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; 755 756 LogSP log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_SYMBOLS)); 757 758 for (i=0; i<m_header.ncmds; ++i) 759 { 760 const uint32_t cmd_offset = offset; 761 // Read in the load command and load command size 762 if (m_data.GetU32(&offset, &symtab_load_command, 2) == NULL) 763 break; 764 // Watch for the symbol table load command 765 if (symtab_load_command.cmd == LoadCommandSymtab) 766 { 767 // Read in the rest of the symtab load command 768 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4)) // fill in symoff, nsyms, stroff, strsize fields 769 { 770 if (symtab_load_command.symoff == 0) 771 { 772 if (log) 773 GetModule()->LogMessage(log.get(), "LC_SYMTAB.symoff == 0"); 774 return 0; 775 } 776 777 if (symtab_load_command.stroff == 0) 778 { 779 if (log) 780 GetModule()->LogMessage(log.get(), "LC_SYMTAB.stroff == 0"); 781 return 0; 782 } 783 784 if (symtab_load_command.nsyms == 0) 785 { 786 if (log) 787 GetModule()->LogMessage(log.get(), "LC_SYMTAB.nsyms == 0"); 788 return 0; 789 } 790 791 if (symtab_load_command.strsize == 0) 792 { 793 if (log) 794 GetModule()->LogMessage(log.get(), "LC_SYMTAB.strsize == 0"); 795 return 0; 796 } 797 798 Symtab *symtab = m_symtab_ap.get(); 799 SectionList *section_list = GetSectionList(); 800 if (section_list == NULL) 801 return 0; 802 803 const size_t addr_byte_size = m_data.GetAddressByteSize(); 804 const ByteOrder byte_order = m_data.GetByteOrder(); 805 bool bit_width_32 = addr_byte_size == 4; 806 const size_t nlist_byte_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 807 808 DataBufferSP symtab_data_sp(m_file.ReadFileContents (m_offset + symtab_load_command.symoff, 809 symtab_load_command.nsyms * nlist_byte_size)); 810 811 if (symtab_data_sp.get() == NULL || 812 symtab_data_sp->GetBytes() == NULL || 813 symtab_data_sp->GetByteSize() == 0) 814 { 815 if (log) 816 GetModule()->LogMessage(log.get(), "failed to read nlist data"); 817 return 0; 818 } 819 820 DataBufferSP strtab_data_sp(m_file.ReadFileContents (m_offset + symtab_load_command.stroff, 821 symtab_load_command.strsize)); 822 823 if (strtab_data_sp.get() == NULL || 824 strtab_data_sp->GetBytes() == NULL || 825 strtab_data_sp->GetByteSize() == 0) 826 { 827 if (log) 828 GetModule()->LogMessage(log.get(), "failed to read strtab data"); 829 return 0; 830 } 831 832 const char *strtab_data = (const char *)strtab_data_sp->GetBytes(); 833 const size_t strtab_data_len = strtab_data_sp->GetByteSize(); 834 835 static ConstString g_segment_name_TEXT ("__TEXT"); 836 static ConstString g_segment_name_DATA ("__DATA"); 837 static ConstString g_segment_name_OBJC ("__OBJC"); 838 static ConstString g_section_name_eh_frame ("__eh_frame"); 839 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT)); 840 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA)); 841 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC)); 842 SectionSP eh_frame_section_sp; 843 if (text_section_sp.get()) 844 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame); 845 else 846 eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame); 847 848 uint8_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() : NListSectionNoSection; 849 //uint32_t symtab_offset = 0; 850 assert (symtab_data_sp->GetByteSize()/nlist_byte_size >= symtab_load_command.nsyms); 851 852 uint32_t nlist_data_offset = 0; 853 DataExtractor nlist_data (symtab_data_sp, byte_order, addr_byte_size); 854 855 uint32_t N_SO_index = UINT32_MAX; 856 857 MachSymtabSectionInfo section_info (section_list); 858 std::vector<uint32_t> N_FUN_indexes; 859 std::vector<uint32_t> N_NSYM_indexes; 860 std::vector<uint32_t> N_INCL_indexes; 861 std::vector<uint32_t> N_BRAC_indexes; 862 std::vector<uint32_t> N_COMM_indexes; 863 typedef std::map <uint64_t, uint32_t> ValueToSymbolIndexMap; 864 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap; 865 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 866 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 867 // Any symbols that get merged into another will get an entry 868 // in this map so we know 869 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 870 uint32_t nlist_idx = 0; 871 Symbol *symbol_ptr = NULL; 872 873 uint32_t sym_idx = 0; 874 Symbol *sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 875 uint32_t num_syms = symtab->GetNumSymbols(); 876 877 //symtab->Reserve (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 878 for (nlist_idx = 0; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) 879 { 880 struct nlist_64 nlist; 881 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 882 break; 883 884 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 885 nlist.n_type = nlist_data.GetU8_unchecked (&nlist_data_offset); 886 nlist.n_sect = nlist_data.GetU8_unchecked (&nlist_data_offset); 887 nlist.n_desc = nlist_data.GetU16_unchecked (&nlist_data_offset); 888 nlist.n_value = nlist_data.GetAddress_unchecked (&nlist_data_offset); 889 890 SymbolType type = eSymbolTypeInvalid; 891 if (nlist.n_strx >= strtab_data_len) 892 { 893 // No symbol should be NULL, even the symbols with no 894 // string values should have an offset zero which points 895 // to an empty C-string 896 fprintf (stderr, 897 "error: symbol[%u] has invalid string table offset 0x%x in %s/%s, ignoring symbol\n", 898 nlist_idx, 899 nlist.n_strx, 900 m_module->GetFileSpec().GetDirectory().GetCString(), 901 m_module->GetFileSpec().GetFilename().GetCString()); 902 continue; 903 } 904 const char *symbol_name = &strtab_data[nlist.n_strx]; 905 const char *symbol_name_non_abi_mangled = NULL; 906 907 if (symbol_name[0] == '\0') 908 symbol_name = NULL; 909 Section* symbol_section = NULL; 910 bool add_nlist = true; 911 bool is_debug = ((nlist.n_type & NlistMaskStab) != 0); 912 913 assert (sym_idx < num_syms); 914 915 sym[sym_idx].SetDebug (is_debug); 916 917 if (is_debug) 918 { 919 switch (nlist.n_type) 920 { 921 case StabGlobalSymbol: 922 // N_GSYM -- global symbol: name,,NO_SECT,type,0 923 // Sometimes the N_GSYM value contains the address. 924 925 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 926 // have the same address, but we want to ensure that we always find only the real symbol, 927 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 928 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 929 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 930 // same address. 931 932 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 933 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 934 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 935 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 936 add_nlist = false; 937 else 938 { 939 sym[sym_idx].SetExternal(true); 940 if (nlist.n_value != 0) 941 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 942 type = eSymbolTypeData; 943 } 944 break; 945 946 case StabFunctionName: 947 // N_FNAME -- procedure name (f77 kludge): name,,NO_SECT,0,0 948 type = eSymbolTypeCompiler; 949 break; 950 951 case StabFunction: 952 // N_FUN -- procedure: name,,n_sect,linenumber,address 953 if (symbol_name) 954 { 955 type = eSymbolTypeCode; 956 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 957 958 N_FUN_addr_to_sym_idx[nlist.n_value] = sym_idx; 959 // We use the current number of symbols in the symbol table in lieu of 960 // using nlist_idx in case we ever start trimming entries out 961 N_FUN_indexes.push_back(sym_idx); 962 } 963 else 964 { 965 type = eSymbolTypeCompiler; 966 967 if ( !N_FUN_indexes.empty() ) 968 { 969 // Copy the size of the function into the original STAB entry so we don't have 970 // to hunt for it later 971 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 972 N_FUN_indexes.pop_back(); 973 // We don't really need the end function STAB as it contains the size which 974 // we already placed with the original symbol, so don't add it if we want a 975 // minimal symbol table 976 if (minimize) 977 add_nlist = false; 978 } 979 } 980 break; 981 982 case StabStaticSymbol: 983 // N_STSYM -- static symbol: name,,n_sect,type,address 984 N_STSYM_addr_to_sym_idx[nlist.n_value] = sym_idx; 985 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 986 type = eSymbolTypeData; 987 break; 988 989 case StabLocalCommon: 990 // N_LCSYM -- .lcomm symbol: name,,n_sect,type,address 991 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 992 type = eSymbolTypeCommonBlock; 993 break; 994 995 case StabBeginSymbol: 996 // N_BNSYM 997 // We use the current number of symbols in the symbol table in lieu of 998 // using nlist_idx in case we ever start trimming entries out 999 if (minimize) 1000 { 1001 // Skip these if we want minimal symbol tables 1002 add_nlist = false; 1003 } 1004 else 1005 { 1006 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1007 N_NSYM_indexes.push_back(sym_idx); 1008 type = eSymbolTypeScopeBegin; 1009 } 1010 break; 1011 1012 case StabEndSymbol: 1013 // N_ENSYM 1014 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 1015 // so that we can always skip the entire symbol if we need to navigate 1016 // more quickly at the source level when parsing STABS 1017 if (minimize) 1018 { 1019 // Skip these if we want minimal symbol tables 1020 add_nlist = false; 1021 } 1022 else 1023 { 1024 if ( !N_NSYM_indexes.empty() ) 1025 { 1026 symbol_ptr = symtab->SymbolAtIndex(N_NSYM_indexes.back()); 1027 symbol_ptr->SetByteSize(sym_idx + 1); 1028 symbol_ptr->SetSizeIsSibling(true); 1029 N_NSYM_indexes.pop_back(); 1030 } 1031 type = eSymbolTypeScopeEnd; 1032 } 1033 break; 1034 1035 1036 case StabSourceFileOptions: 1037 // N_OPT - emitted with gcc2_compiled and in gcc source 1038 type = eSymbolTypeCompiler; 1039 break; 1040 1041 case StabRegisterSymbol: 1042 // N_RSYM - register sym: name,,NO_SECT,type,register 1043 type = eSymbolTypeVariable; 1044 break; 1045 1046 case StabSourceLine: 1047 // N_SLINE - src line: 0,,n_sect,linenumber,address 1048 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1049 type = eSymbolTypeLineEntry; 1050 break; 1051 1052 case StabStructureType: 1053 // N_SSYM - structure elt: name,,NO_SECT,type,struct_offset 1054 type = eSymbolTypeVariableType; 1055 break; 1056 1057 case StabSourceFileName: 1058 // N_SO - source file name 1059 type = eSymbolTypeSourceFile; 1060 if (symbol_name == NULL) 1061 { 1062 if (minimize) 1063 add_nlist = false; 1064 if (N_SO_index != UINT32_MAX) 1065 { 1066 // Set the size of the N_SO to the terminating index of this N_SO 1067 // so that we can always skip the entire N_SO if we need to navigate 1068 // more quickly at the source level when parsing STABS 1069 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 1070 symbol_ptr->SetByteSize(sym_idx + (minimize ? 