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 12 #include "lldb/lldb-private-log.h" 13 #include "lldb/Core/ArchSpec.h" 14 #include "lldb/Core/DataBuffer.h" 15 #include "lldb/Core/Debugger.h" 16 #include "lldb/Core/FileSpecList.h" 17 #include "lldb/Core/Log.h" 18 #include "lldb/Core/Module.h" 19 #include "lldb/Core/ModuleSpec.h" 20 #include "lldb/Core/PluginManager.h" 21 #include "lldb/Core/RangeMap.h" 22 #include "lldb/Core/Section.h" 23 #include "lldb/Core/StreamFile.h" 24 #include "lldb/Core/StreamString.h" 25 #include "lldb/Core/Timer.h" 26 #include "lldb/Core/UUID.h" 27 #include "lldb/Host/Host.h" 28 #include "lldb/Host/FileSpec.h" 29 #include "lldb/Symbol/ClangNamespaceDecl.h" 30 #include "lldb/Symbol/DWARFCallFrameInfo.h" 31 #include "lldb/Symbol/ObjectFile.h" 32 #include "lldb/Target/Platform.h" 33 #include "lldb/Target/Process.h" 34 #include "lldb/Target/SectionLoadList.h" 35 #include "lldb/Target/Target.h" 36 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 37 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h" 38 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 39 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 40 41 #include "lldb/Utility/SafeMachO.h" 42 43 #include "ObjectFileMachO.h" 44 45 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__)) 46 // GetLLDBSharedCacheUUID() needs to call dlsym() 47 #include <dlfcn.h> 48 #endif 49 50 #ifndef __APPLE__ 51 #include "Utility/UuidCompatibility.h" 52 #endif 53 54 using namespace lldb; 55 using namespace lldb_private; 56 using namespace llvm::MachO; 57 58 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 59 { 60 public: 61 RegisterContextDarwin_x86_64_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 62 RegisterContextDarwin_x86_64 (thread, 0) 63 { 64 SetRegisterDataFrom_LC_THREAD (data); 65 } 66 67 virtual void 68 InvalidateAllRegisters () 69 { 70 // Do nothing... registers are always valid... 71 } 72 73 void 74 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 75 { 76 lldb::offset_t offset = 0; 77 SetError (GPRRegSet, Read, -1); 78 SetError (FPURegSet, Read, -1); 79 SetError (EXCRegSet, Read, -1); 80 bool done = false; 81 82 while (!done) 83 { 84 int flavor = data.GetU32 (&offset); 85 if (flavor == 0) 86 done = true; 87 else 88 { 89 uint32_t i; 90 uint32_t count = data.GetU32 (&offset); 91 switch (flavor) 92 { 93 case GPRRegSet: 94 for (i=0; i<count; ++i) 95 (&gpr.rax)[i] = data.GetU64(&offset); 96 SetError (GPRRegSet, Read, 0); 97 done = true; 98 99 break; 100 case FPURegSet: 101 // TODO: fill in FPU regs.... 102 //SetError (FPURegSet, Read, -1); 103 done = true; 104 105 break; 106 case EXCRegSet: 107 exc.trapno = data.GetU32(&offset); 108 exc.err = data.GetU32(&offset); 109 exc.faultvaddr = data.GetU64(&offset); 110 SetError (EXCRegSet, Read, 0); 111 done = true; 112 break; 113 case 7: 114 case 8: 115 case 9: 116 // fancy flavors that encapsulate of the the above 117 // falvors... 118 break; 119 120 default: 121 done = true; 122 break; 123 } 124 } 125 } 126 } 127 protected: 128 virtual int 129 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 130 { 131 return 0; 132 } 133 134 virtual int 135 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 136 { 137 return 0; 138 } 139 140 virtual int 141 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 142 { 143 return 0; 144 } 145 146 virtual int 147 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 148 { 149 return 0; 150 } 151 152 virtual int 153 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 154 { 155 return 0; 156 } 157 158 virtual int 159 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 160 { 161 return 0; 162 } 163 }; 164 165 166 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 167 { 168 public: 169 RegisterContextDarwin_i386_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 170 RegisterContextDarwin_i386 (thread, 0) 171 { 172 SetRegisterDataFrom_LC_THREAD (data); 173 } 174 175 virtual void 176 InvalidateAllRegisters () 177 { 178 // Do nothing... registers are always valid... 179 } 180 181 void 182 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 183 { 184 lldb::offset_t offset = 0; 185 SetError (GPRRegSet, Read, -1); 186 SetError (FPURegSet, Read, -1); 187 SetError (EXCRegSet, Read, -1); 188 bool done = false; 189 190 while (!done) 191 { 192 int flavor = data.GetU32 (&offset); 193 if (flavor == 0) 194 done = true; 195 else 196 { 197 uint32_t i; 198 uint32_t count = data.GetU32 (&offset); 199 switch (flavor) 200 { 201 case GPRRegSet: 202 for (i=0; i<count; ++i) 203 (&gpr.eax)[i] = data.GetU32(&offset); 204 SetError (GPRRegSet, Read, 0); 205 done = true; 206 207 break; 208 case FPURegSet: 209 // TODO: fill in FPU regs.... 210 //SetError (FPURegSet, Read, -1); 211 done = true; 212 213 break; 214 case EXCRegSet: 215 exc.trapno = data.GetU32(&offset); 216 exc.err = data.GetU32(&offset); 217 exc.faultvaddr = data.GetU32(&offset); 218 SetError (EXCRegSet, Read, 0); 219 done = true; 220 break; 221 case 7: 222 case 8: 223 case 9: 224 // fancy flavors that encapsulate of the the above 225 // falvors... 226 break; 227 228 default: 229 done = true; 230 break; 231 } 232 } 233 } 234 } 235 protected: 236 virtual int 237 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 238 { 239 return 0; 240 } 241 242 virtual int 243 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 244 { 245 return 0; 246 } 247 248 virtual int 249 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 250 { 251 return 0; 252 } 253 254 virtual int 255 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 256 { 257 return 0; 258 } 259 260 virtual int 261 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 262 { 263 return 0; 264 } 265 266 virtual int 267 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 268 { 269 return 0; 270 } 271 }; 272 273 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm 274 { 275 public: 276 RegisterContextDarwin_arm_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 277 RegisterContextDarwin_arm (thread, 0) 278 { 279 SetRegisterDataFrom_LC_THREAD (data); 280 } 281 282 virtual void 283 InvalidateAllRegisters () 284 { 285 // Do nothing... registers are always valid... 286 } 287 288 void 289 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 290 { 291 lldb::offset_t offset = 0; 292 SetError (GPRRegSet, Read, -1); 293 SetError (FPURegSet, Read, -1); 294 SetError (EXCRegSet, Read, -1); 295 bool done = false; 296 297 while (!done) 298 { 299 int flavor = data.GetU32 (&offset); 300 uint32_t count = data.GetU32 (&offset); 301 lldb::offset_t next_thread_state = offset + (count * 4); 302 switch (flavor) 303 { 304 case GPRRegSet: 305 for (uint32_t i=0; i<count; ++i) 306 { 307 gpr.r[i] = data.GetU32(&offset); 308 } 309 310 // Note that gpr.cpsr is also copied by the above loop; this loop technically extends 311 // one element past the end of the gpr.r[] array. 312 313 SetError (GPRRegSet, Read, 0); 314 offset = next_thread_state; 315 break; 316 317 case FPURegSet: 318 { 319 uint8_t *fpu_reg_buf = (uint8_t*) &fpu.floats.s[0]; 320 const int fpu_reg_buf_size = sizeof (fpu.floats); 321 if (data.ExtractBytes (offset, fpu_reg_buf_size, eByteOrderLittle, fpu_reg_buf) == fpu_reg_buf_size) 322 { 323 offset += fpu_reg_buf_size; 324 fpu.fpscr = data.GetU32(&offset); 325 SetError (FPURegSet, Read, 0); 326 } 327 else 328 { 329 done = true; 330 } 331 } 332 offset = next_thread_state; 333 break; 334 335 case EXCRegSet: 336 if (count == 3) 337 { 338 exc.exception = data.GetU32(&offset); 339 exc.fsr = data.GetU32(&offset); 340 exc.far = data.GetU32(&offset); 341 SetError (EXCRegSet, Read, 0); 342 } 343 done = true; 344 offset = next_thread_state; 345 break; 346 347 // Unknown register set flavor, stop trying to parse. 348 default: 349 done = true; 350 } 351 } 352 } 353 protected: 354 virtual int 355 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 356 { 357 return -1; 358 } 359 360 virtual int 361 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 362 { 363 return -1; 364 } 365 366 virtual int 367 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 368 { 369 return -1; 370 } 371 372 virtual int 373 DoReadDBG (lldb::tid_t tid, int flavor, DBG &dbg) 374 { 375 return -1; 376 } 377 378 virtual int 379 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 380 { 381 return 0; 382 } 383 384 virtual int 385 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 386 { 387 return 0; 388 } 389 390 virtual int 391 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 392 { 393 return 0; 394 } 395 396 virtual int 397 DoWriteDBG (lldb::tid_t tid, int flavor, const DBG &dbg) 398 { 399 return -1; 400 } 401 }; 402 403 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 404 { 405 public: 406 RegisterContextDarwin_arm64_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 407 RegisterContextDarwin_arm64 (thread, 0) 408 { 409 SetRegisterDataFrom_LC_THREAD (data); 410 } 411 412 virtual void 413 InvalidateAllRegisters () 414 { 415 // Do nothing... registers are always valid... 416 } 417 418 void 419 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 420 { 421 lldb::offset_t offset = 0; 422 SetError (GPRRegSet, Read, -1); 423 SetError (FPURegSet, Read, -1); 424 SetError (EXCRegSet, Read, -1); 425 bool done = false; 426 while (!done) 427 { 428 int flavor = data.GetU32 (&offset); 429 uint32_t count = data.GetU32 (&offset); 430 lldb::offset_t next_thread_state = offset + (count * 4); 431 switch (flavor) 432 { 433 case GPRRegSet: 434 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 32-bit register) 435 if (count >= (33 * 2) + 1) 436 { 437 for (uint32_t i=0; i<33; ++i) 438 gpr.x[i] = data.GetU64(&offset); 439 gpr.cpsr = data.GetU32(&offset); 440 SetError (GPRRegSet, Read, 0); 441 } 442 offset = next_thread_state; 443 break; 444 case FPURegSet: 445 { 446 uint8_t *fpu_reg_buf = (uint8_t*) &fpu.v[0]; 447 const int fpu_reg_buf_size = sizeof (fpu); 448 if (fpu_reg_buf_size == count 449 && data.ExtractBytes (offset, fpu_reg_buf_size, eByteOrderLittle, fpu_reg_buf) == fpu_reg_buf_size) 450 { 451 SetError (FPURegSet, Read, 0); 452 } 453 else 454 { 455 done = true; 456 } 457 } 458 offset = next_thread_state; 459 break; 460 case EXCRegSet: 461 if (count == 4) 462 { 463 exc.far = data.GetU64(&offset); 464 exc.esr = data.GetU32(&offset); 465 exc.exception = data.GetU32(&offset); 466 SetError (EXCRegSet, Read, 0); 467 } 468 offset = next_thread_state; 469 break; 470 default: 471 done = true; 472 break; 473 } 474 } 475 } 476 protected: 477 virtual int 478 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 479 { 480 return -1; 481 } 482 483 virtual int 484 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 485 { 486 return -1; 487 } 488 489 virtual int 490 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 491 { 492 return -1; 493 } 494 495 virtual int 496 DoReadDBG (lldb::tid_t tid, int flavor, DBG &dbg) 497 { 498 return -1; 499 } 500 501 virtual int 502 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 503 { 504 return 0; 505 } 506 507 virtual int 508 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 509 { 510 return 0; 511 } 512 513 virtual int 514 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 515 { 516 return 0; 517 } 518 519 virtual int 520 DoWriteDBG (lldb::tid_t tid, int flavor, const DBG &dbg) 521 { 522 return -1; 523 } 524 }; 525 526 static uint32_t 527 MachHeaderSizeFromMagic(uint32_t magic) 528 { 529 switch (magic) 530 { 531 case MH_MAGIC: 532 case MH_CIGAM: 533 return sizeof(struct mach_header); 534 535 case MH_MAGIC_64: 536 case MH_CIGAM_64: 537 return sizeof(struct mach_header_64); 538 break; 539 540 default: 541 break; 542 } 543 return 0; 544 } 545 546 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 547 548 void 549 ObjectFileMachO::Initialize() 550 { 551 PluginManager::RegisterPlugin (GetPluginNameStatic(), 552 GetPluginDescriptionStatic(), 553 CreateInstance, 554 CreateMemoryInstance, 555 GetModuleSpecifications); 556 } 557 558 void 559 ObjectFileMachO::Terminate() 560 { 561 PluginManager::UnregisterPlugin (CreateInstance); 562 } 563 564 565 lldb_private::ConstString 566 ObjectFileMachO::GetPluginNameStatic() 567 { 568 static ConstString g_name("mach-o"); 569 return g_name; 570 } 571 572 const char * 573 ObjectFileMachO::GetPluginDescriptionStatic() 574 { 575 return "Mach-o object file reader (32 and 64 bit)"; 576 } 577 578 ObjectFile * 579 ObjectFileMachO::CreateInstance (const lldb::ModuleSP &module_sp, 580 DataBufferSP& data_sp, 581 lldb::offset_t data_offset, 582 const FileSpec* file, 583 lldb::offset_t file_offset, 584 lldb::offset_t length) 585 { 586 if (!data_sp) 587 { 588 data_sp = file->MemoryMapFileContents(file_offset, length); 589 data_offset = 0; 590 } 591 592 if (ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 593 { 594 // Update the data to contain the entire file if it doesn't already 595 if (data_sp->GetByteSize() < length) 596 { 597 data_sp = file->MemoryMapFileContents(file_offset, length); 598 data_offset = 0; 599 } 600 std::unique_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module_sp, data_sp, data_offset, file, file_offset, length)); 601 if (objfile_ap.get() && objfile_ap->ParseHeader()) 602 return objfile_ap.release(); 603 } 604 return NULL; 605 } 606 607 ObjectFile * 608 ObjectFileMachO::CreateMemoryInstance (const lldb::ModuleSP &module_sp, 609 DataBufferSP& data_sp, 610 const ProcessSP &process_sp, 611 lldb::addr_t header_addr) 612 { 613 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) 614 { 615 std::unique_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module_sp, data_sp, process_sp, header_addr)); 616 if (objfile_ap.get() && objfile_ap->ParseHeader()) 617 return objfile_ap.release(); 618 } 619 return NULL; 620 } 621 622 size_t 623 ObjectFileMachO::GetModuleSpecifications (const lldb_private::FileSpec& file, 624 lldb::DataBufferSP& data_sp, 625 lldb::offset_t data_offset, 626 lldb::offset_t file_offset, 627 lldb::offset_t length, 628 lldb_private::ModuleSpecList &specs) 629 { 630 const size_t initial_count = specs.GetSize(); 631 632 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) 633 { 634 DataExtractor data; 635 data.SetData(data_sp); 636 llvm::MachO::mach_header header; 637 if (ParseHeader (data, &data_offset, header)) 638 { 639 size_t header_and_load_cmds = header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 640 if (header_and_load_cmds >= data_sp->GetByteSize()) 641 { 642 data_sp = file.ReadFileContents(file_offset, header_and_load_cmds); 643 data.SetData(data_sp); 644 data_offset = MachHeaderSizeFromMagic(header.magic); 645 } 646 if (data_sp) 647 { 648 ModuleSpec spec; 649 spec.GetFileSpec() = file; 650 spec.GetArchitecture().SetArchitecture(eArchTypeMachO, 651 header.cputype, 652 header.cpusubtype); 653 if (header.filetype == MH_PRELOAD) // 0x5u 654 { 655 // Set OS to "unknown" - this is a standalone binary with no dyld et al 656 spec.GetArchitecture().GetTriple().setOS (llvm::Triple::UnknownOS); 657 } 658 if (spec.GetArchitecture().IsValid()) 659 { 660 GetUUID (header, data, data_offset, spec.GetUUID()); 661 specs.Append(spec); 662 } 663 } 664 } 665 } 666 return specs.GetSize() - initial_count; 667 } 668 669 670 671 const ConstString & 672 ObjectFileMachO::GetSegmentNameTEXT() 673 { 674 static ConstString g_segment_name_TEXT ("__TEXT"); 675 return g_segment_name_TEXT; 676 } 677 678 const ConstString & 679 ObjectFileMachO::GetSegmentNameDATA() 680 { 681 static ConstString g_segment_name_DATA ("__DATA"); 682 return g_segment_name_DATA; 683 } 684 685 const ConstString & 686 ObjectFileMachO::GetSegmentNameOBJC() 687 { 688 static ConstString g_segment_name_OBJC ("__OBJC"); 689 return g_segment_name_OBJC; 690 } 691 692 const ConstString & 693 ObjectFileMachO::GetSegmentNameLINKEDIT() 694 { 695 static ConstString g_section_name_LINKEDIT ("__LINKEDIT"); 696 return g_section_name_LINKEDIT; 697 } 698 699 const ConstString & 700 ObjectFileMachO::GetSectionNameEHFrame() 701 { 702 static ConstString g_section_name_eh_frame ("__eh_frame"); 703 return g_section_name_eh_frame; 704 } 705 706 bool 707 ObjectFileMachO::MagicBytesMatch (DataBufferSP& data_sp, 708 lldb::addr_t data_offset, 709 lldb::addr_t data_length) 710 { 711 DataExtractor data; 712 data.SetData (data_sp, data_offset, data_length); 713 lldb::offset_t offset = 0; 714 uint32_t magic = data.GetU32(&offset); 715 return MachHeaderSizeFromMagic(magic) != 0; 716 } 717 718 719 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 720 DataBufferSP& data_sp, 721 lldb::offset_t data_offset, 722 const FileSpec* file, 723 lldb::offset_t file_offset, 724 lldb::offset_t length) : 725 ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 726 m_mach_segments(), 727 m_mach_sections(), 728 m_entry_point_address(), 729 m_thread_context_offsets(), 730 m_thread_context_offsets_valid(false) 731 { 732 ::memset (&m_header, 0, sizeof(m_header)); 733 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 734 } 735 736 ObjectFileMachO::ObjectFileMachO (const lldb::ModuleSP &module_sp, 737 lldb::DataBufferSP& header_data_sp, 738 const lldb::ProcessSP &process_sp, 739 lldb::addr_t header_addr) : 740 ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 741 m_mach_segments(), 742 m_mach_sections(), 743 m_entry_point_address(), 744 m_thread_context_offsets(), 745 m_thread_context_offsets_valid(false) 746 { 747 ::memset (&m_header, 0, sizeof(m_header)); 748 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 749 } 750 751 ObjectFileMachO::~ObjectFileMachO() 752 { 753 } 754 755 bool 756 ObjectFileMachO::ParseHeader (DataExtractor &data, 757 lldb::offset_t *data_offset_ptr, 758 llvm::MachO::mach_header &header) 759 { 760 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 761 // Leave magic in the original byte order 762 header.magic = data.GetU32(data_offset_ptr); 763 bool can_parse = false; 764 bool is_64_bit = false; 765 switch (header.magic) 766 { 767 case MH_MAGIC: 768 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 769 data.SetAddressByteSize(4); 770 can_parse = true; 771 break; 772 773 case MH_MAGIC_64: 774 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 775 data.SetAddressByteSize(8); 776 can_parse = true; 777 is_64_bit = true; 778 break; 779 780 case MH_CIGAM: 781 data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 782 data.SetAddressByteSize(4); 783 can_parse = true; 784 break; 785 786 case MH_CIGAM_64: 787 data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 788 data.SetAddressByteSize(8); 789 is_64_bit = true; 790 can_parse = true; 791 break; 792 793 default: 794 break; 795 } 796 797 if (can_parse) 798 { 799 data.GetU32(data_offset_ptr, &header.cputype, 6); 800 if (is_64_bit) 801 *data_offset_ptr += 4; 802 return true; 803 } 804 else 805 { 806 memset(&header, 0, sizeof(header)); 807 } 808 return false; 809 } 810 811 bool 812 ObjectFileMachO::ParseHeader () 813 { 814 ModuleSP module_sp(GetModule()); 815 if (module_sp) 816 { 817 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 818 bool can_parse = false; 819 lldb::offset_t offset = 0; 820 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 821 // Leave magic in the original byte order 822 m_header.magic = m_data.GetU32(&offset); 823 switch (m_header.magic) 824 { 825 case MH_MAGIC: 826 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 827 m_data.SetAddressByteSize(4); 828 can_parse = true; 829 break; 830 831 case MH_MAGIC_64: 832 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 833 m_data.SetAddressByteSize(8); 834 can_parse = true; 835 break; 836 837 case MH_CIGAM: 838 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 839 m_data.SetAddressByteSize(4); 840 can_parse = true; 841 break; 842 843 case MH_CIGAM_64: 844 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 845 m_data.SetAddressByteSize(8); 846 can_parse = true; 847 break; 848 849 default: 850 break; 851 } 852 853 if (can_parse) 854 { 855 m_data.GetU32(&offset, &m_header.cputype, 6); 856 857 ArchSpec mach_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 858 859 // Check if the module has a required architecture 860 const ArchSpec &module_arch = module_sp->GetArchitecture(); 861 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 862 return false; 863 864 if (SetModulesArchitecture (mach_arch)) 865 { 866 const size_t header_and_lc_size = m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 867 if (m_data.GetByteSize() < header_and_lc_size) 868 { 869 DataBufferSP data_sp; 870 ProcessSP process_sp (m_process_wp.lock()); 871 if (process_sp) 872 { 873 data_sp = ReadMemory (process_sp, m_memory_addr, header_and_lc_size); 874 } 875 else 876 { 877 // Read in all only the load command data from the file on disk 878 data_sp = m_file.ReadFileContents(m_file_offset, header_and_lc_size); 879 if (data_sp->GetByteSize() != header_and_lc_size) 880 return false; 881 } 882 if (data_sp) 883 m_data.SetData (data_sp); 884 } 885 } 886 return true; 887 } 888 else 889 { 890 memset(&m_header, 0, sizeof(struct mach_header)); 891 } 892 } 893 return false; 894 } 895 896 897 ByteOrder 898 ObjectFileMachO::GetByteOrder () const 899 { 900 return m_data.GetByteOrder (); 901 } 902 903 bool 904 ObjectFileMachO::IsExecutable() const 905 { 906 return m_header.filetype == MH_EXECUTE; 907 } 908 909 uint32_t 910 ObjectFileMachO::GetAddressByteSize () const 911 { 912 return m_data.GetAddressByteSize (); 913 } 914 915 AddressClass 916 ObjectFileMachO::GetAddressClass (lldb::addr_t file_addr) 917 { 918 Symtab *symtab = GetSymtab(); 919 if (symtab) 920 { 921 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 922 if (symbol) 923 { 924 if (symbol->ValueIsAddress()) 925 { 926 SectionSP section_sp (symbol->GetAddress().GetSection()); 927 if (section_sp) 928 { 929 const lldb::SectionType section_type = section_sp->GetType(); 930 switch (section_type) 931 { 932 case eSectionTypeInvalid: return eAddressClassUnknown; 933 case eSectionTypeCode: 934 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) 935 { 936 // For ARM we have a bit in the n_desc field of the symbol 937 // that tells us ARM/Thumb which is bit 0x0008. 