSymtab() 733{ 734 ModuleSP module_sp(GetModule()); 735 if (module_sp) 736 { 737 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 738 if (m_symtab_ap.get() == NULL) 739 { 740 m_symtab_ap.reset(new Symtab(this)); 741 Mutex::Locker symtab_locker (m_symtab_ap->GetMutex()); 742 ParseSymtab (true); 743 m_symtab_ap->Finalize (); 744 } 745 } 746 return m_symtab_ap.get(); 747} 748 749 750SectionList * 751ObjectFileMachO::GetSectionList() 752{ 753 ModuleSP module_sp(GetModule()); 754 if (module_sp) 755 { 756 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 757 if (m_sections_ap.get() == NULL) 758 { 759 m_sections_ap.reset(new SectionList()); 760 ParseSections(); 761 } 762 } 763 return m_sections_ap.get(); 764} 765 766 767size_t 768ObjectFileMachO::ParseSections () 769{ 770 lldb::user_id_t segID = 0; 771 lldb::user_id_t sectID = 0; 772 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 773 uint32_t i; 774 const bool is_core = GetType() == eTypeCoreFile; 775 //bool dump_sections = false; 776 ModuleSP module_sp (GetModule()); 777 // First look up any LC_ENCRYPTION_INFO load commands 778 typedef RangeArray<uint32_t, uint32_t, 8> EncryptedFileRanges; 779 EncryptedFileRanges encrypted_file_ranges; 780 encryption_info_command encryption_cmd; 781 for (i=0; i<m_header.ncmds; ++i) 782 { 783 const lldb::offset_t load_cmd_offset = offset; 784 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 785 break; 786 787 if (encryption_cmd.cmd == LoadCommandEncryptionInfo) 788 { 789 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) 790 { 791 if (encryption_cmd.cryptid != 0) 792 { 793 EncryptedFileRanges::Entry entry; 794 entry.SetRangeBase(encryption_cmd.cryptoff); 795 entry.SetByteSize(encryption_cmd.cryptsize); 796 encrypted_file_ranges.Append(entry); 797 } 798 } 799 } 800 offset = load_cmd_offset + encryption_cmd.cmdsize; 801 } 802 803 offset = MachHeaderSizeFromMagic(m_header.magic); 804 805 struct segment_command_64 load_cmd; 806 for (i=0; i<m_header.ncmds; ++i) 807 { 808 const lldb::offset_t load_cmd_offset = offset; 809 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 810 break; 811 812 if (load_cmd.cmd == LoadCommandSegment32 || load_cmd.cmd == LoadCommandSegment64) 813 { 814 if (m_data.GetU8(&offset, (uint8_t*)load_cmd.segname, 16)) 815 { 816 load_cmd.vmaddr = m_data.GetAddress(&offset); 817 load_cmd.vmsize = m_data.GetAddress(&offset); 818 load_cmd.fileoff = m_data.GetAddress(&offset); 819 load_cmd.filesize = m_data.GetAddress(&offset); 820 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 821 { 822 823 const bool segment_is_encrypted = (load_cmd.flags & SegmentCommandFlagBitProtectedVersion1) != 0; 824 825 // Keep a list of mach segments around in case we need to 826 // get at data that isn't stored in the abstracted Sections. 827 m_mach_segments.push_back (load_cmd); 828 829 ConstString segment_name (load_cmd.segname, std::min<size_t>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 830 // Use a segment ID of the segment index shifted left by 8 so they 831 // never conflict with any of the sections. 832 SectionSP segment_sp; 833 if (segment_name || is_core) 834 { 835 segment_sp.reset(new Section (module_sp, // Module to which this section belongs 836 ++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 837 segment_name, // Name of this section 838 eSectionTypeContainer, // This section is a container of other sections. 839 load_cmd.vmaddr, // File VM address == addresses as they are found in the object file 840 load_cmd.vmsize, // VM size in bytes of this section 841 load_cmd.fileoff, // Offset to the data for this section in the file 842 load_cmd.filesize, // Size in bytes of this section as found in the the file 843 load_cmd.flags)); // Flags for this section 844 845 segment_sp->SetIsEncrypted (segment_is_encrypted); 846 m_sections_ap->AddSection(segment_sp); 847 } 848 849 struct section_64 sect64; 850 ::memset (§64, 0, sizeof(sect64)); 851 // Push a section into our mach sections for the section at 852 // index zero (NListSectionNoSection) if we don't have any 853 // mach sections yet... 854 if (m_mach_sections.empty()) 855 m_mach_sections.push_back(sect64); 856 uint32_t segment_sect_idx; 857 const lldb::user_id_t first_segment_sectID = sectID + 1; 858 859 860 const uint32_t num_u32s = load_cmd.cmd == LoadCommandSegment32 ? 7 : 8; 861 for (segment_sect_idx=0; segment_sect_idx<load_cmd.nsects; ++segment_sect_idx) 862 { 863 if (m_data.GetU8(&offset, (uint8_t*)sect64.sectname, sizeof(sect64.sectname)) == NULL) 864 break; 865 if (m_data.GetU8(&offset, (uint8_t*)sect64.segname, sizeof(sect64.segname)) == NULL) 866 break; 867 sect64.addr = m_data.GetAddress(&offset); 868 sect64.size = m_data.GetAddress(&offset); 869 870 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 871 break; 872 873 // Keep a list of mach sections around in case we need to 874 // get at data that isn't stored in the abstracted Sections. 875 m_mach_sections.push_back (sect64); 876 877 ConstString section_name (sect64.sectname, std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 878 if (!segment_name) 879 { 880 // We have a segment with no name so we need to conjure up 881 // segments that correspond to the section's segname if there 882 // isn't already such a section. If there is such a section, 883 // we resize the section so that it spans all sections. 884 // We also mark these sections as fake so address matches don't 885 // hit if they land in the gaps between the child sections. 886 segment_name.SetTrimmedCStringWithLength(sect64.segname, sizeof(sect64.segname)); 887 segment_sp = m_sections_ap->FindSectionByName (segment_name); 888 if (segment_sp.get()) 889 { 890 Section *segment = segment_sp.get(); 891 // Grow the section size as needed. 892 const lldb::addr_t sect64_min_addr = sect64.addr; 893 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 894 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 895 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 896 const lldb::addr_t curr_seg_max_addr = curr_seg_min_addr + curr_seg_byte_size; 897 if (sect64_min_addr >= curr_seg_min_addr) 898 { 899 const lldb::addr_t new_seg_byte_size = sect64_max_addr - curr_seg_min_addr; 900 // Only grow the section size if needed 901 if (new_seg_byte_size > curr_seg_byte_size) 902 segment->SetByteSize (new_seg_byte_size); 903 } 904 else 905 { 906 // We need to change the base address of the segment and 907 // adjust the child section offsets for all existing children. 908 const lldb::addr_t slide_amount = sect64_min_addr - curr_seg_min_addr; 909 segment->Slide(slide_amount, false); 910 segment->GetChildren().Slide(-slide_amount, false); 911 segment->SetByteSize (curr_seg_max_addr - sect64_min_addr); 912 } 913 914 // Grow the section size as needed. 915 if (sect64.offset) 916 { 917 const lldb::addr_t segment_min_file_offset = segment->GetFileOffset(); 918 const lldb::addr_t segment_max_file_offset = segment_min_file_offset + segment->GetFileSize(); 919 920 const lldb::addr_t section_min_file_offset = sect64.offset; 921 const lldb::addr_t section_max_file_offset = section_min_file_offset + sect64.size; 922 const lldb::addr_t new_file_offset = std::min (section_min_file_offset, segment_min_file_offset); 923 const lldb::addr_t new_file_size = std::max (section_max_file_offset, segment_max_file_offset) - new_file_offset; 924 segment->SetFileOffset (new_file_offset); 925 segment->SetFileSize (new_file_size); 926 } 927 } 928 else 929 { 930 // Create a fake section for the section's named segment 931 segment_sp.reset(new Section (segment_sp, // Parent section 932 module_sp, // Module to which this section belongs 933 ++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 934 segment_name, // Name of this section 935 eSectionTypeContainer, // This section is a container of other sections. 936 sect64.addr, // File VM address == addresses as they are found in the object file 937 sect64.size, // VM size in bytes of this section 938 sect64.offset, // Offset to the data for this section in the file 939 sect64.offset ? sect64.size : 0, // Size in bytes of this section as found in the the file 940 load_cmd.flags)); // Flags for this section 941 segment_sp->SetIsFake(true); 942 m_sections_ap->AddSection(segment_sp); 943 segment_sp->SetIsEncrypted (segment_is_encrypted); 944 } 945 } 946 assert (segment_sp.get()); 947 948 uint32_t mach_sect_type = sect64.flags & SectionFlagMaskSectionType; 949 static ConstString g_sect_name_objc_data ("__objc_data"); 950 static ConstString g_sect_name_objc_msgrefs ("__objc_msgrefs"); 951 static ConstString g_sect_name_objc_selrefs ("__objc_selrefs"); 952 static ConstString g_sect_name_objc_classrefs ("__objc_classrefs"); 953 static ConstString g_sect_name_objc_superrefs ("__objc_superrefs"); 954 static ConstString g_sect_name_objc_const ("__objc_const"); 955 static ConstString g_sect_name_objc_classlist ("__objc_classlist"); 956 static ConstString g_sect_name_cfstring ("__cfstring"); 957 958 static ConstString g_sect_name_dwarf_debug_abbrev ("__debug_abbrev"); 959 static ConstString g_sect_name_dwarf_debug_aranges ("__debug_aranges"); 960 static ConstString g_sect_name_dwarf_debug_frame ("__debug_frame"); 961 static ConstString g_sect_name_dwarf_debug_info ("__debug_info"); 962 static ConstString g_sect_name_dwarf_debug_line ("__debug_line"); 963 static ConstString g_sect_name_dwarf_debug_loc ("__debug_loc"); 964 static ConstString g_sect_name_dwarf_debug_macinfo ("__debug_macinfo"); 965 static ConstString g_sect_name_dwarf_debug_pubnames ("__debug_pubnames"); 966 static ConstString g_sect_name_dwarf_debug_pubtypes ("__debug_pubtypes"); 967 static ConstString g_sect_name_dwarf_debug_ranges ("__debug_ranges"); 968 static ConstString g_sect_name_dwarf_debug_str ("__debug_str"); 969 static ConstString g_sect_name_dwarf_apple_names ("__apple_names"); 970 static ConstString g_sect_name_dwarf_apple_types ("__apple_types"); 971 static ConstString g_sect_name_dwarf_apple_namespaces ("__apple_namespac"); 972 static ConstString g_sect_name_dwarf_apple_objc ("__apple_objc"); 973 static ConstString g_sect_name_eh_frame ("__eh_frame"); 974 static ConstString g_sect_name_DATA ("__DATA"); 975 static ConstString g_sect_name_TEXT ("__TEXT"); 976 977 SectionType sect_type = eSectionTypeOther; 978 979 if (section_name == g_sect_name_dwarf_debug_abbrev) 980 sect_type = eSectionTypeDWARFDebugAbbrev; 981 else if (section_name == g_sect_name_dwarf_debug_aranges) 982 sect_type = eSectionTypeDWARFDebugAranges; 983 else if (section_name == g_sect_name_dwarf_debug_frame) 984 sect_type = eSectionTypeDWARFDebugFrame; 985 else if (section_name == g_sect_name_dwarf_debug_info) 986 sect_type = eSectionTypeDWARFDebugInfo; 987 else if (section_name == g_sect_name_dwarf_debug_line) 988 sect_type = eSectionTypeDWARFDebugLine; 989 else if (section_name == g_sect_name_dwarf_debug_loc) 990 sect_type = eSectionTypeDWARFDebugLoc; 991 else if (section_name == g_sect_name_dwarf_debug_macinfo) 992 sect_type = eSectionTypeDWARFDebugMacInfo; 993 else if (section_name == g_sect_name_dwarf_debug_pubnames) 994 sect_type = eSectionTypeDWARFDebugPubNames; 995 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 996 sect_type = eSectionTypeDWARFDebugPubTypes; 997 else if (section_name == g_sect_name_dwarf_debug_ranges) 998 sect_type = eSectionTypeDWARFDebugRanges; 999 else if (section_name == g_sect_name_dwarf_debug_str) 1000 sect_type = eSectionTypeDWARFDebugStr; 1001 else if (section_name == g_sect_name_dwarf_apple_names) 1002 sect_type = eSectionTypeDWARFAppleNames; 1003 else if (section_name == g_sect_name_dwarf_apple_types) 1004 sect_type = eSectionTypeDWARFAppleTypes; 1005 else if (section_name == g_sect_name_dwarf_apple_namespaces) 1006 sect_type = eSectionTypeDWARFAppleNamespaces; 1007 else if (section_name == g_sect_name_dwarf_apple_objc) 1008 sect_type = eSectionTypeDWARFAppleObjC; 1009 else if (section_name == g_sect_name_objc_selrefs) 1010 sect_type = eSectionTypeDataCStringPointers; 1011 else if (section_name == g_sect_name_objc_msgrefs) 1012 sect_type = eSectionTypeDataObjCMessageRefs; 1013 else if (section_name == g_sect_name_eh_frame) 1014 sect_type = eSectionTypeEHFrame; 1015 else if (section_name == g_sect_name_cfstring) 1016 sect_type = eSectionTypeDataObjCCFStrings; 1017 else if (section_name == g_sect_name_objc_data || 1018 section_name == g_sect_name_objc_classrefs || 1019 section_name == g_sect_name_objc_superrefs || 1020 section_name == g_sect_name_objc_const || 1021 section_name == g_sect_name_objc_classlist) 1022 { 1023 sect_type = eSectionTypeDataPointers; 1024 } 1025 1026 if (sect_type == eSectionTypeOther) 1027 { 1028 switch (mach_sect_type) 1029 { 1030 // TODO: categorize sections by other flags for regular sections 1031 case SectionTypeRegular: 1032 if (segment_sp->GetName() == g_sect_name_TEXT) 1033 sect_type = eSectionTypeCode; 1034 else if (segment_sp->GetName() == g_sect_name_DATA) 1035 sect_type = eSectionTypeData; 1036 else 1037 sect_type = eSectionTypeOther; 1038 break; 1039 case SectionTypeZeroFill: sect_type = eSectionTypeZeroFill; break; 1040 case SectionTypeCStringLiterals: sect_type = eSectionTypeDataCString; break; // section with only literal C strings 1041 case SectionType4ByteLiterals: sect_type = eSectionTypeData4; break; // section with only 4 byte literals 1042 case SectionType8ByteLiterals: sect_type = eSectionTypeData8; break; // section with only 8 byte literals 1043 case SectionTypeLiteralPointers: sect_type = eSectionTypeDataPointers; break; // section with only pointers to literals 1044 case SectionTypeNonLazySymbolPointers: sect_type = eSectionTypeDataPointers; break; // section with only non-lazy symbol pointers 1045 case SectionTypeLazySymbolPointers: sect_type = eSectionTypeDataPointers; break; // section with only lazy symbol pointers 1046 case SectionTypeSymbolStubs: sect_type = eSectionTypeCode; break; // section with only symbol stubs, byte size of stub in the reserved2 field 1047 case SectionTypeModuleInitFunctionPointers: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for initialization 1048 case SectionTypeModuleTermFunctionPointers: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for termination 1049 case SectionTypeCoalesced: sect_type = eSectionTypeOther; break; 1050 case SectionTypeZeroFillLarge: sect_type = eSectionTypeZeroFill; break; 1051 case SectionTypeInterposing: sect_type = eSectionTypeCode; break; // section with only pairs of function pointers for interposing 1052 case SectionType16ByteLiterals: sect_type = eSectionTypeData16; break; // section with only 16 byte literals 1053 case SectionTypeDTraceObjectFormat: sect_type = eSectionTypeDebug; break; 1054 case SectionTypeLazyDylibSymbolPointers: sect_type = eSectionTypeDataPointers; break; 1055 default: break; 1056 } 1057 } 1058 1059 SectionSP section_sp(new Section (segment_sp, 1060 module_sp, 1061 ++sectID, 1062 section_name, 1063 sect_type, 1064 sect64.addr - segment_sp->GetFileAddress(), 1065 sect64.size, 1066 sect64.offset, 1067 sect64.offset == 0 ? 