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/Core/FileSpecList.h" 18 #include "lldb/Core/Module.h" 19 #include "lldb/Core/PluginManager.h" 20 #include "lldb/Core/Section.h" 21 #include "lldb/Core/StreamFile.h" 22 #include "lldb/Core/StreamString.h" 23 #include "lldb/Core/Timer.h" 24 #include "lldb/Core/UUID.h" 25 #include "lldb/Host/Host.h" 26 #include "lldb/Host/FileSpec.h" 27 #include "lldb/Symbol/ClangNamespaceDecl.h" 28 #include "lldb/Symbol/ObjectFile.h" 29 30 31 using namespace lldb; 32 using namespace lldb_private; 33 using namespace llvm::MachO; 34 35 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 36 37 void 38 ObjectFileMachO::Initialize() 39 { 40 PluginManager::RegisterPlugin (GetPluginNameStatic(), 41 GetPluginDescriptionStatic(), 42 CreateInstance); 43 } 44 45 void 46 ObjectFileMachO::Terminate() 47 { 48 PluginManager::UnregisterPlugin (CreateInstance); 49 } 50 51 52 const char * 53 ObjectFileMachO::GetPluginNameStatic() 54 { 55 return "object-file.mach-o"; 56 } 57 58 const char * 59 ObjectFileMachO::GetPluginDescriptionStatic() 60 { 61 return "Mach-o object file reader (32 and 64 bit)"; 62 } 63 64 65 ObjectFile * 66 ObjectFileMachO::CreateInstance (Module* module, DataBufferSP& data_sp, const FileSpec* file, addr_t offset, addr_t length) 67 { 68 if (ObjectFileMachO::MagicBytesMatch(data_sp, offset, length)) 69 { 70 std::auto_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module, data_sp, file, offset, length)); 71 if (objfile_ap.get() && objfile_ap->ParseHeader()) 72 return objfile_ap.release(); 73 } 74 return NULL; 75 } 76 77 78 static uint32_t 79 MachHeaderSizeFromMagic(uint32_t magic) 80 { 81 switch (magic) 82 { 83 case HeaderMagic32: 84 case HeaderMagic32Swapped: 85 return sizeof(struct mach_header); 86 87 case HeaderMagic64: 88 case HeaderMagic64Swapped: 89 return sizeof(struct mach_header_64); 90 break; 91 92 default: 93 break; 94 } 95 return 0; 96 } 97 98 99 bool 100 ObjectFileMachO::MagicBytesMatch (DataBufferSP& data_sp, 101 lldb::addr_t data_offset, 102 lldb::addr_t data_length) 103 { 104 DataExtractor data; 105 data.SetData (data_sp, data_offset, data_length); 106 uint32_t offset = 0; 107 uint32_t magic = data.GetU32(&offset); 108 return MachHeaderSizeFromMagic(magic) != 0; 109 } 110 111 112 ObjectFileMachO::ObjectFileMachO(Module* module, DataBufferSP& data_sp, const FileSpec* file, addr_t offset, addr_t length) : 113 ObjectFile(module, file, offset, length, data_sp), 114 m_mutex (Mutex::eMutexTypeRecursive), 115 m_header(), 116 m_sections_ap(), 117 m_symtab_ap(), 118 m_entry_point_address () 119 { 120 ::memset (&m_header, 0, sizeof(m_header)); 121 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 122 } 123 124 125 ObjectFileMachO::~ObjectFileMachO() 126 { 127 } 128 129 130 bool 131 ObjectFileMachO::ParseHeader () 132 { 133 lldb_private::Mutex::Locker locker(m_mutex); 134 bool can_parse = false; 135 uint32_t offset = 0; 136 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 137 // Leave magic in the original byte order 138 m_header.magic = m_data.GetU32(&offset); 139 switch (m_header.magic) 140 { 141 case HeaderMagic32: 142 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 143 m_data.SetAddressByteSize(4); 144 can_parse = true; 145 break; 146 147 case HeaderMagic64: 148 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 149 m_data.SetAddressByteSize(8); 150 can_parse = true; 151 break; 152 153 case HeaderMagic32Swapped: 154 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 155 m_data.SetAddressByteSize(4); 