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