0 : 1)); 1071 symbol_ptr->SetSizeIsSibling(true); 1072 } 1073 N_NSYM_indexes.clear(); 1074 N_INCL_indexes.clear(); 1075 N_BRAC_indexes.clear(); 1076 N_COMM_indexes.clear(); 1077 N_FUN_indexes.clear(); 1078 N_SO_index = UINT32_MAX; 1079 } 1080 else 1081 { 1082 // We use the current number of symbols in the symbol table in lieu of 1083 // using nlist_idx in case we ever start trimming entries out 1084 if (symbol_name[0] == '/') 1085 N_SO_index = sym_idx; 1086 else if (minimize && (N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 1087 { 1088 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 1089 if (so_path && so_path[0]) 1090 { 1091 std::string full_so_path (so_path); 1092 if (*full_so_path.rbegin() != '/') 1093 full_so_path += '/'; 1094 full_so_path += symbol_name; 1095 sym[sym_idx - 1].GetMangled().SetValue(full_so_path.c_str(), false); 1096 add_nlist = false; 1097 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 1098 } 1099 } 1100 } 1101 1102 break; 1103 1104 case StabObjectFileName: 1105 // N_OSO - object file name: name,,0,0,st_mtime 1106 type = eSymbolTypeObjectFile; 1107 break; 1108 1109 case StabLocalSymbol: 1110 // N_LSYM - local sym: name,,NO_SECT,type,offset 1111 type = eSymbolTypeLocal; 1112 break; 1113 1114 //---------------------------------------------------------------------- 1115 // INCL scopes 1116 //---------------------------------------------------------------------- 1117 case StabBeginIncludeFileName: 1118 // N_BINCL - include file beginning: name,,NO_SECT,0,sum 1119 // We use the current number of symbols in the symbol table in lieu of 1120 // using nlist_idx in case we ever start trimming entries out 1121 N_INCL_indexes.push_back(sym_idx); 1122 type = eSymbolTypeScopeBegin; 1123 break; 1124 1125 case StabEndIncludeFile: 1126 // N_EINCL - include file end: name,,NO_SECT,0,0 1127 // Set the size of the N_BINCL to the terminating index of this N_EINCL 1128 // so that we can always skip the entire symbol if we need to navigate 1129 // more quickly at the source level when parsing STABS 1130 if ( !N_INCL_indexes.empty() ) 1131 { 1132 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 1133 symbol_ptr->SetByteSize(sym_idx + 1); 1134 symbol_ptr->SetSizeIsSibling(true); 1135 N_INCL_indexes.pop_back(); 1136 } 1137 type = eSymbolTypeScopeEnd; 1138 break; 1139 1140 case StabIncludeFileName: 1141 // N_SOL - #included file name: name,,n_sect,0,address 1142 type = eSymbolTypeHeaderFile; 1143 1144 // We currently don't use the header files on darwin 1145 if (minimize) 1146 add_nlist = false; 1147 break; 1148 1149 case StabCompilerParameters: 1150 // N_PARAMS - compiler parameters: name,,NO_SECT,0,0 1151 type = eSymbolTypeCompiler; 1152 break; 1153 1154 case StabCompilerVersion: 1155 // N_VERSION - compiler version: name,,NO_SECT,0,0 1156 type = eSymbolTypeCompiler; 1157 break; 1158 1159 case StabCompilerOptLevel: 1160 // N_OLEVEL - compiler -O level: name,,NO_SECT,0,0 1161 type = eSymbolTypeCompiler; 1162 break; 1163 1164 case StabParameter: 1165 // N_PSYM - parameter: name,,NO_SECT,type,offset 1166 type = eSymbolTypeVariable; 1167 break; 1168 1169 case StabAlternateEntry: 1170 // N_ENTRY - alternate entry: name,,n_sect,linenumber,address 1171 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1172 type = eSymbolTypeLineEntry; 1173 break; 1174 1175 //---------------------------------------------------------------------- 1176 // Left and Right Braces 1177 //---------------------------------------------------------------------- 1178 case StabLeftBracket: 1179 // N_LBRAC - left bracket: 0,,NO_SECT,nesting level,address 1180 // We use the current number of symbols in the symbol table in lieu of 1181 // using nlist_idx in case we ever start trimming entries out 1182 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1183 N_BRAC_indexes.push_back(sym_idx); 1184 type = eSymbolTypeScopeBegin; 1185 break; 1186 1187 case StabRightBracket: 1188 // N_RBRAC - right bracket: 0,,NO_SECT,nesting level,address 1189 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 1190 // so that we can always skip the entire symbol if we need to navigate 1191 // more quickly at the source level when parsing STABS 1192 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1193 if ( !N_BRAC_indexes.empty() ) 1194 { 1195 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 1196 symbol_ptr->SetByteSize(sym_idx + 1); 1197 symbol_ptr->SetSizeIsSibling(true); 1198 N_BRAC_indexes.pop_back(); 1199 } 1200 type = eSymbolTypeScopeEnd; 1201 break; 1202 1203 case StabDeletedIncludeFile: 1204 // N_EXCL - deleted include file: name,,NO_SECT,0,sum 1205 type = eSymbolTypeHeaderFile; 1206 break; 1207 1208 //---------------------------------------------------------------------- 1209 // COMM scopes 1210 //---------------------------------------------------------------------- 1211 case StabBeginCommon: 1212 // N_BCOMM - begin common: name,,NO_SECT,0,0 1213 // We use the current number of symbols in the symbol table in lieu