1 //===-- ObjectFileMachO.cpp -------------------------------------*- C++ -*-===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 10 #include "llvm/ADT/StringRef.h" 11 #include "llvm/Support/MachO.h" 12 13 #include "ObjectFileMachO.h" 14 15 #include "lldb/Core/ArchSpec.h" 16 #include "lldb/Core/DataBuffer.h" 17 #include "lldb/Host/FileSpec.h" 18 #include "lldb/Core/FileSpecList.h" 19 #include "lldb/Core/Module.h" 20 #include "lldb/Core/PluginManager.h" 21 #include "lldb/Core/Section.h" 22 #include "lldb/Core/StreamFile.h" 23 #include "lldb/Core/StreamString.h" 24 #include "lldb/Core/Timer.h" 25 #include "lldb/Core/UUID.h" 26 #include "lldb/Symbol/ClangNamespaceDecl.h" 27 #include "lldb/Symbol/ObjectFile.h" 28 29 30 using namespace lldb; 31 using namespace lldb_private; 32 using namespace llvm::MachO; 33 34 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 35 36 void 37 ObjectFileMachO::Initialize() 38 { 39 PluginManager::RegisterPlugin (GetPluginNameStatic(), 40 GetPluginDescriptionStatic(), 41 CreateInstance); 42 } 43 44 void 45 ObjectFileMachO::Terminate() 46 { 47 PluginManager::UnregisterPlugin (CreateInstance); 48 } 49 50 51 const char * 52 ObjectFileMachO::GetPluginNameStatic() 53 { 54 return "object-file.mach-o"; 55 } 56 57 const char * 58 ObjectFileMachO::GetPluginDescriptionStatic() 59 { 60 return "Mach-o object file reader (32 and 64 bit)"; 61 } 62 63 64 ObjectFile * 65 ObjectFileMachO::CreateInstance (Module* module, DataBufferSP& dataSP, const FileSpec* file, addr_t offset, addr_t length) 66 { 67 if (ObjectFileMachO::MagicBytesMatch(dataSP)) 68 { 69 std::auto_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module, dataSP, file, offset, length)); 70 if (objfile_ap.get() && objfile_ap->ParseHeader()) 71 return objfile_ap.release(); 72 } 73 return NULL; 74 } 75 76 77 static uint32_t 78 MachHeaderSizeFromMagic(uint32_t magic) 79 { 80 switch (magic) 81 { 82 case HeaderMagic32: 83 case HeaderMagic32Swapped: 84 return sizeof(struct mach_header); 85 86 case HeaderMagic64: 87 case HeaderMagic64Swapped: 88 return sizeof(struct mach_header_64); 89 break; 90 91 default: 92 break; 93 } 94 return 0; 95 } 96 97 98 bool 99 ObjectFileMachO::MagicBytesMatch (DataBufferSP& dataSP) 100 { 101 DataExtractor data(dataSP, lldb::endian::InlHostByteOrder(), 4); 102 uint32_t offset = 0; 103 uint32_t magic = data.GetU32(&offset); 104 return MachHeaderSizeFromMagic(magic) != 0; 105 } 106 107 108 ObjectFileMachO::ObjectFileMachO(Module* module, DataBufferSP& dataSP, const FileSpec* file, addr_t offset, addr_t length) : 109 ObjectFile(module, file, offset, length, dataSP), 110 m_mutex (Mutex::eMutexTypeRecursive), 111 m_header(), 112 m_sections_ap(), 113 m_symtab_ap(), 114 m_entry_point_address () 115 { 116 ::memset (&m_header, 0, sizeof(m_header)); 117 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 118 } 119 120 121 ObjectFileMachO::~ObjectFileMachO() 122 { 123 } 124 125 126 bool 127 ObjectFileMachO::ParseHeader () 128 { 129 lldb_private::Mutex::Locker locker(m_mutex); 130 bool can_parse = false; 131 uint32_t offset = 0; 132 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 133 // Leave magic in the original byte order 134 m_header.magic = m_data.GetU32(&offset); 135 switch (m_header.magic) 136 { 137 case HeaderMagic32: 138 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 139 m_data.SetAddressByteSize(4); 140 can_parse = true; 141 break; 142 143 case HeaderMagic64: 144 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 145 m_data.SetAddressByteSize(8); 146 can_parse = true; 147 break; 148 149 case HeaderMagic32Swapped: 150 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 151 m_data.SetAddressByteSize(4); 152 can_parse = true; 153 break; 154 155 case HeaderMagic64Swapped: 156 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 157 m_data.SetAddressByteSize(8); 158 can_parse = true; 159 break; 160 161 default: 162 break; 163 } 164 165 if (can_parse) 166 { 167 m_data.GetU32(&offset, &m_header.cputype, 6); 168 169 ArchSpec mach_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 170 171 if (SetModulesArchitecture (mach_arch)) 172 { 173 // Read in all only the load command data 174 DataBufferSP data_sp(m_file.ReadFileContents(m_offset, m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic))); 175 m_data.SetData (data_sp); 176 return true; 177 } 178 } 179 else 180 { 181 memset(&m_header, 0, sizeof(struct mach_header)); 182 } 183 return false; 184 } 185 186 187 ByteOrder 188 ObjectFileMachO::GetByteOrder () const 189 { 190 lldb_private::Mutex::Locker locker(m_mutex); 191 return m_data.GetByteOrder (); 192 } 193 194 bool 195 ObjectFileMachO::IsExecutable() const 196 { 197 return m_header.filetype == HeaderFileTypeExecutable; 198 } 199 200 size_t 201 ObjectFileMachO::GetAddressByteSize () const 202 { 203 lldb_private::Mutex::Locker locker(m_mutex); 204 return m_data.GetAddressByteSize (); 205 } 206 207 AddressClass 208 ObjectFileMachO::GetAddressClass (lldb::addr_t file_addr) 209 { 210 Symtab *symtab = GetSymtab(); 211 if (symtab) 212 { 213 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 214 if (symbol) 215 { 216 const AddressRange *range_ptr = symbol->GetAddressRangePtr(); 217 if (range_ptr) 218 { 219 const Section *section = range_ptr->GetBaseAddress().GetSection(); 220 if (section) 221 { 222 const SectionType section_type = section->GetType(); 223 switch (section_type) 224 { 225 case eSectionTypeInvalid: return eAddressClassUnknown; 226 case eSectionTypeCode: 227 if (m_header.cputype == llvm::MachO::CPUTypeARM) 228 { 229 // For ARM we have a bit in the n_desc field of the symbol 230 // that tells us ARM/Thumb which is bit 0x0008. 231 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 232 return eAddressClassCodeAlternateISA; 233 } 234 return eAddressClassCode; 235 236 case eSectionTypeContainer: return eAddressClassUnknown; 237 case eSectionTypeData: 238 case eSectionTypeDataCString: 239 case eSectionTypeDataCStringPointers: 240 case eSectionTypeDataSymbolAddress: 241 case eSectionTypeData4: 242 case eSectionTypeData8: 243 case eSectionTypeData16: 244 case eSectionTypeDataPointers: 245 case eSectionTypeZeroFill: 246 case eSectionTypeDataObjCMessageRefs: 247 case eSectionTypeDataObjCCFStrings: 248 return eAddressClassData; 249 case eSectionTypeDebug: 250 case eSectionTypeDWARFDebugAbbrev: 251 case eSectionTypeDWARFDebugAranges: 252 case eSectionTypeDWARFDebugFrame: 253 case eSectionTypeDWARFDebugInfo: 254 case eSectionTypeDWARFDebugLine: 255 case eSectionTypeDWARFDebugLoc: 256 case eSectionTypeDWARFDebugMacInfo: 257 case eSectionTypeDWARFDebugPubNames: 258 case eSectionTypeDWARFDebugPubTypes: 259 case eSectionTypeDWARFDebugRanges: 260 case eSectionTypeDWARFDebugStr: 261 case eSectionTypeDWARFAppleNames: 262 case eSectionTypeDWARFAppleTypes: 263 case eSectionTypeDWARFAppleNamespaces: 264 case eSectionTypeDWARFAppleObjC: 265 return eAddressClassDebug; 266 case eSectionTypeEHFrame: return eAddressClassRuntime; 267 case eSectionTypeOther: return eAddressClassUnknown; 268 } 269 } 270 } 271 272 const SymbolType symbol_type = symbol->GetType(); 273 switch (symbol_type) 274 { 275 case eSymbolTypeAny: return eAddressClassUnknown; 276 case eSymbolTypeAbsolute: return eAddressClassUnknown; 277 278 case eSymbolTypeCode: 279 case eSymbolTypeTrampoline: 280 if (m_header.cputype == llvm::MachO::CPUTypeARM) 281 { 282 // For ARM we have a bit in the n_desc field of the symbol 283 // that tells us ARM/Thumb which is bit 0x0008. 