156 can_parse = true; 157 break; 158 159 case HeaderMagic64Swapped: 160 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 161 m_data.SetAddressByteSize(8); 162 can_parse = true; 163 break; 164 165 default: 166 break; 167 } 168 169 if (can_parse) 170 { 171 m_data.GetU32(&offset, &m_header.cputype, 6); 172 173 ArchSpec mach_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 174 175 if (SetModulesArchitecture (mach_arch)) 176 return true; 177 } 178 else 179 { 180 memset(&m_header, 0, sizeof(struct mach_header)); 181 } 182 return false; 183 } 184 185 186 ByteOrder 187 ObjectFileMachO::GetByteOrder () const 188 { 189 lldb_private::Mutex::Locker locker(m_mutex); 190 return m_data.GetByteOrder (); 191 } 192 193 bool 194 ObjectFileMachO::IsExecutable() const 195 { 196 return m_header.filetype == HeaderFileTypeExecutable; 197 } 198 199 size_t 200 ObjectFileMachO::GetAddressByteSize () const 201 { 202 lldb_private::Mutex::Locker locker(m_mutex); 203 return m_data.GetAddressByteSize (); 204 } 205 206 AddressClass 207 ObjectFileMachO::GetAddressClass (lldb::addr_t file_addr) 208 { 209 Symtab *symtab = GetSymtab(); 210 if (symtab) 211 { 212 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 213 if (symbol) 214 { 215 const AddressRange *range_ptr = symbol->GetAddressRangePtr(); 216 if (range_ptr) 217 { 218 const Section *section = range_ptr->GetBaseAddress().GetSection(); 219 if (section) 220 { 221 const SectionType section_type = section->GetType(); 222 switch (section_type) 223 { 224 case eSectionTypeInvalid: return eAddressClassUnknown; 225 case eSectionTypeCode: 226 if (m_header.cputype == llvm::MachO::CPUTypeARM) 227 { 228 // For ARM we have a bit in the n_desc field of the symbol 229 // that tells us ARM/Thumb which is bit 0x0008. 230 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 231 return eAddressClassCodeAlternateISA; 232 } 233 return eAddressClassCode; 234 235 case eSectionTypeContainer: return eAddressClassUnknown; 236 case eSectionTypeData: 237 case eSectionTypeDataCString: 238 case eSectionTypeDataCStringPointers: 239 case eSectionTypeDataSymbolAddress: 240 case eSectionTypeData4: 241 case eSectionTypeData8: 242 case eSectionTypeData16: 243 case eSectionTypeDataPointers: 244 case eSectionTypeZeroFill: 245 case eSectionTypeDataObjCMessageRefs: 246 case eSectionTypeDataObjCCFStrings: 247 return eAddressClassData; 248 case eSectionTypeDebug: 249 case eSectionTypeDWARFDebugAbbrev: 250 case eSectionTypeDWARFDebugAranges: 251 case eSectionTypeDWARFDebugFrame: 252 case eSectionTypeDWARFDebugInfo: 253 case eSectionTypeDWARFDebugLine: 254 case eSectionTypeDWARFDebugLoc: 255 case eSectionTypeDWARFDebugMacInfo: 256 case eSectionTypeDWARFDebugPubNames: 257 case eSectionTypeDWARFDebugPubTypes: 258 case eSectionTypeDWARFDebugRanges: 259 case eSectionTypeDWARFDebugStr: 260 case eSectionTypeDWARFAppleNames: 261 case eSectionTypeDWARFAppleTypes: 262 case eSectionTypeDWARFAppleNamespaces: 263 case eSectionTypeDWARFAppleObjC: 264 return eAddressClassDebug; 265 case eSectionTypeEHFrame: return eAddressClassRuntime; 266 case eSectionTypeOther: return eAddressClassUnknown; 267 } 268 } 269 } 270 271 const SymbolType symbol_type = symbol->GetType(); 272 switch (symbol_type) 273 { 274 case eSymbolTypeAny: return eAddressClassUnknown; 275 case eSymbolTypeAbsolute: return eAddressClassUnknown; 276 277 case eSymbolTypeCode: 278 case eSymbolTypeTrampoline: 279 if (m_header.cputype == llvm::MachO::CPUTypeARM) 280 { 281 // For ARM we have a bit in the n_desc field of the symbol 282 // that tells us ARM/Thumb which is bit 0x0008. 