938 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 939 return eAddressClassCodeAlternateISA; 940 } 941 return eAddressClassCode; 942 943 case eSectionTypeContainer: return eAddressClassUnknown; 944 case eSectionTypeData: 945 case eSectionTypeDataCString: 946 case eSectionTypeDataCStringPointers: 947 case eSectionTypeDataSymbolAddress: 948 case eSectionTypeData4: 949 case eSectionTypeData8: 950 case eSectionTypeData16: 951 case eSectionTypeDataPointers: 952 case eSectionTypeZeroFill: 953 case eSectionTypeDataObjCMessageRefs: 954 case eSectionTypeDataObjCCFStrings: 955 return eAddressClassData; 956 case eSectionTypeDebug: 957 case eSectionTypeDWARFDebugAbbrev: 958 case eSectionTypeDWARFDebugAranges: 959 case eSectionTypeDWARFDebugFrame: 960 case eSectionTypeDWARFDebugInfo: 961 case eSectionTypeDWARFDebugLine: 962 case eSectionTypeDWARFDebugLoc: 963 case eSectionTypeDWARFDebugMacInfo: 964 case eSectionTypeDWARFDebugPubNames: 965 case eSectionTypeDWARFDebugPubTypes: 966 case eSectionTypeDWARFDebugRanges: 967 case eSectionTypeDWARFDebugStr: 968 case eSectionTypeDWARFAppleNames: 969 case eSectionTypeDWARFAppleTypes: 970 case eSectionTypeDWARFAppleNamespaces: 971 case eSectionTypeDWARFAppleObjC: 972 return eAddressClassDebug; 973 case eSectionTypeEHFrame: return eAddressClassRuntime; 974 case eSectionTypeELFSymbolTable: 975 case eSectionTypeELFDynamicSymbols: 976 case eSectionTypeELFRelocationEntries: 977 case eSectionTypeELFDynamicLinkInfo: 978 case eSectionTypeOther: return eAddressClassUnknown; 979 } 980 } 981 } 982 983 const SymbolType symbol_type = symbol->GetType(); 984 switch (symbol_type) 985 { 986 case eSymbolTypeAny: return eAddressClassUnknown; 987 case eSymbolTypeAbsolute: return eAddressClassUnknown; 988 989 case eSymbolTypeCode: 990 case eSymbolTypeTrampoline: 991 case eSymbolTypeResolver: 992 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) 993 { 994 // For ARM we have a bit in the n_desc field of the symbol 995 // that tells us ARM/Thumb which is bit 0x0008. 996 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 997 return eAddressClassCodeAlternateISA; 998 } 999 return eAddressClassCode; 1000 1001 case eSymbolTypeData: return eAddressClassData; 1002 case eSymbolTypeRuntime: return eAddressClassRuntime; 1003 case eSymbolTypeException: return eAddressClassRuntime; 1004 case eSymbolTypeSourceFile: return eAddressClassDebug; 1005 case eSymbolTypeHeaderFile: return eAddressClassDebug; 1006 case eSymbolTypeObjectFile: return eAddressClassDebug; 1007 case eSymbolTypeCommonBlock: return eAddressClassDebug; 1008 case eSymbolTypeBlock: return eAddressClassDebug; 1009 case eSymbolTypeLocal: return eAddressClassData; 1010 case eSymbolTypeParam: return eAddressClassData; 1011 case eSymbolTypeVariable: return eAddressClassData; 1012 case eSymbolTypeVariableType: return eAddressClassDebug; 1013 case eSymbolTypeLineEntry: return eAddressClassDebug; 1014 case eSymbolTypeLineHeader: return eAddressClassDebug; 1015 case eSymbolTypeScopeBegin: return eAddressClassDebug; 1016 case eSymbolTypeScopeEnd: return eAddressClassDebug; 1017 case eSymbolTypeAdditional: return eAddressClassUnknown; 1018 case eSymbolTypeCompiler: return eAddressClassDebug; 1019 case eSymbolTypeInstrumentation:return eAddressClassDebug; 1020 case eSymbolTypeUndefined: return eAddressClassUnknown; 1021 case eSymbolTypeObjCClass: return eAddressClassRuntime; 1022 case eSymbolTypeObjCMetaClass: return eAddressClassRuntime; 1023 case eSymbolTypeObjCIVar: return eAddressClassRuntime; 1024 case eSymbolTypeReExported: return eAddressClassRuntime; 1025 } 1026 } 1027 } 1028 return eAddressClassUnknown; 1029 } 1030 1031 Symtab * 1032 ObjectFileMachO::GetSymtab() 1033 { 1034 ModuleSP module_sp(GetModule()); 1035 if (module_sp) 1036 { 1037 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 1038 if (m_symtab_ap.get() == NULL) 1039 { 1040 m_symtab_ap.reset(new Symtab(this)); 1041 Mutex::Locker symtab_locker (m_symtab_ap->GetMutex()); 1042 ParseSymtab (); 1043 m_symtab_ap->Finalize (); 1044 } 1045 } 1046 return m_symtab_ap.get(); 1047 } 1048 1049 bool 1050 ObjectFileMachO::IsStripped () 1051 { 1052 if (m_dysymtab.cmd == 0) 1053 { 1054 ModuleSP module_sp(GetModule()); 1055 if (module_sp) 1056 { 1057 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1058 for (uint32_t i=0; i<m_header.ncmds; ++i) 1059 { 1060 const lldb::offset_t load_cmd_offset = offset; 1061 1062 load_command lc; 1063 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 1064 break; 1065 if (lc.cmd == LC_DYSYMTAB) 1066 { 1067 m_dysymtab.cmd = lc.cmd; 1068 m_dysymtab.cmdsize = lc.cmdsize; 1069 if (m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == NULL) 1070 { 1071 // Clear m_dysymtab if we were unable to read all items from the load command 1072 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 1073 } 1074 } 1075 offset = load_cmd_offset + lc.cmdsize; 1076 } 1077 } 1078 } 1079 if (m_dysymtab.cmd) 1080 return m_dysymtab.nlocalsym <= 1; 1081 return false; 1082 } 1083 1084 void 1085 ObjectFileMachO::CreateSections (SectionList &unified_section_list) 1086 { 1087 if (!m_sections_ap.get()) 1088 { 1089 m_sections_ap.reset(new SectionList()); 1090 1091 const bool is_dsym = (m_header.filetype == MH_DSYM); 1092 lldb::user_id_t segID = 0; 1093 lldb::user_id_t sectID = 0; 1094 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1095 uint32_t i; 1096 const bool is_core = GetType() == eTypeCoreFile; 1097 //bool dump_sections = false; 1098 ModuleSP module_sp (GetModule()); 1099 // First look up any LC_ENCRYPTION_INFO load commands 1100 typedef RangeArray<uint32_t, uint32_t, 8> EncryptedFileRanges; 1101 EncryptedFileRanges encrypted_file_ranges; 1102 encryption_info_command encryption_cmd; 1103 for (i=0; i<m_header.ncmds; ++i) 1104 { 1105 const lldb::offset_t load_cmd_offset = offset; 1106 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1107 break; 1108 1109 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO) 1110 { 1111 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) 1112 { 1113 if (encryption_cmd.cryptid != 0) 1114 { 1115 EncryptedFileRanges::Entry entry; 1116 entry.SetRangeBase(encryption_cmd.cryptoff); 1117 entry.SetByteSize(encryption_cmd.cryptsize); 1118 encrypted_file_ranges.Append(entry); 1119 } 1120 } 1121 } 1122 offset = load_cmd_offset + encryption_cmd.cmdsize; 1123 } 1124 1125 offset = MachHeaderSizeFromMagic(m_header.magic); 1126 1127 struct segment_command_64 load_cmd; 1128 for (i=0; i<m_header.ncmds; ++i) 1129 { 1130 const lldb::offset_t load_cmd_offset = offset; 1131 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1132 break; 1133 1134 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1135 { 1136 if (m_data.GetU8(&offset, (uint8_t*)load_cmd.segname, 16)) 1137 { 1138 bool add_section = true; 1139 bool add_to_unified = true; 1140 ConstString const_segname (load_cmd.segname, std::min<size_t>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 1141 1142 SectionSP unified_section_sp(unified_section_list.FindSectionByName(const_segname)); 1143 if (is_dsym && unified_section_sp) 1144 { 1145 if (const_segname == GetSegmentNameLINKEDIT()) 1146 { 1147 // We need to keep the __LINKEDIT segment private to this object file only 1148 add_to_unified = false; 1149 } 1150 else 1151 { 1152 // This is the dSYM file and this section has already been created by 1153 // the object file, no need to create it. 1154 add_section = false; 1155 } 1156 } 1157 load_cmd.vmaddr = m_data.GetAddress(&offset); 1158 load_cmd.vmsize = m_data.GetAddress(&offset); 1159 load_cmd.fileoff = m_data.GetAddress(&offset); 1160 load_cmd.filesize = m_data.GetAddress(&offset); 1161 if (m_length != 0 && load_cmd.filesize != 0) 1162 { 1163 if (load_cmd.fileoff > m_length) 1164 { 1165 // We have a load command that says it extends past the end of hte file. This is likely 1166 // a corrupt file. We don't have any way to return an error condition here (this method 1167 // was likely invokved from something like ObjectFile::GetSectionList()) -- all we can do 1168 // is null out the SectionList vector and if a process has been set up, dump a message 1169 // to stdout. The most common case here is core file debugging with a truncated file. 1170 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1171 module_sp->ReportWarning("load command %u %s has a fileoff (0x%" PRIx64 ") that extends beyond the end of the file (0x%" PRIx64 "), ignoring this section", 1172 i, 1173 lc_segment_name, 1174 load_cmd.fileoff, 1175 m_length); 1176 1177 load_cmd.fileoff = 0; 1178 load_cmd.filesize = 0; 1179 } 1180 1181 if (load_cmd.fileoff + load_cmd.filesize > m_length) 1182 { 1183 // We have a load command that says it extends past the end of hte file. This is likely 1184 // a corrupt file. We don't have any way to return an error condition here (this method 1185 // was likely invokved from something like ObjectFile::GetSectionList()) -- all we can do 1186 // is null out the SectionList vector and if a process has been set up, dump a message 1187 // to stdout. The most common case here is core file debugging with a truncated file. 1188 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1189 GetModule()->ReportWarning("load command %u %s has a fileoff + filesize (0x%" PRIx64 ") that extends beyond the end of the file (0x%" PRIx64 "), the segment will be truncated to match", 1190 i, 1191 lc_segment_name, 1192 load_cmd.fileoff + load_cmd.filesize, 1193 m_length); 1194 1195 // Tuncase the length 1196 load_cmd.filesize = m_length - load_cmd.fileoff; 1197 } 1198 } 1199 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1200 { 1201 1202 const bool segment_is_encrypted = (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1203 1204 // Keep a list of mach segments around in case we need to 1205 // get at data that isn't stored in the abstracted Sections. 1206 m_mach_segments.push_back (load_cmd); 1207 1208 // Use a segment ID of the segment index shifted left by 8 so they 1209 // never conflict with any of the sections. 1210 SectionSP segment_sp; 1211 if (add_section && (const_segname || is_core)) 1212 { 1213 segment_sp.reset(new Section (module_sp, // Module to which this section belongs 1214 this, // Object file to which this sections belongs 1215 ++segID << 8, // Section ID is the 1 based segment index shifted right by 8 bits as not to collide with any of the 256 section IDs that are possible 1216 const_segname, // Name of this section 1217 eSectionTypeContainer, // This section is a container of other sections. 1218 load_cmd.vmaddr, // File VM address == addresses as they are found in the object file 1219 load_cmd.vmsize, // VM size in bytes of this section 1220 load_cmd.fileoff, // Offset to the data for this section in the file 1221 load_cmd.filesize, // Size in bytes of this section as found in the the file 1222 load_cmd.flags)); // Flags for this section 1223 1224 segment_sp->SetIsEncrypted (segment_is_encrypted); 1225 m_sections_ap->AddSection(segment_sp); 1226 if (add_to_unified) 1227 unified_section_list.AddSection(segment_sp); 1228 } 1229 else if (unified_section_sp) 1230 { 1231 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) 1232 { 1233 // Check to see if the module was read from memory? 1234 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) 1235 { 1236 // We have a module that is in memory and needs to have its 1237 // file address adjusted. We need to do this because when we 1238 // load a file from memory, its addresses will be slid already, 1239 // yet the addresses in the new symbol file will still be unslid. 1240 // Since everything is stored as section offset, this shouldn't 1241 // cause any problems. 1242 1243 // Make sure we've parsed the symbol table from the 1244 // ObjectFile before we go around changing its Sections. 1245 module_sp->GetObjectFile()->GetSymtab(); 1246 // eh_frame would present the same problems but we parse that on 1247 // a per-function basis as-needed so it's more difficult to 1248 // remove its use of the Sections. Realistically, the environments 1249 // where this code path will be taken will not have eh_frame sections. 1250 1251 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1252 } 1253 } 1254 m_sections_ap->AddSection(unified_section_sp); 1255 } 1256 1257 struct section_64 sect64; 1258 ::memset (§64, 0, sizeof(sect64)); 1259 // Push a section into our mach sections for the section at 1260 // index zero (NO_SECT) if we don't have any mach sections yet... 1261 if (m_mach_sections.empty()) 1262 m_mach_sections.push_back(sect64); 1263 uint32_t segment_sect_idx; 1264 const lldb::user_id_t first_segment_sectID = sectID + 1; 1265 1266 1267 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1268 for (segment_sect_idx=0; segment_sect_idx<load_cmd.nsects; ++segment_sect_idx) 1269 { 1270 if (m_data.GetU8(&offset, (uint8_t*)sect64.sectname, sizeof(sect64.sectname)) == NULL) 1271 break; 1272 if (m_data.GetU8(&offset, (uint8_t*)sect64.segname, sizeof(sect64.segname)) == NULL) 1273 break; 1274 sect64.addr = m_data.GetAddress(&offset); 1275 sect64.size = m_data.GetAddress(&offset); 1276 1277 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1278 break; 1279 1280 // Keep a list of mach sections around in case we need to 1281 // get at data that isn't stored in the abstracted Sections. 1282 m_mach_sections.push_back (sect64); 1283 1284 if (add_section) 1285 { 1286 ConstString section_name (sect64.sectname, std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 1287 if (!const_segname) 1288 { 1289 // We have a segment with no name so we need to conjure up 1290 // segments that correspond to the section's segname if there 1291 // isn't already such a section. If there is such a section, 1292 // we resize the section so that it spans all sections. 1293 // We also mark these sections as fake so address matches don't 1294 // hit if they land in the gaps between the child sections. 1295 const_segname.SetTrimmedCStringWithLength(sect64.segname, sizeof(sect64.segname)); 1296 segment_sp = unified_section_list.FindSectionByName (const_segname); 1297 if (segment_sp.get()) 1298 { 1299 Section *segment = segment_sp.get(); 1300 // Grow the section size as needed. 1301 const lldb::addr_t sect64_min_addr = sect64.addr; 1302 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1303 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1304 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1305 const lldb::addr_t curr_seg_max_addr = curr_seg_min_addr + curr_seg_byte_size; 1306 if (sect64_min_addr >= curr_seg_min_addr) 1307 { 1308 const lldb::addr_t new_seg_byte_size = sect64_max_addr - curr_seg_min_addr; 1309 // Only grow the section size if needed 1310 if (new_seg_byte_size > curr_seg_byte_size) 1311 segment->SetByteSize (new_seg_byte_size); 1312 } 1313 else 1314 { 1315 // We need to change the base address of the segment and 1316 // adjust the child section offsets for all existing children. 1317 const lldb::addr_t slide_amount = sect64_min_addr - curr_seg_min_addr; 1318 segment->Slide(slide_amount, false); 1319 segment->GetChildren().Slide(-slide_amount, false); 1320 segment->SetByteSize (curr_seg_max_addr - sect64_min_addr); 1321 } 1322 1323 // Grow the section size as needed. 1324 if (sect64.offset) 1325 { 1326 const lldb::addr_t segment_min_file_offset = segment->GetFileOffset(); 1327 const lldb::addr_t segment_max_file_offset = segment_min_file_offset + segment->GetFileSize(); 1328 1329 const lldb::addr_t section_min_file_offset = sect64.offset; 1330 const lldb::addr_t section_max_file_offset = section_min_file_offset + sect64.size; 1331 const lldb::addr_t new_file_offset = std::min (section_min_file_offset, segment_min_file_offset); 1332 const lldb::addr_t new_file_size = std::max (section_max_file_offset, segment_max_file_offset) - new_file_offset; 1333 segment->SetFileOffset (new_file_offset); 1334 segment->SetFileSize (new_file_size); 1335 } 1336 } 1337 else 1338 { 1339 // Create a fake section for the section's named segment 1340 segment_sp.reset(new Section (segment_sp, // Parent section 1341 module_sp, // Module to which this section belongs 1342 this, // Object file to which this section belongs 1343 ++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 1344 const_segname, // Name of this section 1345 eSectionTypeContainer, // This section is a container of other sections. 