of 1214 // using nlist_idx in case we ever start trimming entries out 1215 type = eSymbolTypeScopeBegin; 1216 N_COMM_indexes.push_back(sym_idx); 1217 break; 1218 1219 case StabEndCommonLocal: 1220 // N_ECOML - end common (local name): 0,,n_sect,0,address 1221 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1222 // Fall through 1223 1224 case StabEndCommon: 1225 // N_ECOMM - end common: name,,n_sect,0,0 1226 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 1227 // so that we can always skip the entire symbol if we need to navigate 1228 // more quickly at the source level when parsing STABS 1229 if ( !N_COMM_indexes.empty() ) 1230 { 1231 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 1232 symbol_ptr->SetByteSize(sym_idx + 1); 1233 symbol_ptr->SetSizeIsSibling(true); 1234 N_COMM_indexes.pop_back(); 1235 } 1236 type = eSymbolTypeScopeEnd; 1237 break; 1238 1239 case StabLength: 1240 // N_LENG - second stab entry with length information 1241 type = eSymbolTypeAdditional; 1242 break; 1243 1244 default: break; 1245 } 1246 } 1247 else 1248 { 1249 //uint8_t n_pext = NlistMaskPrivateExternal & nlist.n_type; 1250 uint8_t n_type = NlistMaskType & nlist.n_type; 1251 sym[sym_idx].SetExternal((NlistMaskExternal & nlist.n_type) != 0); 1252 1253 switch (n_type) 1254 { 1255 case NListTypeIndirect: // N_INDR - Fall through 1256 case NListTypePreboundUndefined:// N_PBUD - Fall through 1257 case NListTypeUndefined: // N_UNDF 1258 type = eSymbolTypeUndefined; 1259 break; 1260 1261 case NListTypeAbsolute: // N_ABS 1262 type = eSymbolTypeAbsolute; 1263 break; 1264 1265 case NListTypeSection: // N_SECT 1266 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1267 1268 if (symbol_section == NULL) 1269 { 1270 // TODO: warn about this? 1271 add_nlist = false; 1272 break; 1273 } 1274 1275 if (TEXT_eh_frame_sectID == nlist.n_sect) 1276 { 1277 type = eSymbolTypeException; 1278 } 1279 else 1280 { 1281 uint32_t section_type = symbol_section->Get() & SectionFlagMaskSectionType; 1282 1283 switch (section_type) 1284 { 1285 case SectionTypeRegular: break; // regular section 1286 //case SectionTypeZeroFill: type = eSymbolTypeData; break; // zero fill on demand section 1287 case SectionTypeCStringLiterals: type = eSymbolTypeData; break; // section with only literal C strings 1288 case SectionType4ByteLiterals: type = eSymbolTypeData; break; // section with only 4 byte literals 1289 case SectionType8ByteLiterals: type = eSymbolTypeData; break; // section with only 8 byte literals 1290 case SectionTypeLiteralPointers: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 1291 case SectionTypeNonLazySymbolPointers: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 1292 case SectionTypeLazySymbolPointers: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 1293 case SectionTypeSymbolStubs: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 1294 case SectionTypeModuleInitFunctionPointers: type = eSymbolTypeCode; break; // section with only function pointers for initialization 1295 case SectionTypeModuleTermFunctionPointers: type = eSymbolTypeCode; break; // section with only function pointers for termination 1296 //case SectionTypeCoalesced: type = eSymbolType; break; // section contains symbols that are to be coalesced 1297 //case SectionTypeZeroFillLarge: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 1298 case SectionTypeInterposing: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 1299 case SectionType16ByteLiterals: type = eSymbolTypeData; break; // section with only 16 byte literals 1300 case SectionTypeDTraceObjectFormat: type = eSymbolTypeInstrumentation; break; 1301 case SectionTypeLazyDylibSymbolPointers: type = eSymbolTypeTrampoline; break; 1302 default: break; 1303 } 1304 1305 if (type == eSymbolTypeInvalid) 1306 { 1307 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 1308 if (symbol_section->IsDescendant (text_section_sp.get())) 1309 { 1310 if (symbol_section->IsClear(SectionAttrUserPureInstructions | 1311 SectionAttrUserSelfModifyingCode | 1312 SectionAttrSytemSomeInstructions)) 1313 type = eSymbolTypeData; 1314 else 1315 type = eSymbolTypeCode; 1316 } 1317 else 1318 if (symbol_section->IsDescendant(data_section_sp.get())) 1319 { 1320 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 1321 { 1322 type = eSymbolTypeRuntime; 1323 1324 if (symbol_name && 1325 symbol_name[0] == '_' && 1326 symbol_name[1] == 'O' && 1327 symbol_name[2] == 'B') 1328 { 1329 llvm::StringRef symbol_name_ref(symbol_name); 1330 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 1331 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 1332 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 1333 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 1334 { 1335 symbol_name_non_abi_mangled = symbol_name + 1; 1336 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 1337 type = eSymbolTypeObjCClass; 1338 } 1339 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 1340 { 1341 symbol_name_non_abi_mangled = symbol_name + 1; 1342 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 1343 type = eSymbolTypeObjCMetaClass; 1344 } 1345 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 1346 { 1347 symbol_name_non_abi_mangled = symbol_name + 1; 1348 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 1349 type = eSymbolTypeObjCIVar; 1350 } 1351 } 1352 } 1353 else 1354 if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 1355 { 1356 type = eSymbolTypeException; 1357 } 1358 else 1359 { 1360 type = eSymbolTypeData; 1361 } 1362 } 1363 else 1364 if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 1365 { 1366 type = eSymbolTypeTrampoline; 1367 } 1368 else 1369 if (symbol_section->IsDescendant(objc_section_sp.get())) 1370 { 1371 type = eSymbolTypeRuntime; 1372 if (symbol_name && symbol_name[0] == '.') 