283 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 284 return eAddressClassCodeAlternateISA; 285 } 286 return eAddressClassCode; 287 288 case eSymbolTypeData: return eAddressClassData; 289 case eSymbolTypeRuntime: return eAddressClassRuntime; 290 case eSymbolTypeException: return eAddressClassRuntime; 291 case eSymbolTypeSourceFile: return eAddressClassDebug; 292 case eSymbolTypeHeaderFile: return eAddressClassDebug; 293 case eSymbolTypeObjectFile: return eAddressClassDebug; 294 case eSymbolTypeCommonBlock: return eAddressClassDebug; 295 case eSymbolTypeBlock: return eAddressClassDebug; 296 case eSymbolTypeLocal: return eAddressClassData; 297 case eSymbolTypeParam: return eAddressClassData; 298 case eSymbolTypeVariable: return eAddressClassData; 299 case eSymbolTypeVariableType: return eAddressClassDebug; 300 case eSymbolTypeLineEntry: return eAddressClassDebug; 301 case eSymbolTypeLineHeader: return eAddressClassDebug; 302 case eSymbolTypeScopeBegin: return eAddressClassDebug; 303 case eSymbolTypeScopeEnd: return eAddressClassDebug; 304 case eSymbolTypeAdditional: return eAddressClassUnknown; 305 case eSymbolTypeCompiler: return eAddressClassDebug; 306 case eSymbolTypeInstrumentation:return eAddressClassDebug; 307 case eSymbolTypeUndefined: return eAddressClassUnknown; 308 case eSymbolTypeObjCClass: return eAddressClassRuntime; 309 case eSymbolTypeObjCMetaClass: return eAddressClassRuntime; 310 case eSymbolTypeObjCIVar: return eAddressClassRuntime; 311 } 312 } 313 } 314 return eAddressClassUnknown; 315 } 316 317 Symtab * 318 ObjectFileMachO::GetSymtab() 319 { 320 lldb_private::Mutex::Locker symfile_locker(m_mutex); 321 if (m_symtab_ap.get() == NULL) 322 { 323 m_symtab_ap.reset(new Symtab(this)); 324 Mutex::Locker symtab_locker (m_symtab_ap->GetMutex()); 325 ParseSymtab (true); 326 m_symtab_ap->Finalize (); 327 } 328 return m_symtab_ap.get(); 329 } 330 331 332 SectionList * 333 ObjectFileMachO::GetSectionList() 334 { 335 lldb_private::Mutex::Locker locker(m_mutex); 336 if (m_sections_ap.get() == NULL) 337 { 338 m_sections_ap.reset(new SectionList()); 339 ParseSections(); 340 } 341 return m_sections_ap.get(); 342 } 343 344 345 size_t 346 ObjectFileMachO::ParseSections () 347 { 348 lldb::user_id_t segID = 0; 349 lldb::user_id_t sectID = 0; 350 struct segment_command_64 load_cmd; 351 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 352 uint32_t i; 353 //bool dump_sections = false; 354 for (i=0; i<m_header.ncmds; ++i) 355 { 356 const uint32_t load_cmd_offset = offset; 357 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 358 break; 359 360 if (load_cmd.cmd == LoadCommandSegment32 || load_cmd.cmd == LoadCommandSegment64) 361 { 362 if (m_data.GetU8(&offset, (uint8_t*)load_cmd.segname, 16)) 363 { 364 load_cmd.vmaddr = m_data.GetAddress(&offset); 365 load_cmd.vmsize = m_data.GetAddress(&offset); 366 load_cmd.fileoff = m_data.GetAddress(&offset); 367 load_cmd.filesize = m_data.GetAddress(&offset); 368 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 369 { 370 371 const bool segment_is_encrypted = (load_cmd.flags & SegmentCommandFlagBitProtectedVersion1) != 0; 372 373 // Keep a list of mach segments around in case we need to 374 // get at data that isn't stored in the abstracted Sections. 375 m_mach_segments.push_back (load_cmd); 376 377 ConstString segment_name (load_cmd.segname, std::min<int>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 378 // Use a segment ID of the segment index shifted left by 8 so they 379 // never conflict with any of the sections. 