1346 sect64.addr, // File VM address == addresses as they are found in the object file 1347 sect64.size, // VM size in bytes of this section 1348 sect64.offset, // Offset to the data for this section in the file 1349 sect64.offset ? sect64.size : 0, // Size in bytes of this section as found in the the file 1350 load_cmd.flags)); // Flags for this section 1351 segment_sp->SetIsFake(true); 1352 1353 m_sections_ap->AddSection(segment_sp); 1354 if (add_to_unified) 1355 unified_section_list.AddSection(segment_sp); 1356 segment_sp->SetIsEncrypted (segment_is_encrypted); 1357 } 1358 } 1359 assert (segment_sp.get()); 1360 1361 uint32_t mach_sect_type = sect64.flags & SECTION_TYPE; 1362 static ConstString g_sect_name_objc_data ("__objc_data"); 1363 static ConstString g_sect_name_objc_msgrefs ("__objc_msgrefs"); 1364 static ConstString g_sect_name_objc_selrefs ("__objc_selrefs"); 1365 static ConstString g_sect_name_objc_classrefs ("__objc_classrefs"); 1366 static ConstString g_sect_name_objc_superrefs ("__objc_superrefs"); 1367 static ConstString g_sect_name_objc_const ("__objc_const"); 1368 static ConstString g_sect_name_objc_classlist ("__objc_classlist"); 1369 static ConstString g_sect_name_cfstring ("__cfstring"); 1370 1371 static ConstString g_sect_name_dwarf_debug_abbrev ("__debug_abbrev"); 1372 static ConstString g_sect_name_dwarf_debug_aranges ("__debug_aranges"); 1373 static ConstString g_sect_name_dwarf_debug_frame ("__debug_frame"); 1374 static ConstString g_sect_name_dwarf_debug_info ("__debug_info"); 1375 static ConstString g_sect_name_dwarf_debug_line ("__debug_line"); 1376 static ConstString g_sect_name_dwarf_debug_loc ("__debug_loc"); 1377 static ConstString g_sect_name_dwarf_debug_macinfo ("__debug_macinfo"); 1378 static ConstString g_sect_name_dwarf_debug_pubnames ("__debug_pubnames"); 1379 static ConstString g_sect_name_dwarf_debug_pubtypes ("__debug_pubtypes"); 1380 static ConstString g_sect_name_dwarf_debug_ranges ("__debug_ranges"); 1381 static ConstString g_sect_name_dwarf_debug_str ("__debug_str"); 1382 static ConstString g_sect_name_dwarf_apple_names ("__apple_names"); 1383 static ConstString g_sect_name_dwarf_apple_types ("__apple_types"); 1384 static ConstString g_sect_name_dwarf_apple_namespaces ("__apple_namespac"); 1385 static ConstString g_sect_name_dwarf_apple_objc ("__apple_objc"); 1386 static ConstString g_sect_name_eh_frame ("__eh_frame"); 1387 static ConstString g_sect_name_DATA ("__DATA"); 1388 static ConstString g_sect_name_TEXT ("__TEXT"); 1389 1390 lldb::SectionType sect_type = eSectionTypeOther; 1391 1392 if (section_name == g_sect_name_dwarf_debug_abbrev) 1393 sect_type = eSectionTypeDWARFDebugAbbrev; 1394 else if (section_name == g_sect_name_dwarf_debug_aranges) 1395 sect_type = eSectionTypeDWARFDebugAranges; 1396 else if (section_name == g_sect_name_dwarf_debug_frame) 1397 sect_type = eSectionTypeDWARFDebugFrame; 1398 else if (section_name == g_sect_name_dwarf_debug_info) 1399 sect_type = eSectionTypeDWARFDebugInfo; 1400 else if (section_name == g_sect_name_dwarf_debug_line) 1401 sect_type = eSectionTypeDWARFDebugLine; 1402 else if (section_name == g_sect_name_dwarf_debug_loc) 1403 sect_type = eSectionTypeDWARFDebugLoc; 1404 else if (section_name == g_sect_name_dwarf_debug_macinfo) 1405 sect_type = eSectionTypeDWARFDebugMacInfo; 1406 else if (section_name == g_sect_name_dwarf_debug_pubnames) 1407 sect_type = eSectionTypeDWARFDebugPubNames; 1408 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 1409 sect_type = eSectionTypeDWARFDebugPubTypes; 1410 else if (section_name == g_sect_name_dwarf_debug_ranges) 1411 sect_type = eSectionTypeDWARFDebugRanges; 1412 else if (section_name == g_sect_name_dwarf_debug_str) 1413 sect_type = eSectionTypeDWARFDebugStr; 1414 else if (section_name == g_sect_name_dwarf_apple_names) 1415 sect_type = eSectionTypeDWARFAppleNames; 1416 else if (section_name == g_sect_name_dwarf_apple_types) 1417 sect_type = eSectionTypeDWARFAppleTypes; 1418 else if (section_name == g_sect_name_dwarf_apple_namespaces) 1419 sect_type = eSectionTypeDWARFAppleNamespaces; 1420 else if (section_name == g_sect_name_dwarf_apple_objc) 1421 sect_type = eSectionTypeDWARFAppleObjC; 1422 else if (section_name == g_sect_name_objc_selrefs) 1423 sect_type = eSectionTypeDataCStringPointers; 1424 else if (section_name == g_sect_name_objc_msgrefs) 1425 sect_type = eSectionTypeDataObjCMessageRefs; 1426 else if (section_name == g_sect_name_eh_frame) 1427 sect_type = eSectionTypeEHFrame; 1428 else if (section_name == g_sect_name_cfstring) 1429 sect_type = eSectionTypeDataObjCCFStrings; 1430 else if (section_name == g_sect_name_objc_data || 1431 section_name == g_sect_name_objc_classrefs || 1432 section_name == g_sect_name_objc_superrefs || 1433 section_name == g_sect_name_objc_const || 1434 section_name == g_sect_name_objc_classlist) 1435 { 1436 sect_type = eSectionTypeDataPointers; 1437 } 1438 1439 if (sect_type == eSectionTypeOther) 1440 { 1441 switch (mach_sect_type) 1442 { 1443 // TODO: categorize sections by other flags for regular sections 1444 case S_REGULAR: 1445 if (segment_sp->GetName() == g_sect_name_TEXT) 1446 sect_type = eSectionTypeCode; 1447 else if (segment_sp->GetName() == g_sect_name_DATA) 1448 sect_type = eSectionTypeData; 1449 else 1450 sect_type = eSectionTypeOther; 1451 break; 1452 case S_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1453 case S_CSTRING_LITERALS: sect_type = eSectionTypeDataCString; break; // section with only literal C strings 1454 case S_4BYTE_LITERALS: sect_type = eSectionTypeData4; break; // section with only 4 byte literals 1455 case S_8BYTE_LITERALS: sect_type = eSectionTypeData8; break; // section with only 8 byte literals 1456 case S_LITERAL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only pointers to literals 1457 case S_NON_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only non-lazy symbol pointers 1458 case S_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only lazy symbol pointers 1459 case S_SYMBOL_STUBS: sect_type = eSectionTypeCode; break; // section with only symbol stubs, byte size of stub in the reserved2 field 1460 case S_MOD_INIT_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for initialization 1461 case S_MOD_TERM_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for termination 1462 case S_COALESCED: sect_type = eSectionTypeOther; break; 1463 case S_GB_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1464 case S_INTERPOSING: sect_type = eSectionTypeCode; break; // section with only pairs of function pointers for interposing 1465 case S_16BYTE_LITERALS: sect_type = eSectionTypeData16; break; // section with only 16 byte literals 1466 case S_DTRACE_DOF: sect_type = eSectionTypeDebug; break; 1467 case S_LAZY_DYLIB_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; 1468 default: break; 1469 } 1470 } 1471 1472 SectionSP section_sp(new Section (segment_sp, 1473 module_sp, 1474 this, 1475 ++sectID, 1476 section_name, 1477 sect_type, 1478 sect64.addr - segment_sp->GetFileAddress(), 1479 sect64.size, 1480 sect64.offset, 1481 sect64.offset == 0 ? 0 : sect64.size, 1482 sect64.flags)); 1483 // Set the section to be encrypted to match the segment 1484 1485 bool section_is_encrypted = false; 1486 if (!segment_is_encrypted && load_cmd.filesize != 0) 1487 section_is_encrypted = encrypted_file_ranges.FindEntryThatContains(sect64.offset) != NULL; 1488 1489 section_sp->SetIsEncrypted (segment_is_encrypted || section_is_encrypted); 1490 segment_sp->GetChildren().AddSection(section_sp); 1491 1492 if (segment_sp->IsFake()) 1493 { 1494 segment_sp.reset(); 1495 const_segname.Clear(); 1496 } 1497 } 1498 } 1499 if (segment_sp && is_dsym) 1500 { 1501 if (first_segment_sectID <= sectID) 1502 { 1503 lldb::user_id_t sect_uid; 1504 for (sect_uid = first_segment_sectID; sect_uid <= sectID; ++sect_uid) 1505 { 1506 SectionSP curr_section_sp(segment_sp->GetChildren().FindSectionByID (sect_uid)); 1507 SectionSP next_section_sp; 1508 if (sect_uid + 1 <= sectID) 1509 next_section_sp = segment_sp->GetChildren().FindSectionByID (sect_uid+1); 1510 1511 if (curr_section_sp.get()) 1512 { 1513 if (curr_section_sp->GetByteSize() == 0) 1514 { 1515 if (next_section_sp.get() != NULL) 1516 curr_section_sp->SetByteSize ( next_section_sp->GetFileAddress() - curr_section_sp->GetFileAddress() ); 1517 else 1518 curr_section_sp->SetByteSize ( load_cmd.vmsize ); 1519 } 1520 } 1521 } 1522 } 1523 } 1524 } 1525 } 1526 } 1527 else if (load_cmd.cmd == LC_DYSYMTAB) 1528 { 1529 m_dysymtab.cmd = load_cmd.cmd; 1530 m_dysymtab.cmdsize = load_cmd.cmdsize; 1531 m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1532 } 1533 1534 offset = load_cmd_offset + load_cmd.cmdsize; 1535 } 1536 1537 // StreamFile s(stdout, false); // REMOVE THIS LINE 1538 // s.Printf ("Sections for %s:\n", m_file.GetPath().c_str());// REMOVE THIS LINE 1539 // m_sections_ap->Dump(&s, NULL, true, UINT32_MAX);// REMOVE THIS LINE 1540 } 1541 } 1542 1543 class MachSymtabSectionInfo 1544 { 1545 public: 1546 1547 MachSymtabSectionInfo (SectionList *section_list) : 1548 m_section_list (section_list), 1549 m_section_infos() 1550 { 1551 // Get the number of sections down to a depth of 1 to include 1552 // all segments and their sections, but no other sections that 1553 // may be added for debug map or 1554 m_section_infos.resize(section_list->GetNumSections(1)); 1555 } 1556 1557 1558 SectionSP 1559 GetSection (uint8_t n_sect, addr_t file_addr) 1560 { 1561 if (n_sect == 0) 1562 return SectionSP(); 1563 if (n_sect < m_section_infos.size()) 1564 { 1565 if (!m_section_infos[n_sect].section_sp) 1566 { 1567 SectionSP section_sp (m_section_list->FindSectionByID (n_sect)); 1568 m_section_infos[n_sect].section_sp = section_sp; 1569 if (section_sp) 1570 { 1571 m_section_infos[n_sect].vm_range.SetBaseAddress (section_sp->GetFileAddress()); 1572 m_section_infos[n_sect].vm_range.SetByteSize (section_sp->GetByteSize()); 1573 } 1574 else 1575 { 1576 Host::SystemLog (Host::eSystemLogError, "error: unable to find section for section %u\n", n_sect); 1577 } 1578 } 1579 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) 1580 { 1581 // Symbol is in section. 1582 return m_section_infos[n_sect].section_sp; 1583 } 1584 else if (m_section_infos[n_sect].vm_range.GetByteSize () == 0 && 1585 m_section_infos[n_sect].vm_range.GetBaseAddress() == file_addr) 1586 { 1587 // Symbol is in section with zero size, but has the same start 1588 // address as the section. This can happen with linker symbols 1589 // (symbols that start with the letter 'l' or 'L'. 1590 return m_section_infos[n_sect].section_sp; 1591 } 1592 } 1593 return m_section_list->FindSectionContainingFileAddress(file_addr); 1594 } 1595 1596 protected: 1597 struct SectionInfo 1598 { 1599 SectionInfo () : 1600 vm_range(), 1601 section_sp () 1602 { 1603 } 1604 1605 VMRange vm_range; 1606 SectionSP section_sp; 1607 }; 1608 SectionList *m_section_list; 1609 std::vector<SectionInfo> m_section_infos; 1610 }; 1611 1612 struct TrieEntry 1613 { 1614 TrieEntry () : 1615 name(), 1616 address(LLDB_INVALID_ADDRESS), 1617 flags (0), 1618 other(0), 1619 import_name() 1620 { 1621 } 1622 1623 void 1624 Clear () 1625 { 1626 name.Clear(); 1627 address = LLDB_INVALID_ADDRESS; 1628 flags = 0; 1629 other = 0; 1630 import_name.Clear(); 1631 } 1632 1633 void 1634 Dump () const 1635 { 1636 printf ("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1637 static_cast<unsigned long long>(address), 1638 static_cast<unsigned long long>(flags), 1639 static_cast<unsigned long long>(other), name.GetCString()); 1640 if (import_name) 1641 printf (" -> \"%s\"\n", import_name.GetCString()); 1642 else 1643 printf ("\n"); 1644 } 1645 ConstString name; 1646 uint64_t address; 1647 uint64_t flags; 1648 uint64_t other; 1649 ConstString import_name; 1650 }; 1651 1652 struct TrieEntryWithOffset 1653 { 1654 lldb::offset_t nodeOffset; 1655 TrieEntry entry; 1656 1657 TrieEntryWithOffset (lldb::offset_t offset) : 1658 nodeOffset (offset), 1659 entry() 1660 { 1661 } 1662 1663 void 1664 Dump (uint32_t idx) const 1665 { 1666 printf ("[%3u] 0x%16.16llx: ", idx, 1667 static_cast<unsigned long long>(nodeOffset)); 1668 entry.Dump(); 1669 } 1670 1671 bool 1672 operator<(const TrieEntryWithOffset& other) const 1673 { 1674 return ( nodeOffset < other.nodeOffset ); 1675 } 1676 }; 1677 1678 static void 1679 ParseTrieEntries (DataExtractor &data, 1680 lldb::offset_t offset, 1681 std::vector<llvm::StringRef> &nameSlices, 1682 std::set<lldb::addr_t> &resolver_addresses, 1683 std::vector<TrieEntryWithOffset>& output) 1684 { 1685 if (!data.ValidOffset(offset)) 1686 return; 1687 1688 const uint64_t terminalSize = data.GetULEB128(&offset); 1689 lldb::offset_t children_offset = offset + terminalSize; 1690 if ( terminalSize != 0 ) { 1691 TrieEntryWithOffset e (offset); 1692 e.entry.flags = data.GetULEB128(&offset); 1693 const char *import_name = NULL; 1694 if ( e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT ) { 1695 e.entry.address = 0; 1696 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 1697 import_name = data.GetCStr(&offset); 1698 } 1699 else { 1700 e.entry.address = data.GetULEB128(&offset); 1701 if ( e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER ) 1702 { 1703 //resolver_addresses.insert(e.entry.address); 1704 e.entry.other = data.GetULEB128(&offset); 1705 resolver_addresses.insert(e.entry.other); 1706 } 1707 else 1708 e.entry.other = 0; 1709 } 1710 // Only add symbols that are reexport symbols with a valid import name 1711 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && import_name[0]) 1712 { 1713 std::string name; 1714 if (!nameSlices.empty()) 1715 { 1716 for (auto name_slice: nameSlices) 1717 name.append(name_slice.data(), name_slice.size()); 1718 } 1719 if (name.size() > 1) 1720 { 1721 // Skip the leading '_' 1722 e.entry.name.SetCStringWithLength(name.c_str() + 1,name.size() - 1); 1723 } 1724 if (import_name) 1725 { 1726 // Skip the leading '_' 1727 e.entry.import_name.SetCString(import_name+1); 1728 } 1729 output.push_back(e); 1730 } 1731 } 1732 1733 const uint8_t childrenCount = data.GetU8(&children_offset); 1734 for (uint8_t i=0; i < childrenCount; ++i) { 1735 nameSlices.push_back(data.GetCStr(&children_offset)); 1736 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 1737 if (childNodeOffset) 1738 { 1739 ParseTrieEntries(data, 1740 childNodeOffset, 1741 nameSlices, 1742 resolver_addresses, 1743 output); 1744 } 1745 nameSlices.pop_back(); 1746 } 1747 } 1748 1749 size_t 1750 ObjectFileMachO::ParseSymtab () 1751 { 1752 Timer scoped_timer(__PRETTY_FUNCTION__, 1753 "ObjectFileMachO::ParseSymtab () module = %s", 1754 m_file.GetFilename().AsCString("")); 1755 ModuleSP module_sp (GetModule()); 1756 if (!module_sp) 1757 return 0; 1758 1759 struct symtab_command symtab_load_command = { 0, 0, 0, 0, 0, 0 }; 1760 struct linkedit_data_command function_starts_load_command = { 0, 0, 0, 0 }; 1761 struct dyld_info_command dyld_info = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }; 1762 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 1763 FunctionStarts function_starts; 1764 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1765 uint32_t i; 1766 FileSpecList dylib_files; 1767 Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_SYMBOLS)); 1768 1769 for (i=0; i<m_header.ncmds; ++i) 1770 { 1771 const lldb::offset_t cmd_offset = offset; 1772 // Read in the load command and load command size 1773 struct load_command lc; 1774 if (m_data.GetU32(&offset, &lc, 2) == NULL) 1775 break; 1776 // Watch for the symbol table load command 1777 switch (lc.cmd) 1778 { 1779 case LC_SYMTAB: 1780 symtab_load_command.cmd = lc.cmd; 1781 symtab_load_command.cmdsize = lc.cmdsize; 1782 // Read in the rest of the symtab load command 1783 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 0) // fill in symoff, nsyms, stroff, strsize fields 1784 return 0; 1785 if (symtab_load_command.symoff == 0) 1786 { 1787 if (log) 1788 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 1789 return 0; 1790 } 1791 1792 if (symtab_load_command.stroff == 0) 1793 { 1794 if (log) 1795 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 1796 return 0; 1797 } 1798 1799 if (symtab_load_command.nsyms == 0) 1800 { 1801 if (log) 1802 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 1803 return 0; 1804 } 1805 1806 if (symtab_load_command.strsize == 0) 1807 { 1808 if (log) 1809 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 1810 return 0; 1811 } 1812 break; 1813 1814 case LC_DYLD_INFO: 1815 case LC_DYLD_INFO_ONLY: 1816 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) 1817 { 1818 dyld_info.cmd = lc.cmd; 1819 dyld_info.cmdsize = lc.cmdsize; 1820 } 1821 else 1822 { 1823 memset (&dyld_info, 0, sizeof(dyld_info)); 1824 } 1825 break; 1826 1827 case LC_LOAD_DYLIB: 1828 case LC_LOAD_WEAK_DYLIB: 1829 case LC_REEXPORT_DYLIB: 1830 case LC_LOADFVMLIB: 1831 case LC_LOAD_UPWARD_DYLIB: 1832 { 1833 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 1834 const char *path = m_data.PeekCStr(name_offset); 1835 if (path) 1836 { 1837 FileSpec file_spec(path, false); 1838 // Strip the path if there is @rpath, @executanble, etc so we just use the basename 1839 if (path[0] == '@') 1840 file_spec.GetDirectory().Clear(); 1841 1842 dylib_files.Append(file_spec); 1843 } 1844 } 1845 break; 1846 1847 case LC_FUNCTION_STARTS: 1848 function_starts_load_command.cmd = lc.cmd; 1849 function_starts_load_command.cmdsize = lc.cmdsize; 1850 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == NULL) // fill in symoff, nsyms, stroff, strsize fields 1851 memset (&function_starts_load_command, 0, sizeof(function_starts_load_command)); 1852 break; 1853 1854 default: 1855 break; 1856 } 1857 offset = cmd_offset + lc.cmdsize; 1858 } 1859 1860 if (symtab_load_command.cmd) 1861 { 1862 Symtab *symtab = m_symtab_ap.get(); 1863 SectionList *section_list = GetSectionList(); 1864 if (section_list == NULL) 1865 return 0; 1866 1867 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 1868 const ByteOrder byte_order = m_data.GetByteOrder(); 1869 bool bit_width_32 = addr_byte_size == 4; 1870 const size_t nlist_byte_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 1871 1872 DataExtractor nlist_data (NULL, 0, byte_order, addr_byte_size); 1873 DataExtractor strtab_data (NULL, 0, byte_order, addr_byte_size); 1874 DataExtractor function_starts_data (NULL, 0, byte_order, addr_byte_size); 1875 DataExtractor indirect_symbol_index_data (NULL, 0, byte_order, addr_byte_size); 1876 DataExtractor dyld_trie_data (NULL, 0, byte_order, addr_byte_size); 1877 1878 const addr_t nlist_data_byte_size = symtab_load_command.nsyms * nlist_byte_size; 1879 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 1880 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 1881 1882 ProcessSP process_sp (m_process_wp.lock()); 1883 Process *process = process_sp.get(); 1884 1885 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 1886 1887 if (process) 1888 { 1889 Target &target = process->GetTarget(); 1890 1891 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 1892 1893 SectionSP linkedit_section_sp(section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 1894 // Reading mach file from memory in a process or core file... 1895 1896 if (linkedit_section_sp) 1897 { 1898 const addr_t linkedit_load_addr = linkedit_section_sp->GetLoadBaseAddress(&target); 1899 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 1900 const addr_t symoff_addr = linkedit_load_addr + symtab_load_command.symoff - linkedit_file_offset; 1901 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - linkedit_file_offset; 1902 1903 bool data_was_read = false; 1904 1905 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__)) 1906 if (m_header.flags & 0x80000000u && process->GetAddressByteSize() == sizeof (void*)) 1907 { 1908 // This mach-o memory file is in the dyld shared cache. If this 1909 // program is not remote and this is iOS, then this process will 1910 // share the same shared cache as the process we are debugging and 1911 // we can read the entire __LINKEDIT from the address space in this 1912 // process. This is a needed optimization that is used for local iOS 1913 // debugging only since all shared libraries in the shared cache do 1914 // not have corresponding files that exist in the file system of the 1915 // device. They have been combined into a single file. This means we 1916 // always have to load these files from memory. All of the symbol and 1917 // string tables from all of the __LINKEDIT sections from the shared 1918 // libraries in the shared cache have been merged into a single large 1919 // symbol and string table. Reading all of this symbol and string table 1920 // data across can slow down debug launch times, so we optimize this by 1921 // reading the memory for the __LINKEDIT section from this process. 