0 : sect64.size, 1068 sect64.flags)); 1069 // Set the section to be encrypted to match the segment 1070 1071 bool section_is_encrypted = false; 1072 if (!segment_is_encrypted && load_cmd.filesize != 0) 1073 section_is_encrypted = encrypted_file_ranges.FindEntryThatContains(sect64.offset) != NULL; 1074 1075 section_sp->SetIsEncrypted (segment_is_encrypted || section_is_encrypted); 1076 segment_sp->GetChildren().AddSection(section_sp); 1077 1078 if (segment_sp->IsFake()) 1079 { 1080 segment_sp.reset(); 1081 segment_name.Clear(); 1082 } 1083 } 1084 if (segment_sp && m_header.filetype == HeaderFileTypeDSYM) 1085 { 1086 if (first_segment_sectID <= sectID) 1087 { 1088 lldb::user_id_t sect_uid; 1089 for (sect_uid = first_segment_sectID; sect_uid <= sectID; ++sect_uid) 1090 { 1091 SectionSP curr_section_sp(segment_sp->GetChildren().FindSectionByID (sect_uid)); 1092 SectionSP next_section_sp; 1093 if (sect_uid + 1 <= sectID) 1094 next_section_sp = segment_sp->GetChildren().FindSectionByID (sect_uid+1); 1095 1096 if (curr_section_sp.get()) 1097 { 1098 if (curr_section_sp->GetByteSize() == 0) 1099 { 1100 if (next_section_sp.get() != NULL) 1101 curr_section_sp->SetByteSize ( next_section_sp->GetFileAddress() - curr_section_sp->GetFileAddress() ); 1102 else 1103 curr_section_sp->SetByteSize ( load_cmd.vmsize ); 1104 } 1105 } 1106 } 1107 } 1108 } 1109 } 1110 } 1111 } 1112 else if (load_cmd.cmd == LoadCommandDynamicSymtabInfo) 1113 { 1114 m_dysymtab.cmd = load_cmd.cmd; 1115 m_dysymtab.cmdsize = load_cmd.cmdsize; 1116 m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1117 } 1118 1119 offset = load_cmd_offset + load_cmd.cmdsize; 1120 } 1121// if (dump_sections) 1122// { 1123// StreamFile s(stdout); 1124// m_sections_ap->Dump(&s, true); 1125// } 1126 return sectID; // Return the number of sections we registered with the module 1127} 1128 1129class MachSymtabSectionInfo 1130{ 1131public: 1132 1133 MachSymtabSectionInfo (SectionList *section_list) : 1134 m_section_list (section_list), 1135 m_section_infos() 1136 { 1137 // Get the number of sections down to a depth of 1 to include 1138 // all segments and their sections, but no other sections that 1139 // may be added for debug map or 1140 m_section_infos.resize(section_list->GetNumSections(1)); 1141 } 1142 1143 1144 SectionSP 1145 GetSection (uint8_t n_sect, addr_t file_addr) 1146 { 1147 if (n_sect == 0) 1148 return SectionSP(); 1149 if (n_sect < m_section_infos.size()) 1150 { 1151 if (!m_section_infos[n_sect].section_sp) 1152 { 1153 SectionSP section_sp (m_section_list->FindSectionByID (n_sect)); 1154 m_section_infos[n_sect].section_sp = section_sp; 1155 if (section_sp) 1156 { 1157 m_section_infos[n_sect].vm_range.SetBaseAddress (section_sp->GetFileAddress()); 1158 m_section_infos[n_sect].vm_range.SetByteSize (section_sp->GetByteSize()); 1159 } 1160 else 1161 { 1162 Host::SystemLog (Host::eSystemLogError, "error: unable to find section for section %u\n", n_sect); 1163 } 1164 } 1165 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) 1166 { 1167 // Symbol is in section. 1168 return m_section_infos[n_sect].section_sp; 1169 } 1170 else if (m_section_infos[n_sect].vm_range.GetByteSize () == 0 && 1171 m_section_infos[n_sect].vm_range.GetBaseAddress() == file_addr) 1172 { 1173 // Symbol is in section with zero size, but has the same start 1174 // address as the section. This can happen with linker symbols 1175 // (symbols that start with the letter 'l' or 'L'. 1176 return m_section_infos[n_sect].section_sp; 1177 } 1178 } 1179 return m_section_list->FindSectionContainingFileAddress(file_addr); 1180 } 1181 1182protected: 1183 struct SectionInfo 1184 { 1185 SectionInfo () : 1186 vm_range(), 1187 section_sp () 1188 { 1189 } 1190 1191 VMRange vm_range; 1192 SectionSP section_sp; 1193 }; 1194 SectionList *m_section_list; 1195 std::vector<SectionInfo> m_section_infos; 1196}; 1197 1198size_t 1199ObjectFileMachO::ParseSymtab (bool minimize) 1200{ 1201 Timer scoped_timer(__PRETTY_FUNCTION__, 1202 "ObjectFileMachO::ParseSymtab () module = %s", 1203 m_file.GetFilename().AsCString("")); 1204 ModuleSP module_sp (GetModule()); 1205 if (!module_sp) 1206 return 0; 1207 1208 struct symtab_command symtab_load_command = { 0, 0, 0, 0, 0, 0 }; 1209 struct linkedit_data_command function_starts_load_command = { 0, 0, 0, 0 }; 1210 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 1211 FunctionStarts function_starts; 1212 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1213 uint32_t i; 1214 1215 LogSP log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_SYMBOLS)); 1216 1217 for (i=0; i<m_header.ncmds; ++i) 1218 { 1219 const lldb::offset_t cmd_offset = offset; 1220 // Read in the load command and load command size 1221 struct load_command lc; 1222 if (m_data.GetU32(&offset, &lc, 2) == NULL) 1223 break; 1224 // Watch for the symbol table load command 1225 switch (lc.cmd) 1226 { 1227 case LoadCommandSymtab: 1228 symtab_load_command.cmd = lc.cmd; 1229 symtab_load_command.cmdsize = lc.cmdsize; 1230 // Read in the rest of the symtab load command 1231 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 0) // fill in symoff, nsyms, stroff, strsize fields 1232 return 0; 1233 if (symtab_load_command.symoff == 0) 1234 { 1235 if (log) 1236 module_sp->LogMessage(log.get(), "LC_SYMTAB.symoff == 0"); 1237 return 0; 1238 } 1239 1240 if (symtab_load_command.stroff == 0) 1241 { 1242 if (log) 1243 module_sp->LogMessage(log.get(), "LC_SYMTAB.stroff == 0"); 1244 return 0; 1245 } 1246 1247 if (symtab_load_command.nsyms == 0) 1248 { 1249 if (log) 1250 module_sp->LogMessage(log.get(), "LC_SYMTAB.nsyms == 0"); 1251 return 0; 1252 } 1253 1254 if (symtab_load_command.strsize == 0) 1255 { 1256 if (log) 1257 module_sp->LogMessage(log.get(), "LC_SYMTAB.strsize == 0"); 1258 return 0; 1259 } 1260 break; 1261 1262 case LoadCommandFunctionStarts: 1263 function_starts_load_command.cmd = lc.cmd; 1264 function_starts_load_command.cmdsize = lc.cmdsize; 1265 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == NULL) // fill in symoff, nsyms, stroff, strsize fields 1266 bzero (&function_starts_load_command, sizeof(function_starts_load_command)); 1267 break; 1268 1269 default: 1270 break; 1271 } 1272 offset = cmd_offset + lc.cmdsize; 1273 } 1274 1275 if (symtab_load_command.cmd) 1276 { 1277 Symtab *symtab = m_symtab_ap.get(); 1278 SectionList *section_list = GetSectionList(); 1279 if (section_list == NULL) 1280 return 0; 1281 1282 ProcessSP process_sp (m_process_wp.lock()); 1283 Process *process = process_sp.get(); 1284 1285 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 1286 const ByteOrder byte_order = m_data.GetByteOrder(); 1287 bool bit_width_32 = addr_byte_size == 4; 1288 const size_t nlist_byte_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 1289 1290 DataExtractor nlist_data (NULL, 0, byte_order, addr_byte_size); 1291 DataExtractor strtab_data (NULL, 0, byte_order, addr_byte_size); 1292 DataExtractor function_starts_data (NULL, 0, byte_order, addr_byte_size); 1293 DataExtractor indirect_symbol_index_data (NULL, 0, byte_order, addr_byte_size); 1294 1295 const addr_t nlist_data_byte_size = symtab_load_command.nsyms * nlist_byte_size; 1296 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 1297 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 1298 if (process) 1299 { 1300 Target &target = process->GetTarget(); 1301 SectionSP linkedit_section_sp(section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 1302 // Reading mach file from memory in a process or core file... 1303 1304 if (linkedit_section_sp) 1305 { 1306 const addr_t linkedit_load_addr = linkedit_section_sp->GetLoadBaseAddress(&target); 1307 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 1308 const addr_t symoff_addr = linkedit_load_addr + symtab_load_command.symoff - linkedit_file_offset; 1309 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - linkedit_file_offset; 1310 1311 bool data_was_read = false; 1312 1313#if defined (__APPLE__) && defined (__arm__) 1314 if (m_header.flags & 0x80000000u) 1315 { 1316 // This mach-o memory file is in the dyld shared cache. If this 1317 // program is not remote and this is iOS, then this process will 1318 // share the same shared cache as the process we are debugging and 1319 // we can read the entire __LINKEDIT from the address space in this 1320 // process. This is a needed optimization that is used for local iOS 1321 // debugging only since all shared libraries in the shared cache do 1322 // not have corresponding files that exist in the file system of the 1323 // device. They have been combined into a single file. This means we 1324 // always have to load these files from memory. All of the symbol and 1325 // string tables from all of the __LINKEDIT sections from the shared 1326 // libraries in the shared cache have been merged into a single large 1327 // symbol and string table. Reading all of this symbol and string table 1328 // data across can slow down debug launch times, so we optimize this by 1329 // reading the memory for the __LINKEDIT section from this process. 1330 PlatformSP platform_sp (target.GetPlatform()); 1331 if (platform_sp && platform_sp->IsHost()) 1332 { 1333 data_was_read = true; 1334 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, eByteOrderLittle); 1335 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, eByteOrderLittle); 1336 if (function_starts_load_command.cmd) 1337 { 1338 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 1339 function_starts_data.SetData ((void *)func_start_addr, function_starts_load_command.datasize, eByteOrderLittle); 1340 } 1341 } 1342 } 1343#endif 1344 1345 if (!data_was_read) 1346 { 1347 DataBufferSP nlist_data_sp (ReadMemory (process_sp, symoff_addr, nlist_data_byte_size)); 1348 if (nlist_data_sp) 1349 nlist_data.SetData (nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 1350 //DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, strtab_data_byte_size)); 1351 //if (strtab_data_sp) 1352 // strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 1353 if (m_dysymtab.nindirectsyms != 0) 1354 { 1355 const addr_t indirect_syms_addr = linkedit_load_addr + m_dysymtab.indirectsymoff - linkedit_file_offset; 1356 DataBufferSP indirect_syms_data_sp (ReadMemory (process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 1357 if (indirect_syms_data_sp) 1358 indirect_symbol_index_data.SetData (indirect_syms_data_sp, 0, indirect_syms_data_sp->GetByteSize()); 1359 } 1360 if (function_starts_load_command.cmd) 1361 { 1362 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 1363 DataBufferSP func_start_data_sp (ReadMemory (process_sp, func_start_addr, function_starts_load_command.datasize)); 1364 if (func_start_data_sp) 1365 function_starts_data.SetData (func_start_data_sp, 0, func_start_data_sp->GetByteSize()); 1366 } 1367 } 1368 } 1369 } 1370 else 1371 { 1372 nlist_data.SetData (m_data, 1373 symtab_load_command.symoff, 1374 nlist_data_byte_size); 1375 strtab_data.SetData (m_data, 1376 symtab_load_command.stroff, 1377 strtab_data_byte_size); 1378 if (m_dysymtab.nindirectsyms != 0) 1379 { 1380 indirect_symbol_index_data.SetData (m_data, 1381 m_dysymtab.indirectsymoff, 1382 m_dysymtab.nindirectsyms * 4); 1383 } 1384 if (function_starts_load_command.cmd) 1385 { 1386 function_starts_data.SetData (m_data, 1387 function_starts_load_command.dataoff, 1388 function_starts_load_command.datasize); 1389 } 1390 } 1391 1392 if (nlist_data.GetByteSize() == 0) 1393 { 1394 if (log) 1395 module_sp->LogMessage(log.get(), "failed to read nlist data"); 1396 return 0; 1397 } 1398 1399 1400 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 1401 if (!have_strtab_data) 1402 { 1403 if (process) 1404 { 1405 if (strtab_addr == LLDB_INVALID_ADDRESS) 1406 { 1407 if (log) 1408 module_sp->LogMessage(log.get(), "failed to locate the strtab in memory"); 1409 return 0; 1410 } 1411 } 1412 else 1413 { 1414 if (log) 1415 module_sp->LogMessage(log.get(), "failed to read strtab data"); 1416 return 0; 1417 } 1418 } 1419 1420 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 1421 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 1422 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 1423 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 1424 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT)); 1425 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA)); 1426 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC)); 1427 SectionSP eh_frame_section_sp; 