284 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 285 return eAddressClassCodeAlternateISA; 286 } 287 return eAddressClassCode; 288 289 case eSymbolTypeData: return eAddressClassData; 290 case eSymbolTypeRuntime: return eAddressClassRuntime; 291 case eSymbolTypeException: return eAddressClassRuntime; 292 case eSymbolTypeSourceFile: return eAddressClassDebug; 293 case eSymbolTypeHeaderFile: return eAddressClassDebug; 294 case eSymbolTypeObjectFile: return eAddressClassDebug; 295 case eSymbolTypeCommonBlock: return eAddressClassDebug; 296 case eSymbolTypeBlock: return eAddressClassDebug; 297 case eSymbolTypeLocal: return eAddressClassData; 298 case eSymbolTypeParam: return eAddressClassData; 299 case eSymbolTypeVariable: return eAddressClassData; 300 case eSymbolTypeVariableType: return eAddressClassDebug; 301 case eSymbolTypeLineEntry: return eAddressClassDebug; 302 case eSymbolTypeLineHeader: return eAddressClassDebug; 303 case eSymbolTypeScopeBegin: return eAddressClassDebug; 304 case eSymbolTypeScopeEnd: return eAddressClassDebug; 305 case eSymbolTypeAdditional: return eAddressClassUnknown; 306 case eSymbolTypeCompiler: return eAddressClassDebug; 307 case eSymbolTypeInstrumentation:return eAddressClassDebug; 308 case eSymbolTypeUndefined: return eAddressClassUnknown; 309 case eSymbolTypeObjCClass: return eAddressClassRuntime; 310 case eSymbolTypeObjCMetaClass: return eAddressClassRuntime; 311 case eSymbolTypeObjCIVar: return eAddressClassRuntime; 312 } 313 } 314 } 315 return eAddressClassUnknown; 316 } 317 318 Symtab * 319 ObjectFileMachO::GetSymtab() 320 { 321 lldb_private::Mutex::Locker symfile_locker(m_mutex); 322 if (m_symtab_ap.get() == NULL) 323 { 324 m_symtab_ap.reset(new Symtab(this)); 325 Mutex::Locker symtab_locker (m_symtab_ap->GetMutex()); 326 ParseSymtab (true); 327 m_symtab_ap->Finalize (); 328 } 329 return m_symtab_ap.get(); 330 } 331 332 333 SectionList * 334 ObjectFileMachO::GetSectionList() 335 { 336 lldb_private::Mutex::Locker locker(m_mutex); 337 if (m_sections_ap.get() == NULL) 338 { 339 m_sections_ap.reset(new SectionList()); 340 ParseSections(); 341 } 342 return m_sections_ap.get(); 343 } 344 345 346 size_t 347 ObjectFileMachO::ParseSections () 348 { 349 lldb::user_id_t segID = 0; 350 lldb::user_id_t sectID = 0; 351 struct segment_command_64 load_cmd; 352 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 353 uint32_t i; 354 //bool dump_sections = false; 355 for (i=0; i<m_header.ncmds; ++i) 356 { 357 const uint32_t load_cmd_offset = offset; 358 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 359 break; 360 361 if (load_cmd.cmd == LoadCommandSegment32 || load_cmd.cmd == LoadCommandSegment64) 362 { 363 if (m_data.GetU8(&offset, (uint8_t*)load_cmd.segname, 16)) 364 { 365 load_cmd.vmaddr = m_data.GetAddress(&offset); 366 load_cmd.vmsize = m_data.GetAddress(&offset); 367 load_cmd.fileoff = m_data.GetAddress(&offset); 368 load_cmd.filesize = m_data.GetAddress(&offset); 369 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 370 { 371 372 const bool segment_is_encrypted = (load_cmd.flags & SegmentCommandFlagBitProtectedVersion1) != 0; 373 374 // Keep a list of mach segments around in case we need to 375 // get at data that isn't stored in the abstracted Sections. 376 m_mach_segments.push_back (load_cmd); 377 378 ConstString segment_name (load_cmd.segname, std::min<int>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 379 // Use a segment ID of the segment index shifted left by 8 so they 380 // never conflict with any of the sections. 381 SectionSP segment_sp; 382 if (segment_name) 383 { 384 segment_sp.reset(new Section (NULL, 385 GetModule(), // Module to which this section belongs 386 ++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 387 segment_name, // Name of this section 388 eSectionTypeContainer, // This section is a container of other sections. 389 load_cmd.vmaddr, // File VM address == addresses as they are found in the object file 390 load_cmd.vmsize, // VM size in bytes of this section 391 load_cmd.fileoff, // Offset to the data for this section in the file 392 load_cmd.filesize, // Size in bytes of this section as found in the the file 393 load_cmd.flags)); // Flags for this section 394 395 segment_sp->SetIsEncrypted (segment_is_encrypted); 396 m_sections_ap->AddSection(segment_sp); 397 } 398 399 struct section_64 sect64; 400 ::memset (§64, 0, sizeof(sect64)); 401 // Push a section into our mach sections for the section at 402 // index zero (NListSectionNoSection) if we don't have any 403 // mach sections yet... 404 if (m_mach_sections.empty()) 405 m_mach_sections.push_back(sect64); 406 uint32_t segment_sect_idx; 407 const lldb::user_id_t first_segment_sectID = sectID + 1; 408 409 410 const uint32_t num_u32s = load_cmd.cmd == LoadCommandSegment32 ? 7 : 8; 411 for (segment_sect_idx=0; segment_sect_idx<load_cmd.nsects; ++segment_sect_idx) 412 { 413 if (m_data.GetU8(&offset, (uint8_t*)sect64.sectname, sizeof(sect64.sectname)) == NULL) 414 break; 415 if (m_data.GetU8(&offset, (uint8_t*)sect64.segname, sizeof(sect64.segname)) == NULL) 416 break; 417 sect64.addr = m_data.GetAddress(&offset); 418 sect64.size = m_data.GetAddress(&offset); 419 420 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 421 break; 422 423 // Keep a list of mach sections around in case we need to 424 // get at data that isn't stored in the abstracted Sections. 425 m_mach_sections.push_back (sect64); 426 427 ConstString section_name (sect64.sectname, std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 428 if (!segment_name) 429 { 430 // We have a segment with no name so we need to conjure up 431 // segments that correspond to the section's segname if there 432 // isn't already such a section. If there is such a section, 433 // we resize the section so that it spans all sections. 434 // We also mark these sections as fake so address matches don't 435 // hit if they land in the gaps between the child sections. 436 segment_name.SetTrimmedCStringWithLength(sect64.segname, sizeof(sect64.segname)); 437 segment_sp = m_sections_ap->FindSectionByName (segment_name); 438 if (segment_sp.get()) 439 { 440 Section *segment = segment_sp.get(); 441 // Grow the section size as needed. 442 const lldb::addr_t sect64_min_addr = sect64.addr; 443 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 444 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 445 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 446 const lldb::addr_t curr_seg_max_addr = curr_seg_min_addr + curr_seg_byte_size; 447 if (sect64_min_addr >= curr_seg_min_addr) 448 { 449 const lldb::addr_t new_seg_byte_size = sect64_max_addr - curr_seg_min_addr; 450 // Only grow the section size if needed 451 if (new_seg_byte_size > curr_seg_byte_size) 452 segment->SetByteSize (new_seg_byte_size); 453 } 454 else 455 { 456 // We need to change the base address of the segment and 457 // adjust the child section offsets for all existing children. 458 const lldb::addr_t slide_amount = sect64_min_addr - curr_seg_min_addr; 459 segment->Slide(slide_amount, false); 460 segment->GetChildren().Slide (-slide_amount, false); 461 segment->SetByteSize (curr_seg_max_addr - sect64_min_addr); 462 } 463 464 // Grow the section size as needed. 465 if (sect64.offset) 466 { 467 const lldb::addr_t segment_min_file_offset = segment->GetFileOffset(); 468 const lldb::addr_t segment_max_file_offset = segment_min_file_offset + segment->GetFileSize(); 469 470 const lldb::addr_t section_min_file_offset = sect64.offset; 471 const lldb::addr_t section_max_file_offset = section_min_file_offset + sect64.size; 472 const lldb::addr_t new_file_offset = std::min (section_min_file_offset, segment_min_file_offset); 473 const lldb::addr_t new_file_size = std::max (section_max_file_offset, segment_max_file_offset) - new_file_offset; 474 segment->SetFileOffset (new_file_offset); 475 segment->SetFileSize (new_file_size); 476 } 477 } 478 else 479 { 480 // Create a fake section for the section's named segment 481 segment_sp.reset(new Section(segment_sp.get(), // Parent section 482 GetModule(), // Module to which this section belongs 483 ++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 484 segment_name, // Name of this section 485 eSectionTypeContainer, // This section is a container of other sections. 