1922 1923 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 1924 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 1925 bool use_lldb_cache = true; 1926 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && lldb_shared_cache != process_shared_cache) 1927 { 1928 use_lldb_cache = false; 1929 ModuleSP module_sp (GetModule()); 1930 if (module_sp) 1931 module_sp->ReportWarning ("shared cache in process does not match lldb's own shared cache, startup will be slow."); 1932 1933 } 1934 1935 PlatformSP platform_sp (target.GetPlatform()); 1936 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) 1937 { 1938 data_was_read = true; 1939 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, eByteOrderLittle); 1940 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, eByteOrderLittle); 1941 if (function_starts_load_command.cmd) 1942 { 1943 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 1944 function_starts_data.SetData ((void *)func_start_addr, function_starts_load_command.datasize, eByteOrderLittle); 1945 } 1946 } 1947 } 1948 #endif 1949 1950 if (!data_was_read) 1951 { 1952 if (memory_module_load_level == eMemoryModuleLoadLevelComplete) 1953 { 1954 DataBufferSP nlist_data_sp (ReadMemory (process_sp, symoff_addr, nlist_data_byte_size)); 1955 if (nlist_data_sp) 1956 nlist_data.SetData (nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 1957 // Load strings individually from memory when loading from memory since shared cache 1958 // string tables contain strings for all symbols from all shared cached libraries 1959 //DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, strtab_data_byte_size)); 1960 //if (strtab_data_sp) 1961 // strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 1962 if (m_dysymtab.nindirectsyms != 0) 1963 { 1964 const addr_t indirect_syms_addr = linkedit_load_addr + m_dysymtab.indirectsymoff - linkedit_file_offset; 1965 DataBufferSP indirect_syms_data_sp (ReadMemory (process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 1966 if (indirect_syms_data_sp) 1967 indirect_symbol_index_data.SetData (indirect_syms_data_sp, 0, indirect_syms_data_sp->GetByteSize()); 1968 } 1969 } 1970 1971 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) 1972 { 1973 if (function_starts_load_command.cmd) 1974 { 1975 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 1976 DataBufferSP func_start_data_sp (ReadMemory (process_sp, func_start_addr, function_starts_load_command.datasize)); 1977 if (func_start_data_sp) 1978 function_starts_data.SetData (func_start_data_sp, 0, func_start_data_sp->GetByteSize()); 1979 } 1980 } 1981 } 1982 } 1983 } 1984 else 1985 { 1986 nlist_data.SetData (m_data, 1987 symtab_load_command.symoff, 1988 nlist_data_byte_size); 1989 strtab_data.SetData (m_data, 1990 symtab_load_command.stroff, 1991 strtab_data_byte_size); 1992 1993 if (dyld_info.export_size > 0) 1994 { 1995 dyld_trie_data.SetData (m_data, 1996 dyld_info.export_off, 1997 dyld_info.export_size); 1998 } 1999 2000 if (m_dysymtab.nindirectsyms != 0) 2001 { 2002 indirect_symbol_index_data.SetData (m_data, 2003 m_dysymtab.indirectsymoff, 2004 m_dysymtab.nindirectsyms * 4); 2005 } 2006 if (function_starts_load_command.cmd) 2007 { 2008 function_starts_data.SetData (m_data, 2009 function_starts_load_command.dataoff, 2010 function_starts_load_command.datasize); 2011 } 2012 } 2013 2014 if (nlist_data.GetByteSize() == 0 && memory_module_load_level == eMemoryModuleLoadLevelComplete) 2015 { 2016 if (log) 2017 module_sp->LogMessage(log, "failed to read nlist data"); 2018 return 0; 2019 } 2020 2021 2022 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2023 if (!have_strtab_data) 2024 { 2025 if (process) 2026 { 2027 if (strtab_addr == LLDB_INVALID_ADDRESS) 2028 { 2029 if (log) 2030 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2031 return 0; 2032 } 2033 } 2034 else 2035 { 2036 if (log) 2037 module_sp->LogMessage(log, "failed to read strtab data"); 2038 return 0; 2039 } 2040 } 2041 2042 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2043 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2044 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2045 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2046 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT)); 2047 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA)); 2048 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC)); 2049 SectionSP eh_frame_section_sp; 2050 if (text_section_sp.get()) 2051 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame); 2052 else 2053 eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame); 2054 2055 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2056 2057 // lldb works best if it knows the start addresss of all functions in a module. 2058 // Linker symbols or debug info are normally the best source of information for start addr / size but 2059 // they may be stripped in a released binary. 2060 // Two additional sources of information exist in Mach-O binaries: 2061 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each function's start address in the 2062 // binary, relative to the text section. 2063 // eh_frame - the eh_frame FDEs have the start addr & size of each function 2064 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on all modern binaries. 2065 // Binaries built to run on older releases may need to use eh_frame information. 2066 2067 if (text_section_sp && function_starts_data.GetByteSize()) 2068 { 2069 FunctionStarts::Entry function_start_entry; 2070 function_start_entry.data = false; 2071 lldb::offset_t function_start_offset = 0; 2072 function_start_entry.addr = text_section_sp->GetFileAddress(); 2073 uint64_t delta; 2074 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 0) 2075 { 2076 // Now append the current entry 2077 function_start_entry.addr += delta; 2078 function_starts.Append(function_start_entry); 2079 } 2080 } 2081 else 2082 { 2083 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the load command claiming an eh_frame 2084 // but it doesn't actually have the eh_frame content. And if we have a dSYM, we don't need to do any 2085 // of this fill-in-the-missing-symbols works anyway - the debug info should give us all the functions in 2086 // the module. 2087 if (text_section_sp.get() && eh_frame_section_sp.get() && m_type != eTypeDebugInfo) 2088 { 2089 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, eRegisterKindGCC, true); 2090 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2091 eh_frame.GetFunctionAddressAndSizeVector (functions); 2092 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2093 size_t count = functions.GetSize(); 2094 for (size_t i = 0; i < count; ++i) 2095 { 2096 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = functions.GetEntryAtIndex (i); 2097 if (func) 2098 { 2099 FunctionStarts::Entry function_start_entry; 2100 function_start_entry.addr = func->base - text_base_addr; 2101 function_starts.Append(function_start_entry); 2102 } 2103 } 2104 } 2105 } 2106 2107 const size_t function_starts_count = function_starts.GetSize(); 2108 2109 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() : NO_SECT; 2110 2111 lldb::offset_t nlist_data_offset = 0; 2112 2113 uint32_t N_SO_index = UINT32_MAX; 2114 2115 MachSymtabSectionInfo section_info (section_list); 2116 std::vector<uint32_t> N_FUN_indexes; 2117 std::vector<uint32_t> N_NSYM_indexes; 2118 std::vector<uint32_t> N_INCL_indexes; 2119 std::vector<uint32_t> N_BRAC_indexes; 2120 std::vector<uint32_t> N_COMM_indexes; 2121 typedef std::multimap <uint64_t, uint32_t> ValueToSymbolIndexMap; 2122 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2123 typedef std::map <const char *, uint32_t> ConstNameToSymbolIndexMap; 2124 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2125 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2126 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2127 // Any symbols that get merged into another will get an entry 2128 // in this map so we know 2129 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2130 uint32_t nlist_idx = 0; 2131 Symbol *symbol_ptr = NULL; 2132 2133 uint32_t sym_idx = 0; 2134 Symbol *sym = NULL; 2135 size_t num_syms = 0; 2136 std::string memory_symbol_name; 2137 uint32_t unmapped_local_symbols_found = 0; 2138 2139 std::vector<TrieEntryWithOffset> trie_entries; 2140 std::set<lldb::addr_t> resolver_addresses; 2141 2142 if (dyld_trie_data.GetByteSize() > 0) 2143 { 2144 std::vector<llvm::StringRef> nameSlices; 2145 ParseTrieEntries (dyld_trie_data, 2146 0, 2147 nameSlices, 2148 resolver_addresses, 2149 trie_entries); 2150 2151 ConstString text_segment_name ("__TEXT"); 2152 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 2153 if (text_segment_sp) 2154 { 2155 const lldb::addr_t text_segment_file_addr = text_segment_sp->GetFileAddress(); 2156 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) 2157 { 2158 for (auto &e : trie_entries) 2159 e.entry.address += text_segment_file_addr; 2160 } 2161 } 2162 } 2163 2164 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__)) 2165 2166 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been optimized by moving LOCAL 2167 // symbols out of the memory mapped portion of the DSC. The symbol information has all been retained, 2168 // but it isn't available in the normal nlist data. However, there *are* duplicate entries of *some* 2169 // LOCAL symbols in the normal nlist data. To handle this situation correctly, we must first attempt 2170 // to parse any DSC unmapped symbol information. If we find any, we set a flag that tells the normal 2171 // nlist parser to ignore all LOCAL symbols. 2172 2173 if (m_header.flags & 0x80000000u) 2174 { 2175 // Before we can start mapping the DSC, we need to make certain the target process is actually 2176 // using the cache we can find. 2177 2178 // Next we need to determine the correct path for the dyld shared cache. 2179 2180 ArchSpec header_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 2181 char dsc_path[PATH_MAX]; 2182 2183 snprintf(dsc_path, sizeof(dsc_path), "%s%s%s", 2184 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR */ 2185 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2186 header_arch.GetArchitectureName()); 2187 2188 FileSpec dsc_filespec(dsc_path, false); 2189 2190 // We need definitions of two structures in the on-disk DSC, copy them here manually 2191 struct lldb_copy_dyld_cache_header_v0 2192 { 2193 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 2194 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 2195 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 2196 uint32_t imagesOffset; 2197 uint32_t imagesCount; 2198 uint64_t dyldBaseAddress; 2199 uint64_t codeSignatureOffset; 2200 uint64_t codeSignatureSize; 2201 uint64_t slideInfoOffset; 2202 uint64_t slideInfoSize; 2203 uint64_t localSymbolsOffset; // file offset of where local symbols are stored 2204 uint64_t localSymbolsSize; // size of local symbols information 2205 }; 2206 struct lldb_copy_dyld_cache_header_v1 2207 { 2208 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 2209 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 2210 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 2211 uint32_t imagesOffset; 2212 uint32_t imagesCount; 2213 uint64_t dyldBaseAddress; 2214 uint64_t codeSignatureOffset; 2215 uint64_t codeSignatureSize; 2216 uint64_t slideInfoOffset; 2217 uint64_t slideInfoSize; 2218 uint64_t localSymbolsOffset; 2219 uint64_t localSymbolsSize; 2220 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 and later 2221 }; 2222 2223 struct lldb_copy_dyld_cache_mapping_info 2224 { 2225 uint64_t address; 2226 uint64_t size; 2227 uint64_t fileOffset; 2228 uint32_t maxProt; 2229 uint32_t initProt; 2230 }; 2231 2232 struct lldb_copy_dyld_cache_local_symbols_info 2233 { 2234 uint32_t nlistOffset; 2235 uint32_t nlistCount; 2236 uint32_t stringsOffset; 2237 uint32_t stringsSize; 2238 uint32_t entriesOffset; 2239 uint32_t entriesCount; 2240 }; 2241 struct lldb_copy_dyld_cache_local_symbols_entry 2242 { 2243 uint32_t dylibOffset; 2244 uint32_t nlistStartIndex; 2245 uint32_t nlistCount; 2246 }; 2247 2248 /* The dyld_cache_header has a pointer to the dyld_cache_local_symbols_info structure (localSymbolsOffset). 2249 The dyld_cache_local_symbols_info structure gives us three things: 2250 1. The start and count of the nlist records in the dyld_shared_cache file 2251 2. The start and size of the strings for these nlist records 2252 3. The start and count of dyld_cache_local_symbols_entry entries 2253 2254 There is one dyld_cache_local_symbols_entry per dylib/framework in the dyld shared cache. 2255 The "dylibOffset" field is the Mach-O header of this dylib/framework in the dyld shared cache. 2256 The dyld_cache_local_symbols_entry also lists the start of this dylib/framework's nlist records 2257 and the count of how many nlist records there are for this dylib/framework. 2258 */ 2259 2260 // Process the dsc header to find the unmapped symbols 2261 // 2262 // Save some VM space, do not map the entire cache in one shot. 2263 2264 DataBufferSP dsc_data_sp; 2265 dsc_data_sp = dsc_filespec.MemoryMapFileContents(0, sizeof(struct lldb_copy_dyld_cache_header_v1)); 2266 2267 if (dsc_data_sp) 2268 { 2269 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2270 2271 char version_str[17]; 2272 int version = -1; 2273 lldb::offset_t offset = 0; 2274 memcpy (version_str, dsc_header_data.GetData (&offset, 16), 16); 2275 version_str[16] = '\0'; 2276 if (strncmp (version_str, "dyld_v", 6) == 0 && isdigit (version_str[6])) 2277 { 2278 int v; 2279 if (::sscanf (version_str + 6, "%d", &v) == 1) 2280 { 2281 version = v; 2282 } 2283 } 2284 2285 UUID dsc_uuid; 2286 if (version >= 1) 2287 { 2288 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, uuid); 2289 uint8_t uuid_bytes[sizeof (uuid_t)]; 2290 memcpy (uuid_bytes, dsc_header_data.GetData (&offset, sizeof (uuid_t)), sizeof (uuid_t)); 2291 dsc_uuid.SetBytes (uuid_bytes); 2292 } 2293 2294 bool uuid_match = true; 2295 if (dsc_uuid.IsValid() && process) 2296 { 2297 UUID shared_cache_uuid(GetProcessSharedCacheUUID(process)); 2298 2299 if (shared_cache_uuid.IsValid() && dsc_uuid != shared_cache_uuid) 2300 { 2301 // The on-disk dyld_shared_cache file is not the same as the one in this 2302 // process' memory, don't use it. 2303 uuid_match = false; 2304 ModuleSP module_sp (GetModule()); 2305 if (module_sp) 2306 module_sp->ReportWarning ("process shared cache does not match on-disk dyld_shared_cache file, some symbol names will be missing."); 2307 } 2308 } 2309 2310 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2311 2312 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2313 2314 // If the mappingOffset points to a location inside the header, we've 2315 // opened an old dyld shared cache, and should not proceed further. 2316 if (uuid_match && mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v0)) 2317 { 2318 2319 DataBufferSP dsc_mapping_info_data_sp = dsc_filespec.MemoryMapFileContents(mappingOffset, sizeof (struct lldb_copy_dyld_cache_mapping_info)); 2320 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, byte_order, addr_byte_size); 2321 offset = 0; 2322 2323 // The File addresses (from the in-memory Mach-O load commands) for the shared libraries 2324 // in the shared library cache need to be adjusted by an offset to match up with the 2325 // dylibOffset identifying field in the dyld_cache_local_symbol_entry's. This offset is 2326 // recorded in mapping_offset_value. 2327 const uint64_t mapping_offset_value = dsc_mapping_info_data.GetU64(&offset); 2328 2329 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2330 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2331 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2332 2333 if (localSymbolsOffset && localSymbolsSize) 2334 { 2335 // Map the local symbols 2336 if (DataBufferSP dsc_local_symbols_data_sp = dsc_filespec.MemoryMapFileContents(localSymbolsOffset, localSymbolsSize)) 2337 { 2338 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, byte_order, addr_byte_size); 2339 2340 offset = 0; 2341 2342 // Read the local_symbols_infos struct in one shot 2343 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2344 dsc_local_symbols_data.GetU32(&offset, &local_symbols_info.nlistOffset, 6); 2345 2346 SectionSP text_section_sp(section_list->FindSectionByName(GetSegmentNameTEXT())); 2347 2348 uint32_t header_file_offset = (text_section_sp->GetFileAddress() - mapping_offset_value); 2349 2350 offset = local_symbols_info.entriesOffset; 2351 for (uint32_t entry_index = 0; entry_index < local_symbols_info.entriesCount; entry_index++) 2352 { 2353 struct lldb_copy_dyld_cache_local_symbols_entry local_symbols_entry; 2354 local_symbols_entry.dylibOffset = dsc_local_symbols_data.GetU32(&offset); 2355 local_symbols_entry.nlistStartIndex = dsc_local_symbols_data.GetU32(&offset); 2356 local_symbols_entry.nlistCount = dsc_local_symbols_data.GetU32(&offset); 2357 2358 if (header_file_offset == local_symbols_entry.dylibOffset) 2359 { 2360 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2361 2362 // The normal nlist code cannot correctly size the Symbols array, we need to allocate it here. 2363 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms + unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2364 num_syms = symtab->GetNumSymbols(); 2365 2366 nlist_data_offset = local_symbols_info.nlistOffset + (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2367 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2368 2369 for (uint32_t nlist_index = 0; nlist_index < local_symbols_entry.nlistCount; nlist_index++) 2370 { 2371 ///////////////////////////// 2372 { 2373 struct nlist_64 nlist; 2374 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2375 break; 2376 2377 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked(&nlist_data_offset); 2378 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2379 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2380 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked (&nlist_data_offset); 2381 nlist.n_value = dsc_local_symbols_data.GetAddress_unchecked (&nlist_data_offset); 2382 2383 SymbolType type = eSymbolTypeInvalid; 2384 const char *symbol_name = dsc_local_symbols_data.PeekCStr(string_table_offset + nlist.n_strx); 2385 2386 if (symbol_name == NULL) 2387 { 2388 // No symbol should be NULL, even the symbols with no 2389 // string values should have an offset zero which points 2390 // to an empty C-string 2391 Host::SystemLog (Host::eSystemLogError, 2392 "error: DSC unmapped local symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 2393 entry_index, 2394 nlist.n_strx, 2395 module_sp->GetFileSpec().GetPath().c_str()); 2396 continue; 2397 } 2398 if (symbol_name[0] == '\0') 2399 symbol_name = NULL; 2400 2401 const char *symbol_name_non_abi_mangled = NULL; 2402 2403 SectionSP symbol_section; 2404 uint32_t symbol_byte_size = 0; 2405 bool add_nlist = true; 2406 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2407 bool demangled_is_synthesized = false; 2408 bool is_gsym = false; 2409 2410 assert (sym_idx < num_syms); 2411 2412 sym[sym_idx].SetDebug (is_debug); 2413 2414 if (is_debug) 2415 { 2416 switch (nlist.n_type) 2417 { 2418 case N_GSYM: 2419 // global symbol: name,,NO_SECT,type,0 2420 // Sometimes the N_GSYM value contains the address. 2421 2422 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 2423 // have the same address, but we want to ensure that we always find only the real symbol, 2424 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 2425 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 2426 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 2427 // same address. 