1428 if (text_section_sp.get()) 1429 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame); 1430 else 1431 eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame); 1432 1433 const bool is_arm = (m_header.cputype == llvm::MachO::CPUTypeARM); 1434 if (text_section_sp && function_starts_data.GetByteSize()) 1435 { 1436 FunctionStarts::Entry function_start_entry; 1437 function_start_entry.data = false; 1438 lldb::offset_t function_start_offset = 0; 1439 function_start_entry.addr = text_section_sp->GetFileAddress(); 1440 uint64_t delta; 1441 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 0) 1442 { 1443 // Now append the current entry 1444 function_start_entry.addr += delta; 1445 function_starts.Append(function_start_entry); 1446 } 1447 } 1448 1449 const size_t function_starts_count = function_starts.GetSize(); 1450 1451 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() : NListSectionNoSection; 1452 1453 lldb::offset_t nlist_data_offset = 0; 1454 1455 uint32_t N_SO_index = UINT32_MAX; 1456 1457 MachSymtabSectionInfo section_info (section_list); 1458 std::vector<uint32_t> N_FUN_indexes; 1459 std::vector<uint32_t> N_NSYM_indexes; 1460 std::vector<uint32_t> N_INCL_indexes; 1461 std::vector<uint32_t> N_BRAC_indexes; 1462 std::vector<uint32_t> N_COMM_indexes; 1463 typedef std::map <uint64_t, uint32_t> ValueToSymbolIndexMap; 1464 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap; 1465 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 1466 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 1467 // Any symbols that get merged into another will get an entry 1468 // in this map so we know 1469 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 1470 uint32_t nlist_idx = 0; 1471 Symbol *symbol_ptr = NULL; 1472 1473 uint32_t sym_idx = 0; 1474 Symbol *sym = NULL; 1475 size_t num_syms = 0; 1476 std::string memory_symbol_name; 1477 uint32_t unmapped_local_symbols_found = 0; 1478 1479#if defined (__APPLE__) && defined (__arm__) 1480 1481 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been optimized by moving LOCAL 1482 // symbols out of the memory mapped portion of the DSC. The symbol information has all been retained, 1483 // but it isn't available in the normal nlist data. However, there *are* duplicate entries of *some* 1484 // LOCAL symbols in the normal nlist data. To handle this situation correctly, we must first attempt 1485 // to parse any DSC unmapped symbol information. If we find any, we set a flag that tells the normal 1486 // nlist parser to ignore all LOCAL symbols. 1487 1488 if (m_header.flags & 0x80000000u) 1489 { 1490 // Before we can start mapping the DSC, we need to make certain the target process is actually 1491 // using the cache we can find. 1492 1493 /* 1494 * TODO (FIXME!) 1495 * 1496 * Consider the case of testing with a separate DSC file. 1497 * If we go through the normal code paths, we will give symbols for the wrong DSC, and 1498 * that is bad. We need to read the target process' all_image_infos struct, and look 1499 * at the values of the processDetachedFromSharedRegion field. If that is set, we should skip 1500 * this code section. 1501 */ 1502 1503 // Next we need to determine the correct path for the dyld shared cache. 1504 1505 ArchSpec header_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 1506 char dsc_path[PATH_MAX]; 1507 1508 snprintf(dsc_path, sizeof(dsc_path), "%s%s%s", 1509 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR */ 1510 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 1511 header_arch.GetArchitectureName()); 1512 1513 FileSpec dsc_filespec(dsc_path, false); 1514 1515 // We need definitions of two structures in the on-disk DSC, copy them here manually 1516 struct lldb_copy_dyld_cache_header 1517 { 1518 char magic[16]; 1519 uint32_t mappingOffset; 1520 uint32_t mappingCount; 1521 uint32_t imagesOffset; 1522 uint32_t imagesCount; 1523 uint64_t dyldBaseAddress; 1524 uint64_t codeSignatureOffset; 1525 uint64_t codeSignatureSize; 1526 uint64_t slideInfoOffset; 1527 uint64_t slideInfoSize; 1528 uint64_t localSymbolsOffset; 1529 uint64_t localSymbolsSize; 1530 }; 1531 struct lldb_copy_dyld_cache_local_symbols_info 1532 { 1533 uint32_t nlistOffset; 1534 uint32_t nlistCount; 1535 uint32_t stringsOffset; 1536 uint32_t stringsSize; 1537 uint32_t entriesOffset; 1538 uint32_t entriesCount; 1539 }; 1540 struct lldb_copy_dyld_cache_local_symbols_entry 1541 { 1542 uint32_t dylibOffset; 1543 uint32_t nlistStartIndex; 1544 uint32_t nlistCount; 1545 }; 1546 1547 /* The dyld_cache_header has a pointer to the dyld_cache_local_symbols_info structure (localSymbolsOffset). 1548 The dyld_cache_local_symbols_info structure gives us three things: 1549 1. The start and count of the nlist records in the dyld_shared_cache file 1550 2. The start and size of the strings for these nlist records 1551 3. The start and count of dyld_cache_local_symbols_entry entries 1552 1553 There is one dyld_cache_local_symbols_entry per dylib/framework in the dyld shared cache. 1554 The "dylibOffset" field is the Mach-O header of this dylib/framework in the dyld shared cache. 1555 The dyld_cache_local_symbols_entry also lists the start of this dylib/framework's nlist records 1556 and the count of how many nlist records there are for this dylib/framework. 1557 */ 1558 1559 // Process the dsc header to find the unmapped symbols 1560 // 1561 // Save some VM space, do not map the entire cache in one shot. 1562 1563 if (DataBufferSP dsc_data_sp = dsc_filespec.MemoryMapFileContents(0, sizeof(struct lldb_copy_dyld_cache_header))) 1564 { 1565 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 1566 1567 lldb::offset_t offset = offsetof (struct lldb_copy_dyld_cache_header, mappingOffset); 1568 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 1569 1570 // If the mappingOffset points to a location inside the header, we've 1571 // opened an old dyld shared cache, and should not proceed further. 1572 if (mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header)) 1573 { 1574 1575 offset = offsetof (struct lldb_copy_dyld_cache_header, localSymbolsOffset); 1576 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 1577 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 1578 1579 if (localSymbolsOffset && localSymbolsSize) 1580 { 1581 // Map the local symbols 1582 if (DataBufferSP dsc_local_symbols_data_sp = dsc_filespec.MemoryMapFileContents(localSymbolsOffset, localSymbolsSize)) 1583 { 1584 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, byte_order, addr_byte_size); 1585 1586 offset = 0; 1587 1588 // Read the local_symbols_infos struct in one shot 1589 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 1590 dsc_local_symbols_data.GetU32(&offset, &local_symbols_info.nlistOffset, 6); 1591 1592 // The local_symbols_infos offsets are offsets into local symbols memory, NOT file offsets! 1593 // We first need to identify the local "entry" that matches the current header. 1594 // The "entry" is stored as a file offset in the dyld_shared_cache, so we need to 1595 // adjust the raw m_header value by slide and 0x30000000. 1596 1597 SectionSP text_section_sp(section_list->FindSectionByName(GetSegmentNameTEXT())); 1598 1599 uint32_t header_file_offset = (text_section_sp->GetFileAddress() - 0x30000000); 1600 1601 offset = local_symbols_info.entriesOffset; 1602 for (uint32_t entry_index = 0; entry_index < local_symbols_info.entriesCount; entry_index++) 1603 { 1604 struct lldb_copy_dyld_cache_local_symbols_entry local_symbols_entry; 1605 local_symbols_entry.dylibOffset = dsc_local_symbols_data.GetU32(&offset); 1606 local_symbols_entry.nlistStartIndex = dsc_local_symbols_data.GetU32(&offset); 1607 local_symbols_entry.nlistCount = dsc_local_symbols_data.GetU32(&offset); 1608 1609 if (header_file_offset == local_symbols_entry.dylibOffset) 1610 { 1611 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 1612 1613 // The normal nlist code cannot correctly size the Symbols array, we need to allocate it here. 1614 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms + unmapped_local_symbols_found - m_dysymtab.nlocalsym); 1615 num_syms = symtab->GetNumSymbols(); 1616 1617 nlist_data_offset = local_symbols_info.nlistOffset + (nlist_byte_size * local_symbols_entry.nlistStartIndex); 1618 uint32_t string_table_offset = local_symbols_info.stringsOffset; 1619 1620 for (uint32_t nlist_index = 0; nlist_index < local_symbols_entry.nlistCount; nlist_index++) 1621 { 1622 ///////////////////////////// 1623 { 1624 struct nlist_64 nlist; 1625 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 1626 break; 1627 1628 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked(&nlist_data_offset); 1629 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 1630 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 1631 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked (&nlist_data_offset); 1632 nlist.n_value = dsc_local_symbols_data.GetAddress_unchecked (&nlist_data_offset); 1633 1634 SymbolType type = eSymbolTypeInvalid; 1635 const char *symbol_name = dsc_local_symbols_data.PeekCStr(string_table_offset + nlist.n_strx); 1636 1637 if (symbol_name == NULL) 1638 { 1639 // No symbol should be NULL, even the symbols with no 1640 // string values should have an offset zero which points 1641 // to an empty C-string 1642 Host::SystemLog (Host::eSystemLogError, 1643 "error: DSC unmapped local symbol[%u] has invalid string table offset 0x%x in %s/%s, ignoring symbol\n", 1644 entry_index, 1645 nlist.n_strx, 1646 module_sp->GetFileSpec().GetDirectory().GetCString(), 1647 module_sp->GetFileSpec().GetFilename().GetCString()); 1648 continue; 1649 } 1650 if (symbol_name[0] == '\0') 1651 symbol_name = NULL; 1652 1653 const char *symbol_name_non_abi_mangled = NULL; 1654 1655 SectionSP symbol_section; 1656 uint32_t symbol_byte_size = 0; 1657 bool add_nlist = true; 1658 bool is_debug = ((nlist.n_type & NlistMaskStab) != 0); 1659 bool demangled_is_synthesized = false; 1660 1661 assert (sym_idx < num_syms); 1662 1663 sym[sym_idx].SetDebug (is_debug); 1664 1665 if (is_debug) 1666 { 1667 switch (nlist.n_type) 1668 { 1669 case StabGlobalSymbol: 1670 // N_GSYM -- global symbol: name,,NO_SECT,type,0 1671 // Sometimes the N_GSYM value contains the address. 1672 1673 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 1674 // have the same address, but we want to ensure that we always find only the real symbol, 1675 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 1676 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 1677 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 1678 // same address. 1679 1680 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 1681 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 1682 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 1683 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 1684 add_nlist = false; 1685 else 1686 { 1687 sym[sym_idx].SetExternal(true); 1688 if (nlist.n_value != 0) 1689 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1690 type = eSymbolTypeData; 1691 } 1692 break; 1693 1694 case StabFunctionName: 1695 // N_FNAME -- procedure name (f77 kludge): name,,NO_SECT,0,0 1696 type = eSymbolTypeCompiler; 1697 break; 1698 1699 case StabFunction: 1700 // N_FUN -- procedure: name,,n_sect,linenumber,address 1701 if (symbol_name) 1702 { 1703 type = eSymbolTypeCode; 1704 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1705 1706 N_FUN_addr_to_sym_idx[nlist.n_value] = sym_idx; 1707 // We use the current number of symbols in the symbol table in lieu of 1708 // using nlist_idx in case we ever start trimming entries out 1709 N_FUN_indexes.push_back(sym_idx); 1710 } 1711 else 1712 { 1713 type = eSymbolTypeCompiler; 1714 1715 if ( !N_FUN_indexes.empty() ) 1716 { 1717 // Copy the size of the function into the original STAB entry so we don't have 1718 // to hunt for it later 1719 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 1720 N_FUN_indexes.pop_back(); 1721 // We don't really need the end function STAB as it contains the size which 1722 // we already placed with the original symbol, so don't add it if we want a 1723 // minimal symbol table 1724 if (minimize) 1725 add_nlist = false; 1726 } 1727 } 1728 break; 1729 1730 case StabStaticSymbol: 1731 // N_STSYM -- static symbol: name,,n_sect,type,address 1732 N_STSYM_addr_to_sym_idx[nlist.n_value] = sym_idx; 1733 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1734 type = eSymbolTypeData; 1735 break; 1736 1737 case StabLocalCommon: 1738 // N_LCSYM -- .lcomm symbol: name,,n_sect,type,address 1739 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1740 type = eSymbolTypeCommonBlock; 1741 break; 1742 1743 case StabBeginSymbol: 1744 // N_BNSYM 1745 // We use the current number of symbols in the symbol table in lieu of 1746 // using nlist_idx in case we ever start trimming entries out 1747 if (minimize) 1748 { 1749 // Skip these if we want minimal symbol tables 1750 add_nlist = false; 1751 } 1752 else 1753 { 1754 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1755 N_NSYM_indexes.push_back(sym_idx); 1756 type = eSymbolTypeScopeBegin; 1757 } 1758 break; 1759 1760 case StabEndSymbol: 1761 // N_ENSYM 1762 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 1763 // so that we can always skip the entire symbol if we need to navigate 1764 // more quickly at the source level when parsing STABS 1765 if (minimize) 1766 { 1767 // Skip these if we want minimal symbol tables 1768 add_nlist = false; 1769 } 1770 else 1771 { 1772 if ( !N_NSYM_indexes.empty() ) 1773 { 1774 symbol_ptr = symtab->SymbolAtIndex(N_NSYM_indexes.back()); 1775 symbol_ptr->SetByteSize(sym_idx + 1); 1776 symbol_ptr->SetSizeIsSibling(true); 1777 N_NSYM_indexes.pop_back(); 1778 } 1779 type = eSymbolTypeScopeEnd; 1780 } 1781 break; 1782 1783 1784 case StabSourceFileOptions: 1785 // N_OPT - emitted with gcc2_compiled and in gcc source 1786 type = eSymbolTypeCompiler; 1787 break; 1788 1789 case StabRegisterSymbol: 1790 // N_RSYM - register sym: name,,NO_SECT,type,register 1791 type = eSymbolTypeVariable; 1792 break; 1793 1794 case StabSourceLine: 1795 // N_SLINE - src line: 0,,n_sect,linenumber,address 1796 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1797 type = eSymbolTypeLineEntry; 1798 break; 1799 1800 case StabStructureType: 1801 // N_SSYM - structure elt: name,,NO_SECT,type,struct_offset 1802 type = eSymbolTypeVariableType; 1803 break; 1804 1805 case StabSourceFileName: 1806 // N_SO - source file name 1807 type = eSymbolTypeSourceFile; 1808 if (symbol_name == NULL) 1809 { 1810 if (minimize) 1811 add_nlist = false; 1812 if (N_SO_index != UINT32_MAX) 1813 { 1814 // Set the size of the N_SO to the terminating index of this N_SO 1815 // so that we can always skip the entire N_SO if we need to navigate 1816 // more quickly at the source level when parsing STABS 1817 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 1818 symbol_ptr->SetByteSize(sym_idx + (minimize ? 