486 sect64.addr, // File VM address == addresses as they are found in the object file 487 sect64.size, // VM size in bytes of this section 488 sect64.offset, // Offset to the data for this section in the file 489 sect64.offset ? sect64.size : 0, // Size in bytes of this section as found in the the file 490 load_cmd.flags)); // Flags for this section 491 segment_sp->SetIsFake(true); 492 m_sections_ap->AddSection(segment_sp); 493 segment_sp->SetIsEncrypted (segment_is_encrypted); 494 } 495 } 496 assert (segment_sp.get()); 497 498 uint32_t mach_sect_type = sect64.flags & SectionFlagMaskSectionType; 499 static ConstString g_sect_name_objc_data ("__objc_data"); 500 static ConstString g_sect_name_objc_msgrefs ("__objc_msgrefs"); 501 static ConstString g_sect_name_objc_selrefs ("__objc_selrefs"); 502 static ConstString g_sect_name_objc_classrefs ("__objc_classrefs"); 503 static ConstString g_sect_name_objc_superrefs ("__objc_superrefs"); 504 static ConstString g_sect_name_objc_const ("__objc_const"); 505 static ConstString g_sect_name_objc_classlist ("__objc_classlist"); 506 static ConstString g_sect_name_cfstring ("__cfstring"); 507 508 static ConstString g_sect_name_dwarf_debug_abbrev ("__debug_abbrev"); 509 static ConstString g_sect_name_dwarf_debug_aranges ("__debug_aranges"); 510 static ConstString g_sect_name_dwarf_debug_frame ("__debug_frame"); 511 static ConstString g_sect_name_dwarf_debug_info ("__debug_info"); 512 static ConstString g_sect_name_dwarf_debug_line ("__debug_line"); 513 static ConstString g_sect_name_dwarf_debug_loc ("__debug_loc"); 514 static ConstString g_sect_name_dwarf_debug_macinfo ("__debug_macinfo"); 515 static ConstString g_sect_name_dwarf_debug_pubnames ("__debug_pubnames"); 516 static ConstString g_sect_name_dwarf_debug_pubtypes ("__debug_pubtypes"); 517 static ConstString g_sect_name_dwarf_debug_ranges ("__debug_ranges"); 518 static ConstString g_sect_name_dwarf_debug_str ("__debug_str"); 519 static ConstString g_sect_name_dwarf_apple_names ("__apple_names"); 520 static ConstString g_sect_name_dwarf_apple_types ("__apple_types"); 521 static ConstString g_sect_name_dwarf_apple_namespaces ("__apple_namespac"); 522 static ConstString g_sect_name_dwarf_apple_objc ("__apple_objc"); 523 static ConstString g_sect_name_eh_frame ("__eh_frame"); 524 static ConstString g_sect_name_DATA ("__DATA"); 525 static ConstString g_sect_name_TEXT ("__TEXT"); 526 527 SectionType sect_type = eSectionTypeOther; 528 529 if (section_name == g_sect_name_dwarf_debug_abbrev) 530 sect_type = eSectionTypeDWARFDebugAbbrev; 531 else if (section_name == g_sect_name_dwarf_debug_aranges) 532 sect_type = eSectionTypeDWARFDebugAranges; 533 else if (section_name == g_sect_name_dwarf_debug_frame) 534 sect_type = eSectionTypeDWARFDebugFrame; 535 else if (section_name == g_sect_name_dwarf_debug_info) 536 sect_type = eSectionTypeDWARFDebugInfo; 537 else if (section_name == g_sect_name_dwarf_debug_line) 538 sect_type = eSectionTypeDWARFDebugLine; 539 else if (section_name == g_sect_name_dwarf_debug_loc) 540 sect_type = eSectionTypeDWARFDebugLoc; 541 else if (section_name == g_sect_name_dwarf_debug_macinfo) 542 sect_type = eSectionTypeDWARFDebugMacInfo; 543 else if (section_name == g_sect_name_dwarf_debug_pubnames) 544 sect_type = eSectionTypeDWARFDebugPubNames; 545 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 546 sect_type = eSectionTypeDWARFDebugPubTypes; 547 else if (section_name == g_sect_name_dwarf_debug_ranges) 548 sect_type = eSectionTypeDWARFDebugRanges; 549 else if (section_name == g_sect_name_dwarf_debug_str) 550 sect_type = eSectionTypeDWARFDebugStr; 551 else if (section_name == g_sect_name_dwarf_apple_names) 552 sect_type = eSectionTypeDWARFAppleNames; 553 else if (section_name == g_sect_name_dwarf_apple_types) 554 sect_type = eSectionTypeDWARFAppleTypes; 555 else if (section_name == g_sect_name_dwarf_apple_namespaces) 556 sect_type = eSectionTypeDWARFAppleNamespaces; 557 else if (section_name == g_sect_name_dwarf_apple_objc) 558 sect_type = eSectionTypeDWARFAppleObjC; 559 else if (section_name == g_sect_name_objc_selrefs) 560 sect_type = eSectionTypeDataCStringPointers; 561 else if (section_name == g_sect_name_objc_msgrefs) 562 sect_type = eSectionTypeDataObjCMessageRefs; 563 else if (section_name == g_sect_name_eh_frame) 564 sect_type = eSectionTypeEHFrame; 565 else if (section_name == g_sect_name_cfstring) 566 sect_type = eSectionTypeDataObjCCFStrings; 567 else if (section_name == g_sect_name_objc_data || 568 section_name == g_sect_name_objc_classrefs || 569 section_name == g_sect_name_objc_superrefs || 570 section_name == g_sect_name_objc_const || 571 section_name == g_sect_name_objc_classlist) 572 { 573 sect_type = eSectionTypeDataPointers; 574 } 575 576 if (sect_type == eSectionTypeOther) 577 { 578 switch (mach_sect_type) 579 { 580 // TODO: categorize sections by other flags for regular sections 581 case SectionTypeRegular: 582 if (segment_sp->GetName() == g_sect_name_TEXT) 583 sect_type = eSectionTypeCode; 584 else if (segment_sp->GetName() == g_sect_name_DATA) 585 sect_type = eSectionTypeData; 586 else 587 sect_type = eSectionTypeOther; 588 break; 589 case SectionTypeZeroFill: sect_type = eSectionTypeZeroFill; break; 590 case SectionTypeCStringLiterals: sect_type = eSectionTypeDataCString; break; // section with only literal C strings 591 case SectionType4ByteLiterals: sect_type = eSectionTypeData4; break; // section with only 4 byte literals 592 case SectionType8ByteLiterals: sect_type = eSectionTypeData8; break; // section with only 8 byte literals 593 case SectionTypeLiteralPointers: sect_type = eSectionTypeDataPointers; break; // section with only pointers to literals 594 case SectionTypeNonLazySymbolPointers: sect_type = eSectionTypeDataPointers; break; // section with only non-lazy symbol pointers 595 case SectionTypeLazySymbolPointers: sect_type = eSectionTypeDataPointers; break; // section with only lazy symbol pointers 596 case SectionTypeSymbolStubs: sect_type = eSectionTypeCode; break; // section with only symbol stubs, byte size of stub in the reserved2 field 597 case SectionTypeModuleInitFunctionPointers: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for initialization 598 case SectionTypeModuleTermFunctionPointers: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for termination 599 case SectionTypeCoalesced: sect_type = eSectionTypeOther; break; 600 case SectionTypeZeroFillLarge: sect_type = eSectionTypeZeroFill; break; 601 case SectionTypeInterposing: sect_type = eSectionTypeCode; break; // section with only pairs of function pointers for interposing 602 case SectionType16ByteLiterals: sect_type = eSectionTypeData16; break; // section with only 16 byte literals 603 case SectionTypeDTraceObjectFormat: sect_type = eSectionTypeDebug; break; 604 case SectionTypeLazyDylibSymbolPointers: sect_type = eSectionTypeDataPointers; break; 605 default: break; 606 } 607 } 608 609 SectionSP section_sp(new Section(segment_sp.get(), 610 GetModule(), 611 ++sectID, 612 section_name, 613 sect_type, 614 sect64.addr - segment_sp->GetFileAddress(), 615 sect64.size, 616 sect64.offset, 617 sect64.offset == 0 ? 0 : sect64.size, 618 sect64.flags)); 619 // Set the section to be encrypted to match the segment 620 section_sp->SetIsEncrypted (segment_is_encrypted); 621 622 segment_sp->GetChildren().AddSection(section_sp); 623 624 if (segment_sp->IsFake()) 625 { 626 segment_sp.reset(); 627 segment_name.Clear(); 628 } 629 } 630 if (segment_sp && m_header.filetype == HeaderFileTypeDSYM) 631 { 632 if (first_segment_sectID <= sectID) 633 { 634 lldb::user_id_t sect_uid; 635 for (sect_uid = first_segment_sectID; sect_uid <= sectID; ++sect_uid) 636 { 637 SectionSP curr_section_sp(segment_sp->GetChildren().FindSectionByID (sect_uid)); 638 SectionSP next_section_sp; 639 if (sect_uid + 1 <= sectID) 640 next_section_sp = segment_sp->GetChildren().FindSectionByID (sect_uid+1); 641 642 if (curr_section_sp.get()) 643 { 644 if (curr_section_sp->GetByteSize() == 0) 645 { 646 if (next_section_sp.get() != NULL) 647 curr_section_sp->SetByteSize ( next_section_sp->GetFileAddress() - curr_section_sp->GetFileAddress() ); 648 else 649 curr_section_sp->SetByteSize ( load_cmd.vmsize ); 650 } 651 } 652 } 653 } 654 } 655 } 656 } 657 } 658 else if (load_cmd.cmd == LoadCommandDynamicSymtabInfo) 659 { 660 m_dysymtab.cmd = load_cmd.cmd; 661 m_dysymtab.cmdsize = load_cmd.cmdsize; 662 m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 663 } 664 665 offset = load_cmd_offset + load_cmd.cmdsize; 666 } 667 // if (dump_sections) 668 // { 669 // StreamFile s(stdout); 670 // m_sections_ap->Dump(&s, true); 671 // } 672 return sectID; // Return the number of sections we registered with the module 673 } 674 675 class MachSymtabSectionInfo 676 { 677 public: 678 679 MachSymtabSectionInfo (SectionList *section_list) : 680 m_section_list (section_list), 681 m_section_infos() 682 { 683 // Get the number of sections down to a depth of 1 to include 684 // all segments and their sections, but no other sections that 685 // may be added for debug map or 686 m_section_infos.resize(section_list->GetNumSections(1)); 687 } 688 689 690 Section * 691 GetSection (uint8_t n_sect, addr_t file_addr) 692 { 693 if (n_sect == 0) 694 return NULL; 695 if (n_sect < m_section_infos.size()) 696 { 697 if (m_section_infos[n_sect].section == NULL) 698 { 699 Section *section = m_section_list->FindSectionByID (n_sect).get(); 700 m_section_infos[n_sect].section = section; 701 if (section != NULL) 702 { 703 m_section_infos[n_sect].vm_range.SetBaseAddress (section->GetFileAddress()); 704 m_section_infos[n_sect].vm_range.SetByteSize (section->GetByteSize()); 705 } 706 else 707 { 708 fprintf (stderr, "error: unable to find section for section %u\n", n_sect); 709 } 710 } 711 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) 712 { 713 // Symbol is in section. 