380 SectionSP segment_sp; 381 if (segment_name) 382 { 383 segment_sp.reset(new Section (NULL, 384 GetModule(), // Module to which this section belongs 385 ++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 386 segment_name, // Name of this section 387 eSectionTypeContainer, // This section is a container of other sections. 388 load_cmd.vmaddr, // File VM address == addresses as they are found in the object file 389 load_cmd.vmsize, // VM size in bytes of this section 390 load_cmd.fileoff, // Offset to the data for this section in the file 391 load_cmd.filesize, // Size in bytes of this section as found in the the file 392 load_cmd.flags)); // Flags for this section 393 394 segment_sp->SetIsEncrypted (segment_is_encrypted); 395 m_sections_ap->AddSection(segment_sp); 396 } 397 398 struct section_64 sect64; 399 ::memset (§64, 0, sizeof(sect64)); 400 // Push a section into our mach sections for the section at 401 // index zero (NListSectionNoSection) if we don't have any 402 // mach sections yet... 403 if (m_mach_sections.empty()) 404 m_mach_sections.push_back(sect64); 405 uint32_t segment_sect_idx; 406 const lldb::user_id_t first_segment_sectID = sectID + 1; 407 408 409 const uint32_t num_u32s = load_cmd.cmd == LoadCommandSegment32 ? 7 : 8; 410 for (segment_sect_idx=0; segment_sect_idx<load_cmd.nsects; ++segment_sect_idx) 411 { 412 if (m_data.GetU8(&offset, (uint8_t*)sect64.sectname, sizeof(sect64.sectname)) == NULL) 413 break; 414 if (m_data.GetU8(&offset, (uint8_t*)sect64.segname, sizeof(sect64.segname)) == NULL) 415 break; 416 sect64.addr = m_data.GetAddress(&offset); 417 sect64.size = m_data.GetAddress(&offset); 418 419 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 420 break; 421 422 // Keep a list of mach sections around in case we need to 423 // get at data that isn't stored in the abstracted Sections. 424 m_mach_sections.push_back (sect64); 425 426 ConstString section_name (sect64.sectname, std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 427 if (!segment_name) 428 { 429 // We have a segment with no name so we need to conjure up 430 // segments that correspond to the section's segname if there 431 // isn't already such a section. If there is such a section, 432 // we resize the section so that it spans all sections. 433 // We also mark these sections as fake so address matches don't 434 // hit if they land in the gaps between the child sections. 435 segment_name.SetTrimmedCStringWithLength(sect64.segname, sizeof(sect64.segname)); 436 segment_sp = m_sections_ap->FindSectionByName (segment_name); 437 if (segment_sp.get()) 438 { 439 Section *segment = segment_sp.get(); 440 // Grow the section size as needed. 441 const lldb::addr_t sect64_min_addr = sect64.addr; 442 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 443 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 444 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 445 const lldb::addr_t curr_seg_max_addr = curr_seg_min_addr + curr_seg_byte_size; 446 if (sect64_min_addr >= curr_seg_min_addr) 447 { 448 const lldb::addr_t new_seg_byte_size = sect64_max_addr - curr_seg_min_addr; 449 // Only grow the section size if needed 450 if (new_seg_byte_size > curr_seg_byte_size) 451 segment->SetByteSize (new_seg_byte_size); 452 } 453 else 454 { 455 // We need to change the base address of the segment and 456 // adjust the child section offsets for all existing children. 