1373 { 1374 llvm::StringRef symbol_name_ref(symbol_name); 1375 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 1376 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 1377 { 1378 symbol_name_non_abi_mangled = symbol_name; 1379 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 1380 type = eSymbolTypeObjCClass; 1381 } 1382 } 1383 } 1384 } 1385 } 1386 break; 1387 } 1388 } 1389 1390 if (add_nlist) 1391 { 1392 uint64_t symbol_value = nlist.n_value; 1393 bool symbol_name_is_mangled = false; 1394 1395 if (symbol_name_non_abi_mangled) 1396 { 1397 sym[sym_idx].GetMangled().SetMangledName (symbol_name_non_abi_mangled); 1398 sym[sym_idx].GetMangled().SetDemangledName (symbol_name); 1399 } 1400 else 1401 { 1402 if (symbol_name && symbol_name[0] == '_') 1403 { 1404 symbol_name_is_mangled = symbol_name[1] == '_'; 1405 symbol_name++; // Skip the leading underscore 1406 } 1407 1408 if (symbol_name) 1409 { 1410 sym[sym_idx].GetMangled().SetValue(symbol_name, symbol_name_is_mangled); 1411 } 1412 } 1413 1414 if (is_debug == false) 1415 { 1416 if (type == eSymbolTypeCode) 1417 { 1418 // See if we can find a N_FUN entry for any code symbols. 1419 // If we do find a match, and the name matches, then we 1420 // can merge the two into just the function symbol to avoid 1421 // duplicate entries in the symbol table 1422 ValueToSymbolIndexMap::const_iterator pos = N_FUN_addr_to_sym_idx.find (nlist.n_value); 1423 if (pos != N_FUN_addr_to_sym_idx.end()) 1424 { 1425 if ((symbol_name_is_mangled == true && sym[sym_idx].GetMangled().GetMangledName() == sym[pos->second].GetMangled().GetMangledName()) || 1426 (symbol_name_is_mangled == false && sym[sym_idx].GetMangled().GetDemangledName() == sym[pos->second].GetMangled().GetDemangledName())) 1427 { 1428 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 1429 // We just need the flags from the linker symbol, so put these flags 1430 // into the N_FUN flags to avoid duplicate symbols in the symbol table 1431 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 1432 sym[sym_idx].Clear(); 1433 continue; 1434 } 1435 } 1436 } 1437 else if (type == eSymbolTypeData) 1438 { 1439 // See if we can find a N_STSYM entry for any data symbols. 1440 // If we do find a match, and the name matches, then we 1441 // can merge the two into just the Static symbol to avoid 1442 // duplicate entries in the symbol table 1443 ValueToSymbolIndexMap::const_iterator pos = N_STSYM_addr_to_sym_idx.find (nlist.n_value); 1444 if (pos != N_STSYM_addr_to_sym_idx.end()) 1445 { 1446 if ((symbol_name_is_mangled == true && sym[sym_idx].GetMangled().GetMangledName() == sym[pos->second].GetMangled().GetMangledName()) || 1447 (symbol_name_is_mangled == false && sym[sym_idx].GetMangled().GetDemangledName() == sym[pos->second].GetMangled().GetDemangledName())) 1448 { 1449 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 1450 // We just need the flags from the linker symbol, so put these flags 1451 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 1452 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 1453 sym[sym_idx].Clear(); 1454 continue; 1455 } 1456 } 1457 } 1458 } 1459 if (symbol_section != NULL) 1460 symbol_value -= symbol_section->GetFileAddress(); 1461 1462 sym[sym_idx].SetID (nlist_idx); 1463 sym[sym_idx].SetType (type); 1464 sym[sym_idx].GetAddressRangeRef().GetBaseAddress().SetSection (symbol_section); 1465 sym[sym_idx].GetAddressRangeRef().GetBaseAddress().SetOffset (symbol_value); 1466 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 1467 1468 ++sym_idx; 1469 } 1470 else 1471 { 1472 sym[sym_idx].Clear(); 1473 } 1474 1475 } 1476 1477 // STAB N_GSYM entries end up having a symbol type eSymbolTypeGlobal and when the symbol value 1478 // is zero, the address of the global ends up being in a non-STAB entry. Try and fix up all 1479 // such entries by figuring out what the address for the global is by looking up this non-STAB 1480 // entry and copying the value into the debug symbol's value to save us the hassle in the 1481 // debug symbol parser. 1482 1483 Symbol *global_symbol = NULL; 1484 for (nlist_idx = 0; 1485 nlist_idx < symtab_load_command.nsyms && (global_symbol = symtab->FindSymbolWithType (eSymbolTypeData, Symtab::eDebugYes, Symtab::eVisibilityAny, nlist_idx)) != NULL; 1486 nlist_idx++) 1487 { 1488 if (global_symbol->GetValue().GetFileAddress() == 0) 1489 { 1490 std::vector<uint32_t> indexes; 1491 if (symtab->AppendSymbolIndexesWithName (global_symbol->GetMangled().GetName(), indexes) > 0) 1492 { 1493 std::vector<uint32_t>::const_iterator pos; 1494 std::vector<uint32_t>::const_iterator end = indexes.end(); 1495 for (pos = indexes.begin(); pos != end; ++pos) 1496 { 1497 symbol_ptr = symtab->SymbolAtIndex(*pos); 1498 if (symbol_ptr != global_symbol && symbol_ptr->IsDebug() == false) 1499 { 1500 global_symbol->SetValue(symbol_ptr->GetValue()); 1501 break; 1502 } 1503 } 1504 } 1505 } 1506 } 1507 1508 // Trim our symbols down to just what we ended up with after 1509 // removing any symbols. 