714 return m_section_infos[n_sect].section; 715 } 716 else if (m_section_infos[n_sect].vm_range.GetByteSize () == 0 && 717 m_section_infos[n_sect].vm_range.GetBaseAddress() == file_addr) 718 { 719 // Symbol is in section with zero size, but has the same start 720 // address as the section. This can happen with linker symbols 721 // (symbols that start with the letter 'l' or 'L'. 722 return m_section_infos[n_sect].section; 723 } 724 } 725 return m_section_list->FindSectionContainingFileAddress(file_addr).get(); 726 } 727 728 protected: 729 struct SectionInfo 730 { 731 SectionInfo () : 732 vm_range(), 733 section (NULL) 734 { 735 } 736 737 VMRange vm_range; 738 Section *section; 739 }; 740 SectionList *m_section_list; 741 std::vector<SectionInfo> m_section_infos; 742 }; 743 744 745 746 size_t 747 ObjectFileMachO::ParseSymtab (bool minimize) 748 { 749 Timer scoped_timer(__PRETTY_FUNCTION__, 750 "ObjectFileMachO::ParseSymtab () module = %s", 751 m_file.GetFilename().AsCString("")); 752 struct symtab_command symtab_load_command; 753 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 754 uint32_t i; 755 for (i=0; i<m_header.ncmds; ++i) 756 { 757 const uint32_t cmd_offset = offset; 758 // Read in the load command and load command size 759 if (m_data.GetU32(&offset, &symtab_load_command, 2) == NULL) 760 break; 761 // Watch for the symbol table load command 762 if (symtab_load_command.cmd == LoadCommandSymtab) 763 { 764 // Read in the rest of the symtab load command 765 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4)) // fill in symoff, nsyms, stroff, strsize fields 766 { 767 Symtab *symtab = m_symtab_ap.get(); 768 SectionList *section_list = GetSectionList(); 769 assert(section_list); 770 const size_t addr_size = m_data.GetAddressByteSize(); 771 const ByteOrder endian = m_data.GetByteOrder(); 772 bool bit_width_32 = addr_size == 4; 773 const size_t nlist_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 774 775 DataBufferSP symtab_data_sp(m_file.ReadFileContents(m_offset + symtab_load_command.symoff, symtab_load_command.nsyms * nlist_size)); 776 DataBufferSP strtab_data_sp(m_file.ReadFileContents(m_offset + symtab_load_command.stroff, symtab_load_command.strsize)); 777 778 const char *strtab_data = (const char *)strtab_data_sp->GetBytes(); 779 const size_t strtab_data_len = strtab_data_sp->GetByteSize(); 780 781 static ConstString g_segment_name_TEXT ("__TEXT"); 782 static ConstString g_segment_name_DATA ("__DATA"); 783 static ConstString g_segment_name_OBJC ("__OBJC"); 784 static ConstString g_section_name_eh_frame ("__eh_frame"); 785 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT)); 786 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA)); 787 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC)); 788 SectionSP eh_frame_section_sp; 789 if (text_section_sp.get()) 790 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame); 791 else 792 eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame); 793 794 uint8_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() : NListSectionNoSection; 795 //uint32_t symtab_offset = 0; 796 const uint8_t* nlist_data = symtab_data_sp->GetBytes(); 797 assert (symtab_data_sp->GetByteSize()/nlist_size >= symtab_load_command.nsyms); 798 799 800 if (endian != lldb::endian::InlHostByteOrder()) 801 { 802 // ... 803 assert (!"UNIMPLEMENTED: Swap all nlist entries"); 804 } 805 uint32_t N_SO_index = UINT32_MAX; 806 807 MachSymtabSectionInfo section_info (section_list); 808 std::vector<uint32_t> N_FUN_indexes; 809 std::vector<uint32_t> N_NSYM_indexes; 810 std::vector<uint32_t> N_INCL_indexes; 811 std::vector<uint32_t> N_BRAC_indexes; 812 std::vector<uint32_t> N_COMM_indexes; 813 typedef std::map <uint64_t, uint32_t> ValueToSymbolIndexMap; 814 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap; 815 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 816 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 817 // Any symbols that get merged into another will get an entry 818 // in this map so we know 819 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 820 uint32_t nlist_idx = 0; 821 Symbol *symbol_ptr = NULL; 822 823 uint32_t sym_idx = 0; 824 Symbol *sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 825 uint32_t num_syms = symtab->GetNumSymbols(); 826 827 //symtab->Reserve (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 828 for (nlist_idx = 0; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) 829 { 830 struct nlist_64 nlist; 831 if (bit_width_32) 832 { 833 struct nlist* nlist32_ptr = (struct nlist*)(nlist_data + (nlist_idx * nlist_size)); 834 nlist.n_strx = nlist32_ptr->n_strx; 835 nlist.n_type = nlist32_ptr->n_type; 836 nlist.n_sect = nlist32_ptr->n_sect; 837 nlist.n_desc = nlist32_ptr->n_desc; 838 nlist.n_value = nlist32_ptr->n_value; 839 } 840 else 841 { 842 nlist = *((struct nlist_64*)(nlist_data + (nlist_idx * nlist_size))); 843 } 844 845 SymbolType type = eSymbolTypeInvalid; 846 if (nlist.n_strx >= strtab_data_len) 847 { 848 // No symbol should be NULL, even the symbols with no 849 // string values should have an offset zero which points 850 // to an empty C-string 851 fprintf (stderr, 852 "error: symbol[%u] has invalid string table offset 0x%x in %s/%s, ignoring symbol\n", 853 nlist_idx, 854 nlist.n_strx, 855 m_module->GetFileSpec().GetDirectory().GetCString(), 856 m_module->GetFileSpec().GetFilename().GetCString()); 857 continue; 858 } 859 const char *symbol_name = &strtab_data[nlist.n_strx]; 860 const char *symbol_name_non_abi_mangled = NULL; 861 862 if (symbol_name[0] == '\0') 863 symbol_name = NULL; 864 Section* symbol_section = NULL; 865 bool add_nlist = true; 866 bool is_debug = ((nlist.n_type & NlistMaskStab) != 0); 867 868 assert (sym_idx < num_syms); 869 870 sym[sym_idx].SetDebug (is_debug); 871 872 if (is_debug) 873 { 874 switch (nlist.n_type) 875 { 876 case StabGlobalSymbol: 877 // N_GSYM -- global symbol: name,,NO_SECT,type,0 878 // Sometimes the N_GSYM value contains the address. 879 880 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 881 // have the same address, but we want to ensure that we always find only the real symbol, 882 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 883 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 884 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 885 // same address. 886 887 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 888 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 889 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 890 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 891 add_nlist = false; 892 else 893 { 894 sym[sym_idx].SetExternal(true); 895 if (nlist.n_value != 0) 896 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 897 type = eSymbolTypeData; 898 } 899 break; 900 901 case StabFunctionName: 902 // N_FNAME -- procedure name (f77 kludge): name,,NO_SECT,0,0 903 type = eSymbolTypeCompiler; 904 break; 905 906 case StabFunction: 907 // N_FUN -- procedure: name,,n_sect,linenumber,address 908 if (symbol_name) 909 { 910 type = eSymbolTypeCode; 911 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 912 913 N_FUN_addr_to_sym_idx[nlist.n_value] = sym_idx; 914 // We use the current number of symbols in the symbol table in lieu of 915 // using nlist_idx in case we ever start trimming entries out 916 N_FUN_indexes.push_back(sym_idx); 917 } 918 else 919 { 920 type = eSymbolTypeCompiler; 921 922 if ( !N_FUN_indexes.empty() ) 923 { 924 // Copy the size of the function into the original STAB entry so we don't have 925 // to hunt for it later 926 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 927 N_FUN_indexes.pop_back(); 928 // We don't really need the end function STAB as it contains the size which 929 // we already placed with the original symbol, so don't add it if we want a 930 // minimal symbol table 931 if (minimize) 932 add_nlist = false; 933 } 934 } 935 break; 936 937 case StabStaticSymbol: 938 // N_STSYM -- static symbol: name,,n_sect,type,address 939 N_STSYM_addr_to_sym_idx[nlist.n_value] = sym_idx; 