0 : 1)); 1819 symbol_ptr->SetSizeIsSibling(true); 1820 } 1821 N_NSYM_indexes.clear(); 1822 N_INCL_indexes.clear(); 1823 N_BRAC_indexes.clear(); 1824 N_COMM_indexes.clear(); 1825 N_FUN_indexes.clear(); 1826 N_SO_index = UINT32_MAX; 1827 } 1828 else 1829 { 1830 // We use the current number of symbols in the symbol table in lieu of 1831 // using nlist_idx in case we ever start trimming entries out 1832 const bool N_SO_has_full_path = symbol_name[0] == '/'; 1833 if (N_SO_has_full_path) 1834 { 1835 if (minimize && (N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 1836 { 1837 // We have two consecutive N_SO entries where the first contains a directory 1838 // and the second contains a full path. 1839 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 1840 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 1841 add_nlist = false; 1842 } 1843 else 1844 { 1845 // This is the first entry in a N_SO that contains a directory or 1846 // a full path to the source file 1847 N_SO_index = sym_idx; 1848 } 1849 } 1850 else if (minimize && (N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 1851 { 1852 // This is usually the second N_SO entry that contains just the filename, 1853 // so here we combine it with the first one if we are minimizing the symbol table 1854 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 1855 if (so_path && so_path[0]) 1856 { 1857 std::string full_so_path (so_path); 1858 const size_t double_slash_pos = full_so_path.find("//"); 1859 if (double_slash_pos != std::string::npos) 1860 { 1861 // The linker has been generating bad N_SO entries with doubled up paths 1862 // in the format "%s%s" where the first stirng in the DW_AT_comp_dir, 1863 // and the second is the directory for the source file so you end up with 1864 // a path that looks like "/tmp/src//tmp/src/" 1865 FileSpec so_dir(so_path, false); 1866 if (!so_dir.Exists()) 1867 { 1868 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 1869 if (so_dir.Exists()) 1870 { 1871 // Trim off the incorrect path 1872 full_so_path.erase(0, double_slash_pos + 1); 1873 } 1874 } 1875 } 1876 if (*full_so_path.rbegin() != '/') 1877 full_so_path += '/'; 1878 full_so_path += symbol_name; 1879 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 1880 add_nlist = false; 1881 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 1882 } 1883 } 1884 else 1885 { 1886 // This could be a relative path to a N_SO 1887 N_SO_index = sym_idx; 1888 } 1889 } 1890 break; 1891 1892 case StabObjectFileName: 1893 // N_OSO - object file name: name,,0,0,st_mtime 1894 type = eSymbolTypeObjectFile; 1895 break; 1896 1897 case StabLocalSymbol: 1898 // N_LSYM - local sym: name,,NO_SECT,type,offset 1899 type = eSymbolTypeLocal; 1900 break; 1901 1902 //---------------------------------------------------------------------- 1903 // INCL scopes 1904 //---------------------------------------------------------------------- 1905 case StabBeginIncludeFileName: 1906 // N_BINCL - include file beginning: name,,NO_SECT,0,sum 1907 // We use the current number of symbols in the symbol table in lieu of 1908 // using nlist_idx in case we ever start trimming entries out 1909 N_INCL_indexes.push_back(sym_idx); 1910 type = eSymbolTypeScopeBegin; 1911 break; 1912 1913 case StabEndIncludeFile: 1914 // N_EINCL - include file end: name,,NO_SECT,0,0 1915 // Set the size of the N_BINCL to the terminating index of this N_EINCL 1916 // so that we can always skip the entire symbol if we need to navigate 1917 // more quickly at the source level when parsing STABS 1918 if ( !N_INCL_indexes.empty() ) 1919 { 1920 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 1921 symbol_ptr->SetByteSize(sym_idx + 1); 1922 symbol_ptr->SetSizeIsSibling(true); 1923 N_INCL_indexes.pop_back(); 1924 } 1925 type = eSymbolTypeScopeEnd; 1926 break; 1927 1928 case StabIncludeFileName: 1929 // N_SOL - #included file name: name,,n_sect,0,address 1930 type = eSymbolTypeHeaderFile; 1931 1932 // We currently don't use the header files on darwin 1933 if (minimize) 1934 add_nlist = false; 1935 break; 1936 1937 case StabCompilerParameters: 1938 // N_PARAMS - compiler parameters: name,,NO_SECT,0,0 1939 type = eSymbolTypeCompiler; 1940 break; 1941 1942 case StabCompilerVersion: 1943 // N_VERSION - compiler version: name,,NO_SECT,0,0 1944 type = eSymbolTypeCompiler; 1945 break; 1946 1947 case StabCompilerOptLevel: 1948 // N_OLEVEL - compiler -O level: name,,NO_SECT,0,0 1949 type = eSymbolTypeCompiler; 1950 break; 1951 1952 case StabParameter: 1953 // N_PSYM - parameter: name,,NO_SECT,type,offset 1954 type = eSymbolTypeVariable; 1955 break; 1956 1957 case StabAlternateEntry: 1958 // N_ENTRY - alternate entry: name,,n_sect,linenumber,address 1959 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1960 type = eSymbolTypeLineEntry; 1961 break; 1962 1963 //---------------------------------------------------------------------- 1964 // Left and Right Braces 1965 //---------------------------------------------------------------------- 1966 case StabLeftBracket: 1967 // N_LBRAC - left bracket: 0,,NO_SECT,nesting level,address 1968 // We use the current number of symbols in the symbol table in lieu of 1969 // using nlist_idx in case we ever start trimming entries out 1970 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1971 N_BRAC_indexes.push_back(sym_idx); 1972 type = eSymbolTypeScopeBegin; 1973 break; 1974 1975 case StabRightBracket: 1976 // N_RBRAC - right bracket: 0,,NO_SECT,nesting level,address 1977 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 1978 // so that we can always skip the entire symbol if we need to navigate 1979 // more quickly at the source level when parsing STABS 1980 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1981 if ( !N_BRAC_indexes.empty() ) 1982 { 1983 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 1984 symbol_ptr->SetByteSize(sym_idx + 1); 1985 symbol_ptr->SetSizeIsSibling(true); 1986 N_BRAC_indexes.pop_back(); 1987 } 1988 type = eSymbolTypeScopeEnd; 1989 break; 1990 1991 case StabDeletedIncludeFile: 1992 // N_EXCL - deleted include file: name,,NO_SECT,0,sum 1993 type = eSymbolTypeHeaderFile; 1994 break; 1995 1996 //---------------------------------------------------------------------- 1997 // COMM scopes 1998 //---------------------------------------------------------------------- 1999 case StabBeginCommon: 2000 // N_BCOMM - begin common: name,,NO_SECT,0,0 2001 // We use the current number of symbols in the symbol table in lieu of 2002 // using nlist_idx in case we ever start trimming entries out 2003 type = eSymbolTypeScopeBegin; 2004 N_COMM_indexes.push_back(sym_idx); 2005 break; 2006 2007 case StabEndCommonLocal: 2008 // N_ECOML - end common (local name): 0,,n_sect,0,address 2009 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2010 // Fall through 2011 2012 case StabEndCommon: 2013 // N_ECOMM - end common: name,,n_sect,0,0 2014 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 2015 // so that we can always skip the entire symbol if we need to navigate 2016 // more quickly at the source level when parsing STABS 2017 if ( !N_COMM_indexes.empty() ) 2018 { 2019 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 2020 symbol_ptr->SetByteSize(sym_idx + 1); 2021 symbol_ptr->SetSizeIsSibling(true); 2022 N_COMM_indexes.pop_back(); 2023 } 2024 type = eSymbolTypeScopeEnd; 2025 break; 2026 2027 case StabLength: 2028 // N_LENG - second stab entry with length information 2029 type = eSymbolTypeAdditional; 2030 break; 2031 2032 default: break; 2033 } 2034 } 2035 else 2036 { 2037 //uint8_t n_pext = NlistMaskPrivateExternal & nlist.n_type; 2038 uint8_t n_type = NlistMaskType & nlist.n_type; 2039 sym[sym_idx].SetExternal((NlistMaskExternal & nlist.n_type) != 0); 2040 2041 switch (n_type) 2042 { 2043 case NListTypeIndirect: // N_INDR - Fall through 2044 case NListTypePreboundUndefined:// N_PBUD - Fall through 2045 case NListTypeUndefined: // N_UNDF 2046 type = eSymbolTypeUndefined; 2047 break; 2048 2049 case NListTypeAbsolute: // N_ABS 2050 type = eSymbolTypeAbsolute; 2051 break; 2052 2053 case NListTypeSection: // N_SECT 2054 { 2055 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2056 2057 if (symbol_section == NULL) 2058 { 2059 // TODO: warn about this? 2060 add_nlist = false; 2061 break; 2062 } 2063 2064 if (TEXT_eh_frame_sectID == nlist.n_sect) 2065 { 2066 type = eSymbolTypeException; 2067 } 2068 else 2069 { 2070 uint32_t section_type = symbol_section->Get() & SectionFlagMaskSectionType; 2071 2072 switch (section_type) 2073 { 2074 case SectionTypeRegular: break; // regular section 2075 //case SectionTypeZeroFill: type = eSymbolTypeData; break; // zero fill on demand section 2076 case SectionTypeCStringLiterals: type = eSymbolTypeData; break; // section with only literal C strings 2077 case SectionType4ByteLiterals: type = eSymbolTypeData; break; // section with only 4 byte literals 2078 case SectionType8ByteLiterals: type = eSymbolTypeData; break; // section with only 8 byte literals 2079 case SectionTypeLiteralPointers: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 2080 case SectionTypeNonLazySymbolPointers: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 2081 case SectionTypeLazySymbolPointers: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 2082 case SectionTypeSymbolStubs: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 2083 case SectionTypeModuleInitFunctionPointers: type = eSymbolTypeCode; break; // section with only function pointers for initialization 2084 case SectionTypeModuleTermFunctionPointers: type = eSymbolTypeCode; break; // section with only function pointers for termination 2085 //case SectionTypeCoalesced: type = eSymbolType; break; // section contains symbols that are to be coalesced 2086 //case SectionTypeZeroFillLarge: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 2087 case SectionTypeInterposing: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 2088 case SectionType16ByteLiterals: type = eSymbolTypeData; break; // section with only 16 byte literals 2089 case SectionTypeDTraceObjectFormat: type = eSymbolTypeInstrumentation; break; 2090 case SectionTypeLazyDylibSymbolPointers: type = eSymbolTypeTrampoline; break; 2091 default: break; 2092 } 2093 2094 if (type == eSymbolTypeInvalid) 2095 { 2096 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 2097 if (symbol_section->IsDescendant (text_section_sp.get())) 2098 { 2099 if (symbol_section->IsClear(SectionAttrUserPureInstructions | 2100 SectionAttrUserSelfModifyingCode | 2101 SectionAttrSytemSomeInstructions)) 2102 type = eSymbolTypeData; 2103 else 2104 type = eSymbolTypeCode; 2105 } 2106 else if (symbol_section->IsDescendant(data_section_sp.get())) 2107 { 2108 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 2109 { 2110 type = eSymbolTypeRuntime; 2111 2112 if (symbol_name && 2113 symbol_name[0] == '_' && 2114 symbol_name[1] == 'O' && 2115 symbol_name[2] == 'B') 2116 { 2117 llvm::StringRef symbol_name_ref(symbol_name); 2118 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 2119 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 2120 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 2121 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 2122 { 2123 symbol_name_non_abi_mangled = symbol_name + 1; 2124 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2125 type = eSymbolTypeObjCClass; 2126 demangled_is_synthesized = true; 2127 } 2128 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 2129 { 2130 symbol_name_non_abi_mangled = symbol_name + 1; 2131 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2132 type = eSymbolTypeObjCMetaClass; 2133 demangled_is_synthesized = true; 2134 } 2135 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 2136 { 2137 symbol_name_non_abi_mangled = symbol_name + 1; 2138 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2139 type = eSymbolTypeObjCIVar; 2140 demangled_is_synthesized = true; 2141 } 2142 } 2143 } 2144 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 2145 { 2146 type = eSymbolTypeException; 2147 } 2148 else 2149 { 2150 type = eSymbolTypeData; 2151 } 2152 } 2153 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 2154 { 2155 type = eSymbolTypeTrampoline; 2156 } 2157 else if (symbol_section->IsDescendant(objc_section_sp.get())) 2158 { 2159 type = eSymbolTypeRuntime; 2160 if (symbol_name && symbol_name[0] == '.') 