2428 2429 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 2430 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 2431 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 2432 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 2433 add_nlist = false; 2434 else 2435 { 2436 is_gsym = true; 2437 sym[sym_idx].SetExternal(true); 2438 if (nlist.n_value != 0) 2439 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2440 type = eSymbolTypeData; 2441 } 2442 break; 2443 2444 case N_FNAME: 2445 // procedure name (f77 kludge): name,,NO_SECT,0,0 2446 type = eSymbolTypeCompiler; 2447 break; 2448 2449 case N_FUN: 2450 // procedure: name,,n_sect,linenumber,address 2451 if (symbol_name) 2452 { 2453 type = eSymbolTypeCode; 2454 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2455 2456 N_FUN_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 2457 // We use the current number of symbols in the symbol table in lieu of 2458 // using nlist_idx in case we ever start trimming entries out 2459 N_FUN_indexes.push_back(sym_idx); 2460 } 2461 else 2462 { 2463 type = eSymbolTypeCompiler; 2464 2465 if ( !N_FUN_indexes.empty() ) 2466 { 2467 // Copy the size of the function into the original STAB entry so we don't have 2468 // to hunt for it later 2469 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 2470 N_FUN_indexes.pop_back(); 2471 // We don't really need the end function STAB as it contains the size which 2472 // we already placed with the original symbol, so don't add it if we want a 2473 // minimal symbol table 2474 add_nlist = false; 2475 } 2476 } 2477 break; 2478 2479 case N_STSYM: 2480 // static symbol: name,,n_sect,type,address 2481 N_STSYM_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 2482 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2483 type = eSymbolTypeData; 2484 break; 2485 2486 case N_LCSYM: 2487 // .lcomm symbol: name,,n_sect,type,address 2488 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2489 type = eSymbolTypeCommonBlock; 2490 break; 2491 2492 case N_BNSYM: 2493 // We use the current number of symbols in the symbol table in lieu of 2494 // using nlist_idx in case we ever start trimming entries out 2495 // Skip these if we want minimal symbol tables 2496 add_nlist = false; 2497 break; 2498 2499 case N_ENSYM: 2500 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 2501 // so that we can always skip the entire symbol if we need to navigate 2502 // more quickly at the source level when parsing STABS 2503 // Skip these if we want minimal symbol tables 2504 add_nlist = false; 2505 break; 2506 2507 2508 case N_OPT: 2509 // emitted with gcc2_compiled and in gcc source 2510 type = eSymbolTypeCompiler; 2511 break; 2512 2513 case N_RSYM: 2514 // register sym: name,,NO_SECT,type,register 2515 type = eSymbolTypeVariable; 2516 break; 2517 2518 case N_SLINE: 2519 // src line: 0,,n_sect,linenumber,address 2520 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2521 type = eSymbolTypeLineEntry; 2522 break; 2523 2524 case N_SSYM: 2525 // structure elt: name,,NO_SECT,type,struct_offset 2526 type = eSymbolTypeVariableType; 2527 break; 2528 2529 case N_SO: 2530 // source file name 2531 type = eSymbolTypeSourceFile; 2532 if (symbol_name == NULL) 2533 { 2534 add_nlist = false; 2535 if (N_SO_index != UINT32_MAX) 2536 { 2537 // Set the size of the N_SO to the terminating index of this N_SO 2538 // so that we can always skip the entire N_SO if we need to navigate 2539 // more quickly at the source level when parsing STABS 2540 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2541 symbol_ptr->SetByteSize(sym_idx); 2542 symbol_ptr->SetSizeIsSibling(true); 2543 } 2544 N_NSYM_indexes.clear(); 2545 N_INCL_indexes.clear(); 2546 N_BRAC_indexes.clear(); 2547 N_COMM_indexes.clear(); 2548 N_FUN_indexes.clear(); 2549 N_SO_index = UINT32_MAX; 2550 } 2551 else 2552 { 2553 // We use the current number of symbols in the symbol table in lieu of 2554 // using nlist_idx in case we ever start trimming entries out 2555 const bool N_SO_has_full_path = symbol_name[0] == '/'; 2556 if (N_SO_has_full_path) 2557 { 2558 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2559 { 2560 // We have two consecutive N_SO entries where the first contains a directory 2561 // and the second contains a full path. 2562 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 2563 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2564 add_nlist = false; 2565 } 2566 else 2567 { 2568 // This is the first entry in a N_SO that contains a directory or 2569 // a full path to the source file 2570 N_SO_index = sym_idx; 2571 } 2572 } 2573 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2574 { 2575 // This is usually the second N_SO entry that contains just the filename, 2576 // so here we combine it with the first one if we are minimizing the symbol table 2577 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 2578 if (so_path && so_path[0]) 2579 { 2580 std::string full_so_path (so_path); 2581 const size_t double_slash_pos = full_so_path.find("//"); 2582 if (double_slash_pos != std::string::npos) 2583 { 2584 // The linker has been generating bad N_SO entries with doubled up paths 2585 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 2586 // and the second is the directory for the source file so you end up with 2587 // a path that looks like "/tmp/src//tmp/src/" 2588 FileSpec so_dir(so_path, false); 2589 if (!so_dir.Exists()) 2590 { 2591 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 2592 if (so_dir.Exists()) 2593 { 2594 // Trim off the incorrect path 2595 full_so_path.erase(0, double_slash_pos + 1); 2596 } 2597 } 2598 } 2599 if (*full_so_path.rbegin() != '/') 2600 full_so_path += '/'; 2601 full_so_path += symbol_name; 2602 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 2603 add_nlist = false; 2604 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2605 } 2606 } 2607 else 2608 { 2609 // This could be a relative path to a N_SO 2610 N_SO_index = sym_idx; 2611 } 2612 } 2613 break; 2614 2615 case N_OSO: 2616 // object file name: name,,0,0,st_mtime 2617 type = eSymbolTypeObjectFile; 2618 break; 2619 2620 case N_LSYM: 2621 // local sym: name,,NO_SECT,type,offset 2622 type = eSymbolTypeLocal; 2623 break; 2624 2625 //---------------------------------------------------------------------- 2626 // INCL scopes 2627 //---------------------------------------------------------------------- 2628 case N_BINCL: 2629 // include file beginning: name,,NO_SECT,0,sum 2630 // We use the current number of symbols in the symbol table in lieu of 2631 // using nlist_idx in case we ever start trimming entries out 2632 N_INCL_indexes.push_back(sym_idx); 2633 type = eSymbolTypeScopeBegin; 2634 break; 2635 2636 case N_EINCL: 2637 // include file end: name,,NO_SECT,0,0 2638 // Set the size of the N_BINCL to the terminating index of this N_EINCL 2639 // so that we can always skip the entire symbol if we need to navigate 2640 // more quickly at the source level when parsing STABS 2641 if ( !N_INCL_indexes.empty() ) 2642 { 2643 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 2644 symbol_ptr->SetByteSize(sym_idx + 1); 2645 symbol_ptr->SetSizeIsSibling(true); 2646 N_INCL_indexes.pop_back(); 2647 } 2648 type = eSymbolTypeScopeEnd; 2649 break; 2650 2651 case N_SOL: 2652 // #included file name: name,,n_sect,0,address 2653 type = eSymbolTypeHeaderFile; 2654 2655 // We currently don't use the header files on darwin 2656 add_nlist = false; 2657 break; 2658 2659 case N_PARAMS: 2660 // compiler parameters: name,,NO_SECT,0,0 2661 type = eSymbolTypeCompiler; 2662 break; 2663 2664 case N_VERSION: 2665 // compiler version: name,,NO_SECT,0,0 2666 type = eSymbolTypeCompiler; 2667 break; 2668 2669 case N_OLEVEL: 2670 // compiler -O level: name,,NO_SECT,0,0 2671 type = eSymbolTypeCompiler; 2672 break; 2673 2674 case N_PSYM: 2675 // parameter: name,,NO_SECT,type,offset 2676 type = eSymbolTypeVariable; 2677 break; 2678 2679 case N_ENTRY: 2680 // alternate entry: name,,n_sect,linenumber,address 2681 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2682 type = eSymbolTypeLineEntry; 2683 break; 2684 2685 //---------------------------------------------------------------------- 2686 // Left and Right Braces 2687 //---------------------------------------------------------------------- 2688 case N_LBRAC: 2689 // left bracket: 0,,NO_SECT,nesting level,address 2690 // We use the current number of symbols in the symbol table in lieu of 2691 // using nlist_idx in case we ever start trimming entries out 2692 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2693 N_BRAC_indexes.push_back(sym_idx); 2694 type = eSymbolTypeScopeBegin; 2695 break; 2696 2697 case N_RBRAC: 2698 // right bracket: 0,,NO_SECT,nesting level,address 2699 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 2700 // so that we can always skip the entire symbol if we need to navigate 2701 // more quickly at the source level when parsing STABS 2702 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2703 if ( !N_BRAC_indexes.empty() ) 2704 { 2705 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 2706 symbol_ptr->SetByteSize(sym_idx + 1); 2707 symbol_ptr->SetSizeIsSibling(true); 2708 N_BRAC_indexes.pop_back(); 2709 } 2710 type = eSymbolTypeScopeEnd; 2711 break; 2712 2713 case N_EXCL: 2714 // deleted include file: name,,NO_SECT,0,sum 2715 type = eSymbolTypeHeaderFile; 2716 break; 2717 2718 //---------------------------------------------------------------------- 2719 // COMM scopes 2720 //---------------------------------------------------------------------- 2721 case N_BCOMM: 2722 // begin common: name,,NO_SECT,0,0 2723 // We use the current number of symbols in the symbol table in lieu of 2724 // using nlist_idx in case we ever start trimming entries out 2725 type = eSymbolTypeScopeBegin; 2726 N_COMM_indexes.push_back(sym_idx); 2727 break; 2728 2729 case N_ECOML: 2730 // end common (local name): 0,,n_sect,0,address 2731 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2732 // Fall through 2733 2734 case N_ECOMM: 2735 // end common: name,,n_sect,0,0 2736 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 2737 // so that we can always skip the entire symbol if we need to navigate 2738 // more quickly at the source level when parsing STABS 2739 if ( !N_COMM_indexes.empty() ) 2740 { 2741 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 2742 symbol_ptr->SetByteSize(sym_idx + 1); 2743 symbol_ptr->SetSizeIsSibling(true); 2744 N_COMM_indexes.pop_back(); 2745 } 2746 type = eSymbolTypeScopeEnd; 2747 break; 2748 2749 case N_LENG: 2750 // second stab entry with length information 2751 type = eSymbolTypeAdditional; 2752 break; 2753 2754 default: break; 2755 } 2756 } 2757 else 2758 { 2759 //uint8_t n_pext = N_PEXT & nlist.n_type; 2760 uint8_t n_type = N_TYPE & nlist.n_type; 2761 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 2762 2763 switch (n_type) 2764 { 2765 case N_INDR: // Fall through 2766 case N_PBUD: // Fall through 2767 case N_UNDF: 2768 type = eSymbolTypeUndefined; 2769 break; 2770 2771 case N_ABS: 2772 type = eSymbolTypeAbsolute; 2773 break; 2774 2775 case N_SECT: 2776 { 2777 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2778 2779 if (symbol_section == NULL) 2780 { 2781 // TODO: warn about this? 2782 add_nlist = false; 2783 break; 2784 } 2785 2786 if (TEXT_eh_frame_sectID == nlist.n_sect) 2787 { 2788 type = eSymbolTypeException; 2789 } 2790 else 2791 { 2792 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 2793 2794 switch (section_type) 2795 { 2796 case S_REGULAR: break; // regular section 2797 //case S_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section 2798 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 2799 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 2800 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 2801 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 2802 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 2803 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 2804 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 2805 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 2806 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 2807 //case S_COALESCED: type = eSymbolType; break; // section contains symbols that are to be coalesced 2808 //case S_GB_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 2809 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 2810 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 2811 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 2812 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 2813 default: break; 2814 } 2815 2816 if (type == eSymbolTypeInvalid) 2817 { 2818 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 2819 if (symbol_section->IsDescendant (text_section_sp.get())) 2820 { 2821 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 2822 S_ATTR_SELF_MODIFYING_CODE | 2823 S_ATTR_SOME_INSTRUCTIONS)) 2824 type = eSymbolTypeData; 2825 else 2826 type = eSymbolTypeCode; 2827 } 2828 else if (symbol_section->IsDescendant(data_section_sp.get())) 2829 { 2830 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 2831 { 2832 type = eSymbolTypeRuntime; 2833 2834 if (symbol_name && 2835 symbol_name[0] == '_' && 2836 symbol_name[1] == 'O' && 2837 symbol_name[2] == 'B') 2838 { 2839 llvm::StringRef symbol_name_ref(symbol_name); 2840 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 2841 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 2842 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 2843 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 2844 { 2845 symbol_name_non_abi_mangled = symbol_name + 1; 2846 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2847 type = eSymbolTypeObjCClass; 2848 demangled_is_synthesized = true; 2849 } 2850 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 2851 { 2852 symbol_name_non_abi_mangled = symbol_name + 1; 2853 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2854 type = eSymbolTypeObjCMetaClass; 2855 demangled_is_synthesized = true; 2856 } 2857 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 2858 { 2859 symbol_name_non_abi_mangled = symbol_name + 1; 2860 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2861 type = eSymbolTypeObjCIVar; 2862 demangled_is_synthesized = true; 2863 } 2864 } 2865 } 2866 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 2867 { 2868 type = eSymbolTypeException; 2869 } 2870 else 2871 { 2872 type = eSymbolTypeData; 2873 } 2874 } 2875 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 2876 { 2877 type = eSymbolTypeTrampoline; 2878 } 2879 else if (symbol_section->IsDescendant(objc_section_sp.get())) 2880 { 2881 type = eSymbolTypeRuntime; 2882 if (symbol_name && symbol_name[0] == '.') 2883 { 2884 llvm::StringRef symbol_name_ref(symbol_name); 2885 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 2886 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 2887 { 2888 symbol_name_non_abi_mangled = symbol_name; 2889 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 2890 type = eSymbolTypeObjCClass; 2891 demangled_is_synthesized = true; 2892 } 2893 } 2894 } 2895 } 2896 } 2897 } 2898 break; 2899 } 2900 } 2901 2902 if (add_nlist) 2903 { 2904 uint64_t symbol_value = nlist.n_value; 2905 if (symbol_name_non_abi_mangled) 2906 { 2907 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 2908 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 2909 } 2910 else 2911 { 2912 bool symbol_name_is_mangled = false; 2913 2914 if (symbol_name && symbol_name[0] == '_') 2915 { 2916 symbol_name_is_mangled = symbol_name[1] == '_'; 2917 symbol_name++; // Skip the leading underscore 2918 } 2919 2920 if (symbol_name) 2921 { 2922 ConstString const_symbol_name(symbol_name); 2923 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 2924 if (is_gsym && is_debug) 2925 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 2926 } 2927 } 2928 if (symbol_section) 2929 { 2930 const addr_t section_file_addr = symbol_section->GetFileAddress(); 2931 if (symbol_byte_size == 0 && function_starts_count > 0) 2932 { 2933 addr_t symbol_lookup_file_addr = nlist.n_value; 2934 // Do an exact address match for non-ARM addresses, else get the closest since 2935 // the symbol might be a thumb symbol which has an address with bit zero set 2936 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 2937 if (is_arm && func_start_entry) 2938 { 2939 // Verify that the function start address is the symbol address (ARM) 2940 // or the symbol address + 1 (thumb) 2941 if (func_start_entry->addr != symbol_lookup_file_addr && 2942 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 2943 { 2944 // Not the right entry, NULL it out... 2945 func_start_entry = NULL; 2946 } 2947 } 2948 if (func_start_entry) 2949 { 2950 func_start_entry->data = true; 2951 2952 addr_t symbol_file_addr = func_start_entry->addr; 2953 uint32_t symbol_flags = 0; 2954 if (is_arm) 2955 { 2956 if (symbol_file_addr & 1) 2957 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 2958 symbol_file_addr &= 0xfffffffffffffffeull; 2959 } 2960 2961 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 2962 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 2963 if (next_func_start_entry) 2964 { 2965 addr_t next_symbol_file_addr = next_func_start_entry->addr; 2966 // Be sure the clear the Thumb address bit when we calculate the size 2967 // from the current and next address 2968 if (is_arm) 2969 next_symbol_file_addr &= 0xfffffffffffffffeull; 2970 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 2971 } 2972 else 2973 { 2974 symbol_byte_size = section_end_file_addr - symbol_file_addr; 2975 } 2976 } 2977 } 2978 symbol_value -= section_file_addr; 2979 } 2980 2981 if (is_debug == false) 2982 { 2983 if (type == eSymbolTypeCode) 2984 { 2985 // See if we can find a N_FUN entry for any code symbols. 2986 // If we do find a match, and the name matches, then we 2987 // can merge the two into just the function symbol to avoid 2988 // duplicate entries in the symbol table 2989 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 2990 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 2991 if (range.first != range.second) 2992 { 2993 bool found_it = false; 2994 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 2995 { 2996 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 2997 { 2998 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 2999 // We just need the flags from the linker symbol, so put these flags 3000 // into the N_FUN flags to avoid duplicate symbols in the symbol table 3001 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3002 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3003 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3004 sym[pos->second].SetType (eSymbolTypeResolver); 3005 sym[sym_idx].Clear(); 3006 found_it = true; 3007 break; 3008 } 3009 } 3010 if (found_it) 3011 continue; 3012 } 3013 else 3014 { 3015 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3016 type = eSymbolTypeResolver; 3017 } 3018 } 3019 else if (type == eSymbolTypeData) 3020 { 3021 // See if we can find a N_STSYM entry for any data symbols. 3022 // If we do find a match, and the name matches, then we 3023 // can merge the two into just the Static symbol to avoid 3024 // duplicate entries in the symbol table 3025 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3026 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 3027 if (range.first != range.second) 3028 { 3029 bool found_it = false; 3030 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3031 { 3032 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3033 { 3034 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3035 // We just need the flags from the linker symbol, so put these flags 3036 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3037 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3038 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3039 sym[sym_idx].Clear(); 3040 found_it = true; 3041 break; 3042 } 3043 } 3044 if (found_it) 3045 continue; 3046 } 3047 else 3048 { 3049 // Combine N_GSYM stab entries with the non stab symbol 3050 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 3051 if (pos != N_GSYM_name_to_sym_idx.end()) 3052 { 3053 const uint32_t GSYM_sym_idx = pos->second; 3054 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 3055 // Copy the address, because often the N_GSYM address has an invalid address of zero 3056 // when the global is a common symbol 3057 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 3058 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 3059 // We just need the flags from the linker symbol, so put these flags 3060 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3061 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3062 sym[sym_idx].Clear(); 3063 continue; 3064 } 3065 } 3066 } 3067 } 3068 3069 sym[sym_idx].SetID (nlist_idx); 3070 sym[sym_idx].SetType (type); 3071 sym[sym_idx].GetAddress().SetSection (symbol_section); 3072 sym[sym_idx].GetAddress().SetOffset (symbol_value); 3073 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3074 3075 if (symbol_byte_size > 0) 3076 sym[sym_idx].SetByteSize(symbol_byte_size); 3077 3078 if (demangled_is_synthesized) 3079 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3080 ++sym_idx; 3081 } 3082 else 3083 { 3084 sym[sym_idx].Clear(); 3085 } 3086 3087 } 3088 ///////////////////////////// 3089 } 3090 break; // No more entries to consider 3091 } 3092 } 3093 } 3094 } 3095 } 3096 } 3097 } 3098 3099 // Must reset this in case it was mutated above! 