940 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 941 type = eSymbolTypeData; 942 break; 943 944 case StabLocalCommon: 945 // N_LCSYM -- .lcomm symbol: name,,n_sect,type,address 946 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 947 type = eSymbolTypeCommonBlock; 948 break; 949 950 case StabBeginSymbol: 951 // N_BNSYM 952 // We use the current number of symbols in the symbol table in lieu of 953 // using nlist_idx in case we ever start trimming entries out 954 if (minimize) 955 { 956 // Skip these if we want minimal symbol tables 957 add_nlist = false; 958 } 959 else 960 { 961 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 962 N_NSYM_indexes.push_back(sym_idx); 963 type = eSymbolTypeScopeBegin; 964 } 965 break; 966 967 case StabEndSymbol: 968 // N_ENSYM 969 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 970 // so that we can always skip the entire symbol if we need to navigate 971 // more quickly at the source level when parsing STABS 972 if (minimize) 973 { 974 // Skip these if we want minimal symbol tables 975 add_nlist = false; 976 } 977 else 978 { 979 if ( !N_NSYM_indexes.empty() ) 980 { 981 symbol_ptr = symtab->SymbolAtIndex(N_NSYM_indexes.back()); 982 symbol_ptr->SetByteSize(sym_idx + 1); 983 symbol_ptr->SetSizeIsSibling(true); 984 N_NSYM_indexes.pop_back(); 985 } 986 type = eSymbolTypeScopeEnd; 987 } 988 break; 989 990 991 case StabSourceFileOptions: 992 // N_OPT - emitted with gcc2_compiled and in gcc source 993 type = eSymbolTypeCompiler; 994 break; 995 996 case StabRegisterSymbol: 997 // N_RSYM - register sym: name,,NO_SECT,type,register 998 type = eSymbolTypeVariable; 999 break; 1000 1001 case StabSourceLine: 1002 // N_SLINE - src line: 0,,n_sect,linenumber,address 1003 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1004 type = eSymbolTypeLineEntry; 1005 break; 1006 1007 case StabStructureType: 1008 // N_SSYM - structure elt: name,,NO_SECT,type,struct_offset 1009 type = eSymbolTypeVariableType; 1010 break; 1011 1012 case StabSourceFileName: 1013 // N_SO - source file name 1014 type = eSymbolTypeSourceFile; 1015 if (symbol_name == NULL) 1016 { 1017 if (minimize) 1018 add_nlist = false; 1019 if (N_SO_index != UINT32_MAX) 1020 { 1021 // Set the size of the N_SO to the terminating index of this N_SO 1022 // so that we can always skip the entire N_SO if we need to navigate 1023 // more quickly at the source level when parsing STABS 1024 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 1025 symbol_ptr->SetByteSize(sym_idx + (minimize ? 0 : 1)); 1026 symbol_ptr->SetSizeIsSibling(true); 1027 } 1028 N_NSYM_indexes.clear(); 1029 N_INCL_indexes.clear(); 1030 N_BRAC_indexes.clear(); 1031 N_COMM_indexes.clear(); 1032 N_FUN_indexes.clear(); 1033 N_SO_index = UINT32_MAX; 1034 } 1035 else 1036 { 1037 // We use the current number of symbols in the symbol table in lieu of 1038 // using nlist_idx in case we ever start trimming entries out 1039 if (symbol_name[0] == '/') 1040 N_SO_index = sym_idx; 1041 else if (minimize && (N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 1042 { 1043 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 1044 if (so_path && so_path[0]) 1045 { 1046 std::string full_so_path (so_path); 1047 if (*full_so_path.rbegin() != '/') 1048 full_so_path += '/'; 1049 full_so_path += symbol_name; 1050 sym[sym_idx - 1].GetMangled().SetValue(full_so_path.c_str(), false); 1051 add_nlist = false; 1052 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 1053 } 1054 } 1055 } 1056 1057 break; 1058 1059 case StabObjectFileName: 1060 // N_OSO - object file name: name,,0,0,st_mtime 1061 type = eSymbolTypeObjectFile; 1062 break; 1063 1064 case StabLocalSymbol: 1065 // N_LSYM - local sym: name,,NO_SECT,type,offset 1066 type = eSymbolTypeLocal; 1067 break; 1068 1069 //---------------------------------------------------------------------- 1070 // INCL scopes 1071 //---------------------------------------------------------------------- 1072 case StabBeginIncludeFileName: 1073 // N_BINCL - include file beginning: name,,NO_SECT,0,sum 1074 // We use the current number of symbols in the symbol table in lieu of 1075 // using nlist_idx in case we ever start trimming entries out 1076 N_INCL_indexes.push_back(sym_idx); 1077 type = eSymbolTypeScopeBegin; 1078 break; 1079 1080 case StabEndIncludeFile: 1081 // N_EINCL - include file end: name,,NO_SECT,0,0 1082 // Set the size of the N_BINCL to the terminating index of this N_EINCL 1083 // so that we can always skip the entire symbol if we need to navigate 1084 // more quickly at the source level when parsing STABS 1085 if ( !N_INCL_indexes.empty() ) 1086 { 1087 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 1088 symbol_ptr->SetByteSize(sym_idx + 1); 1089 symbol_ptr->SetSizeIsSibling(true); 1090 N_INCL_indexes.pop_back(); 1091 } 1092 type = eSymbolTypeScopeEnd; 1093 break; 1094 1095 case StabIncludeFileName: 1096 // N_SOL - #included file name: name,,n_sect,0,address 1097 type = eSymbolTypeHeaderFile; 1098 1099 // We currently don't use the header files on darwin 1100 if (minimize) 1101 add_nlist = false; 1102 break; 1103 1104 case StabCompilerParameters: 1105 // N_PARAMS - compiler parameters: name,,NO_SECT,0,0 1106 type = eSymbolTypeCompiler; 1107 break; 1108 1109 case StabCompilerVersion: 1110 // N_VERSION - compiler version: name,,NO_SECT,0,0 1111 type = eSymbolTypeCompiler; 1112 break; 1113 1114 case StabCompilerOptLevel: 1115 // N_OLEVEL - compiler -O level: name,,NO_SECT,0,0 1116 type = eSymbolTypeCompiler; 1117 break; 1118 1119 case StabParameter: 1120 // N_PSYM - parameter: name,,NO_SECT,type,offset 1121 type = eSymbolTypeVariable; 1122 break; 1123 1124 case StabAlternateEntry: 1125 // N_ENTRY - alternate entry: name,,n_sect,linenumber,address 1126 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1127 type = eSymbolTypeLineEntry; 1128 break; 1129 1130 //---------------------------------------------------------------------- 1131 // Left and Right Braces 1132 //---------------------------------------------------------------------- 1133 case StabLeftBracket: 1134 // N_LBRAC - left bracket: 0,,NO_SECT,nesting level,address 1135 // We use the current number of symbols in the symbol table in lieu of 1136 // using nlist_idx in case we ever start trimming entries out 1137 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1138 N_BRAC_indexes.push_back(sym_idx); 1139 type = eSymbolTypeScopeBegin; 1140 break; 1141 1142 case StabRightBracket: 1143 // N_RBRAC - right bracket: 0,,NO_SECT,nesting level,address 1144 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 1145 // so that we can always skip the entire symbol if we need to navigate 1146 // more quickly at the source level when parsing STABS 1147 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1148 if ( !N_BRAC_indexes.empty() ) 1149 { 1150 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 1151 symbol_ptr->SetByteSize(sym_idx + 1); 1152 symbol_ptr->SetSizeIsSibling(true); 1153 N_BRAC_indexes.pop_back(); 1154 } 1155 type = eSymbolTypeScopeEnd; 1156 break; 1157 1158 case StabDeletedIncludeFile: 1159 // N_EXCL - deleted include file: name,,NO_SECT,0,sum 1160 type = eSymbolTypeHeaderFile; 1161 break; 1162 1163 //---------------------------------------------------------------------- 1164 // COMM scopes 1165 //---------------------------------------------------------------------- 1166 case StabBeginCommon: 1167 // N_BCOMM - begin common: name,,NO_SECT,0,0 1168 // We use the current number of symbols in the symbol table in lieu of 1169 // using nlist_idx in case we ever start trimming entries out 1170 type = eSymbolTypeScopeBegin; 1171 N_COMM_indexes.push_back(sym_idx); 1172 break; 1173 1174 case StabEndCommonLocal: 1175 // N_ECOML - end common (local name): 0,,n_sect,0,address 1176 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1177 // Fall through 1178 1179 case StabEndCommon: 1180 // N_ECOMM - end common: name,,n_sect,0,0 1181 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 1182 // so that we can always skip the entire symbol if we need to navigate 1183 // more quickly at the source level when parsing STABS 1184 if ( !N_COMM_indexes.empty() ) 1185 { 1186 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 1187 symbol_ptr->SetByteSize(sym_idx + 1); 1188 symbol_ptr->SetSizeIsSibling(true); 1189 N_COMM_indexes.pop_back(); 1190 } 1191 type = eSymbolTypeScopeEnd; 1192 break; 1193 