2161 { 2162 llvm::StringRef symbol_name_ref(symbol_name); 2163 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 2164 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 2165 { 2166 symbol_name_non_abi_mangled = symbol_name; 2167 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 2168 type = eSymbolTypeObjCClass; 2169 demangled_is_synthesized = true; 2170 } 2171 } 2172 } 2173 } 2174 } 2175 } 2176 break; 2177 } 2178 } 2179 2180 if (add_nlist) 2181 { 2182 uint64_t symbol_value = nlist.n_value; 2183 bool symbol_name_is_mangled = false; 2184 2185 if (symbol_name_non_abi_mangled) 2186 { 2187 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 2188 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 2189 } 2190 else 2191 { 2192 if (symbol_name && symbol_name[0] == '_') 2193 { 2194 symbol_name_is_mangled = symbol_name[1] == '_'; 2195 symbol_name++; // Skip the leading underscore 2196 } 2197 2198 if (symbol_name) 2199 { 2200 sym[sym_idx].GetMangled().SetValue(ConstString(symbol_name), symbol_name_is_mangled); 2201 } 2202 } 2203 2204 if (is_debug == false) 2205 { 2206 if (type == eSymbolTypeCode) 2207 { 2208 // See if we can find a N_FUN entry for any code symbols. 2209 // If we do find a match, and the name matches, then we 2210 // can merge the two into just the function symbol to avoid 2211 // duplicate entries in the symbol table 2212 ValueToSymbolIndexMap::const_iterator pos = N_FUN_addr_to_sym_idx.find (nlist.n_value); 2213 if (pos != N_FUN_addr_to_sym_idx.end()) 2214 { 2215 if ((symbol_name_is_mangled == true && sym[sym_idx].GetMangled().GetMangledName() == sym[pos->second].GetMangled().GetMangledName()) || 2216 (symbol_name_is_mangled == false && sym[sym_idx].GetMangled().GetDemangledName() == sym[pos->second].GetMangled().GetDemangledName())) 2217 { 2218 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 2219 // We just need the flags from the linker symbol, so put these flags 2220 // into the N_FUN flags to avoid duplicate symbols in the symbol table 2221 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2222 sym[sym_idx].Clear(); 2223 continue; 2224 } 2225 } 2226 } 2227 else if (type == eSymbolTypeData) 2228 { 2229 // See if we can find a N_STSYM entry for any data symbols. 2230 // If we do find a match, and the name matches, then we 2231 // can merge the two into just the Static symbol to avoid 2232 // duplicate entries in the symbol table 2233 ValueToSymbolIndexMap::const_iterator pos = N_STSYM_addr_to_sym_idx.find (nlist.n_value); 2234 if (pos != N_STSYM_addr_to_sym_idx.end()) 2235 { 2236 if ((symbol_name_is_mangled == true && sym[sym_idx].GetMangled().GetMangledName() == sym[pos->second].GetMangled().GetMangledName()) || 2237 (symbol_name_is_mangled == false && sym[sym_idx].GetMangled().GetDemangledName() == sym[pos->second].GetMangled().GetDemangledName())) 2238 { 2239 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 2240 // We just need the flags from the linker symbol, so put these flags 2241 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 2242 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2243 sym[sym_idx].Clear(); 2244 continue; 2245 } 2246 } 2247 } 2248 } 2249 if (symbol_section) 2250 { 2251 const addr_t section_file_addr = symbol_section->GetFileAddress(); 2252 if (symbol_byte_size == 0 && function_starts_count > 0) 2253 { 2254 addr_t symbol_lookup_file_addr = nlist.n_value; 2255 // Do an exact address match for non-ARM addresses, else get the closest since 2256 // the symbol might be a thumb symbol which has an address with bit zero set 2257 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 2258 if (is_arm && func_start_entry) 2259 { 2260 // Verify that the function start address is the symbol address (ARM) 2261 // or the symbol address + 1 (thumb) 2262 if (func_start_entry->addr != symbol_lookup_file_addr && 2263 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 2264 { 2265 // Not the right entry, NULL it out... 2266 func_start_entry = NULL; 2267 } 2268 } 2269 if (func_start_entry) 2270 { 2271 func_start_entry->data = true; 2272 2273 addr_t symbol_file_addr = func_start_entry->addr; 2274 uint32_t symbol_flags = 0; 2275 if (is_arm) 2276 { 2277 if (symbol_file_addr & 1) 2278 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 2279 symbol_file_addr &= 0xfffffffffffffffeull; 2280 } 2281 2282 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 2283 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 2284 if (next_func_start_entry) 2285 { 2286 addr_t next_symbol_file_addr = next_func_start_entry->addr; 2287 // Be sure the clear the Thumb address bit when we calculate the size 2288 // from the current and next address 2289 if (is_arm) 2290 next_symbol_file_addr &= 0xfffffffffffffffeull; 2291 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 2292 } 2293 else 2294 { 2295 symbol_byte_size = section_end_file_addr - symbol_file_addr; 2296 } 2297 } 2298 } 2299 symbol_value -= section_file_addr; 2300 } 2301 2302 sym[sym_idx].SetID (nlist_idx); 2303 sym[sym_idx].SetType (type); 2304 sym[sym_idx].GetAddress().SetSection (symbol_section); 2305 sym[sym_idx].GetAddress().SetOffset (symbol_value); 2306 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2307 2308 if (symbol_byte_size > 0) 2309 sym[sym_idx].SetByteSize(symbol_byte_size); 2310 2311 if (demangled_is_synthesized) 2312 sym[sym_idx].SetDemangledNameIsSynthesized(true); 2313 ++sym_idx; 2314 } 2315 else 2316 { 2317 sym[sym_idx].Clear(); 2318 } 2319 2320 } 2321 ///////////////////////////// 2322 } 2323 break; // No more entries to consider 2324 } 2325 } 2326 } 2327 } 2328 } 2329 } 2330 } 2331 2332 // Must reset this in case it was mutated above! 2333 nlist_data_offset = 0; 2334#endif 2335 2336 // If the sym array was not created while parsing the DSC unmapped 2337 // symbols, create it now. 2338 if (sym == NULL) 2339 { 2340 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 2341 num_syms = symtab->GetNumSymbols(); 2342 } 2343 2344 if (unmapped_local_symbols_found) 2345 { 2346 assert(m_dysymtab.ilocalsym == 0); 2347 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 2348 nlist_idx = m_dysymtab.nlocalsym; 2349 } 2350 else 2351 { 2352 nlist_idx = 0; 2353 } 2354 2355 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) 2356 { 2357 struct nlist_64 nlist; 2358 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2359 break; 2360 2361 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 2362 nlist.n_type = nlist_data.GetU8_unchecked (&nlist_data_offset); 2363 nlist.n_sect = nlist_data.GetU8_unchecked (&nlist_data_offset); 2364 nlist.n_desc = nlist_data.GetU16_unchecked (&nlist_data_offset); 2365 nlist.n_value = nlist_data.GetAddress_unchecked (&nlist_data_offset); 2366 2367 SymbolType type = eSymbolTypeInvalid; 2368 const char *symbol_name = NULL; 2369 2370 if (have_strtab_data) 2371 { 2372 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 2373 2374 if (symbol_name == NULL) 2375 { 2376 // No symbol should be NULL, even the symbols with no 2377 // string values should have an offset zero which points 2378 // to an empty C-string 2379 Host::SystemLog (Host::eSystemLogError, 2380 "error: symbol[%u] has invalid string table offset 0x%x in %s/%s, ignoring symbol\n", 2381 nlist_idx, 2382 nlist.n_strx, 2383 module_sp->GetFileSpec().GetDirectory().GetCString(), 2384 module_sp->GetFileSpec().GetFilename().GetCString()); 2385 continue; 2386 } 2387 if (symbol_name[0] == '\0') 2388 symbol_name = NULL; 2389 } 2390 else 2391 { 2392 const addr_t str_addr = strtab_addr + nlist.n_strx; 2393 Error str_error; 2394 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, str_error)) 2395 symbol_name = memory_symbol_name.c_str(); 2396 } 2397 const char *symbol_name_non_abi_mangled = NULL; 2398 2399 SectionSP symbol_section; 2400 lldb::addr_t symbol_byte_size = 0; 2401 bool add_nlist = true; 2402 bool is_debug = ((nlist.n_type & NlistMaskStab) != 0); 2403 bool demangled_is_synthesized = false; 2404 2405 assert (sym_idx < num_syms); 2406 2407 sym[sym_idx].SetDebug (is_debug); 2408 2409 if (is_debug) 2410 { 2411 switch (nlist.n_type) 2412 { 2413 case StabGlobalSymbol: 2414 // N_GSYM -- global symbol: name,,NO_SECT,type,0 2415 // Sometimes the N_GSYM value contains the address. 2416 2417 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 2418 // have the same address, but we want to ensure that we always find only the real symbol, 2419 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 2420 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 2421 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 2422 // same address. 2423 2424 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 2425 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 2426 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 2427 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 2428 add_nlist = false; 2429 else 2430 { 2431 sym[sym_idx].SetExternal(true); 2432 if (nlist.n_value != 0) 2433 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2434 type = eSymbolTypeData; 2435 } 2436 break; 2437 2438 case StabFunctionName: 2439 // N_FNAME -- procedure name (f77 kludge): name,,NO_SECT,0,0 2440 type = eSymbolTypeCompiler; 2441 break; 2442 2443 case StabFunction: 2444 // N_FUN -- procedure: name,,n_sect,linenumber,address 2445 if (symbol_name) 2446 { 2447 type = eSymbolTypeCode; 2448 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2449 2450 N_FUN_addr_to_sym_idx[nlist.n_value] = sym_idx; 2451 // We use the current number of symbols in the symbol table in lieu of 2452 // using nlist_idx in case we ever start trimming entries out 2453 N_FUN_indexes.push_back(sym_idx); 2454 } 2455 else 2456 { 2457 type = eSymbolTypeCompiler; 2458 2459 if ( !N_FUN_indexes.empty() ) 2460 { 2461 // Copy the size of the function into the original STAB entry so we don't have 2462 // to hunt for it later 2463 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 2464 N_FUN_indexes.pop_back(); 2465 // We don't really need the end function STAB as it contains the size which 2466 // we already placed with the original symbol, so don't add it if we want a 2467 // minimal symbol table 2468 if (minimize) 2469 add_nlist = false; 2470 } 2471 } 2472 break; 2473 2474 case StabStaticSymbol: 2475 // N_STSYM -- static symbol: name,,n_sect,type,address 2476 N_STSYM_addr_to_sym_idx[nlist.n_value] = sym_idx; 2477 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2478 type = eSymbolTypeData; 2479 break; 2480 2481 case StabLocalCommon: 2482 // N_LCSYM -- .lcomm symbol: name,,n_sect,type,address 2483 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2484 type = eSymbolTypeCommonBlock; 2485 break; 2486 2487 case StabBeginSymbol: 2488 // N_BNSYM 2489 // We use the current number of symbols in the symbol table in lieu of 2490 // using nlist_idx in case we ever start trimming entries out 2491 if (minimize) 2492 { 2493 // Skip these if we want minimal symbol tables 2494 add_nlist = false; 2495 } 2496 else 2497 { 2498 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2499 N_NSYM_indexes.push_back(sym_idx); 2500 type = eSymbolTypeScopeBegin; 2501 } 2502 break; 2503 2504 case StabEndSymbol: 2505 // N_ENSYM 2506 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 2507 // so that we can always skip the entire symbol if we need to navigate 2508 // more quickly at the source level when parsing STABS 2509 if (minimize) 2510 { 2511 // Skip these if we want minimal symbol tables 2512 add_nlist = false; 2513 } 2514 else 2515 { 2516 if ( !N_NSYM_indexes.empty() ) 2517 { 2518 symbol_ptr = symtab->SymbolAtIndex(N_NSYM_indexes.back()); 2519 symbol_ptr->SetByteSize(sym_idx + 1); 2520 symbol_ptr->SetSizeIsSibling(true); 2521 N_NSYM_indexes.pop_back(); 2522 } 2523 type = eSymbolTypeScopeEnd; 2524 } 2525 break; 2526 2527 2528 case StabSourceFileOptions: 2529 // N_OPT - emitted with gcc2_compiled and in gcc source 2530 type = eSymbolTypeCompiler; 2531 break; 2532 2533 case StabRegisterSymbol: 2534 // N_RSYM - register sym: name,,NO_SECT,type,register 2535 type = eSymbolTypeVariable; 2536 break; 2537 2538 case StabSourceLine: 2539 // N_SLINE - src line: 0,,n_sect,linenumber,address 2540 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2541 type = eSymbolTypeLineEntry; 2542 break; 2543 2544 case StabStructureType: 2545 // N_SSYM - structure elt: name,,NO_SECT,type,struct_offset 2546 type = eSymbolTypeVariableType; 2547 break; 2548 2549 case StabSourceFileName: 2550 // N_SO - source file name 2551 type = eSymbolTypeSourceFile; 2552 if (symbol_name == NULL) 2553 { 2554 if (minimize) 2555 add_nlist = false; 2556 if (N_SO_index != UINT32_MAX) 2557 { 2558 // Set the size of the N_SO to the terminating index of this N_SO 2559 // so that we can always skip the entire N_SO if we need to navigate 2560 // more quickly at the source level when parsing STABS 2561 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2562 symbol_ptr->SetByteSize(sym_idx + (minimize ? 