1510 if (sym_idx < num_syms) 1511 { 1512 num_syms = sym_idx; 1513 sym = symtab->Resize (num_syms); 1514 } 1515 1516 // Now synthesize indirect symbols 1517 if (m_dysymtab.nindirectsyms != 0) 1518 { 1519 DataBufferSP indirect_symbol_indexes_sp(m_file.ReadFileContents(m_offset + m_dysymtab.indirectsymoff, m_dysymtab.nindirectsyms * 4)); 1520 1521 if (indirect_symbol_indexes_sp && indirect_symbol_indexes_sp->GetByteSize()) 1522 { 1523 NListIndexToSymbolIndexMap::const_iterator end_index_pos = m_nlist_idx_to_sym_idx.end(); 1524 DataExtractor indirect_symbol_index_data (indirect_symbol_indexes_sp, m_data.GetByteOrder(), m_data.GetAddressByteSize()); 1525 1526 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); ++sect_idx) 1527 { 1528 if ((m_mach_sections[sect_idx].flags & SectionFlagMaskSectionType) == SectionTypeSymbolStubs) 1529 { 1530 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 1531 if (symbol_stub_byte_size == 0) 1532 continue; 1533 1534 const uint32_t num_symbol_stubs = m_mach_sections[sect_idx].size / symbol_stub_byte_size; 1535 1536 if (num_symbol_stubs == 0) 1537 continue; 1538 1539 const uint32_t symbol_stub_index_offset = m_mach_sections[sect_idx].reserved1; 1540 uint32_t synthetic_stub_sym_id = symtab_load_command.nsyms; 1541 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) 1542 { 1543 const uint32_t symbol_stub_index = symbol_stub_index_offset + stub_idx; 1544 const lldb::addr_t symbol_stub_addr = m_mach_sections[sect_idx].addr + (stub_idx * symbol_stub_byte_size); 1545 uint32_t symbol_stub_offset = symbol_stub_index * 4; 1546 if (indirect_symbol_index_data.ValidOffsetForDataOfSize(symbol_stub_offset, 4)) 1547 { 1548 const uint32_t stub_sym_id = indirect_symbol_index_data.GetU32 (&symbol_stub_offset); 1549 if (stub_sym_id & (IndirectSymbolAbsolute | IndirectSymbolLocal)) 1550 continue; 1551 1552 NListIndexToSymbolIndexMap::const_iterator index_pos = m_nlist_idx_to_sym_idx.find (stub_sym_id); 1553 Symbol *stub_symbol = NULL; 1554 if (index_pos != end_index_pos) 1555 { 1556 // We have a remapping from the original nlist index to 1557 // a current symbol index, so just look this up by index 1558 stub_symbol = symtab->SymbolAtIndex (index_pos->second); 1559 } 1560 else 1561 { 1562 // We need to lookup a symbol using the original nlist 1563 // symbol index since this index is coming from the 1564 // S_SYMBOL_STUBS 1565 stub_symbol = symtab->FindSymbolByID (stub_sym_id); 1566 } 1567 1568 assert (stub_symbol); 1569 if (stub_symbol) 1570 { 1571 Address so_addr(symbol_stub_addr, section_list); 1572 1573 if (stub_symbol->GetType() == eSymbolTypeUndefined) 1574 { 1575 // Change the external symbol into a trampoline that makes sense 1576 // These symbols were N_UNDF N_EXT, and are useless to us, so we 1577 // can re-use them so we don't have to make up a synthetic symbol 1578 // for no good reason. 1579 stub_symbol->SetType (eSymbolTypeTrampoline); 1580 stub_symbol->SetExternal (false); 1581 stub_symbol->GetAddressRangeRef().GetBaseAddress() = so_addr; 1582 stub_symbol->GetAddressRangeRef().SetByteSize (symbol_stub_byte_size); 1583 } 1584 else 1585 { 1586 // Make a synthetic symbol to describe the trampoline stub 1587 if (sym_idx >= num_syms) 1588 sym = symtab->Resize (++num_syms); 1589 sym[sym_idx].SetID (synthetic_stub_sym_id++); 1590 sym[sym_idx].GetMangled() = stub_symbol->GetMangled(); 1591 sym[sym_idx].SetType (eSymbolTypeTrampoline); 1592 sym[sym_idx].SetIsSynthetic (true); 1593 sym[sym_idx].GetAddressRangeRef().GetBaseAddress() = so_addr; 1594 sym[sym_idx].GetAddressRangeRef().SetByteSize (symbol_stub_byte_size); 1595 ++sym_idx; 1596 } 1597 } 1598 } 1599 } 1600 } 1601 } 1602 } 1603 } 1604 return symtab->GetNumSymbols(); 1605 } 1606 } 1607 offset = cmd_offset + symtab_load_command.cmdsize; 1608 } 1609 return 0; 1610 } 1611 1612 1613 void 1614 ObjectFileMachO::Dump (Stream *s) 1615 { 1616 lldb_private::Mutex::Locker locker(m_mutex); 1617 s->Printf("%p: ", this); 1618 s->Indent(); 1619 if (m_header.magic == HeaderMagic64 || m_header.magic == HeaderMagic64Swapped) 1620 s->PutCString("ObjectFileMachO64"); 1621 else 1622 s->PutCString("ObjectFileMachO32"); 1623 1624 ArchSpec header_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 1625 1626 *s << ", file = '" << m_file << "', arch = " << header_arch.GetArchitectureName() << "\n"; 1627 1628 if (m_sections_ap.get()) 1629 m_sections_ap->Dump(s, NULL, true, UINT32_MAX); 1630 1631 if (m_symtab_ap.get()) 1632 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 1633 } 1634 1635 1636 bool 1637 ObjectFileMachO::GetUUID (lldb_private::UUID* uuid) 1638 { 1639 lldb_private::Mutex::Locker locker(m_mutex); 1640 struct uuid_command load_cmd; 1641 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 1642 uint32_t i; 1643 for (i=0; i<m_header.ncmds; ++i) 1644 { 1645 const uint32_t cmd_offset = offset; 1646 