1194 case StabLength: 1195 // N_LENG - second stab entry with length information 1196 type = eSymbolTypeAdditional; 1197 break; 1198 1199 default: break; 1200 } 1201 } 1202 else 1203 { 1204 //uint8_t n_pext = NlistMaskPrivateExternal & nlist.n_type; 1205 uint8_t n_type = NlistMaskType & nlist.n_type; 1206 sym[sym_idx].SetExternal((NlistMaskExternal & nlist.n_type) != 0); 1207 1208 switch (n_type) 1209 { 1210 case NListTypeIndirect: // N_INDR - Fall through 1211 case NListTypePreboundUndefined:// N_PBUD - Fall through 1212 case NListTypeUndefined: // N_UNDF 1213 type = eSymbolTypeUndefined; 1214 break; 1215 1216 case NListTypeAbsolute: // N_ABS 1217 type = eSymbolTypeAbsolute; 1218 break; 1219 1220 case NListTypeSection: // N_SECT 1221 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1222 1223 if (symbol_section == NULL) 1224 { 1225 // TODO: warn about this? 1226 add_nlist = false; 1227 break; 1228 } 1229 1230 if (TEXT_eh_frame_sectID == nlist.n_sect) 1231 { 1232 type = eSymbolTypeException; 1233 } 1234 else 1235 { 1236 uint32_t section_type = symbol_section->Get() & SectionFlagMaskSectionType; 1237 1238 switch (section_type) 1239 { 1240 case SectionTypeRegular: break; // regular section 1241 //case SectionTypeZeroFill: type = eSymbolTypeData; break; // zero fill on demand section 1242 case SectionTypeCStringLiterals: type = eSymbolTypeData; break; // section with only literal C strings 1243 case SectionType4ByteLiterals: type = eSymbolTypeData; break; // section with only 4 byte literals 1244 case SectionType8ByteLiterals: type = eSymbolTypeData; break; // section with only 8 byte literals 1245 case SectionTypeLiteralPointers: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 1246 case SectionTypeNonLazySymbolPointers: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 1247 case SectionTypeLazySymbolPointers: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 1248 case SectionTypeSymbolStubs: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 1249 case SectionTypeModuleInitFunctionPointers: type = eSymbolTypeCode; break; // section with only function pointers for initialization 1250 case SectionTypeModuleTermFunctionPointers: type = eSymbolTypeCode; break; // section with only function pointers for termination 1251 //case SectionTypeCoalesced: type = eSymbolType; break; // section contains symbols that are to be coalesced 1252 //case SectionTypeZeroFillLarge: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 1253 case SectionTypeInterposing: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 1254 case SectionType16ByteLiterals: type = eSymbolTypeData; break; // section with only 16 byte literals 1255 case SectionTypeDTraceObjectFormat: type = eSymbolTypeInstrumentation; break; 1256 case SectionTypeLazyDylibSymbolPointers: type = eSymbolTypeTrampoline; break; 1257 default: break; 1258 } 1259 1260 if (type == eSymbolTypeInvalid) 1261 { 1262 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 1263 if (symbol_section->IsDescendant (text_section_sp.get())) 1264 { 1265 if (symbol_section->IsClear(SectionAttrUserPureInstructions | 1266 SectionAttrUserSelfModifyingCode | 1267 SectionAttrSytemSomeInstructions)) 1268 type = eSymbolTypeData; 1269 else 1270 type = eSymbolTypeCode; 1271 } 1272 else 1273 if (symbol_section->IsDescendant(data_section_sp.get())) 1274 { 1275 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 1276 { 1277 type = eSymbolTypeRuntime; 1278 1279 if (symbol_name && 1280 symbol_name[0] == '_' && 1281 symbol_name[1] == 'O' && 1282 symbol_name[2] == 'B') 1283 { 1284 llvm::StringRef symbol_name_ref(symbol_name); 1285 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 1286 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 1287 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 1288 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 1289 { 1290 symbol_name_non_abi_mangled = symbol_name + 1; 1291 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 1292 type = eSymbolTypeObjCClass; 1293 } 1294 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 1295 { 1296 symbol_name_non_abi_mangled = symbol_name + 1; 1297 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 1298 type = eSymbolTypeObjCMetaClass; 1299 } 1300 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 1301 { 1302 symbol_name_non_abi_mangled = symbol_name + 1; 1303 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 1304 type = eSymbolTypeObjCIVar; 1305 } 1306 } 1307 } 1308 else 1309 if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 1310 { 1311 type = eSymbolTypeException; 1312 } 1313 else 1314 { 1315 type = eSymbolTypeData; 1316 } 1317 } 1318 else 1319 if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 1320 { 1321 type = eSymbolTypeTrampoline; 1322 } 1323 else 1324 if (symbol_section->IsDescendant(objc_section_sp.get())) 1325 { 1326 type = eSymbolTypeRuntime; 1327 if (symbol_name && symbol_name[0] == '.') 1328 { 1329 llvm::StringRef symbol_name_ref(symbol_name); 1330 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 1331 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 1332 { 1333 symbol_name_non_abi_mangled = symbol_name; 1334 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 1335 type = eSymbolTypeObjCClass; 1336 } 1337 } 1338 } 1339 } 1340 } 1341 break; 1342 } 1343 } 1344 1345 if (add_nlist) 1346 { 1347 uint64_t symbol_value = nlist.n_value; 1348 bool symbol_name_is_mangled = false; 1349 1350 if (symbol_name_non_abi_mangled) 1351 { 1352 sym[sym_idx].GetMangled().SetMangledName (symbol_name_non_abi_mangled); 1353 sym[sym_idx].GetMangled().SetDemangledName (symbol_name); 1354 } 1355 else 1356 { 1357 if (symbol_name && symbol_name[0] == '_') 1358 { 1359 symbol_name_is_mangled = symbol_name[1] == '_'; 1360 symbol_name++; // Skip the leading underscore 1361 } 1362 1363 if (symbol_name) 1364 { 1365 sym[sym_idx].GetMangled().SetValue(symbol_name, symbol_name_is_mangled); 1366 } 1367 } 1368 1369 if (is_debug == false) 1370 { 1371 if (type == eSymbolTypeCode) 1372 { 1373 // See if we can find a N_FUN entry for any code symbols. 1374 // If we do find a match, and the name matches, then we 1375 // can merge the two into just the function symbol to avoid 1376 // duplicate entries in the symbol table 1377 ValueToSymbolIndexMap::const_iterator pos = N_FUN_addr_to_sym_idx.find (nlist.n_value); 1378 if (pos != N_FUN_addr_to_sym_idx.end()) 1379 { 1380 if ((symbol_name_is_mangled == true && sym[sym_idx].GetMangled().GetMangledName() == sym[pos->second].GetMangled().GetMangledName()) || 1381 (symbol_name_is_mangled == false && sym[sym_idx].GetMangled().GetDemangledName() == sym[pos->second].GetMangled().GetDemangledName())) 1382 { 1383 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 1384 // We just need the flags from the linker symbol, so put these flags 1385 // into the N_FUN flags to avoid duplicate symbols in the symbol table 1386 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 1387 sym[sym_idx].Clear(); 1388 continue; 1389 } 1390 } 1391 } 1392 else if (type == eSymbolTypeData) 1393 { 1394 // See if we can find a N_STSYM entry for any data symbols. 1395 // If we do find a match, and the name matches, then we 1396 // can merge the two into just the Static symbol to avoid 1397 // duplicate entries in the symbol table 1398 ValueToSymbolIndexMap::const_iterator pos = N_STSYM_addr_to_sym_idx.find (nlist.n_value); 1399 if (pos != N_STSYM_addr_to_sym_idx.end()) 1400 { 1401 if ((symbol_name_is_mangled == true && sym[sym_idx].GetMangled().GetMangledName() == sym[pos->second].GetMangled().GetMangledName()) || 1402 (symbol_name_is_mangled == false && sym[sym_idx].GetMangled().GetDemangledName() == sym[pos->second].GetMangled().GetDemangledName())) 1403 { 1404 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 1405 // We just need the flags from the linker symbol, so put these flags 1406 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 1407 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 1408 sym[sym_idx].Clear(); 1409 continue; 1410 } 1411 } 1412 } 1413 } 1414 if (symbol_section != NULL) 1415 symbol_value -= symbol_section->GetFileAddress(); 1416 1417 sym[sym_idx].SetID (nlist_idx); 1418 sym[sym_idx].SetType (type); 1419 sym[sym_idx].GetAddressRangeRef().GetBaseAddress().SetSection (symbol_section); 1420 sym[sym_idx].GetAddressRangeRef().GetBaseAddress().SetOffset (symbol_value); 1421 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 1422 1423 ++sym_idx; 1424 } 1425 else 1426 { 1427 sym[sym_idx].Clear(); 1428 } 1429 1430 } 1431 1432 // STAB N_GSYM entries end up having a symbol type eSymbolTypeGlobal and when the symbol value 1433 // is zero, the address of the global ends up being in a non-STAB entry. Try and fix up all 1434 // such entries by figuring out what the address for the global is by looking up this non-STAB 1435 // entry and copying the value into the debug symbol's value to save us the hassle in the 1436 // debug symbol parser. 