0 : 1)); 2563 symbol_ptr->SetSizeIsSibling(true); 2564 } 2565 N_NSYM_indexes.clear(); 2566 N_INCL_indexes.clear(); 2567 N_BRAC_indexes.clear(); 2568 N_COMM_indexes.clear(); 2569 N_FUN_indexes.clear(); 2570 N_SO_index = UINT32_MAX; 2571 } 2572 else 2573 { 2574 // We use the current number of symbols in the symbol table in lieu of 2575 // using nlist_idx in case we ever start trimming entries out 2576 const bool N_SO_has_full_path = symbol_name[0] == '/'; 2577 if (N_SO_has_full_path) 2578 { 2579 if (minimize && (N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2580 { 2581 // We have two consecutive N_SO entries where the first contains a directory 2582 // and the second contains a full path. 2583 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 2584 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2585 add_nlist = false; 2586 } 2587 else 2588 { 2589 // This is the first entry in a N_SO that contains a directory or 2590 // a full path to the source file 2591 N_SO_index = sym_idx; 2592 } 2593 } 2594 else if (minimize && (N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2595 { 2596 // This is usually the second N_SO entry that contains just the filename, 2597 // so here we combine it with the first one if we are minimizing the symbol table 2598 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 2599 if (so_path && so_path[0]) 2600 { 2601 std::string full_so_path (so_path); 2602 const size_t double_slash_pos = full_so_path.find("//"); 2603 if (double_slash_pos != std::string::npos) 2604 { 2605 // The linker has been generating bad N_SO entries with doubled up paths 2606 // in the format "%s%s" where the first stirng in the DW_AT_comp_dir, 2607 // and the second is the directory for the source file so you end up with 2608 // a path that looks like "/tmp/src//tmp/src/" 2609 FileSpec so_dir(so_path, false); 2610 if (!so_dir.Exists()) 2611 { 2612 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 2613 if (so_dir.Exists()) 2614 { 2615 // Trim off the incorrect path 2616 full_so_path.erase(0, double_slash_pos + 1); 2617 } 2618 } 2619 } 2620 if (*full_so_path.rbegin() != '/') 2621 full_so_path += '/'; 2622 full_so_path += symbol_name; 2623 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 2624 add_nlist = false; 2625 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2626 } 2627 } 2628 else 2629 { 2630 // This could be a relative path to a N_SO 2631 N_SO_index = sym_idx; 2632 } 2633 } 2634 2635 break; 2636 2637 case StabObjectFileName: 2638 // N_OSO - object file name: name,,0,0,st_mtime 2639 type = eSymbolTypeObjectFile; 2640 break; 2641 2642 case StabLocalSymbol: 2643 // N_LSYM - local sym: name,,NO_SECT,type,offset 2644 type = eSymbolTypeLocal; 2645 break; 2646 2647 //---------------------------------------------------------------------- 2648 // INCL scopes 2649 //---------------------------------------------------------------------- 2650 case StabBeginIncludeFileName: 2651 // N_BINCL - include file beginning: name,,NO_SECT,0,sum 2652 // We use the current number of symbols in the symbol table in lieu of 2653 // using nlist_idx in case we ever start trimming entries out 2654 N_INCL_indexes.push_back(sym_idx); 2655 type = eSymbolTypeScopeBegin; 2656 break; 2657 2658 case StabEndIncludeFile: 2659 // N_EINCL - include file end: name,,NO_SECT,0,0 2660 // Set the size of the N_BINCL to the terminating index of this N_EINCL 2661 // so that we can always skip the entire symbol if we need to navigate 2662 // more quickly at the source level when parsing STABS 2663 if ( !N_INCL_indexes.empty() ) 2664 { 2665 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 2666 symbol_ptr->SetByteSize(sym_idx + 1); 2667 symbol_ptr->SetSizeIsSibling(true); 2668 N_INCL_indexes.pop_back(); 2669 } 2670 type = eSymbolTypeScopeEnd; 2671 break; 2672 2673 case StabIncludeFileName: 2674 // N_SOL - #included file name: name,,n_sect,0,address 2675 type = eSymbolTypeHeaderFile; 2676 2677 // We currently don't use the header files on darwin 2678 if (minimize) 2679 add_nlist = false; 2680 break; 2681 2682 case StabCompilerParameters: 2683 // N_PARAMS - compiler parameters: name,,NO_SECT,0,0 2684 type = eSymbolTypeCompiler; 2685 break; 2686 2687 case StabCompilerVersion: 2688 // N_VERSION - compiler version: name,,NO_SECT,0,0 2689 type = eSymbolTypeCompiler; 2690 break; 2691 2692 case StabCompilerOptLevel: 2693 // N_OLEVEL - compiler -O level: name,,NO_SECT,0,0 2694 type = eSymbolTypeCompiler; 2695 break; 2696 2697 case StabParameter: 2698 // N_PSYM - parameter: name,,NO_SECT,type,offset 2699 type = eSymbolTypeVariable; 2700 break; 2701 2702 case StabAlternateEntry: 2703 // N_ENTRY - alternate entry: name,,n_sect,linenumber,address 2704 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2705 type = eSymbolTypeLineEntry; 2706 break; 2707 2708 //---------------------------------------------------------------------- 2709 // Left and Right Braces 2710 //---------------------------------------------------------------------- 2711 case StabLeftBracket: 2712 // N_LBRAC - left bracket: 0,,NO_SECT,nesting level,address 2713 // We use the current number of symbols in the symbol table in lieu of 2714 // using nlist_idx in case we ever start trimming entries out 2715 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2716 N_BRAC_indexes.push_back(sym_idx); 2717 type = eSymbolTypeScopeBegin; 2718 break; 2719 2720 case StabRightBracket: 2721 // N_RBRAC - right bracket: 0,,NO_SECT,nesting level,address 2722 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 2723 // so that we can always skip the entire symbol if we need to navigate 2724 // more quickly at the source level when parsing STABS 2725 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2726 if ( !N_BRAC_indexes.empty() ) 2727 { 2728 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 2729 symbol_ptr->SetByteSize(sym_idx + 1); 2730 symbol_ptr->SetSizeIsSibling(true); 2731 N_BRAC_indexes.pop_back(); 2732 } 2733 type = eSymbolTypeScopeEnd; 2734 break; 2735 2736 case StabDeletedIncludeFile: 2737 // N_EXCL - deleted include file: name,,NO_SECT,0,sum 2738 type = eSymbolTypeHeaderFile; 2739 break; 2740 2741 //---------------------------------------------------------------------- 2742 // COMM scopes 2743 //---------------------------------------------------------------------- 2744 case StabBeginCommon: 2745 // N_BCOMM - begin common: name,,NO_SECT,0,0 2746 // We use the current number of symbols in the symbol table in lieu of 2747 // using nlist_idx in case we ever start trimming entries out 2748 type = eSymbolTypeScopeBegin; 2749 N_COMM_indexes.push_back(sym_idx); 2750 break; 2751 2752 case StabEndCommonLocal: 2753 // N_ECOML - end common (local name): 0,,n_sect,0,address 2754 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2755 // Fall through 2756 2757 case StabEndCommon: 2758 // N_ECOMM - end common: name,,n_sect,0,0 2759 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 2760 // so that we can always skip the entire symbol if we need to navigate 2761 // more quickly at the source level when parsing STABS 2762 if ( !N_COMM_indexes.empty() ) 2763 { 2764 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 2765 symbol_ptr->SetByteSize(sym_idx + 1); 2766 symbol_ptr->SetSizeIsSibling(true); 2767 N_COMM_indexes.pop_back(); 2768 } 2769 type = eSymbolTypeScopeEnd; 2770 break; 2771 2772 case StabLength: 2773 // N_LENG - second stab entry with length information 2774 type = eSymbolTypeAdditional; 2775 break; 2776 2777 default: break; 2778 } 2779 } 2780 else 2781 { 2782 //uint8_t n_pext = NlistMaskPrivateExternal & nlist.n_type; 2783 uint8_t n_type = NlistMaskType & nlist.n_type; 2784 sym[sym_idx].SetExternal((NlistMaskExternal & nlist.n_type) != 0); 2785 2786 switch (n_type) 2787 { 2788 case NListTypeIndirect: // N_INDR - Fall through 2789 case NListTypePreboundUndefined:// N_PBUD - Fall through 2790 case NListTypeUndefined: // N_UNDF 2791 type = eSymbolTypeUndefined; 2792 break; 2793 2794 case NListTypeAbsolute: // N_ABS 2795 type = eSymbolTypeAbsolute; 2796 break; 2797 2798 case NListTypeSection: // N_SECT 2799 { 2800 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2801 2802 if (!symbol_section) 2803 { 2804 // TODO: warn about this? 2805 add_nlist = false; 2806 break; 2807 } 2808 2809 if (TEXT_eh_frame_sectID == nlist.n_sect) 2810 { 2811 type = eSymbolTypeException; 2812 } 2813 else 2814 { 2815 uint32_t section_type = symbol_section->Get() & SectionFlagMaskSectionType; 2816 2817 switch (section_type) 2818 { 2819 case SectionTypeRegular: break; // regular section 2820 //case SectionTypeZeroFill: type = eSymbolTypeData; break; // zero fill on demand section 2821 case SectionTypeCStringLiterals: type = eSymbolTypeData; break; // section with only literal C strings 2822 case SectionType4ByteLiterals: type = eSymbolTypeData; break; // section with only 4 byte literals 2823 case SectionType8ByteLiterals: type = eSymbolTypeData; break; // section with only 8 byte literals 2824 case SectionTypeLiteralPointers: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 2825 case SectionTypeNonLazySymbolPointers: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 2826 case SectionTypeLazySymbolPointers: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 2827 case SectionTypeSymbolStubs: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 2828 case SectionTypeModuleInitFunctionPointers: type = eSymbolTypeCode; break; // section with only function pointers for initialization 2829 case SectionTypeModuleTermFunctionPointers: type = eSymbolTypeCode; break; // section with only function pointers for termination 2830 //case SectionTypeCoalesced: type = eSymbolType; break; // section contains symbols that are to be coalesced 2831 //case SectionTypeZeroFillLarge: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 2832 case SectionTypeInterposing: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 2833 case SectionType16ByteLiterals: type = eSymbolTypeData; break; // section with only 16 byte literals 2834 case SectionTypeDTraceObjectFormat: type = eSymbolTypeInstrumentation; break; 2835 case SectionTypeLazyDylibSymbolPointers: type = eSymbolTypeTrampoline; break; 2836 default: break; 2837 } 2838 2839 if (type == eSymbolTypeInvalid) 2840 { 2841 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 2842 if (symbol_section->IsDescendant (text_section_sp.get())) 2843 { 2844 if (symbol_section->IsClear(SectionAttrUserPureInstructions | 2845 SectionAttrUserSelfModifyingCode | 2846 SectionAttrSytemSomeInstructions)) 2847 type = eSymbolTypeData; 2848 else 2849 type = eSymbolTypeCode; 2850 } 2851 else 2852 if (symbol_section->IsDescendant(data_section_sp.get())) 2853 { 2854 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 2855 { 2856 type = eSymbolTypeRuntime; 2857 2858 if (symbol_name && 2859 symbol_name[0] == '_' && 2860 symbol_name[1] == 'O' && 2861 symbol_name[2] == 'B') 2862 { 2863 llvm::StringRef symbol_name_ref(symbol_name); 2864 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 2865 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 2866 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 2867 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 2868 { 2869 symbol_name_non_abi_mangled = symbol_name + 1; 2870 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2871 type = eSymbolTypeObjCClass; 2872 demangled_is_synthesized = true; 2873 } 2874 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 2875 { 2876 symbol_name_non_abi_mangled = symbol_name + 1; 2877 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2878 type = eSymbolTypeObjCMetaClass; 2879 demangled_is_synthesized = true; 2880 } 2881 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 2882 { 2883 symbol_name_non_abi_mangled = symbol_name + 1; 2884 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2885 type = eSymbolTypeObjCIVar; 2886 demangled_is_synthesized = true; 2887 } 2888 } 2889 } 2890 else 2891 if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 2892 { 2893 type = eSymbolTypeException; 2894 } 2895 else 2896 { 2897 type = eSymbolTypeData; 2898 } 2899 } 2900 else 2901 if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 2902 { 2903 type = eSymbolTypeTrampoline; 2904 } 2905 else 2906 if (symbol_section->IsDescendant(objc_section_sp.get())) 2907 { 2908 type = eSymbolTypeRuntime; 2909 if (symbol_name && symbol_name[0] == '.') 2910 { 2911 llvm::StringRef symbol_name_ref(symbol_name); 2912 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 2913 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 2914 { 2915 symbol_name_non_abi_mangled = symbol_name; 2916 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 2917 type = eSymbolTypeObjCClass; 2918 demangled_is_synthesized = true; 2919 } 2920 } 2921 } 2922 } 2923 } 2924 } 2925 break; 2926 } 2927 } 2928 2929 if (add_nlist) 2930 { 2931 uint64_t symbol_value = nlist.n_value; 2932 bool symbol_name_is_mangled = false; 2933 2934 if (symbol_name_non_abi_mangled) 2935 { 2936 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 2937 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 2938 } 2939 else 2940 { 2941 if (symbol_name && symbol_name[0] == '_') 2942 { 2943 symbol_name_is_mangled = symbol_name[1] == '_'; 2944 symbol_name++; // Skip the leading underscore 2945 } 2946 2947 if (symbol_name) 2948 { 2949 sym[sym_idx].GetMangled().SetValue(ConstString(symbol_name), symbol_name_is_mangled); 2950 } 2951 } 2952 2953 if (is_debug == false) 2954 { 2955 if (type == eSymbolTypeCode) 2956 { 2957 // See if we can find a N_FUN entry for any code symbols. 2958 // If we do find a match, and the name matches, then we 2959 // can merge the two into just the function symbol to avoid 2960 // duplicate entries in the symbol table 2961 ValueToSymbolIndexMap::const_iterator pos = N_FUN_addr_to_sym_idx.find (nlist.n_value); 2962 if (pos != N_FUN_addr_to_sym_idx.end()) 2963 { 2964 if ((symbol_name_is_mangled == true && sym[sym_idx].GetMangled().GetMangledName() == sym[pos->second].GetMangled().GetMangledName()) || 2965 (symbol_name_is_mangled == false && sym[sym_idx].GetMangled().GetDemangledName() == sym[pos->second].GetMangled().GetDemangledName())) 2966 { 2967 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 2968 // We just need the flags from the linker symbol, so put these flags 2969 // into the N_FUN flags to avoid duplicate symbols in the symbol table 2970 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2971 sym[sym_idx].Clear(); 2972 continue; 2973 } 2974 } 2975 } 2976 else if (type == eSymbolTypeData) 2977 { 2978 // See if we can find a N_STSYM entry for any data symbols. 