3100 nlist_data_offset = 0; 3101 #endif 3102 3103 if (nlist_data.GetByteSize() > 0) 3104 { 3105 3106 // If the sym array was not created while parsing the DSC unmapped 3107 // symbols, create it now. 3108 if (sym == NULL) 3109 { 3110 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3111 num_syms = symtab->GetNumSymbols(); 3112 } 3113 3114 if (unmapped_local_symbols_found) 3115 { 3116 assert(m_dysymtab.ilocalsym == 0); 3117 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3118 nlist_idx = m_dysymtab.nlocalsym; 3119 } 3120 else 3121 { 3122 nlist_idx = 0; 3123 } 3124 3125 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) 3126 { 3127 struct nlist_64 nlist; 3128 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 3129 break; 3130 3131 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3132 nlist.n_type = nlist_data.GetU8_unchecked (&nlist_data_offset); 3133 nlist.n_sect = nlist_data.GetU8_unchecked (&nlist_data_offset); 3134 nlist.n_desc = nlist_data.GetU16_unchecked (&nlist_data_offset); 3135 nlist.n_value = nlist_data.GetAddress_unchecked (&nlist_data_offset); 3136 3137 SymbolType type = eSymbolTypeInvalid; 3138 const char *symbol_name = NULL; 3139 3140 if (have_strtab_data) 3141 { 3142 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3143 3144 if (symbol_name == NULL) 3145 { 3146 // No symbol should be NULL, even the symbols with no 3147 // string values should have an offset zero which points 3148 // to an empty C-string 3149 Host::SystemLog (Host::eSystemLogError, 3150 "error: symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 3151 nlist_idx, 3152 nlist.n_strx, 3153 module_sp->GetFileSpec().GetPath().c_str()); 3154 continue; 3155 } 3156 if (symbol_name[0] == '\0') 3157 symbol_name = NULL; 3158 } 3159 else 3160 { 3161 const addr_t str_addr = strtab_addr + nlist.n_strx; 3162 Error str_error; 3163 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, str_error)) 3164 symbol_name = memory_symbol_name.c_str(); 3165 } 3166 const char *symbol_name_non_abi_mangled = NULL; 3167 3168 SectionSP symbol_section; 3169 lldb::addr_t symbol_byte_size = 0; 3170 bool add_nlist = true; 3171 bool is_gsym = false; 3172 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3173 bool demangled_is_synthesized = false; 3174 3175 assert (sym_idx < num_syms); 3176 3177 sym[sym_idx].SetDebug (is_debug); 3178 3179 if (is_debug) 3180 { 3181 switch (nlist.n_type) 3182 { 3183 case N_GSYM: 3184 // global symbol: name,,NO_SECT,type,0 3185 // Sometimes the N_GSYM value contains the address. 3186 3187 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 3188 // have the same address, but we want to ensure that we always find only the real symbol, 3189 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 3190 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 3191 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 3192 // same address. 3193 3194 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 3195 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 3196 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 3197 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 3198 add_nlist = false; 3199 else 3200 { 3201 is_gsym = true; 3202 sym[sym_idx].SetExternal(true); 3203 if (nlist.n_value != 0) 3204 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3205 type = eSymbolTypeData; 3206 } 3207 break; 3208 3209 case N_FNAME: 3210 // procedure name (f77 kludge): name,,NO_SECT,0,0 3211 type = eSymbolTypeCompiler; 3212 break; 3213 3214 case N_FUN: 3215 // procedure: name,,n_sect,linenumber,address 3216 if (symbol_name) 3217 { 3218 type = eSymbolTypeCode; 3219 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3220 3221 N_FUN_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3222 // We use the current number of symbols in the symbol table in lieu of 3223 // using nlist_idx in case we ever start trimming entries out 3224 N_FUN_indexes.push_back(sym_idx); 3225 } 3226 else 3227 { 3228 type = eSymbolTypeCompiler; 3229 3230 if ( !N_FUN_indexes.empty() ) 3231 { 3232 // Copy the size of the function into the original STAB entry so we don't have 3233 // to hunt for it later 3234 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 3235 N_FUN_indexes.pop_back(); 3236 // We don't really need the end function STAB as it contains the size which 3237 // we already placed with the original symbol, so don't add it if we want a 3238 // minimal symbol table 3239 add_nlist = false; 3240 } 3241 } 3242 break; 3243 3244 case N_STSYM: 3245 // static symbol: name,,n_sect,type,address 3246 N_STSYM_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3247 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3248 type = eSymbolTypeData; 3249 break; 3250 3251 case N_LCSYM: 3252 // .lcomm symbol: name,,n_sect,type,address 3253 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3254 type = eSymbolTypeCommonBlock; 3255 break; 3256 3257 case N_BNSYM: 3258 // We use the current number of symbols in the symbol table in lieu of 3259 // using nlist_idx in case we ever start trimming entries out 3260 // Skip these if we want minimal symbol tables 3261 add_nlist = false; 3262 break; 3263 3264 case N_ENSYM: 3265 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 3266 // so that we can always skip the entire symbol if we need to navigate 3267 // more quickly at the source level when parsing STABS 3268 // Skip these if we want minimal symbol tables 3269 add_nlist = false; 3270 break; 3271 3272 3273 case N_OPT: 3274 // emitted with gcc2_compiled and in gcc source 3275 type = eSymbolTypeCompiler; 3276 break; 3277 3278 case N_RSYM: 3279 // register sym: name,,NO_SECT,type,register 3280 type = eSymbolTypeVariable; 3281 break; 3282 3283 case N_SLINE: 3284 // src line: 0,,n_sect,linenumber,address 3285 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3286 type = eSymbolTypeLineEntry; 3287 break; 3288 3289 case N_SSYM: 3290 // structure elt: name,,NO_SECT,type,struct_offset 3291 type = eSymbolTypeVariableType; 3292 break; 3293 3294 case N_SO: 3295 // source file name 3296 type = eSymbolTypeSourceFile; 3297 if (symbol_name == NULL) 3298 { 3299 add_nlist = false; 3300 if (N_SO_index != UINT32_MAX) 3301 { 3302 // Set the size of the N_SO to the terminating index of this N_SO 3303 // so that we can always skip the entire N_SO if we need to navigate 3304 // more quickly at the source level when parsing STABS 3305 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3306 symbol_ptr->SetByteSize(sym_idx); 3307 symbol_ptr->SetSizeIsSibling(true); 3308 } 3309 N_NSYM_indexes.clear(); 3310 N_INCL_indexes.clear(); 3311 N_BRAC_indexes.clear(); 3312 N_COMM_indexes.clear(); 3313 N_FUN_indexes.clear(); 3314 N_SO_index = UINT32_MAX; 3315 } 3316 else 3317 { 3318 // We use the current number of symbols in the symbol table in lieu of 3319 // using nlist_idx in case we ever start trimming entries out 3320 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3321 if (N_SO_has_full_path) 3322 { 3323 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3324 { 3325 // We have two consecutive N_SO entries where the first contains a directory 3326 // and the second contains a full path. 3327 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 3328 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3329 add_nlist = false; 3330 } 3331 else 3332 { 3333 // This is the first entry in a N_SO that contains a directory or 3334 // a full path to the source file 3335 N_SO_index = sym_idx; 3336 } 3337 } 3338 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3339 { 3340 // This is usually the second N_SO entry that contains just the filename, 3341 // so here we combine it with the first one if we are minimizing the symbol table 3342 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3343 if (so_path && so_path[0]) 3344 { 3345 std::string full_so_path (so_path); 3346 const size_t double_slash_pos = full_so_path.find("//"); 3347 if (double_slash_pos != std::string::npos) 3348 { 3349 // The linker has been generating bad N_SO entries with doubled up paths 3350 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 3351 // and the second is the directory for the source file so you end up with 3352 // a path that looks like "/tmp/src//tmp/src/" 3353 FileSpec so_dir(so_path, false); 3354 if (!so_dir.Exists()) 3355 { 3356 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 3357 if (so_dir.Exists()) 3358 { 3359 // Trim off the incorrect path 3360 full_so_path.erase(0, double_slash_pos + 1); 3361 } 3362 } 3363 } 3364 if (*full_so_path.rbegin() != '/') 3365 full_so_path += '/'; 3366 full_so_path += symbol_name; 3367 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 3368 add_nlist = false; 3369 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3370 } 3371 } 3372 else 3373 { 3374 // This could be a relative path to a N_SO 3375 N_SO_index = sym_idx; 3376 } 3377 } 3378 3379 break; 3380 3381 case N_OSO: 3382 // object file name: name,,0,0,st_mtime 3383 type = eSymbolTypeObjectFile; 3384 break; 3385 3386 case N_LSYM: 3387 // local sym: name,,NO_SECT,type,offset 3388 type = eSymbolTypeLocal; 3389 break; 3390 3391 //---------------------------------------------------------------------- 3392 // INCL scopes 3393 //---------------------------------------------------------------------- 3394 case N_BINCL: 3395 // include file beginning: name,,NO_SECT,0,sum 3396 // We use the current number of symbols in the symbol table in lieu of 3397 // using nlist_idx in case we ever start trimming entries out 3398 N_INCL_indexes.push_back(sym_idx); 3399 type = eSymbolTypeScopeBegin; 3400 break; 3401 3402 case N_EINCL: 3403 // include file end: name,,NO_SECT,0,0 3404 // Set the size of the N_BINCL to the terminating index of this N_EINCL 3405 // so that we can always skip the entire symbol if we need to navigate 3406 // more quickly at the source level when parsing STABS 3407 if ( !N_INCL_indexes.empty() ) 3408 { 3409 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3410 symbol_ptr->SetByteSize(sym_idx + 1); 3411 symbol_ptr->SetSizeIsSibling(true); 3412 N_INCL_indexes.pop_back(); 3413 } 3414 type = eSymbolTypeScopeEnd; 3415 break; 3416 3417 case N_SOL: 3418 // #included file name: name,,n_sect,0,address 3419 type = eSymbolTypeHeaderFile; 3420 3421 // We currently don't use the header files on darwin 3422 add_nlist = false; 3423 break; 3424 3425 case N_PARAMS: 3426 // compiler parameters: name,,NO_SECT,0,0 3427 type = eSymbolTypeCompiler; 3428 break; 3429 3430 case N_VERSION: 3431 // compiler version: name,,NO_SECT,0,0 3432 type = eSymbolTypeCompiler; 3433 break; 3434 3435 case N_OLEVEL: 3436 // compiler -O level: name,,NO_SECT,0,0 3437 type = eSymbolTypeCompiler; 3438 break; 3439 3440 case N_PSYM: 3441 // parameter: name,,NO_SECT,type,offset 3442 type = eSymbolTypeVariable; 3443 break; 3444 3445 case N_ENTRY: 3446 // alternate entry: name,,n_sect,linenumber,address 3447 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3448 type = eSymbolTypeLineEntry; 3449 break; 3450 3451 //---------------------------------------------------------------------- 3452 // Left and Right Braces 3453 //---------------------------------------------------------------------- 3454 case N_LBRAC: 3455 // left bracket: 0,,NO_SECT,nesting level,address 3456 // We use the current number of symbols in the symbol table in lieu of 3457 // using nlist_idx in case we ever start trimming entries out 3458 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3459 N_BRAC_indexes.push_back(sym_idx); 3460 type = eSymbolTypeScopeBegin; 3461 break; 3462 3463 case N_RBRAC: 3464 // right bracket: 0,,NO_SECT,nesting level,address 3465 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 3466 // so that we can always skip the entire symbol if we need to navigate 3467 // more quickly at the source level when parsing STABS 3468 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3469 if ( !N_BRAC_indexes.empty() ) 3470 { 3471 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3472 symbol_ptr->SetByteSize(sym_idx + 1); 3473 symbol_ptr->SetSizeIsSibling(true); 3474 N_BRAC_indexes.pop_back(); 3475 } 3476 type = eSymbolTypeScopeEnd; 3477 break; 3478 3479 case N_EXCL: 3480 // deleted include file: name,,NO_SECT,0,sum 3481 type = eSymbolTypeHeaderFile; 3482 break; 3483 3484 //---------------------------------------------------------------------- 3485 // COMM scopes 3486 //---------------------------------------------------------------------- 3487 case N_BCOMM: 3488 // begin common: name,,NO_SECT,0,0 3489 // We use the current number of symbols in the symbol table in lieu of 3490 // using nlist_idx in case we ever start trimming entries out 3491 type = eSymbolTypeScopeBegin; 3492 N_COMM_indexes.push_back(sym_idx); 3493 break; 3494 3495 case N_ECOML: 3496 // end common (local name): 0,,n_sect,0,address 3497 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3498 // Fall through 3499 3500 case N_ECOMM: 3501 // end common: name,,n_sect,0,0 3502 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 3503 // so that we can always skip the entire symbol if we need to navigate 3504 // more quickly at the source level when parsing STABS 3505 if ( !N_COMM_indexes.empty() ) 3506 { 3507 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 3508 symbol_ptr->SetByteSize(sym_idx + 1); 3509 symbol_ptr->SetSizeIsSibling(true); 3510 N_COMM_indexes.pop_back(); 3511 } 3512 type = eSymbolTypeScopeEnd; 3513 break; 3514 3515 case N_LENG: 3516 // second stab entry with length information 3517 type = eSymbolTypeAdditional; 3518 break; 3519 3520 default: break; 3521 } 3522 } 3523 else 3524 { 3525 //uint8_t n_pext = N_PEXT & nlist.n_type; 3526 uint8_t n_type = N_TYPE & nlist.n_type; 3527 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3528 3529 switch (n_type) 3530 { 3531 case N_INDR:// Fall through 3532 case N_PBUD:// Fall through 3533 case N_UNDF: 3534 type = eSymbolTypeUndefined; 3535 break; 3536 3537 case N_ABS: 3538 type = eSymbolTypeAbsolute; 3539 break; 3540 3541 case N_SECT: 3542 { 3543 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3544 3545 if (!symbol_section) 3546 { 3547 // TODO: warn about this? 3548 add_nlist = false; 3549 break; 3550 } 3551 3552 if (TEXT_eh_frame_sectID == nlist.n_sect) 3553 { 3554 type = eSymbolTypeException; 3555 } 3556 else 3557 { 3558 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 3559 3560 switch (section_type) 3561 { 3562 case S_REGULAR: break; // regular section 3563 //case S_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section 3564 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 3565 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 3566 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 3567 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 3568 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 3569 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 3570 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 3571 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 3572 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 3573 //case S_COALESCED: type = eSymbolType; break; // section contains symbols that are to be coalesced 3574 //case S_GB_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 3575 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 3576 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 3577 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 3578 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 3579 default: break; 3580 } 3581 3582 if (type == eSymbolTypeInvalid) 3583 { 3584 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 3585 if (symbol_section->IsDescendant (text_section_sp.get())) 3586 { 3587 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 3588 S_ATTR_SELF_MODIFYING_CODE | 3589 S_ATTR_SOME_INSTRUCTIONS)) 3590 type = eSymbolTypeData; 3591 else 3592 type = eSymbolTypeCode; 3593 } 3594 else 3595 if (symbol_section->IsDescendant(data_section_sp.get())) 3596 { 3597 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 3598 { 3599 type = eSymbolTypeRuntime; 3600 3601 if (symbol_name && 3602 symbol_name[0] == '_' && 3603 symbol_name[1] == 'O' && 3604 symbol_name[2] == 'B') 3605 { 3606 llvm::StringRef symbol_name_ref(symbol_name); 3607 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 3608 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 3609 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 3610 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 3611 { 3612 symbol_name_non_abi_mangled = symbol_name + 1; 3613 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3614 type = eSymbolTypeObjCClass; 3615 demangled_is_synthesized = true; 3616 } 3617 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 3618 { 3619 symbol_name_non_abi_mangled = symbol_name + 1; 3620 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3621 type = eSymbolTypeObjCMetaClass; 3622 demangled_is_synthesized = true; 3623 } 3624 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 3625 { 3626 symbol_name_non_abi_mangled = symbol_name + 1; 3627 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3628 type = eSymbolTypeObjCIVar; 3629 demangled_is_synthesized = true; 3630 } 3631 } 3632 } 3633 else 3634 if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 3635 { 3636 type = eSymbolTypeException; 3637 } 3638 else 3639 { 3640 type = eSymbolTypeData; 3641 } 3642 } 3643 else 3644 if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 3645 { 3646 type = eSymbolTypeTrampoline; 3647 } 3648 else 3649 if (symbol_section->IsDescendant(objc_section_sp.get())) 3650 { 3651 type = eSymbolTypeRuntime; 3652 if (symbol_name && symbol_name[0] == '.') 