1437 1438 Symbol *global_symbol = NULL; 1439 for (nlist_idx = 0; 1440 nlist_idx < symtab_load_command.nsyms && (global_symbol = symtab->FindSymbolWithType (eSymbolTypeData, Symtab::eDebugYes, Symtab::eVisibilityAny, nlist_idx)) != NULL; 1441 nlist_idx++) 1442 { 1443 if (global_symbol->GetValue().GetFileAddress() == 0) 1444 { 1445 std::vector<uint32_t> indexes; 1446 if (symtab->AppendSymbolIndexesWithName (global_symbol->GetMangled().GetName(), indexes) > 0) 1447 { 1448 std::vector<uint32_t>::const_iterator pos; 1449 std::vector<uint32_t>::const_iterator end = indexes.end(); 1450 for (pos = indexes.begin(); pos != end; ++pos) 1451 { 1452 symbol_ptr = symtab->SymbolAtIndex(*pos); 1453 if (symbol_ptr != global_symbol && symbol_ptr->IsDebug() == false) 1454 { 1455 global_symbol->SetValue(symbol_ptr->GetValue()); 1456 break; 1457 } 1458 } 1459 } 1460 } 1461 } 1462 1463 // Trim our symbols down to just what we ended up with after 1464 // removing any symbols. 1465 if (sym_idx < num_syms) 1466 { 1467 num_syms = sym_idx; 1468 sym = symtab->Resize (num_syms); 1469 } 1470 1471 // Now synthesize indirect symbols 1472 if (m_dysymtab.nindirectsyms != 0) 1473 { 1474 DataBufferSP indirect_symbol_indexes_sp(m_file.ReadFileContents(m_offset + m_dysymtab.indirectsymoff, m_dysymtab.nindirectsyms * 4)); 1475 1476 if (indirect_symbol_indexes_sp && indirect_symbol_indexes_sp->GetByteSize()) 1477 { 1478 NListIndexToSymbolIndexMap::const_iterator end_index_pos = m_nlist_idx_to_sym_idx.end(); 1479 DataExtractor indirect_symbol_index_data (indirect_symbol_indexes_sp, m_data.GetByteOrder(), m_data.GetAddressByteSize()); 1480 1481 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); ++sect_idx) 1482 { 1483 if ((m_mach_sections[sect_idx].flags & SectionFlagMaskSectionType) == SectionTypeSymbolStubs) 1484 { 1485 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 1486 if (symbol_stub_byte_size == 0) 1487 continue; 1488 1489 const uint32_t num_symbol_stubs = m_mach_sections[sect_idx].size / symbol_stub_byte_size; 1490 1491 if (num_symbol_stubs == 0) 1492 continue; 1493 1494 const uint32_t symbol_stub_index_offset = m_mach_sections[sect_idx].reserved1; 1495 uint32_t synthetic_stub_sym_id = symtab_load_command.nsyms; 1496 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) 1497 { 1498 const uint32_t symbol_stub_index = symbol_stub_index_offset + stub_idx; 1499 const lldb::addr_t symbol_stub_addr = m_mach_sections[sect_idx].addr + (stub_idx * symbol_stub_byte_size); 1500 uint32_t symbol_stub_offset = symbol_stub_index * 4; 1501 if (indirect_symbol_index_data.ValidOffsetForDataOfSize(symbol_stub_offset, 4)) 1502 { 1503 const uint32_t stub_sym_id = indirect_symbol_index_data.GetU32 (&symbol_stub_offset); 1504 if (stub_sym_id & (IndirectSymbolAbsolute | IndirectSymbolLocal)) 1505 continue; 1506 1507 NListIndexToSymbolIndexMap::const_iterator index_pos = m_nlist_idx_to_sym_idx.find (stub_sym_id); 1508 Symbol *stub_symbol = NULL; 1509 if (index_pos != end_index_pos) 1510 { 1511 // We have a remapping from the original nlist index to 1512 // a current symbol index, so just look this up by index 1513 stub_symbol = symtab->SymbolAtIndex (index_pos->second); 1514 } 1515 else 1516 { 1517 // We need to lookup a symbol using the original nlist 1518 // symbol index since this index is coming from the 1519 // S_SYMBOL_STUBS 1520 stub_symbol = symtab->FindSymbolByID (stub_sym_id); 1521 } 1522 1523 assert (stub_symbol); 1524 if (stub_symbol) 1525 { 1526 Address so_addr(symbol_stub_addr, section_list); 1527 1528 if (stub_symbol->GetType() == eSymbolTypeUndefined) 1529 { 1530 // Change the external symbol into a trampoline that makes sense 1531 // These symbols were N_UNDF N_EXT, and are useless to us, so we 1532 // can re-use them so we don't have to make up a synthetic symbol 1533 // for no good reason. 1534 stub_symbol->SetType (eSymbolTypeTrampoline); 1535 stub_symbol->SetExternal (false); 1536 stub_symbol->GetAddressRangeRef().GetBaseAddress() = so_addr; 1537 stub_symbol->GetAddressRangeRef().SetByteSize (symbol_stub_byte_size); 1538 } 1539 else 1540 { 1541 // Make a synthetic symbol to describe the trampoline stub 1542 if (sym_idx >= num_syms) 1543 sym = symtab->Resize (++num_syms); 1544 sym[sym_idx].SetID (synthetic_stub_sym_id++); 1545 sym[sym_idx].GetMangled() = stub_symbol->GetMangled(); 1546 sym[sym_idx].SetType (eSymbolTypeTrampoline); 1547 sym[sym_idx].SetIsSynthetic (true); 1548 sym[sym_idx].GetAddressRangeRef().GetBaseAddress() = so_addr; 1549 sym[sym_idx].GetAddressRangeRef().SetByteSize (symbol_stub_byte_size); 1550 ++sym_idx; 1551 } 1552 } 1553 } 1554 } 1555 } 1556 } 1557 } 1558 } 1559 1560 1561 1562 return symtab->GetNumSymbols(); 1563 } 1564 } 1565 offset = cmd_offset + symtab_load_command.cmdsize; 1566 } 1567 return 0; 1568 } 1569 1570 1571 void 1572 ObjectFileMachO::Dump (Stream *s) 1573 { 1574 lldb_private::Mutex::Locker locker(m_mutex); 1575 s->Printf("%p: ", this); 1576 s->Indent(); 1577 if (m_header.magic == HeaderMagic64 || m_header.magic == HeaderMagic64Swapped) 1578 s->PutCString("ObjectFileMachO64"); 1579 else 1580 s->PutCString("ObjectFileMachO32"); 1581 1582 ArchSpec header_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 1583 1584 *s << ", file = '" << m_file << "', arch = " << header_arch.GetArchitectureName() << "\n"; 1585 1586 if (m_sections_ap.get()) 1587 m_sections_ap->Dump(s, NULL, true, UINT32_MAX); 1588 1589 if (m_symtab_ap.get()) 1590 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 1591 } 1592 1593 1594 bool 1595 ObjectFileMachO::GetUUID (lldb_private::UUID* uuid) 1596 { 1597 lldb_private::Mutex::Locker locker(m_mutex); 1598 struct uuid_command load_cmd; 1599 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 1600 uint32_t i; 1601 for (i=0; i<m_header.ncmds; ++i) 1602 { 1603 const uint32_t cmd_offset = offset; 1604 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1605 break; 1606 1607 if (load_cmd.cmd == LoadCommandUUID) 1608 { 1609 const uint8_t *uuid_bytes = m_data.PeekData(offset, 16); 1610 if (uuid_bytes) 1611 { 1612 uuid->SetBytes (uuid_bytes); 1613 return true; 1614 } 1615 return false; 1616 } 1617 offset = cmd_offset + load_cmd.cmdsize; 1618 } 1619 return false; 1620 } 1621 1622 1623 uint32_t 1624 ObjectFileMachO::GetDependentModules (FileSpecList& files) 1625 { 1626 lldb_private::Mutex::Locker locker(m_mutex); 1627 struct load_command load_cmd; 1628 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 1629 uint32_t count = 0; 1630 const bool resolve_path = false; // Don't resolve the dependend file paths since they may not reside on this system 1631 uint32_t i; 1632 for (i=0; i<m_header.ncmds; ++i) 1633 { 1634 const uint32_t cmd_offset = offset; 1635 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1636 break; 1637 1638 switch (load_cmd.cmd) 1639 { 1640 case LoadCommandDylibLoad: 1641 case LoadCommandDylibLoadWeak: 1642 case LoadCommandDylibReexport: 1643 case LoadCommandDynamicLinkerLoad: 1644 case LoadCommandFixedVMShlibLoad: 1645 case LoadCommandDylibLoadUpward: 1646 { 1647 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 1648 const char *path = m_data.PeekCStr(name_offset); 1649 // Skip any path that starts with '@' since these are usually: 1650 // @executable_path/.../file 1651 // @rpath/.../file 1652 if (path && path[0] != '@') 1653 { 1654 FileSpec file_spec(path, resolve_path); 1655 if (files.AppendIfUnique(file_spec)) 1656 count++; 1657 } 1658 } 1659 break; 1660 1661 default: 1662 break; 1663 } 1664 offset = cmd_offset + load_cmd.cmdsize; 1665 } 1666 return count; 1667 } 1668 1669 lldb_private::Address 1670 ObjectFileMachO::GetEntryPointAddress () 1671 { 1672 // If the object file is not an executable it can't hold the entry point. m_entry_point_address 1673 // is initialized to an invalid address, so we can just return that. 1674 // If m_entry_point_address is valid it means we've found it already, so return the cached value. 