2979 // If we do find a match, and the name matches, then we 2980 // can merge the two into just the Static symbol to avoid 2981 // duplicate entries in the symbol table 2982 ValueToSymbolIndexMap::const_iterator pos = N_STSYM_addr_to_sym_idx.find (nlist.n_value); 2983 if (pos != N_STSYM_addr_to_sym_idx.end()) 2984 { 2985 if ((symbol_name_is_mangled == true && sym[sym_idx].GetMangled().GetMangledName() == sym[pos->second].GetMangled().GetMangledName()) || 2986 (symbol_name_is_mangled == false && sym[sym_idx].GetMangled().GetDemangledName() == sym[pos->second].GetMangled().GetDemangledName())) 2987 { 2988 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 2989 // We just need the flags from the linker symbol, so put these flags 2990 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 2991 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 2992 sym[sym_idx].Clear(); 2993 continue; 2994 } 2995 } 2996 } 2997 } 2998 if (symbol_section) 2999 { 3000 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3001 if (symbol_byte_size == 0 && function_starts_count > 0) 3002 { 3003 addr_t symbol_lookup_file_addr = nlist.n_value; 3004 // Do an exact address match for non-ARM addresses, else get the closest since 3005 // the symbol might be a thumb symbol which has an address with bit zero set 3006 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 3007 if (is_arm && func_start_entry) 3008 { 3009 // Verify that the function start address is the symbol address (ARM) 3010 // or the symbol address + 1 (thumb) 3011 if (func_start_entry->addr != symbol_lookup_file_addr && 3012 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 3013 { 3014 // Not the right entry, NULL it out... 3015 func_start_entry = NULL; 3016 } 3017 } 3018 if (func_start_entry) 3019 { 3020 func_start_entry->data = true; 3021 3022 addr_t symbol_file_addr = func_start_entry->addr; 3023 if (is_arm) 3024 symbol_file_addr &= 0xfffffffffffffffeull; 3025 3026 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3027 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3028 if (next_func_start_entry) 3029 { 3030 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3031 // Be sure the clear the Thumb address bit when we calculate the size 3032 // from the current and next address 3033 if (is_arm) 3034 next_symbol_file_addr &= 0xfffffffffffffffeull; 3035 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3036 } 3037 else 3038 { 3039 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3040 } 3041 } 3042 } 3043 symbol_value -= section_file_addr; 3044 } 3045 3046 sym[sym_idx].SetID (nlist_idx); 3047 sym[sym_idx].SetType (type); 3048 sym[sym_idx].GetAddress().SetSection (symbol_section); 3049 sym[sym_idx].GetAddress().SetOffset (symbol_value); 3050 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3051 3052 if (symbol_byte_size > 0) 3053 sym[sym_idx].SetByteSize(symbol_byte_size); 3054 3055 if (demangled_is_synthesized) 3056 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3057 3058 ++sym_idx; 3059 } 3060 else 3061 { 3062 sym[sym_idx].Clear(); 3063 } 3064 3065 } 3066 3067 // STAB N_GSYM entries end up having a symbol type eSymbolTypeGlobal and when the symbol value 3068 // is zero, the address of the global ends up being in a non-STAB entry. Try and fix up all 3069 // such entries by figuring out what the address for the global is by looking up this non-STAB 3070 // entry and copying the value into the debug symbol's value to save us the hassle in the 3071 // debug symbol parser. 3072 3073 Symbol *global_symbol = NULL; 3074 for (nlist_idx = 0; 3075 nlist_idx < symtab_load_command.nsyms && (global_symbol = symtab->FindSymbolWithType (eSymbolTypeData, Symtab::eDebugYes, Symtab::eVisibilityAny, nlist_idx)) != NULL; 3076 nlist_idx++) 3077 { 3078 if (global_symbol->GetAddress().GetFileAddress() == 0) 3079 { 3080 std::vector<uint32_t> indexes; 3081 if (symtab->AppendSymbolIndexesWithName (global_symbol->GetMangled().GetName(), indexes) > 0) 3082 { 3083 std::vector<uint32_t>::const_iterator pos; 3084 std::vector<uint32_t>::const_iterator end = indexes.end(); 3085 for (pos = indexes.begin(); pos != end; ++pos) 3086 { 3087 symbol_ptr = symtab->SymbolAtIndex(*pos); 3088 if (symbol_ptr != global_symbol && symbol_ptr->IsDebug() == false) 3089 { 3090 global_symbol->GetAddress() = symbol_ptr->GetAddress(); 3091 break; 3092 } 3093 } 3094 } 3095 } 3096 } 3097 3098 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 3099 3100 if (function_starts_count > 0) 3101 { 3102 char synthetic_function_symbol[PATH_MAX]; 3103 uint32_t num_synthetic_function_symbols = 0; 3104 for (i=0; i<function_starts_count; ++i) 3105 { 3106 if (function_starts.GetEntryRef (i).data == false) 3107 ++num_synthetic_function_symbols; 3108 } 3109 3110 if (num_synthetic_function_symbols > 0) 3111 { 3112 if (num_syms < sym_idx + num_synthetic_function_symbols) 3113 { 3114 num_syms = sym_idx + num_synthetic_function_symbols; 3115 sym = symtab->Resize (num_syms); 3116 } 3117 uint32_t synthetic_function_symbol_idx = 0; 3118 for (i=0; i<function_starts_count; ++i) 3119 { 3120 const FunctionStarts::Entry *func_start_entry = function_starts.GetEntryAtIndex (i); 3121 if (func_start_entry->data == false) 3122 { 3123 addr_t symbol_file_addr = func_start_entry->addr; 3124 uint32_t symbol_flags = 0; 3125 if (is_arm) 3126 { 3127 if (symbol_file_addr & 1) 3128 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3129 symbol_file_addr &= 0xfffffffffffffffeull; 3130 } 3131 Address symbol_addr; 3132 if (module_sp->ResolveFileAddress (symbol_file_addr, symbol_addr)) 3133 { 3134 SectionSP symbol_section (symbol_addr.GetSection()); 3135 uint32_t symbol_byte_size = 0; 3136 if (symbol_section) 3137 { 3138 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3139 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3140 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3141 if (next_func_start_entry) 3142 { 3143 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3144 if (is_arm) 3145 next_symbol_file_addr &= 0xfffffffffffffffeull; 3146 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3147 } 3148 else 3149 { 3150 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3151 } 3152 snprintf (synthetic_function_symbol, 3153 sizeof(synthetic_function_symbol), 3154 "___lldb_unnamed_function%u$$%s", 3155 ++synthetic_function_symbol_idx, 3156 module_sp->GetFileSpec().GetFilename().GetCString()); 3157 sym[sym_idx].SetID (synthetic_sym_id++); 3158 sym[sym_idx].GetMangled().SetDemangledName(ConstString(synthetic_function_symbol)); 3159 sym[sym_idx].SetType (eSymbolTypeCode); 3160 sym[sym_idx].SetIsSynthetic (true); 3161 sym[sym_idx].GetAddress() = symbol_addr; 3162 if (symbol_flags) 3163 sym[sym_idx].SetFlags (symbol_flags); 3164 if (symbol_byte_size) 3165 sym[sym_idx].SetByteSize (symbol_byte_size); 3166 ++sym_idx; 3167 } 3168 } 3169 } 3170 } 3171 } 3172 } 3173 3174 // Trim our symbols down to just what we ended up with after 3175 // removing any symbols. 3176 if (sym_idx < num_syms) 3177 { 3178 num_syms = sym_idx; 3179 sym = symtab->Resize (num_syms); 3180 } 3181 3182 // Now synthesize indirect symbols 3183 if (m_dysymtab.nindirectsyms != 0) 3184 { 3185 if (indirect_symbol_index_data.GetByteSize()) 3186 { 3187 NListIndexToSymbolIndexMap::const_iterator end_index_pos = m_nlist_idx_to_sym_idx.end(); 3188 3189 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); ++sect_idx) 3190 { 3191 if ((m_mach_sections[sect_idx].flags & SectionFlagMaskSectionType) == SectionTypeSymbolStubs) 3192 { 3193 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 3194 if (symbol_stub_byte_size == 0) 3195 continue; 3196 3197 const uint32_t num_symbol_stubs = m_mach_sections[sect_idx].size / symbol_stub_byte_size; 3198 3199 if (num_symbol_stubs == 0) 3200 continue; 3201 3202 const uint32_t symbol_stub_index_offset = m_mach_sections[sect_idx].reserved1; 3203 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) 3204 { 3205 const uint32_t symbol_stub_index = symbol_stub_index_offset + stub_idx; 3206 const lldb::addr_t symbol_stub_addr = m_mach_sections[sect_idx].addr + (stub_idx * symbol_stub_byte_size); 3207 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 3208 if (indirect_symbol_index_data.ValidOffsetForDataOfSize(symbol_stub_offset, 4)) 3209 { 3210 const uint32_t stub_sym_id = indirect_symbol_index_data.GetU32 (&symbol_stub_offset); 3211 if (stub_sym_id & (IndirectSymbolAbsolute | IndirectSymbolLocal)) 3212 continue; 3213 3214 NListIndexToSymbolIndexMap::const_iterator index_pos = m_nlist_idx_to_sym_idx.find (stub_sym_id); 3215 Symbol *stub_symbol = NULL; 3216 if (index_pos != end_index_pos) 3217 { 3218 // We have a remapping from the original nlist index to 3219 // a current symbol index, so just look this up by index 3220 stub_symbol = symtab->SymbolAtIndex (index_pos->second); 3221 } 3222 else 3223 { 3224 // We need to lookup a symbol using the original nlist 3225 // symbol index since this index is coming from the 3226 // S_SYMBOL_STUBS 3227 stub_symbol = symtab->FindSymbolByID (stub_sym_id); 3228 } 3229 3230 if (stub_symbol) 3231 { 3232 Address so_addr(symbol_stub_addr, section_list); 3233 3234 if (stub_symbol->GetType() == eSymbolTypeUndefined) 3235 { 3236 // Change the external symbol into a trampoline that makes sense 3237 // These symbols were N_UNDF N_EXT, and are useless to us, so we 3238 // can re-use them so we don't have to make up a synthetic symbol 3239 // for no good reason. 3240 stub_symbol->SetType (eSymbolTypeTrampoline); 3241 stub_symbol->SetExternal (false); 3242 stub_symbol->GetAddress() = so_addr; 3243 stub_symbol->SetByteSize (symbol_stub_byte_size); 3244 } 3245 else 3246 { 3247 // Make a synthetic symbol to describe the trampoline stub 3248 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 3249 if (sym_idx >= num_syms) 3250 { 3251 sym = symtab->Resize (++num_syms); 3252 stub_symbol = NULL; // this pointer no longer valid 3253 } 3254 sym[sym_idx].SetID (synthetic_sym_id++); 3255 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 3256 sym[sym_idx].SetType (eSymbolTypeTrampoline); 3257 sym[sym_idx].SetIsSynthetic (true); 3258 sym[sym_idx].GetAddress() = so_addr; 3259 sym[sym_idx].SetByteSize (symbol_stub_byte_size); 3260 ++sym_idx; 3261 } 3262 } 3263 else 3264 { 3265 if (log) 3266 log->Warning ("symbol stub referencing symbol table symbol %u that isn't in our minimal symbol table, fix this!!!", stub_sym_id); 3267 } 3268 } 3269 } 3270 } 3271 } 3272 } 3273 } 3274 return symtab->GetNumSymbols(); 3275 } 3276 return 0; 3277} 3278 3279 3280void 3281ObjectFileMachO::Dump (Stream *s) 3282{ 3283 ModuleSP module_sp(GetModule()); 3284 if (module_sp) 3285 { 3286 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 3287 s->Printf("%p: ", this); 3288 s->Indent(); 3289 if (m_header.magic == HeaderMagic64 || m_header.magic == HeaderMagic64Swapped) 3290 s->PutCString("ObjectFileMachO64"); 3291 else 3292 s->PutCString("ObjectFileMachO32"); 3293 3294 ArchSpec header_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 3295 3296 *s << ", file = '" << m_file << "', arch = " << header_arch.GetArchitectureName() << "\n"; 3297 3298 if (m_sections_ap.get()) 3299 m_sections_ap->Dump(s, NULL, true, UINT32_MAX); 3300 3301 if (m_symtab_ap.get()) 3302 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 3303 } 3304} 3305 3306 3307bool 3308ObjectFileMachO::GetUUID (lldb_private::UUID* uuid) 3309{ 3310 ModuleSP module_sp(GetModule()); 3311 if (module_sp) 3312 { 3313 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 3314 struct uuid_command load_cmd; 3315 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 3316 uint32_t i; 3317 for (i=0; i<m_header.ncmds; ++i) 3318 { 3319 const lldb::offset_t cmd_offset = offset; 3320 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 3321 break; 3322 3323 if (load_cmd.cmd == LoadCommandUUID) 3324 { 3325 const uint8_t *uuid_bytes = m_data.PeekData(offset, 16); 3326 3327 if (uuid_bytes) 3328 { 3329 // OpenCL on Mac OS X uses the same UUID for each of its object files. 3330 // We pretend these object files have no UUID to prevent crashing. 3331 3332 const uint8_t opencl_uuid[] = { 0x8c, 0x8e, 0xb3, 0x9b, 3333 0x3b, 0xa8, 3334 0x4b, 0x16, 3335 0xb6, 0xa4, 3336 0x27, 0x63, 0xbb, 0x14, 0xf0, 0x0d }; 3337 3338 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 3339 return false; 3340 3341 uuid->SetBytes (uuid_bytes); 3342 return true; 3343 } 3344 return false; 3345 } 3346 offset = cmd_offset + load_cmd.cmdsize; 3347 } 3348 } 3349 return false; 3350} 3351 3352 3353uint32_t 3354ObjectFileMachO::GetDependentModules (FileSpecList& files) 3355{ 3356 uint32_t count = 0; 3357 ModuleSP module_sp(GetModule()); 3358 if (module_sp) 3359 { 3360 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 3361 struct load_command load_cmd; 3362 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 3363 const bool resolve_path = false; // Don't resolve the dependend file paths since they may not reside on this system 3364 uint32_t i; 3365 for (i=0; i<m_header.ncmds; ++i) 3366 { 3367 const uint32_t cmd_offset = offset; 3368 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 3369 break; 3370 3371 switch (load_cmd.cmd) 3372 { 3373 case LoadCommandDylibLoad: 3374 case LoadCommandDylibLoadWeak: 3375 case LoadCommandDylibReexport: 3376 case LoadCommandDynamicLinkerLoad: 3377 case LoadCommandFixedVMShlibLoad: 3378 case LoadCommandDylibLoadUpward: 3379 { 3380 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 3381 const char *path = m_data.PeekCStr(name_offset); 3382 // Skip any path that starts with '@' since these are usually: 3383 // @executable_path/.../file 3384 // @rpath/.../file 3385 if (path && path[0] != '@') 3386 { 3387 FileSpec file_spec(path, resolve_path); 3388 if (files.AppendIfUnique(file_spec)) 3389 count++; 3390 } 3391 } 3392 break; 3393 3394 default: 3395 break; 3396 } 3397 offset = cmd_offset + load_cmd.cmdsize; 3398 } 3399 } 3400 return count; 3401} 3402 3403lldb_private::Address 3404ObjectFileMachO::GetEntryPointAddress () 3405{ 3406 // If the object file is not an executable it can't hold the entry point. m_entry_point_address 3407 // is initialized to an invalid address, so we can just return that. 3408 // If m_entry_point_address is valid it means we've found it already, so return the cached value. 