3653 { 3654 llvm::StringRef symbol_name_ref(symbol_name); 3655 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 3656 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 3657 { 3658 symbol_name_non_abi_mangled = symbol_name; 3659 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 3660 type = eSymbolTypeObjCClass; 3661 demangled_is_synthesized = true; 3662 } 3663 } 3664 } 3665 } 3666 } 3667 } 3668 break; 3669 } 3670 } 3671 3672 if (add_nlist) 3673 { 3674 uint64_t symbol_value = nlist.n_value; 3675 3676 if (symbol_name_non_abi_mangled) 3677 { 3678 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 3679 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 3680 } 3681 else 3682 { 3683 bool symbol_name_is_mangled = false; 3684 3685 if (symbol_name && symbol_name[0] == '_') 3686 { 3687 symbol_name_is_mangled = symbol_name[1] == '_'; 3688 symbol_name++; // Skip the leading underscore 3689 } 3690 3691 if (symbol_name) 3692 { 3693 ConstString const_symbol_name(symbol_name); 3694 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 3695 if (is_gsym && is_debug) 3696 { 3697 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 3698 } 3699 } 3700 } 3701 if (symbol_section) 3702 { 3703 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3704 if (symbol_byte_size == 0 && function_starts_count > 0) 3705 { 3706 addr_t symbol_lookup_file_addr = nlist.n_value; 3707 // Do an exact address match for non-ARM addresses, else get the closest since 3708 // the symbol might be a thumb symbol which has an address with bit zero set 3709 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 3710 if (is_arm && func_start_entry) 3711 { 3712 // Verify that the function start address is the symbol address (ARM) 3713 // or the symbol address + 1 (thumb) 3714 if (func_start_entry->addr != symbol_lookup_file_addr && 3715 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 3716 { 3717 // Not the right entry, NULL it out... 3718 func_start_entry = NULL; 3719 } 3720 } 3721 if (func_start_entry) 3722 { 3723 func_start_entry->data = true; 3724 3725 addr_t symbol_file_addr = func_start_entry->addr; 3726 if (is_arm) 3727 symbol_file_addr &= 0xfffffffffffffffeull; 3728 3729 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3730 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3731 if (next_func_start_entry) 3732 { 3733 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3734 // Be sure the clear the Thumb address bit when we calculate the size 3735 // from the current and next address 3736 if (is_arm) 3737 next_symbol_file_addr &= 0xfffffffffffffffeull; 3738 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3739 } 3740 else 3741 { 3742 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3743 } 3744 } 3745 } 3746 symbol_value -= section_file_addr; 3747 } 3748 3749 if (is_debug == false) 3750 { 3751 if (type == eSymbolTypeCode) 3752 { 3753 // See if we can find a N_FUN entry for any code symbols. 3754 // If we do find a match, and the name matches, then we 3755 // can merge the two into just the function symbol to avoid 3756 // duplicate entries in the symbol table 3757 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3758 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3759 if (range.first != range.second) 3760 { 3761 bool found_it = false; 3762 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3763 { 3764 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3765 { 3766 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3767 // We just need the flags from the linker symbol, so put these flags 3768 // into the N_FUN flags to avoid duplicate symbols in the symbol table 3769 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3770 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3771 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3772 sym[pos->second].SetType (eSymbolTypeResolver); 3773 sym[sym_idx].Clear(); 3774 found_it = true; 3775 break; 3776 } 3777 } 3778 if (found_it) 3779 continue; 3780 } 3781 else 3782 { 3783 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3784 type = eSymbolTypeResolver; 3785 } 3786 } 3787 else if (type == eSymbolTypeData) 3788 { 3789 // See if we can find a N_STSYM entry for any data symbols. 3790 // If we do find a match, and the name matches, then we 3791 // can merge the two into just the Static symbol to avoid 3792 // duplicate entries in the symbol table 3793 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3794 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 3795 if (range.first != range.second) 3796 { 3797 bool found_it = false; 3798 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3799 { 3800 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3801 { 3802 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3803 // We just need the flags from the linker symbol, so put these flags 3804 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3805 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3806 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3807 sym[sym_idx].Clear(); 3808 found_it = true; 3809 break; 3810 } 3811 } 3812 if (found_it) 3813 continue; 3814 } 3815 else 3816 { 3817 // Combine N_GSYM stab entries with the non stab symbol 3818 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 3819 if (pos != N_GSYM_name_to_sym_idx.end()) 3820 { 3821 const uint32_t GSYM_sym_idx = pos->second; 3822 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 3823 // Copy the address, because often the N_GSYM address has an invalid address of zero 3824 // when the global is a common symbol 3825 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 3826 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 3827 // We just need the flags from the linker symbol, so put these flags 3828 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3829 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3830 sym[sym_idx].Clear(); 3831 continue; 3832 } 3833 } 3834 } 3835 } 3836 3837 sym[sym_idx].SetID (nlist_idx); 3838 sym[sym_idx].SetType (type); 3839 sym[sym_idx].GetAddress().SetSection (symbol_section); 3840 sym[sym_idx].GetAddress().SetOffset (symbol_value); 3841 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3842 3843 if (symbol_byte_size > 0) 3844 sym[sym_idx].SetByteSize(symbol_byte_size); 3845 3846 if (demangled_is_synthesized) 3847 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3848 3849 ++sym_idx; 3850 } 3851 else 3852 { 3853 sym[sym_idx].Clear(); 3854 } 3855 } 3856 } 3857 3858 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 3859 3860 if (function_starts_count > 0) 3861 { 3862 char synthetic_function_symbol[PATH_MAX]; 3863 uint32_t num_synthetic_function_symbols = 0; 3864 for (i=0; i<function_starts_count; ++i) 3865 { 3866 if (function_starts.GetEntryRef (i).data == false) 3867 ++num_synthetic_function_symbols; 3868 } 3869 3870 if (num_synthetic_function_symbols > 0) 3871 { 3872 if (num_syms < sym_idx + num_synthetic_function_symbols) 3873 { 3874 num_syms = sym_idx + num_synthetic_function_symbols; 3875 sym = symtab->Resize (num_syms); 3876 } 3877 uint32_t synthetic_function_symbol_idx = 0; 3878 for (i=0; i<function_starts_count; ++i) 3879 { 3880 const FunctionStarts::Entry *func_start_entry = function_starts.GetEntryAtIndex (i); 3881 if (func_start_entry->data == false) 3882 { 3883 addr_t symbol_file_addr = func_start_entry->addr; 3884 uint32_t symbol_flags = 0; 3885 if (is_arm) 3886 { 3887 if (symbol_file_addr & 1) 3888 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3889 symbol_file_addr &= 0xfffffffffffffffeull; 3890 } 3891 Address symbol_addr; 3892 if (module_sp->ResolveFileAddress (symbol_file_addr, symbol_addr)) 3893 { 3894 SectionSP symbol_section (symbol_addr.GetSection()); 3895 uint32_t symbol_byte_size = 0; 3896 if (symbol_section) 3897 { 3898 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3899 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3900 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3901 if (next_func_start_entry) 3902 { 3903 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3904 if (is_arm) 3905 next_symbol_file_addr &= 0xfffffffffffffffeull; 3906 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3907 } 3908 else 3909 { 3910 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3911 } 3912 snprintf (synthetic_function_symbol, 3913 sizeof(synthetic_function_symbol), 3914 "___lldb_unnamed_function%u$$%s", 3915 ++synthetic_function_symbol_idx, 3916 module_sp->GetFileSpec().GetFilename().GetCString()); 3917 sym[sym_idx].SetID (synthetic_sym_id++); 3918 sym[sym_idx].GetMangled().SetDemangledName(ConstString(synthetic_function_symbol)); 3919 sym[sym_idx].SetType (eSymbolTypeCode); 3920 sym[sym_idx].SetIsSynthetic (true); 3921 sym[sym_idx].GetAddress() = symbol_addr; 3922 if (symbol_flags) 3923 sym[sym_idx].SetFlags (symbol_flags); 3924 if (symbol_byte_size) 3925 sym[sym_idx].SetByteSize (symbol_byte_size); 3926 ++sym_idx; 3927 } 3928 } 3929 } 3930 } 3931 } 3932 } 3933 3934 // Trim our symbols down to just what we ended up with after 3935 // removing any symbols. 3936 if (sym_idx < num_syms) 3937 { 3938 num_syms = sym_idx; 3939 sym = symtab->Resize (num_syms); 3940 } 3941 3942 // Now synthesize indirect symbols 3943 if (m_dysymtab.nindirectsyms != 0) 3944 { 3945 if (indirect_symbol_index_data.GetByteSize()) 3946 { 3947 NListIndexToSymbolIndexMap::const_iterator end_index_pos = m_nlist_idx_to_sym_idx.end(); 3948 3949 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); ++sect_idx) 3950 { 3951 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == S_SYMBOL_STUBS) 3952 { 3953 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 3954 if (symbol_stub_byte_size == 0) 3955 continue; 3956 3957 const uint32_t num_symbol_stubs = m_mach_sections[sect_idx].size / symbol_stub_byte_size; 3958 3959 if (num_symbol_stubs == 0) 3960 continue; 3961 3962 const uint32_t symbol_stub_index_offset = m_mach_sections[sect_idx].reserved1; 3963 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) 3964 { 3965 const uint32_t symbol_stub_index = symbol_stub_index_offset + stub_idx; 3966 const lldb::addr_t symbol_stub_addr = m_mach_sections[sect_idx].addr + (stub_idx * symbol_stub_byte_size); 3967 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 3968 if (indirect_symbol_index_data.ValidOffsetForDataOfSize(symbol_stub_offset, 4)) 3969 { 3970 const uint32_t stub_sym_id = indirect_symbol_index_data.GetU32 (&symbol_stub_offset); 3971 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 3972 continue; 3973 3974 NListIndexToSymbolIndexMap::const_iterator index_pos = m_nlist_idx_to_sym_idx.find (stub_sym_id); 3975 Symbol *stub_symbol = NULL; 3976 if (index_pos != end_index_pos) 3977 { 3978 // We have a remapping from the original nlist index to 3979 // a current symbol index, so just look this up by index 3980 stub_symbol = symtab->SymbolAtIndex (index_pos->second); 3981 } 3982 else 3983 { 3984 // We need to lookup a symbol using the original nlist 3985 // symbol index since this index is coming from the 3986 // S_SYMBOL_STUBS 3987 stub_symbol = symtab->FindSymbolByID (stub_sym_id); 3988 } 3989 3990 if (stub_symbol) 3991 { 3992 Address so_addr(symbol_stub_addr, section_list); 3993 3994 if (stub_symbol->GetType() == eSymbolTypeUndefined) 3995 { 3996 // Change the external symbol into a trampoline that makes sense 3997 // These symbols were N_UNDF N_EXT, and are useless to us, so we 3998 // can re-use them so we don't have to make up a synthetic symbol 3999 // for no good reason. 4000 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 4001 stub_symbol->SetType (eSymbolTypeTrampoline); 4002 else 4003 stub_symbol->SetType (eSymbolTypeResolver); 4004 stub_symbol->SetExternal (false); 4005 stub_symbol->GetAddress() = so_addr; 4006 stub_symbol->SetByteSize (symbol_stub_byte_size); 4007 } 4008 else 4009 { 4010 // Make a synthetic symbol to describe the trampoline stub 4011 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4012 if (sym_idx >= num_syms) 4013 { 4014 sym = symtab->Resize (++num_syms); 4015 stub_symbol = NULL; // this pointer no longer valid 4016 } 4017 sym[sym_idx].SetID (synthetic_sym_id++); 4018 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4019 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 4020 sym[sym_idx].SetType (eSymbolTypeTrampoline); 4021 else 4022 sym[sym_idx].SetType (eSymbolTypeResolver); 4023 sym[sym_idx].SetIsSynthetic (true); 4024 sym[sym_idx].GetAddress() = so_addr; 4025 sym[sym_idx].SetByteSize (symbol_stub_byte_size); 4026 ++sym_idx; 4027 } 4028 } 4029 else 4030 { 4031 if (log) 4032 log->Warning ("symbol stub referencing symbol table symbol %u that isn't in our minimal symbol table, fix this!!!", stub_sym_id); 4033 } 4034 } 4035 } 4036 } 4037 } 4038 } 4039 } 4040 4041 4042 if (!trie_entries.empty()) 4043 { 4044 for (const auto &e : trie_entries) 4045 { 4046 if (e.entry.import_name) 4047 { 4048 // Make a synthetic symbol to describe re-exported symbol. 4049 if (sym_idx >= num_syms) 4050 sym = symtab->Resize (++num_syms); 4051 sym[sym_idx].SetID (synthetic_sym_id++); 4052 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4053 sym[sym_idx].SetType (eSymbolTypeReExported); 4054 sym[sym_idx].SetIsSynthetic (true); 4055 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4056 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) 4057 { 4058 sym[sym_idx].SetReExportedSymbolSharedLibrary(dylib_files.GetFileSpecAtIndex(e.entry.other-1)); 4059 } 4060 ++sym_idx; 4061 } 4062 } 4063 } 4064 4065 4066 4067 // StreamFile s(stdout, false); 4068 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4069 // symtab->Dump(&s, NULL, eSortOrderNone); 4070 // Set symbol byte sizes correctly since mach-o nlist entries don't have sizes 4071 symtab->CalculateSymbolSizes(); 4072 4073 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4074 // symtab->Dump(&s, NULL, eSortOrderNone); 4075 4076 return symtab->GetNumSymbols(); 4077 } 4078 return 0; 4079 } 4080 4081 4082 void 4083 ObjectFileMachO::Dump (Stream *s) 4084 { 4085 ModuleSP module_sp(GetModule()); 4086 if (module_sp) 4087 { 4088 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4089 s->Printf("%p: ", static_cast<void*>(this)); 4090 s->Indent(); 4091 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4092 s->PutCString("ObjectFileMachO64"); 4093 else 4094 s->PutCString("ObjectFileMachO32"); 4095 4096 ArchSpec header_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 4097 4098 *s << ", file = '" << m_file << "', arch = " << header_arch.GetArchitectureName() << "\n"; 4099 4100 SectionList *sections = GetSectionList(); 4101 if (sections) 4102 sections->Dump(s, NULL, true, UINT32_MAX); 4103 4104 if (m_symtab_ap.get()) 4105 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4106 } 4107 } 4108 4109 bool 4110 ObjectFileMachO::GetUUID (const llvm::MachO::mach_header &header, 4111 const lldb_private::DataExtractor &data, 4112 lldb::offset_t lc_offset, 4113 lldb_private::UUID& uuid) 4114 { 4115 uint32_t i; 4116 struct uuid_command load_cmd; 4117 4118 lldb::offset_t offset = lc_offset; 4119 for (i=0; i<header.ncmds; ++i) 4120 { 4121 const lldb::offset_t cmd_offset = offset; 4122 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4123 break; 4124 4125 if (load_cmd.cmd == LC_UUID) 4126 { 4127 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4128 4129 if (uuid_bytes) 4130 { 4131 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4132 // We pretend these object files have no UUID to prevent crashing. 4133 4134 const uint8_t opencl_uuid[] = { 0x8c, 0x8e, 0xb3, 0x9b, 4135 0x3b, 0xa8, 4136 0x4b, 0x16, 4137 0xb6, 0xa4, 4138 0x27, 0x63, 0xbb, 0x14, 0xf0, 0x0d }; 4139 4140 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4141 return false; 4142 4143 uuid.SetBytes (uuid_bytes); 4144 return true; 4145 } 4146 return false; 4147 } 4148 offset = cmd_offset + load_cmd.cmdsize; 4149 } 4150 return false; 4151 } 4152 4153 bool 4154 ObjectFileMachO::GetUUID (lldb_private::UUID* uuid) 4155 { 4156 ModuleSP module_sp(GetModule()); 4157 if (module_sp) 4158 { 4159 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4160 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4161 return GetUUID (m_header, m_data, offset, *uuid); 4162 } 4163 return false; 4164 } 4165 4166 4167 uint32_t 4168 ObjectFileMachO::GetDependentModules (FileSpecList& files) 4169 { 4170 uint32_t count = 0; 4171 ModuleSP module_sp(GetModule()); 4172 if (module_sp) 4173 { 4174 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4175 struct load_command load_cmd; 4176 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4177 const bool resolve_path = false; // Don't resolve the dependend file paths since they may not reside on this system 4178 uint32_t i; 4179 for (i=0; i<m_header.ncmds; ++i) 4180 { 4181 const uint32_t cmd_offset = offset; 4182 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4183 break; 4184 4185 switch (load_cmd.cmd) 4186 { 4187 case LC_LOAD_DYLIB: 4188 case LC_LOAD_WEAK_DYLIB: 4189 case LC_REEXPORT_DYLIB: 4190 case LC_LOAD_DYLINKER: 4191 case LC_LOADFVMLIB: 4192 case LC_LOAD_UPWARD_DYLIB: 4193 { 4194 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 4195 const char *path = m_data.PeekCStr(name_offset); 4196 // Skip any path that starts with '@' since these are usually: 4197 // @executable_path/.../file 4198 // @rpath/.../file 4199 if (path && path[0] != '@') 4200 { 4201 FileSpec file_spec(path, resolve_path); 4202 if (files.AppendIfUnique(file_spec)) 4203 count++; 4204 } 4205 } 4206 break; 4207 4208 default: 4209 break; 4210 } 4211 offset = cmd_offset + load_cmd.cmdsize; 4212 } 4213 } 4214 return count; 4215 } 4216 4217 lldb_private::Address 4218 ObjectFileMachO::GetEntryPointAddress () 4219 { 4220 // If the object file is not an executable it can't hold the entry point. m_entry_point_address 4221 // is initialized to an invalid address, so we can just return that. 4222 // If m_entry_point_address is valid it means we've found it already, so return the cached value. 4223 4224 if (!IsExecutable() || m_entry_point_address.IsValid()) 4225 return m_entry_point_address; 4226 4227 // Otherwise, look for the UnixThread or Thread command. The data for the Thread command is given in 4228 // /usr/include/mach-o.h, but it is basically: 4229 // 4230 // uint32_t flavor - this is the flavor argument you would pass to thread_get_state 4231 // uint32_t count - this is the count of longs in the thread state data 4232 // struct XXX_thread_state state - this is the structure from <machine/thread_status.h> corresponding to the flavor. 4233 // <repeat this trio> 4234 // 4235 // So we just keep reading the various register flavors till we find the GPR one, then read the PC out of there. 4236 // FIXME: We will need to have a "RegisterContext data provider" class at some point that can get all the registers 4237 // out of data in this form & attach them to a given thread. That should underlie the MacOS X User process plugin, 4238 // and we'll also need it for the MacOS X Core File process plugin. When we have that we can also use it here. 4239 // 4240 // For now we hard-code the offsets and flavors we need: 4241 // 4242 // 4243 4244 ModuleSP module_sp(GetModule()); 4245 if (module_sp) 4246 { 4247 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4248 struct load_command load_cmd; 4249 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4250 uint32_t i; 4251 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 4252 bool done = false; 4253 4254 for (i=0; i<m_header.ncmds; ++i) 4255 { 4256 const lldb::offset_t cmd_offset = offset; 4257 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4258 break; 4259 4260 switch (load_cmd.cmd) 4261 { 4262 case LC_UNIXTHREAD: 4263 case LC_THREAD: 4264 { 4265 while (offset < cmd_offset + load_cmd.cmdsize) 4266 { 4267 uint32_t flavor = m_data.GetU32(&offset); 4268 uint32_t count = m_data.GetU32(&offset); 4269 if (count == 0) 4270 { 4271 // We've gotten off somehow, log and exit; 4272 return m_entry_point_address; 4273 } 4274 4275 switch (m_header.cputype) 4276 { 4277 case llvm::MachO::CPU_TYPE_ARM: 4278 if (flavor == 1) // ARM_THREAD_STATE from mach/arm/thread_status.h 4279 { 4280 offset += 60; // This is the offset of pc in the GPR thread state data structure. 4281 start_address = m_data.GetU32(&offset); 4282 done = true; 4283 } 4284 break; 4285 case llvm::MachO::CPU_TYPE_ARM64: 4286 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 4287 { 4288 offset += 256; // This is the offset of pc in the GPR thread state data structure. 4289 start_address = m_data.GetU64(&offset); 4290 done = true; 4291 } 4292 break; 4293 case llvm::MachO::CPU_TYPE_I386: 4294 if (flavor == 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 4295 { 4296 offset += 40; // This is the offset of eip in the GPR thread state data structure. 4297 start_address = m_data.GetU32(&offset); 4298 done = true; 4299 } 4300 break; 4301 case llvm::MachO::CPU_TYPE_X86_64: 4302 if (flavor == 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 4303 { 4304 offset += 16 * 8; // This is the offset of rip in the GPR thread state data structure. 