1675 1676 if (!IsExecutable() || m_entry_point_address.IsValid()) 1677 return m_entry_point_address; 1678 1679 // Otherwise, look for the UnixThread or Thread command. The data for the Thread command is given in 1680 // /usr/include/mach-o.h, but it is basically: 1681 // 1682 // uint32_t flavor - this is the flavor argument you would pass to thread_get_state 1683 // uint32_t count - this is the count of longs in the thread state data 1684 // struct XXX_thread_state state - this is the structure from <machine/thread_status.h> corresponding to the flavor. 1685 // <repeat this trio> 1686 // 1687 // So we just keep reading the various register flavors till we find the GPR one, then read the PC out of there. 1688 // FIXME: We will need to have a "RegisterContext data provider" class at some point that can get all the registers 1689 // out of data in this form & attach them to a given thread. That should underlie the MacOS X User process plugin, 1690 // and we'll also need it for the MacOS X Core File process plugin. When we have that we can also use it here. 1691 // 1692 // For now we hard-code the offsets and flavors we need: 1693 // 1694 // 1695 1696 lldb_private::Mutex::Locker locker(m_mutex); 1697 struct load_command load_cmd; 1698 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 1699 uint32_t i; 1700 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 1701 bool done = false; 1702 1703 for (i=0; i<m_header.ncmds; ++i) 1704 { 1705 const uint32_t cmd_offset = offset; 1706 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1707 break; 1708 1709 switch (load_cmd.cmd) 1710 { 1711 case LoadCommandUnixThread: 1712 case LoadCommandThread: 1713 { 1714 while (offset < cmd_offset + load_cmd.cmdsize) 1715 { 1716 uint32_t flavor = m_data.GetU32(&offset); 1717 uint32_t count = m_data.GetU32(&offset); 1718 if (count == 0) 1719 { 1720 // We've gotten off somehow, log and exit; 1721 return m_entry_point_address; 1722 } 1723 1724 switch (m_header.cputype) 1725 { 1726 case llvm::MachO::CPUTypeARM: 1727 if (flavor == 1) // ARM_THREAD_STATE from mach/arm/thread_status.h 1728 { 1729 offset += 60; // This is the offset of pc in the GPR thread state data structure. 1730 start_address = m_data.GetU32(&offset); 1731 done = true; 1732 } 1733 break; 1734 case llvm::MachO::CPUTypeI386: 1735 if (flavor == 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 1736 { 1737 offset += 40; // This is the offset of eip in the GPR thread state data structure. 1738 start_address = m_data.GetU32(&offset); 1739 done = true; 1740 } 1741 break; 1742 case llvm::MachO::CPUTypeX86_64: 1743 if (flavor == 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 1744 { 1745 offset += 16 * 8; // This is the offset of rip in the GPR thread state data structure. 1746 start_address = m_data.GetU64(&offset); 1747 done = true; 1748 } 1749 break; 1750 default: 1751 return m_entry_point_address; 1752 } 1753 // Haven't found the GPR flavor yet, skip over the data for this flavor: 1754 if (done) 1755 break; 1756 offset += count * 4; 1757 } 1758 } 1759 break; 1760 1761 default: 1762 break; 1763 } 1764 if (done) 1765 break; 1766 1767 // Go to the next load command: 1768 offset = cmd_offset + load_cmd.cmdsize; 1769 } 1770 1771 if (start_address != LLDB_INVALID_ADDRESS) 1772 { 1773 // We got the start address from the load commands, so now resolve that address in the sections 1774 // of this ObjectFile: 1775 if (!m_entry_point_address.ResolveAddressUsingFileSections (start_address, GetSectionList())) 1776 { 1777 m_entry_point_address.Clear(); 1778 } 1779 } 1780 else 1781 { 1782 // We couldn't read the UnixThread load command - maybe it wasn't there. As a fallback look for the 1783 // "start" symbol in the main executable. 1784 1785 SymbolContextList contexts; 1786 SymbolContext context; 1787 if (!m_module->FindSymbolsWithNameAndType(ConstString ("start"), eSymbolTypeCode, contexts)) 1788 return m_entry_point_address; 1789 1790 contexts.GetContextAtIndex(0, context); 1791 1792 m_entry_point_address = context.symbol->GetValue(); 1793 } 1794 1795 return m_entry_point_address; 1796 1797 } 1798 1799 ObjectFile::Type 1800 ObjectFileMachO::CalculateType() 1801 { 1802 switch (m_header.filetype) 1803 { 1804 case HeaderFileTypeObject: // 0x1u MH_OBJECT 1805 if (GetAddressByteSize () == 4) 1806 { 1807 // 32 bit kexts are just object files, but they do have a valid 1808 // UUID load command. 1809 UUID uuid; 1810 if (GetUUID(&uuid)) 1811 { 1812 // this checking for the UUID load command is not enough 1813 // we could eventually look for the symbol named 1814 // "OSKextGetCurrentIdentifier" as this is required of kexts 1815 if (m_strata == eStrataInvalid) 1816 m_strata = eStrataKernel; 1817 return eTypeSharedLibrary; 1818 } 1819 } 1820 return eTypeObjectFile; 1821 1822 case HeaderFileTypeExecutable: return eTypeExecutable; // 0x2u MH_EXECUTE 1823 case HeaderFileTypeFixedVMShlib: return eTypeSharedLibrary; // 0x3u MH_FVMLIB 1824 case HeaderFileTypeCore: return eTypeCoreFile; // 0x4u MH_CORE 1825 case HeaderFileTypePreloadedExecutable: return eTypeSharedLibrary; // 0x5u MH_PRELOAD 1826 case HeaderFileTypeDynamicShlib: return eTypeSharedLibrary; // 0x6u MH_DYLIB 1827 case HeaderFileTypeDynamicLinkEditor: return eTypeDynamicLinker; // 0x7u MH_DYLINKER 1828 case HeaderFileTypeBundle: return eTypeSharedLibrary; // 0x8u MH_BUNDLE 1829 case HeaderFileTypeDynamicShlibStub: return eTypeStubLibrary; // 0x9u MH_DYLIB_STUB 1830 case HeaderFileTypeDSYM: return eTypeDebugInfo; // 0xAu MH_DSYM 1831 case HeaderFileTypeKextBundle: return eTypeSharedLibrary; // 0xBu MH_KEXT_BUNDLE 1832 default: 1833 break; 1834 } 1835 return eTypeUnknown; 1836 } 1837 1838 ObjectFile::Strata 1839 ObjectFileMachO::CalculateStrata() 1840 { 1841 switch (m_header.filetype) 1842 { 1843 case HeaderFileTypeObject: // 0x1u MH_OBJECT 1844 { 1845 // 32 bit kexts are just object files, but they do have a valid 1846 // UUID load command. 1847 UUID uuid; 1848 if (GetUUID(&uuid)) 1849 { 1850 // this checking for the UUID load command is not enough 1851 // we could eventually look for the symbol named 1852 // "OSKextGetCurrentIdentifier" as this is required of kexts 1853 if (m_type == eTypeInvalid) 1854 m_type = eTypeSharedLibrary; 1855 1856 return eStrataKernel; 1857 } 1858 } 1859 return eStrataUnknown; 1860 1861 case HeaderFileTypeExecutable: // 0x2u MH_EXECUTE 1862 // Check for the MH_DYLDLINK bit in the flags 1863 if (m_header.flags & HeaderFlagBitIsDynamicLinkObject) 1864 return eStrataUser; 1865 return eStrataKernel; 1866 1867 case HeaderFileTypeFixedVMShlib: return eStrataUser; // 0x3u MH_FVMLIB 1868 case HeaderFileTypeCore: return eStrataUnknown; // 0x4u MH_CORE 1869 case HeaderFileTypePreloadedExecutable: return eStrataUser; // 0x5u MH_PRELOAD 1870 case HeaderFileTypeDynamicShlib: return eStrataUser; // 0x6u MH_DYLIB 1871 case HeaderFileTypeDynamicLinkEditor: return eStrataUser; // 0x7u MH_DYLINKER 1872 case HeaderFileTypeBundle: return eStrataUser; // 0x8u MH_BUNDLE 1873 case HeaderFileTypeDynamicShlibStub: return eStrataUser; // 0x9u MH_DYLIB_STUB 1874 case HeaderFileTypeDSYM: return eStrataUnknown; // 0xAu MH_DSYM 1875 case HeaderFileTypeKextBundle: return eStrataKernel; // 0xBu MH_KEXT_BUNDLE 1876 default: 1877 break; 1878 } 1879 return eStrataUnknown; 1880 } 1881 1882 1883 bool 1884 ObjectFileMachO::GetArchitecture (ArchSpec &arch) 1885 { 1886 lldb_private::Mutex::Locker locker(m_mutex); 1887 arch.SetArchitecture (eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 1888 1889 // Files with type MH_PRELOAD are currently used in cases where the image 1890 // debugs at the addresses in the file itself. Below we set the OS to 1891 // unknown to make sure we use the DynamicLoaderStatic()... 1892 if (m_header.filetype == HeaderFileTypePreloadedExecutable) 1893 { 1894 arch.GetTriple().setOS (llvm::Triple::UnknownOS); 1895 } 1896 1897 return true; 1898 } 1899 1900 1901 //------------------------------------------------------------------ 1902 // PluginInterface protocol 1903 //------------------------------------------------------------------ 1904 const char * 1905 ObjectFileMachO::GetPluginName() 1906 { 1907 return "ObjectFileMachO"; 1908 } 1909 1910 const char * 1911 ObjectFileMachO::GetShortPluginName() 1912 { 1913 return GetPluginNameStatic(); 1914 } 1915 1916 uint32_t 1917 ObjectFileMachO::GetPluginVersion() 1918 { 1919 return 1; 1920 } 1921 1922