3409 3410 if (!IsExecutable() || m_entry_point_address.IsValid()) 3411 return m_entry_point_address; 3412 3413 // Otherwise, look for the UnixThread or Thread command. The data for the Thread command is given in 3414 // /usr/include/mach-o.h, but it is basically: 3415 // 3416 // uint32_t flavor - this is the flavor argument you would pass to thread_get_state 3417 // uint32_t count - this is the count of longs in the thread state data 3418 // struct XXX_thread_state state - this is the structure from <machine/thread_status.h> corresponding to the flavor. 3419 // <repeat this trio> 3420 // 3421 // So we just keep reading the various register flavors till we find the GPR one, then read the PC out of there. 3422 // FIXME: We will need to have a "RegisterContext data provider" class at some point that can get all the registers 3423 // out of data in this form & attach them to a given thread. That should underlie the MacOS X User process plugin, 3424 // and we'll also need it for the MacOS X Core File process plugin. When we have that we can also use it here. 3425 // 3426 // For now we hard-code the offsets and flavors we need: 3427 // 3428 // 3429 3430 ModuleSP module_sp(GetModule()); 3431 if (module_sp) 3432 { 3433 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 3434 struct load_command load_cmd; 3435 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 3436 uint32_t i; 3437 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 3438 bool done = false; 3439 3440 for (i=0; i<m_header.ncmds; ++i) 3441 { 3442 const lldb::offset_t cmd_offset = offset; 3443 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 3444 break; 3445 3446 switch (load_cmd.cmd) 3447 { 3448 case LoadCommandUnixThread: 3449 case LoadCommandThread: 3450 { 3451 while (offset < cmd_offset + load_cmd.cmdsize) 3452 { 3453 uint32_t flavor = m_data.GetU32(&offset); 3454 uint32_t count = m_data.GetU32(&offset); 3455 if (count == 0) 3456 { 3457 // We've gotten off somehow, log and exit; 3458 return m_entry_point_address; 3459 } 3460 3461 switch (m_header.cputype) 3462 { 3463 case llvm::MachO::CPUTypeARM: 3464 if (flavor == 1) // ARM_THREAD_STATE from mach/arm/thread_status.h 3465 { 3466 offset += 60; // This is the offset of pc in the GPR thread state data structure. 3467 start_address = m_data.GetU32(&offset); 3468 done = true; 3469 } 3470 break; 3471 case llvm::MachO::CPUTypeI386: 3472 if (flavor == 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 3473 { 3474 offset += 40; // This is the offset of eip in the GPR thread state data structure. 3475 start_address = m_data.GetU32(&offset); 3476 done = true; 3477 } 3478 break; 3479 case llvm::MachO::CPUTypeX86_64: 3480 if (flavor == 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 3481 { 3482 offset += 16 * 8; // This is the offset of rip in the GPR thread state data structure. 3483 start_address = m_data.GetU64(&offset); 3484 done = true; 3485 } 3486 break; 3487 default: 3488 return m_entry_point_address; 3489 } 3490 // Haven't found the GPR flavor yet, skip over the data for this flavor: 3491 if (done) 3492 break; 3493 offset += count * 4; 3494 } 3495 } 3496 break; 3497 case LoadCommandMain: 3498 { 3499 ConstString text_segment_name ("__TEXT"); 3500 uint64_t entryoffset = m_data.GetU64(&offset); 3501 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 3502 if (text_segment_sp) 3503 { 3504 done = true; 3505 start_address = text_segment_sp->GetFileAddress() + entryoffset; 3506 } 3507 } 3508 3509 default: 3510 break; 3511 } 3512 if (done) 3513 break; 3514 3515 // Go to the next load command: 3516 offset = cmd_offset + load_cmd.cmdsize; 3517 } 3518 3519 if (start_address != LLDB_INVALID_ADDRESS) 3520 { 3521 // We got the start address from the load commands, so now resolve that address in the sections 3522 // of this ObjectFile: 3523 if (!m_entry_point_address.ResolveAddressUsingFileSections (start_address, GetSectionList())) 3524 { 3525 m_entry_point_address.Clear(); 3526 } 3527 } 3528 else 3529 { 3530 // We couldn't read the UnixThread load command - maybe it wasn't there. As a fallback look for the 3531 // "start" symbol in the main executable. 3532 3533 ModuleSP module_sp (GetModule()); 3534 3535 if (module_sp) 3536 { 3537 SymbolContextList contexts; 3538 SymbolContext context; 3539 if (module_sp->FindSymbolsWithNameAndType(ConstString ("start"), eSymbolTypeCode, contexts)) 3540 { 3541 if (contexts.GetContextAtIndex(0, context)) 3542 m_entry_point_address = context.symbol->GetAddress(); 3543 } 3544 } 3545 } 3546 } 3547 3548 return m_entry_point_address; 3549 3550} 3551 3552lldb_private::Address 3553ObjectFileMachO::GetHeaderAddress () 3554{ 3555 lldb_private::Address header_addr; 3556 SectionList *section_list = GetSectionList(); 3557 if (section_list) 3558 { 3559 SectionSP text_segment_sp (section_list->FindSectionByName (GetSegmentNameTEXT())); 3560 if (text_segment_sp) 3561 { 3562 header_addr.SetSection (text_segment_sp); 3563 header_addr.SetOffset (0); 3564 } 3565 } 3566 return header_addr; 3567} 3568 3569uint32_t 3570ObjectFileMachO::GetNumThreadContexts () 3571{ 3572 ModuleSP module_sp(GetModule()); 3573 if (module_sp) 3574 { 3575 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 3576 if (!m_thread_context_offsets_valid) 3577 { 3578 m_thread_context_offsets_valid = true; 3579 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 3580 FileRangeArray::Entry file_range; 3581 thread_command thread_cmd; 3582 for (uint32_t i=0; i<m_header.ncmds; ++i) 3583 { 3584 const uint32_t cmd_offset = offset; 3585 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 3586 break; 3587 3588 if (thread_cmd.cmd == LoadCommandThread) 3589 { 3590 file_range.SetRangeBase (offset); 3591 file_range.SetByteSize (thread_cmd.cmdsize - 8); 3592 m_thread_context_offsets.Append (file_range); 3593 } 3594 offset = cmd_offset + thread_cmd.cmdsize; 3595 } 3596 } 3597 } 3598 return m_thread_context_offsets.GetSize(); 3599} 3600 3601lldb::RegisterContextSP 3602ObjectFileMachO::GetThreadContextAtIndex (uint32_t idx, lldb_private::Thread &thread) 3603{ 3604 lldb::RegisterContextSP reg_ctx_sp; 3605 3606 ModuleSP module_sp(GetModule()); 3607 if (module_sp) 3608 { 3609 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 3610 if (!m_thread_context_offsets_valid) 3611 GetNumThreadContexts (); 3612 3613 const FileRangeArray::Entry *thread_context_file_range = m_thread_context_offsets.GetEntryAtIndex (idx); 3614 if (thread_context_file_range) 3615 { 3616 3617 DataExtractor data (m_data, 3618 thread_context_file_range->GetRangeBase(), 3619 thread_context_file_range->GetByteSize()); 3620 3621 switch (m_header.cputype) 3622 { 3623 case llvm::MachO::CPUTypeARM: 3624 reg_ctx_sp.reset (new RegisterContextDarwin_arm_Mach (thread, data)); 3625 break; 3626 3627 case llvm::MachO::CPUTypeI386: 3628 reg_ctx_sp.reset (new RegisterContextDarwin_i386_Mach (thread, data)); 3629 break; 3630 3631 case llvm::MachO::CPUTypeX86_64: 3632 reg_ctx_sp.reset (new RegisterContextDarwin_x86_64_Mach (thread, data)); 3633 break; 3634 } 3635 } 3636 } 3637 return reg_ctx_sp; 3638} 3639 3640 3641ObjectFile::Type 3642ObjectFileMachO::CalculateType() 3643{ 3644 switch (m_header.filetype) 3645 { 3646 case HeaderFileTypeObject: // 0x1u MH_OBJECT 3647 if (GetAddressByteSize () == 4) 3648 { 3649 // 32 bit kexts are just object files, but they do have a valid 3650 // UUID load command. 3651 UUID uuid; 3652 if (GetUUID(&uuid)) 3653 { 3654 // this checking for the UUID load command is not enough 3655 // we could eventually look for the symbol named 3656 // "OSKextGetCurrentIdentifier" as this is required of kexts 3657 if (m_strata == eStrataInvalid) 3658 m_strata = eStrataKernel; 3659 return eTypeSharedLibrary; 3660 } 3661 } 3662 return eTypeObjectFile; 3663 3664 case HeaderFileTypeExecutable: return eTypeExecutable; // 0x2u MH_EXECUTE 3665 case HeaderFileTypeFixedVMShlib: return eTypeSharedLibrary; // 0x3u MH_FVMLIB 3666 case HeaderFileTypeCore: return eTypeCoreFile; // 0x4u MH_CORE 3667 case HeaderFileTypePreloadedExecutable: return eTypeSharedLibrary; // 0x5u MH_PRELOAD 3668 case HeaderFileTypeDynamicShlib: return eTypeSharedLibrary; // 0x6u MH_DYLIB 3669 case HeaderFileTypeDynamicLinkEditor: return eTypeDynamicLinker; // 0x7u MH_DYLINKER 3670 case HeaderFileTypeBundle: return eTypeSharedLibrary; // 0x8u MH_BUNDLE 3671 case HeaderFileTypeDynamicShlibStub: return eTypeStubLibrary; // 0x9u MH_DYLIB_STUB 3672 case HeaderFileTypeDSYM: return eTypeDebugInfo; // 0xAu MH_DSYM 3673 case HeaderFileTypeKextBundle: return eTypeSharedLibrary; // 0xBu MH_KEXT_BUNDLE 3674 default: 3675 break; 3676 } 3677 return eTypeUnknown; 3678} 3679 3680ObjectFile::Strata 3681ObjectFileMachO::CalculateStrata() 3682{ 3683 switch (m_header.filetype) 3684 { 3685 case HeaderFileTypeObject: // 0x1u MH_OBJECT 3686 { 3687 // 32 bit kexts are just object files, but they do have a valid 3688 // UUID load command. 3689 UUID uuid; 3690 if (GetUUID(&uuid)) 3691 { 3692 // this checking for the UUID load command is not enough 3693 // we could eventually look for the symbol named 3694 // "OSKextGetCurrentIdentifier" as this is required of kexts 3695 if (m_type == eTypeInvalid) 3696 m_type = eTypeSharedLibrary; 3697 3698 return eStrataKernel; 3699 } 3700 } 3701 return eStrataUnknown; 3702 3703 case HeaderFileTypeExecutable: // 0x2u MH_EXECUTE 3704 // Check for the MH_DYLDLINK bit in the flags 3705 if (m_header.flags & HeaderFlagBitIsDynamicLinkObject) 3706 { 3707 return eStrataUser; 3708 } 3709 else 3710 { 3711 SectionList *section_list = GetSectionList(); 3712 if (section_list) 3713 { 3714 static ConstString g_kld_section_name ("__KLD"); 3715 if (section_list->FindSectionByName(g_kld_section_name)) 3716 return eStrataKernel; 3717 } 3718 } 3719 return eStrataRawImage; 3720 3721 case HeaderFileTypeFixedVMShlib: return eStrataUser; // 0x3u MH_FVMLIB 3722 case HeaderFileTypeCore: return eStrataUnknown; // 0x4u MH_CORE 3723 case HeaderFileTypePreloadedExecutable: return eStrataRawImage; // 0x5u MH_PRELOAD 3724 case HeaderFileTypeDynamicShlib: return eStrataUser; // 0x6u MH_DYLIB 3725 case HeaderFileTypeDynamicLinkEditor: return eStrataUser; // 0x7u MH_DYLINKER 3726 case HeaderFileTypeBundle: return eStrataUser; // 0x8u MH_BUNDLE 3727 case HeaderFileTypeDynamicShlibStub: return eStrataUser; // 0x9u MH_DYLIB_STUB 3728 case HeaderFileTypeDSYM: return eStrataUnknown; // 0xAu MH_DSYM 3729 case HeaderFileTypeKextBundle: return eStrataKernel; // 0xBu MH_KEXT_BUNDLE 3730 default: 3731 break; 3732 } 3733 return eStrataUnknown; 3734} 3735 3736 3737uint32_t 3738ObjectFileMachO::GetVersion (uint32_t *versions, uint32_t num_versions) 3739{ 3740 ModuleSP module_sp(GetModule()); 3741 if (module_sp) 3742 { 3743 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 3744 struct dylib_command load_cmd; 3745 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 3746 uint32_t version_cmd = 0; 3747 uint64_t version = 0; 3748 uint32_t i; 3749 for (i=0; i<m_header.ncmds; ++i) 3750 { 3751 const lldb::offset_t cmd_offset = offset; 3752 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 3753 break; 3754 3755 if (load_cmd.cmd == LoadCommandDylibIdent) 3756 { 3757 if (version_cmd == 0) 3758 { 3759 version_cmd = load_cmd.cmd; 3760 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 3761 break; 3762 version = load_cmd.dylib.current_version; 3763 } 3764 break; // Break for now unless there is another more complete version 3765 // number load command in the future. 3766 } 3767 offset = cmd_offset + load_cmd.cmdsize; 3768 } 3769 3770 if (version_cmd == LoadCommandDylibIdent) 3771 { 3772 if (versions != NULL && num_versions > 0) 3773 { 3774 if (num_versions > 0) 3775 versions[0] = (version & 0xFFFF0000ull) >> 16; 3776 if (num_versions > 1) 3777 versions[1] = (version & 0x0000FF00ull) >> 8; 3778 if (num_versions > 2) 3779 versions[2] = (version & 0x000000FFull); 3780 // Fill in an remaining version numbers with invalid values 3781 for (i=3; i<num_versions; ++i) 3782 versions[i] = UINT32_MAX; 3783 } 3784 // The LC_ID_DYLIB load command has a version with 3 version numbers 3785 // in it, so always return 3 3786 return 3; 3787 } 3788 } 3789 return false; 3790} 3791 3792bool 3793ObjectFileMachO::GetArchitecture (ArchSpec &arch) 3794{ 3795 ModuleSP module_sp(GetModule()); 3796 if (module_sp) 3797 { 3798 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 3799 arch.SetArchitecture (eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 3800 3801 // Files with type MH_PRELOAD are currently used in cases where the image 3802 // debugs at the addresses in the file itself. Below we set the OS to 3803 // unknown to make sure we use the DynamicLoaderStatic()... 3804 if (m_header.filetype == HeaderFileTypePreloadedExecutable) 3805 { 3806 arch.GetTriple().setOS (llvm::Triple::UnknownOS); 3807 } 3808 return true; 3809 } 3810 return false; 3811} 3812 3813 3814//------------------------------------------------------------------ 3815// PluginInterface protocol 3816//------------------------------------------------------------------ 3817const char * 3818ObjectFileMachO::GetPluginName() 3819{ 3820 return "ObjectFileMachO"; 3821} 3822 3823const char * 3824ObjectFileMachO::GetShortPluginName() 3825{ 3826 return GetPluginNameStatic(); 3827} 3828 3829uint32_t 3830ObjectFileMachO::GetPluginVersion() 3831{ 3832 return 1; 3833} 3834 3835