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1647 break; 1648 1649 if (load_cmd.cmd == LoadCommandUUID) 1650 { 1651 const uint8_t *uuid_bytes = m_data.PeekData(offset, 16); 1652 if (uuid_bytes) 1653 { 1654 uuid->SetBytes (uuid_bytes); 1655 return true; 1656 } 1657 return false; 1658 } 1659 offset = cmd_offset + load_cmd.cmdsize; 1660 } 1661 return false; 1662 } 1663 1664 1665 uint32_t 1666 ObjectFileMachO::GetDependentModules (FileSpecList& files) 1667 { 1668 lldb_private::Mutex::Locker locker(m_mutex); 1669 struct load_command load_cmd; 1670 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 1671 uint32_t count = 0; 1672 const bool resolve_path = false; // Don't resolve the dependend file paths since they may not reside on this system 1673 uint32_t i; 1674 for (i=0; i<m_header.ncmds; ++i) 1675 { 1676 const uint32_t cmd_offset = offset; 1677 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1678 break; 1679 1680 switch (load_cmd.cmd) 1681 { 1682 case LoadCommandDylibLoad: 1683 case LoadCommandDylibLoadWeak: 1684 case LoadCommandDylibReexport: 1685 case LoadCommandDynamicLinkerLoad: 1686 case LoadCommandFixedVMShlibLoad: 1687 case LoadCommandDylibLoadUpward: 1688 { 1689 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 1690 const char *path = m_data.PeekCStr(name_offset); 1691 // Skip any path that starts with '@' since these are usually: 1692 // @executable_path/.../file 1693 // @rpath/.../file 1694 if (path && path[0] != '@') 1695 { 1696 FileSpec file_spec(path, resolve_path); 1697 if (files.AppendIfUnique(file_spec)) 1698 count++; 1699 } 1700 } 1701 break; 1702 1703 default: 1704 break; 1705 } 1706 offset = cmd_offset + load_cmd.cmdsize; 1707 } 1708 return count; 1709 } 1710 1711 lldb_private::Address 1712 ObjectFileMachO::GetEntryPointAddress () 1713 { 1714 // If the object file is not an executable it can't hold the entry point. m_entry_point_address 1715 // is initialized to an invalid address, so we can just return that. 1716 // If m_entry_point_address is valid it means we've found it already, so return the cached value. 1717 1718 if (!IsExecutable() || m_entry_point_address.IsValid()) 1719 return m_entry_point_address; 1720 1721 // Otherwise, look for the UnixThread or Thread command. The data for the Thread command is given in 1722 // /usr/include/mach-o.h, but it is basically: 1723 // 1724 // uint32_t flavor - this is the flavor argument you would pass to thread_get_state 1725 // uint32_t count - this is the count of longs in the thread state data 1726 // struct XXX_thread_state state - this is the structure from <machine/thread_status.h> corresponding to the flavor. 1727 // <repeat this trio> 1728 // 1729 // So we just keep reading the various register flavors till we find the GPR one, then read the PC out of there. 1730 // FIXME: We will need to have a "RegisterContext data provider" class at some point that can get all the registers 1731 // out of data in this form & attach them to a given thread. That should underlie the MacOS X User process plugin, 1732 // and we'll also need it for the MacOS X Core File process plugin. When we have that we can also use it here. 1733 // 1734 // For now we hard-code the offsets and flavors we need: 1735 // 1736 // 1737 1738 lldb_private::Mutex::Locker locker(m_mutex); 1739 struct load_command load_cmd; 1740 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 1741 uint32_t i; 1742 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 1743 bool done = false; 1744 1745 for (i=0; i<m_header.ncmds; ++i) 1746 { 1747 const uint32_t cmd_offset = offset; 1748 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1749 break; 1750 1751 switch (load_cmd.cmd) 1752 { 1753 case LoadCommandUnixThread: 1754 case LoadCommandThread: 1755 { 1756 while (offset < cmd_offset + load_cmd.cmdsize) 1757 { 1758 uint32_t flavor = m_data.GetU32(&offset); 1759 uint32_t count = m_data.GetU32(&offset); 1760 if (count == 0) 1761 { 1762 // We've gotten off somehow, log and exit; 1763 return m_entry_point_address; 1764 } 1765 1766 switch (m_header.cputype) 1767 { 1768 case llvm::MachO::CPUTypeARM: 1769 if (flavor == 1) // ARM_THREAD_STATE from mach/arm/thread_status.h 1770 { 1771 offset += 60; // This is the offset of pc in the GPR thread state data structure. 1772 start_address = m_data.GetU32(&offset); 1773 done = true; 1774 } 1775 break; 1776 case llvm::MachO::CPUTypeI386: 1777 if (flavor == 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 1778 { 1779 offset += 40; // This is the offset of eip in the GPR thread state data structure. 1780 start_address = m_data.GetU32(&offset); 1781 done = true; 1782 } 1783 break; 1784 case llvm::MachO::CPUTypeX86_64: 1785 if (flavor == 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 1786 { 1787 offset += 16 * 8; // This is the offset of rip in the GPR thread state data structure. 