4305 start_address = m_data.GetU64(&offset); 4306 done = true; 4307 } 4308 break; 4309 default: 4310 return m_entry_point_address; 4311 } 4312 // Haven't found the GPR flavor yet, skip over the data for this flavor: 4313 if (done) 4314 break; 4315 offset += count * 4; 4316 } 4317 } 4318 break; 4319 case LC_MAIN: 4320 { 4321 ConstString text_segment_name ("__TEXT"); 4322 uint64_t entryoffset = m_data.GetU64(&offset); 4323 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 4324 if (text_segment_sp) 4325 { 4326 done = true; 4327 start_address = text_segment_sp->GetFileAddress() + entryoffset; 4328 } 4329 } 4330 4331 default: 4332 break; 4333 } 4334 if (done) 4335 break; 4336 4337 // Go to the next load command: 4338 offset = cmd_offset + load_cmd.cmdsize; 4339 } 4340 4341 if (start_address != LLDB_INVALID_ADDRESS) 4342 { 4343 // We got the start address from the load commands, so now resolve that address in the sections 4344 // of this ObjectFile: 4345 if (!m_entry_point_address.ResolveAddressUsingFileSections (start_address, GetSectionList())) 4346 { 4347 m_entry_point_address.Clear(); 4348 } 4349 } 4350 else 4351 { 4352 // We couldn't read the UnixThread load command - maybe it wasn't there. As a fallback look for the 4353 // "start" symbol in the main executable. 4354 4355 ModuleSP module_sp (GetModule()); 4356 4357 if (module_sp) 4358 { 4359 SymbolContextList contexts; 4360 SymbolContext context; 4361 if (module_sp->FindSymbolsWithNameAndType(ConstString ("start"), eSymbolTypeCode, contexts)) 4362 { 4363 if (contexts.GetContextAtIndex(0, context)) 4364 m_entry_point_address = context.symbol->GetAddress(); 4365 } 4366 } 4367 } 4368 } 4369 4370 return m_entry_point_address; 4371 4372 } 4373 4374 lldb_private::Address 4375 ObjectFileMachO::GetHeaderAddress () 4376 { 4377 lldb_private::Address header_addr; 4378 SectionList *section_list = GetSectionList(); 4379 if (section_list) 4380 { 4381 SectionSP text_segment_sp (section_list->FindSectionByName (GetSegmentNameTEXT())); 4382 if (text_segment_sp) 4383 { 4384 header_addr.SetSection (text_segment_sp); 4385 header_addr.SetOffset (0); 4386 } 4387 } 4388 return header_addr; 4389 } 4390 4391 uint32_t 4392 ObjectFileMachO::GetNumThreadContexts () 4393 { 4394 ModuleSP module_sp(GetModule()); 4395 if (module_sp) 4396 { 4397 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4398 if (!m_thread_context_offsets_valid) 4399 { 4400 m_thread_context_offsets_valid = true; 4401 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4402 FileRangeArray::Entry file_range; 4403 thread_command thread_cmd; 4404 for (uint32_t i=0; i<m_header.ncmds; ++i) 4405 { 4406 const uint32_t cmd_offset = offset; 4407 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 4408 break; 4409 4410 if (thread_cmd.cmd == LC_THREAD) 4411 { 4412 file_range.SetRangeBase (offset); 4413 file_range.SetByteSize (thread_cmd.cmdsize - 8); 4414 m_thread_context_offsets.Append (file_range); 4415 } 4416 offset = cmd_offset + thread_cmd.cmdsize; 4417 } 4418 } 4419 } 4420 return m_thread_context_offsets.GetSize(); 4421 } 4422 4423 lldb::RegisterContextSP 4424 ObjectFileMachO::GetThreadContextAtIndex (uint32_t idx, lldb_private::Thread &thread) 4425 { 4426 lldb::RegisterContextSP reg_ctx_sp; 4427 4428 ModuleSP module_sp(GetModule()); 4429 if (module_sp) 4430 { 4431 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4432 if (!m_thread_context_offsets_valid) 4433 GetNumThreadContexts (); 4434 4435 const FileRangeArray::Entry *thread_context_file_range = m_thread_context_offsets.GetEntryAtIndex (idx); 4436 if (thread_context_file_range) 4437 { 4438 4439 DataExtractor data (m_data, 4440 thread_context_file_range->GetRangeBase(), 4441 thread_context_file_range->GetByteSize()); 4442 4443 switch (m_header.cputype) 4444 { 4445 case llvm::MachO::CPU_TYPE_ARM64: 4446 reg_ctx_sp.reset (new RegisterContextDarwin_arm64_Mach (thread, data)); 4447 break; 4448 4449 case llvm::MachO::CPU_TYPE_ARM: 4450 reg_ctx_sp.reset (new RegisterContextDarwin_arm_Mach (thread, data)); 4451 break; 4452 4453 case llvm::MachO::CPU_TYPE_I386: 4454 reg_ctx_sp.reset (new RegisterContextDarwin_i386_Mach (thread, data)); 4455 break; 4456 4457 case llvm::MachO::CPU_TYPE_X86_64: 4458 reg_ctx_sp.reset (new RegisterContextDarwin_x86_64_Mach (thread, data)); 4459 break; 4460 } 4461 } 4462 } 4463 return reg_ctx_sp; 4464 } 4465 4466 4467 ObjectFile::Type 4468 ObjectFileMachO::CalculateType() 4469 { 4470 switch (m_header.filetype) 4471 { 4472 case MH_OBJECT: // 0x1u 4473 if (GetAddressByteSize () == 4) 4474 { 4475 // 32 bit kexts are just object files, but they do have a valid 4476 // UUID load command. 4477 UUID uuid; 4478 if (GetUUID(&uuid)) 4479 { 4480 // this checking for the UUID load command is not enough 4481 // we could eventually look for the symbol named 4482 // "OSKextGetCurrentIdentifier" as this is required of kexts 4483 if (m_strata == eStrataInvalid) 4484 m_strata = eStrataKernel; 4485 return eTypeSharedLibrary; 4486 } 4487 } 4488 return eTypeObjectFile; 4489 4490 case MH_EXECUTE: return eTypeExecutable; // 0x2u 4491 case MH_FVMLIB: return eTypeSharedLibrary; // 0x3u 4492 case MH_CORE: return eTypeCoreFile; // 0x4u 4493 case MH_PRELOAD: return eTypeSharedLibrary; // 0x5u 4494 case MH_DYLIB: return eTypeSharedLibrary; // 0x6u 4495 case MH_DYLINKER: return eTypeDynamicLinker; // 0x7u 4496 case MH_BUNDLE: return eTypeSharedLibrary; // 0x8u 4497 case MH_DYLIB_STUB: return eTypeStubLibrary; // 0x9u 4498 case MH_DSYM: return eTypeDebugInfo; // 0xAu 4499 case MH_KEXT_BUNDLE: return eTypeSharedLibrary; // 0xBu 4500 default: 4501 break; 4502 } 4503 return eTypeUnknown; 4504 } 4505 4506 ObjectFile::Strata 4507 ObjectFileMachO::CalculateStrata() 4508 { 4509 switch (m_header.filetype) 4510 { 4511 case MH_OBJECT: // 0x1u 4512 { 4513 // 32 bit kexts are just object files, but they do have a valid 4514 // UUID load command. 4515 UUID uuid; 4516 if (GetUUID(&uuid)) 4517 { 4518 // this checking for the UUID load command is not enough 4519 // we could eventually look for the symbol named 4520 // "OSKextGetCurrentIdentifier" as this is required of kexts 4521 if (m_type == eTypeInvalid) 4522 m_type = eTypeSharedLibrary; 4523 4524 return eStrataKernel; 4525 } 4526 } 4527 return eStrataUnknown; 4528 4529 case MH_EXECUTE: // 0x2u 4530 // Check for the MH_DYLDLINK bit in the flags 4531 if (m_header.flags & MH_DYLDLINK) 4532 { 4533 return eStrataUser; 4534 } 4535 else 4536 { 4537 SectionList *section_list = GetSectionList(); 4538 if (section_list) 4539 { 4540 static ConstString g_kld_section_name ("__KLD"); 4541 if (section_list->FindSectionByName(g_kld_section_name)) 4542 return eStrataKernel; 4543 } 4544 } 4545 return eStrataRawImage; 4546 4547 case MH_FVMLIB: return eStrataUser; // 0x3u 4548 case MH_CORE: return eStrataUnknown; // 0x4u 4549 case MH_PRELOAD: return eStrataRawImage; // 0x5u 4550 case MH_DYLIB: return eStrataUser; // 0x6u 4551 case MH_DYLINKER: return eStrataUser; // 0x7u 4552 case MH_BUNDLE: return eStrataUser; // 0x8u 4553 case MH_DYLIB_STUB: return eStrataUser; // 0x9u 4554 case MH_DSYM: return eStrataUnknown; // 0xAu 4555 case MH_KEXT_BUNDLE: return eStrataKernel; // 0xBu 4556 default: 4557 break; 4558 } 4559 return eStrataUnknown; 4560 } 4561 4562 4563 uint32_t 4564 ObjectFileMachO::GetVersion (uint32_t *versions, uint32_t num_versions) 4565 { 4566 ModuleSP module_sp(GetModule()); 4567 if (module_sp) 4568 { 4569 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4570 struct dylib_command load_cmd; 4571 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4572 uint32_t version_cmd = 0; 4573 uint64_t version = 0; 4574 uint32_t i; 4575 for (i=0; i<m_header.ncmds; ++i) 4576 { 4577 const lldb::offset_t cmd_offset = offset; 4578 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4579 break; 4580 4581 if (load_cmd.cmd == LC_ID_DYLIB) 4582 { 4583 if (version_cmd == 0) 4584 { 4585 version_cmd = load_cmd.cmd; 4586 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 4587 break; 4588 version = load_cmd.dylib.current_version; 4589 } 4590 break; // Break for now unless there is another more complete version 4591 // number load command in the future. 4592 } 4593 offset = cmd_offset + load_cmd.cmdsize; 4594 } 4595 4596 if (version_cmd == LC_ID_DYLIB) 4597 { 4598 if (versions != NULL && num_versions > 0) 4599 { 4600 if (num_versions > 0) 4601 versions[0] = (version & 0xFFFF0000ull) >> 16; 4602 if (num_versions > 1) 4603 versions[1] = (version & 0x0000FF00ull) >> 8; 4604 if (num_versions > 2) 4605 versions[2] = (version & 0x000000FFull); 4606 // Fill in an remaining version numbers with invalid values 4607 for (i=3; i<num_versions; ++i) 4608 versions[i] = UINT32_MAX; 4609 } 4610 // The LC_ID_DYLIB load command has a version with 3 version numbers 4611 // in it, so always return 3 4612 return 3; 4613 } 4614 } 4615 return false; 4616 } 4617 4618 bool 4619 ObjectFileMachO::GetArchitecture (ArchSpec &arch) 4620 { 4621 ModuleSP module_sp(GetModule()); 4622 if (module_sp) 4623 { 4624 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4625 arch.SetArchitecture (eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 4626 4627 // Files with type MH_PRELOAD are currently used in cases where the image 4628 // debugs at the addresses in the file itself. Below we set the OS to 4629 // unknown to make sure we use the DynamicLoaderStatic()... 4630 if (m_header.filetype == MH_PRELOAD) 4631 { 4632 arch.GetTriple().setOS (llvm::Triple::UnknownOS); 4633 } 4634 return true; 4635 } 4636 return false; 4637 } 4638 4639 4640 UUID 4641 ObjectFileMachO::GetProcessSharedCacheUUID (Process *process) 4642 { 4643 UUID uuid; 4644 if (process) 4645 { 4646 addr_t all_image_infos = process->GetImageInfoAddress(); 4647 4648 // The address returned by GetImageInfoAddress may be the address of dyld (don't want) 4649 // or it may be the address of the dyld_all_image_infos structure (want). The first four 4650 // bytes will be either the version field (all_image_infos) or a Mach-O file magic constant. 4651 // Version 13 and higher of dyld_all_image_infos is required to get the sharedCacheUUID field. 4652 4653 Error err; 4654 uint32_t version_or_magic = process->ReadUnsignedIntegerFromMemory (all_image_infos, 4, -1, err); 4655 if (version_or_magic != static_cast<uint32_t>(-1) 4656 && version_or_magic != MH_MAGIC 4657 && version_or_magic != MH_CIGAM 4658 && version_or_magic != MH_MAGIC_64 4659 && version_or_magic != MH_CIGAM_64 4660 && version_or_magic >= 13) 4661 { 4662 addr_t sharedCacheUUID_address = LLDB_INVALID_ADDRESS; 4663 int wordsize = process->GetAddressByteSize(); 4664 if (wordsize == 8) 4665 { 4666 sharedCacheUUID_address = all_image_infos + 160; // sharedCacheUUID <mach-o/dyld_images.h> 4667 } 4668 if (wordsize == 4) 4669 { 4670 sharedCacheUUID_address = all_image_infos + 84; // sharedCacheUUID <mach-o/dyld_images.h> 4671 } 4672 if (sharedCacheUUID_address != LLDB_INVALID_ADDRESS) 4673 { 4674 uuid_t shared_cache_uuid; 4675 if (process->ReadMemory (sharedCacheUUID_address, shared_cache_uuid, sizeof (uuid_t), err) == sizeof (uuid_t)) 4676 { 4677 uuid.SetBytes (shared_cache_uuid); 4678 } 4679 } 4680 } 4681 } 4682 return uuid; 4683 } 4684 4685 UUID 4686 ObjectFileMachO::GetLLDBSharedCacheUUID () 4687 { 4688 UUID uuid; 4689 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__)) 4690 uint8_t *(*dyld_get_all_image_infos)(void); 4691 dyld_get_all_image_infos = (uint8_t*(*)()) dlsym (RTLD_DEFAULT, "_dyld_get_all_image_infos"); 4692 if (dyld_get_all_image_infos) 4693 { 4694 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 4695 if (dyld_all_image_infos_address) 4696 { 4697 uint32_t *version = (uint32_t*) dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 4698 if (*version >= 13) 4699 { 4700 uuid_t *sharedCacheUUID_address = 0; 4701 int wordsize = sizeof (uint8_t *); 4702 if (wordsize == 8) 4703 { 4704 sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 160); // sharedCacheUUID <mach-o/dyld_images.h> 4705 } 4706 else 4707 { 4708 sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 84); // sharedCacheUUID <mach-o/dyld_images.h> 4709 } 4710 uuid.SetBytes (sharedCacheUUID_address); 4711 } 4712 } 4713 } 4714 #endif 4715 return uuid; 4716 } 4717 4718 uint32_t 4719 ObjectFileMachO::GetMinimumOSVersion (uint32_t *versions, uint32_t num_versions) 4720 { 4721 if (m_min_os_versions.empty()) 4722 { 4723 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4724 bool success = false; 4725 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 4726 { 4727 const lldb::offset_t load_cmd_offset = offset; 4728 4729 version_min_command lc; 4730 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 4731 break; 4732 if (lc.cmd == LC_VERSION_MIN_MACOSX || lc.cmd == LC_VERSION_MIN_IPHONEOS) 4733 { 4734 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 4735 { 4736 const uint32_t xxxx = lc.version >> 16; 4737 const uint32_t yy = (lc.version >> 8) & 0xffu; 4738 const uint32_t zz = lc.version & 0xffu; 4739 if (xxxx) 4740 { 4741 m_min_os_versions.push_back(xxxx); 4742 if (yy) 4743 { 4744 m_min_os_versions.push_back(yy); 4745 if (zz) 4746 m_min_os_versions.push_back(zz); 4747 } 4748 } 4749 success = true; 4750 } 4751 } 4752 offset = load_cmd_offset + lc.cmdsize; 4753 } 4754 4755 if (success == false) 4756 { 4757 // Push an invalid value so we don't keep trying to 4758 m_min_os_versions.push_back(UINT32_MAX); 4759 } 4760 } 4761 4762 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) 4763 { 4764 if (versions != NULL && num_versions > 0) 4765 { 4766 for (size_t i=0; i<num_versions; ++i) 4767 { 4768 if (i < m_min_os_versions.size()) 4769 versions[i] = m_min_os_versions[i]; 4770 else 4771 versions[i] = 0; 4772 } 4773 } 4774 return m_min_os_versions.size(); 4775 } 4776 // Call the superclasses version that will empty out the data 4777 return ObjectFile::GetMinimumOSVersion (versions, num_versions); 4778 } 4779 4780 uint32_t 4781 ObjectFileMachO::GetSDKVersion(uint32_t *versions, uint32_t num_versions) 4782 { 4783 if (m_sdk_versions.empty()) 4784 { 4785 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4786 bool success = false; 4787 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 4788 { 4789 const lldb::offset_t load_cmd_offset = offset; 4790 4791 version_min_command lc; 4792 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 4793 break; 4794 if (lc.cmd == LC_VERSION_MIN_MACOSX || lc.cmd == LC_VERSION_MIN_IPHONEOS) 4795 { 4796 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 4797 { 4798 const uint32_t xxxx = lc.reserved >> 16; 4799 const uint32_t yy = (lc.reserved >> 8) & 0xffu; 4800 const uint32_t zz = lc.reserved & 0xffu; 4801 if (xxxx) 4802 { 4803 m_sdk_versions.push_back(xxxx); 4804 if (yy) 4805 { 4806 m_sdk_versions.push_back(yy); 4807 if (zz) 4808 m_sdk_versions.push_back(zz); 4809 } 4810 } 4811 success = true; 4812 } 4813 } 4814 offset = load_cmd_offset + lc.cmdsize; 4815 } 4816 4817 if (success == false) 4818 { 4819 // Push an invalid value so we don't keep trying to 4820 m_sdk_versions.push_back(UINT32_MAX); 4821 } 4822 } 4823 4824 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) 4825 { 4826 if (versions != NULL && num_versions > 0) 4827 { 4828 for (size_t i=0; i<num_versions; ++i) 4829 { 4830 if (i < m_sdk_versions.size()) 4831 versions[i] = m_sdk_versions[i]; 4832 else 4833 versions[i] = 0; 4834 } 4835 } 4836 return m_sdk_versions.size(); 4837 } 4838 // Call the superclasses version that will empty out the data 4839 return ObjectFile::GetSDKVersion (versions, num_versions); 4840 } 4841 4842 4843 //------------------------------------------------------------------ 4844 // PluginInterface protocol 4845 //------------------------------------------------------------------ 4846 lldb_private::ConstString 4847 ObjectFileMachO::GetPluginName() 4848 { 4849 return GetPluginNameStatic(); 4850 } 4851 4852 uint32_t 4853 ObjectFileMachO::GetPluginVersion() 4854 { 4855 return 1; 4856 } 4857 4858 4859 bool 4860 ObjectFileMachO::SetLoadAddress (Target &target, 4861 lldb::addr_t value, 4862 bool value_is_offset) 4863 { 4864 bool changed = false; 4865 ModuleSP module_sp = GetModule(); 4866 if (module_sp) 4867 { 4868 size_t num_loaded_sections = 0; 4869 SectionList *section_list = GetSectionList (); 4870 if (section_list) 4871 { 4872 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 4873 const size_t num_sections = section_list->GetSize(); 4874 4875 const bool is_memory_image = (bool)m_process_wp.lock(); 4876 const Strata strata = GetStrata(); 4877 static ConstString g_linkedit_segname ("__LINKEDIT"); 4878 if (value_is_offset) 4879 { 4880 // "value" is an offset to apply to each top level segment 4881 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) 4882 { 4883 // Iterate through the object file sections to find all 4884 // of the sections that size on disk (to avoid __PAGEZERO) 4885 // and load them 4886 SectionSP section_sp (section_list->GetSectionAtIndex (sect_idx)); 4887 if (section_sp && 4888 section_sp->GetFileSize() > 0 && 4889 section_sp->IsThreadSpecific() == false && 4890 module_sp.get() == section_sp->GetModule().get()) 4891 { 4892 // Ignore __LINKEDIT and __DWARF segments 4893 if (section_sp->GetName() == g_linkedit_segname) 4894 { 4895 // Only map __LINKEDIT if we have an in memory image and this isn't 4896 // a kernel binary like a kext or mach_kernel. 4897 if (is_memory_image == false || strata == eStrataKernel) 4898 continue; 4899 } 4900 if (target.GetSectionLoadList().SetSectionLoadAddress (section_sp, section_sp->GetFileAddress() + value)) 4901 ++num_loaded_sections; 4902 } 4903 } 4904 } 4905 else 4906 { 4907 // "value" is the new base address of the mach_header, adjust each 4908 // section accordingly 4909 4910 // First find the address of the mach header which is the first non-zero 4911 // file sized section whose file offset is zero as this will be subtracted 4912 // from each other valid section's vmaddr and then get "base_addr" added to 4913 // it when loading the module in the target 4914 for (size_t sect_idx = 0; 4915 sect_idx < num_sections && mach_base_file_addr == LLDB_INVALID_ADDRESS; 4916 ++sect_idx) 4917 { 4918 // Iterate through the object file sections to find all 4919 // of the sections that size on disk (to avoid __PAGEZERO) 4920 // and load them 4921 Section *section = section_list->GetSectionAtIndex (sect_idx).get(); 4922 if (section && 4923 section->GetFileSize() > 0 && 4924 section->GetFileOffset() == 0 && 4925 section->IsThreadSpecific() == false && 4926 module_sp.get() == section->GetModule().get()) 4927 { 4928 // Ignore __LINKEDIT and __DWARF segments 4929 if (section->GetName() == g_linkedit_segname) 4930 { 4931 // Only map __LINKEDIT if we have an in memory image and this isn't 4932 // a kernel binary like a kext or mach_kernel. 4933 if (is_memory_image == false || strata == eStrataKernel) 4934 continue; 4935 } 4936 mach_base_file_addr = section->GetFileAddress(); 4937 } 4938 } 4939 4940 if (mach_base_file_addr != LLDB_INVALID_ADDRESS) 4941 { 4942 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) 4943 { 4944 // Iterate through the object file sections to find all 4945 // of the sections that size on disk (to avoid __PAGEZERO) 4946 // and load them 4947 SectionSP section_sp (section_list->GetSectionAtIndex (sect_idx)); 4948 if (section_sp && 4949 section_sp->GetFileSize() > 0 && 4950 section_sp->IsThreadSpecific() == false && 4951 module_sp.get() == section_sp->GetModule().get()) 4952 { 4953 // Ignore __LINKEDIT and __DWARF segments 4954 if (section_sp->GetName() == g_linkedit_segname) 4955 { 4956 // Only map __LINKEDIT if we have an in memory image and this isn't 4957 // a kernel binary like a kext or mach_kernel. 4958 if (is_memory_image == false || strata == eStrataKernel) 4959 continue; 4960 } 4961 if (target.GetSectionLoadList().SetSectionLoadAddress (section_sp, section_sp->GetFileAddress() - mach_base_file_addr + value)) 4962 ++num_loaded_sections; 4963 } 4964 } 4965 } 4966 } 4967 } 4968 changed = num_loaded_sections > 0; 4969 return num_loaded_sections > 0; 4970 } 4971 return changed; 4972 } 4973 4974