1788 start_address = m_data.GetU64(&offset); 1789 done = true; 1790 } 1791 break; 1792 default: 1793 return m_entry_point_address; 1794 } 1795 // Haven't found the GPR flavor yet, skip over the data for this flavor: 1796 if (done) 1797 break; 1798 offset += count * 4; 1799 } 1800 } 1801 break; 1802 1803 default: 1804 break; 1805 } 1806 if (done) 1807 break; 1808 1809 // Go to the next load command: 1810 offset = cmd_offset + load_cmd.cmdsize; 1811 } 1812 1813 if (start_address != LLDB_INVALID_ADDRESS) 1814 { 1815 // We got the start address from the load commands, so now resolve that address in the sections 1816 // of this ObjectFile: 1817 if (!m_entry_point_address.ResolveAddressUsingFileSections (start_address, GetSectionList())) 1818 { 1819 m_entry_point_address.Clear(); 1820 } 1821 } 1822 else 1823 { 1824 // We couldn't read the UnixThread load command - maybe it wasn't there. As a fallback look for the 1825 // "start" symbol in the main executable. 1826 1827 SymbolContextList contexts; 1828 SymbolContext context; 1829 if (!m_module->FindSymbolsWithNameAndType(ConstString ("start"), eSymbolTypeCode, contexts)) 1830 return m_entry_point_address; 1831 1832 contexts.GetContextAtIndex(0, context); 1833 1834 m_entry_point_address = context.symbol->GetValue(); 1835 } 1836 1837 return m_entry_point_address; 1838 1839 } 1840 1841 ObjectFile::Type 1842 ObjectFileMachO::CalculateType() 1843 { 1844 switch (m_header.filetype) 1845 { 1846 case HeaderFileTypeObject: // 0x1u MH_OBJECT 1847 if (GetAddressByteSize () == 4) 1848 { 1849 // 32 bit kexts are just object files, but they do have a valid 1850 // UUID load command. 1851 UUID uuid; 1852 if (GetUUID(&uuid)) 1853 { 1854 // this checking for the UUID load command is not enough 1855 // we could eventually look for the symbol named 1856 // "OSKextGetCurrentIdentifier" as this is required of kexts 1857 if (m_strata == eStrataInvalid) 1858 m_strata = eStrataKernel; 1859 return eTypeSharedLibrary; 1860 } 1861 } 1862 return eTypeObjectFile; 1863 1864 case HeaderFileTypeExecutable: return eTypeExecutable; // 0x2u MH_EXECUTE 1865 case HeaderFileTypeFixedVMShlib: return eTypeSharedLibrary; // 0x3u MH_FVMLIB 1866 case HeaderFileTypeCore: return eTypeCoreFile; // 0x4u MH_CORE 1867 case HeaderFileTypePreloadedExecutable: return eTypeSharedLibrary; // 0x5u MH_PRELOAD 1868 case HeaderFileTypeDynamicShlib: return eTypeSharedLibrary; // 0x6u MH_DYLIB 1869 case HeaderFileTypeDynamicLinkEditor: return eTypeDynamicLinker; // 0x7u MH_DYLINKER 1870 case HeaderFileTypeBundle: return eTypeSharedLibrary; // 0x8u MH_BUNDLE 1871 case HeaderFileTypeDynamicShlibStub: return eTypeStubLibrary; // 0x9u MH_DYLIB_STUB 1872 case HeaderFileTypeDSYM: return eTypeDebugInfo; // 0xAu MH_DSYM 1873 case HeaderFileTypeKextBundle: return eTypeSharedLibrary; // 0xBu MH_KEXT_BUNDLE 1874 default: 1875 break; 1876 } 1877 return eTypeUnknown; 1878 } 1879 1880 ObjectFile::Strata 1881 ObjectFileMachO::CalculateStrata() 1882 { 1883 switch (m_header.filetype) 1884 { 1885 case HeaderFileTypeObject: // 0x1u MH_OBJECT 1886 { 1887 // 32 bit kexts are just object files, but they do have a valid 1888 // UUID load command. 1889 UUID uuid; 1890 if (GetUUID(&uuid)) 1891 { 1892 // this checking for the UUID load command is not enough 1893 // we could eventually look for the symbol named 1894 // "OSKextGetCurrentIdentifier" as this is required of kexts 1895 if (m_type == eTypeInvalid) 1896 m_type = eTypeSharedLibrary; 1897 1898 return eStrataKernel; 1899 } 1900 } 1901 return eStrataUnknown; 1902 1903 case HeaderFileTypeExecutable: // 0x2u MH_EXECUTE 1904 // Check for the MH_DYLDLINK bit in the flags 1905 if (m_header.flags & HeaderFlagBitIsDynamicLinkObject) 1906 return eStrataUser; 1907 return eStrataKernel; 1908 1909 case HeaderFileTypeFixedVMShlib: return eStrataUser; // 0x3u MH_FVMLIB 1910 case HeaderFileTypeCore: return eStrataUnknown; // 0x4u MH_CORE 1911 case HeaderFileTypePreloadedExecutable: return eStrataUser; // 0x5u MH_PRELOAD 1912 case HeaderFileTypeDynamicShlib: return eStrataUser; // 0x6u MH_DYLIB 1913 case HeaderFileTypeDynamicLinkEditor: return eStrataUser; // 0x7u MH_DYLINKER 1914 case HeaderFileTypeBundle: return eStrataUser; // 0x8u MH_BUNDLE 1915 case HeaderFileTypeDynamicShlibStub: return eStrataUser; // 0x9u MH_DYLIB_STUB 1916 case HeaderFileTypeDSYM: return eStrataUnknown; // 0xAu MH_DSYM 1917 case HeaderFileTypeKextBundle: return eStrataKernel; // 0xBu MH_KEXT_BUNDLE 1918 default: 1919 break; 1920 } 1921 return eStrataUnknown; 1922 } 1923 1924 1925 bool 1926 ObjectFileMachO::GetArchitecture (ArchSpec &arch) 1927 { 1928 lldb_private::Mutex::Locker locker(m_mutex); 1929 arch.SetArchitecture (eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 1930 1931 // Files with type MH_PRELOAD are currently used in cases where the image 1932 // debugs at the addresses in the file itself. Below we set the OS to 1933 // unknown to make sure we use the DynamicLoaderStatic()... 1934 if (m_header.filetype == HeaderFileTypePreloadedExecutable) 1935 { 1936 arch.GetTriple().setOS (llvm::Triple::UnknownOS); 1937 } 1938 1939 return true; 1940 } 1941 1942 1943 //------------------------------------------------------------------ 1944 // PluginInterface protocol 1945 //------------------------------------------------------------------ 1946 const char * 1947 ObjectFileMachO::GetPluginName() 1948 { 1949 return "ObjectFileMachO"; 1950 } 1951 1952 const char * 1953 ObjectFileMachO::GetShortPluginName() 1954 { 1955 return GetPluginNameStatic(); 1956 } 1957 1958 uint32_t 1959 ObjectFileMachO::GetPluginVersion() 1960 { 1961 return 1; 1962 } 1963 1964