457 const lldb::addr_t slide_amount = sect64_min_addr - curr_seg_min_addr; 458 segment->Slide(slide_amount, false); 459 segment->GetChildren().Slide (-slide_amount, false); 460 segment->SetByteSize (curr_seg_max_addr - sect64_min_addr); 461 } 462 463 // Grow the section size as needed. 464 if (sect64.offset) 465 { 466 const lldb::addr_t segment_min_file_offset = segment->GetFileOffset(); 467 const lldb::addr_t segment_max_file_offset = segment_min_file_offset + segment->GetFileSize(); 468 469 const lldb::addr_t section_min_file_offset = sect64.offset; 470 const lldb::addr_t section_max_file_offset = section_min_file_offset + sect64.size; 471 const lldb::addr_t new_file_offset = std::min (section_min_file_offset, segment_min_file_offset); 472 const lldb::addr_t new_file_size = std::max (section_max_file_offset, segment_max_file_offset) - new_file_offset; 473 segment->SetFileOffset (new_file_offset); 474 segment->SetFileSize (new_file_size); 475 } 476 } 477 else 478 { 479 // Create a fake section for the section's named segment 480 segment_sp.reset(new Section(segment_sp.get(), // Parent section 481 GetModule(), // Module to which this section belongs 482 ++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 483 segment_name, // Name of this section 484 eSectionTypeContainer, // This section is a container of other sections. 485 sect64.addr, // File VM address == addresses as they are found in the object file 486 sect64.size, // VM size in bytes of this section 487 sect64.offset, // Offset to the data for this section in the file 488 sect64.offset ? sect64.size : 0, // Size in bytes of this section as found in the the file 489 load_cmd.flags)); // Flags for this section 490 segment_sp->SetIsFake(true); 491 m_sections_ap->AddSection(segment_sp); 492 segment_sp->SetIsEncrypted (segment_is_encrypted); 493 } 494 } 495 assert (segment_sp.get()); 496 497 uint32_t mach_sect_type = sect64.flags & SectionFlagMaskSectionType; 498 static ConstString g_sect_name_objc_data ("__objc_data"); 499 static ConstString g_sect_name_objc_msgrefs ("__objc_msgrefs"); 500 static ConstString g_sect_name_objc_selrefs ("__objc_selrefs"); 501 static ConstString g_sect_name_objc_classrefs ("__objc_classrefs"); 502 static ConstString g_sect_name_objc_superrefs ("__objc_superrefs"); 503 static ConstString g_sect_name_objc_const ("__objc_const"); 504 static ConstString g_sect_name_objc_classlist ("__objc_classlist"); 505 static ConstString g_sect_name_cfstring ("__cfstring"); 506 507 static ConstString g_sect_name_dwarf_debug_abbrev ("__debug_abbrev"); 508 static ConstString g_sect_name_dwarf_debug_aranges ("__debug_aranges"); 509 static ConstString g_sect_name_dwarf_debug_frame ("__debug_frame"); 510 static ConstString g_sect_name_dwarf_debug_info ("__debug_info"); 511 static ConstString g_sect_name_dwarf_debug_line ("__debug_line"); 512 static ConstString g_sect_name_dwarf_debug_loc ("__debug_loc"); 513 static ConstString g_sect_name_dwarf_debug_macinfo ("__debug_macinfo"); 514 static ConstString g_sect_name_dwarf_debug_pubnames ("__debug_pubnames"); 515 static ConstString g_sect_name_dwarf_debug_pubtypes ("__debug_pubtypes"); 516 static ConstString g_sect_name_dwarf_debug_ranges ("__debug_ranges"); 517 static ConstString g_sect_name_dwarf_debug_str ("__debug_str"); 518 static ConstString g_sect_name_dwarf_apple_names ("__apple_names"); 519 static ConstString g_sect_name_dwarf_apple_types ("__apple_types"); 520 static ConstString g_sect_name_dwarf_apple_namespaces ("__apple_namespac"); 521 static ConstString g_sect_name_dwarf_apple_objc ("__apple_objc"); 522 static ConstString g_sect_name_eh_frame ("__eh_frame"); 523 static ConstString g_sect_name_DATA ("__DATA"); 524 static ConstString g_sect_name_TEXT ("__TEXT"); 525 526 SectionType sect_type = eSectionTypeOther; 527 528 if (section_name == g_sect_name_dwarf_debug_abbrev) 529 sect_type = eSectionTypeDWARFDebugAbbrev; 530 else if (section_name == g_sect_name_dwarf_debug_aranges) 531 sect_type = eSectionTypeDWARFDebugAranges; 532 else if (section_name == g_sect_name_dwarf_debug_frame) 533 sect_type = eSectionTypeDWARFDebugFrame; 534 else if (section_name == g_sect_name_dwarf_debug_info) 535 sect_type = eSectionTypeDWARFDebugInfo; 536 else if (section_name == g_sect_name_dwarf_debug_line) 537 sect_type = eSectionTypeDWARFDebugLine; 538 else if (section_name == g_sect_name_dwarf_debug_loc) 539 sect_type = eSectionTypeDWARFDebugLoc; 540 else if (section_name == g_sect_name_dwarf_debug_macinfo) 541 sect_type = eSectionTypeDWARFDebugMacInfo; 542 else if (section_name == g_sect_name_dwarf_debug_pubnames) 543 sect_type = eSectionTypeDWARFDebugPubNames; 544 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 545 sect_type = eSectionTypeDWARFDebugPubTypes; 546 else if (section_name == g_sect_name_dwarf_debug_ranges) 547 sect_type = eSectionTypeDWARFDebugRanges; 548 else if (section_name == g_sect_name_dwarf_debug_str) 549 sect_type = eSectionTypeDWARFDebugStr; 550 else if (section_name == g_sect_name_dwarf_apple_names) 551 sect_type = eSectionTypeDWARFAppleNames; 552 else if (section_name == g_sect_name_dwarf_apple_types) 553 sect_type = eSectionTypeDWARFAppleTypes; 554 else if (section_name == g_sect_name_dwarf_apple_namespaces) 555 sect_type = eSectionTypeDWARFAppleNamespaces; 556 else if (section_name == g_sect_name_dwarf_apple_objc) 557 sect_type = eSectionTypeDWARFAppleObjC; 558 else if (section_name == g_sect_name_objc_selrefs) 559 sect_type = eSectionTypeDataCStringPointers; 560 else if (section_name == g_sect_name_objc_msgrefs) 561 sect_type = eSectionTypeDataObjCMessageRefs; 562 else if (section_name == g_sect_name_eh_frame) 563 sect_type = eSectionTypeEHFrame; 564 else if (section_name == g_sect_name_cfstring) 565 sect_type = eSectionTypeDataObjCCFStrings; 566 else if (section_name == g_sect_name_objc_data || 567 section_name == g_sect_name_objc_classrefs || 568 section_name == g_sect_name_objc_superrefs || 569 section_name == g_sect_name_objc_const || 570 section_name == g_sect_name_objc_classlist) 571 { 572 sect_type = eSectionTypeDataPointers; 573 } 574 575 if (sect_type == eSectionTypeOther) 576 { 577 switch (mach_sect_type) 578 { 579 // TODO: categorize sections by other flags for regular sections 580 case SectionTypeRegular: 581 if (segment_sp->GetName() == g_sect_name_TEXT) 582 sect_type = eSectionTypeCode; 583 else if (segment_sp->GetName() == g_sect_name_DATA) 584 sect_type = eSectionTypeData; 585 else 586 sect_type = eSectionTypeOther; 587 break; 588 case SectionTypeZeroFill: sect_type = eSectionTypeZeroFill; break; 589 case SectionTypeCStringLiterals: sect_type = eSectionTypeDataCString; break; // section with only literal C strings 590 case SectionType4ByteLiterals: sect_type = eSectionTypeData4; break; // section with only 4 byte literals 591 case SectionType8ByteLiterals: sect_type = eSectionTypeData8; break; // section with only 8 byte literals 592 case SectionTypeLiteralPointers: sect_type = eSectionTypeDataPointers; break; // section with only pointers to literals 593 case SectionTypeNonLazySymbolPointers: sect_type = eSectionTypeDataPointers; break; // section with only non-lazy symbol pointers 594 case SectionTypeLazySymbolPointers: sect_type = eSectionTypeDataPointers; break; // section with only lazy symbol pointers 595 case SectionTypeSymbolStubs: sect_type = eSectionTypeCode; break; // section with only symbol stubs, byte size of stub in the reserved2 field 596 case SectionTypeModuleInitFunctionPointers: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for initialization 597 case SectionTypeModuleTermFunctionPointers: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for termination 598 case SectionTypeCoalesced: sect_type = eSectionTypeOther; break; 599 case SectionTypeZeroFillLarge: sect_type = eSectionTypeZeroFill; break; 600 case SectionTypeInterposing: sect_type = eSectionTypeCode; break; // section with only pairs of function pointers for interposing 601 case SectionType16ByteLiterals: sect_type = eSectionTypeData16; break; // section with only 16 byte literals 602 case SectionTypeDTraceObjectFormat: sect_type = eSectionTypeDebug; break; 603 case SectionTypeLazyDylibSymbolPointers: sect_type = eSectionTypeDataPointers; break; 604 default: break; 605 } 606 } 607 608 SectionSP section_sp(new Section(segment_sp.get(), 609 GetModule(), 610 ++sectID, 611 section_name, 612 sect_type, 613 sect64.addr - segment_sp->GetFileAddress(), 614 sect64.size, 615 sect64.offset, 616 sect64.offset == 0 ? 0 : sect64.size, 617 sect64.flags)); 618 // Set the section to be encrypted to match the segment 619 section_sp->SetIsEncrypted (segment_is_encrypted); 620 621 segment_sp->GetChildren().AddSection(section_sp); 622 623 if (segment_sp->IsFake()) 624 { 625 segment_sp.reset(); 626 segment_name.Clear(); 627 } 628 } 629 if (segment_sp && m_header.filetype == HeaderFileTypeDSYM) 630 { 631 if (first_segment_sectID <= sectID) 632 { 633 lldb::user_id_t sect_uid; 634 for (sect_uid = first_segment_sectID; sect_uid <= sectID; ++sect_uid) 635 { 636 SectionSP curr_section_sp(segment_sp->GetChildren().FindSectionByID (sect_uid)); 637 SectionSP next_section_sp; 638 if (sect_uid + 1 <= sectID) 639 next_section_sp = segment_sp->GetChildren().FindSectionByID (sect_uid+1); 640 641 if (curr_section_sp.get()) 642 { 643 if (curr_section_sp->GetByteSize() == 0) 644 { 645 if (next_section_sp.get() != NULL) 646 curr_section_sp->SetByteSize ( next_section_sp->GetFileAddress() - curr_section_sp->GetFileAddress() ); 647 else 648 curr_section_sp->SetByteSize ( load_cmd.vmsize ); 649 } 650 } 651 } 652 } 653 } 654 } 655 } 656 } 657 else if (load_cmd.cmd == LoadCommandDynamicSymtabInfo) 658 { 659 m_dysymtab.cmd = load_cmd.cmd; 660 m_dysymtab.cmdsize = load_cmd.cmdsize; 661 m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 662 } 663 664 offset = load_cmd_offset + load_cmd.cmdsize; 665 } 666 // if (dump_sections) 667 // { 668 // StreamFile s(stdout); 669 // m_sections_ap->Dump(&s, true); 670 // } 671 return sectID; // Return the number of sections we registered with the module 672 } 673 674 class MachSymtabSectionInfo 675 { 676 public: 677 678 MachSymtabSectionInfo (SectionList *section_list) : 679 m_section_list (section_list), 680 m_section_infos() 681 { 682 // Get the number of sections down to a depth of 1 to include 683 // all segments and their sections, but no other sections that 684 // may be added for debug map or 685 m_section_infos.resize(section_list->GetNumSections(1)); 686 } 687 688 689 Section * 690 GetSection (uint8_t n_sect, addr_t file_addr) 691 { 692 if (n_sect == 0) 693 return NULL; 694 if (n_sect < m_section_infos.size()) 695 { 696 if (m_section_infos[n_sect].section == NULL) 697 { 698 Section *section = m_section_list->FindSectionByID (n_sect).get(); 699 m_section_infos[n_sect].section = section; 700 if (section != NULL) 701 { 702 m_section_infos[n_sect].vm_range.SetBaseAddress (section->GetFileAddress()); 703 m_section_infos[n_sect].vm_range.SetByteSize (section->GetByteSize()); 704 } 705 else 706 { 707 Host::SystemLog (Host::eSystemLogError, "error: unable to find section for section %u\n", n_sect); 708 } 709 } 710 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) 711 { 712 // Symbol is in section. 713 return m_section_infos[n_sect].section; 714 } 715 else if (m_section_infos[n_sect].vm_range.GetByteSize () == 0 && 716 m_section_infos[n_sect].vm_range.GetBaseAddress() == file_addr) 717 { 718 // Symbol is in section with zero size, but has the same start 719 // address as the section. This can happen with linker symbols 720 // (symbols that start with the letter 'l' or 'L'. 721 return m_section_infos[n_sect].section; 722 } 723 } 724 return m_section_list->FindSectionContainingFileAddress(file_addr).get(); 725 } 726 727 protected: 728 struct SectionInfo 729 { 730 SectionInfo () : 731 vm_range(), 732 section (NULL) 733 { 734 } 735 736 VMRange vm_range; 737 Section *section; 738 }; 739 SectionList *m_section_list; 740 std::vector<SectionInfo> m_section_infos; 741 }; 742 743 744 745 size_t 746 ObjectFileMachO::ParseSymtab (bool minimize) 747 { 748 Timer scoped_timer(__PRETTY_FUNCTION__, 749 "ObjectFileMachO::ParseSymtab () module = %s", 750 m_file.GetFilename().AsCString("")); 751 struct symtab_command symtab_load_command; 752 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 753 uint32_t i; 754 755 LogSP log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_SYMBOLS)); 756 757 for (i=0; i<m_header.ncmds; ++i) 758 { 759 const uint32_t cmd_offset = offset; 760 // Read in the load command and load command size 761 if (m_data.GetU32(&offset, &symtab_load_command, 2) == NULL) 762 break; 763 // Watch for the symbol table load command 764 if (symtab_load_command.cmd == LoadCommandSymtab) 765 { 766 // Read in the rest of the symtab load command 767 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4)) // fill in symoff, nsyms, stroff, strsize fields 768 { 769 if (symtab_load_command.symoff == 0) 770 { 771 if (log) 772 GetModule()->LogMessage(log.get(), "LC_SYMTAB.symoff == 0"); 773 return 0; 774 } 775 776 if (symtab_load_command.stroff == 0) 777 { 778 if (log) 779 GetModule()->LogMessage(log.get(), "LC_SYMTAB.stroff == 0"); 780 return 0; 781 } 782 783 if (symtab_load_command.nsyms == 0) 784 { 785 if (log) 786 GetModule()->LogMessage(log.get(), "LC_SYMTAB.nsyms == 0"); 787 return 0; 788 } 789 790 if (symtab_load_command.strsize == 0) 791 { 792 if (log) 793 GetModule()->LogMessage(log.get(), "LC_SYMTAB.strsize == 0"); 794 return 0; 795 } 796 797 Symtab *symtab = m_symtab_ap.get(); 798 SectionList *section_list = GetSectionList(); 799 if (section_list == NULL) 800 return 0; 801 802 const size_t addr_byte_size = m_data.GetAddressByteSize(); 803 bool bit_width_32 = addr_byte_size == 4; 804 const size_t nlist_byte_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 805 806 DataExtractor nlist_data (m_data, symtab_load_command.symoff, symtab_load_command.nsyms * nlist_byte_size); 807 808 if (nlist_data.GetByteSize() == 0) 809 { 810 if (log) 811 GetModule()->LogMessage(log.get(), "failed to read nlist data"); 812 return 0; 813 } 814 815 DataExtractor strtab_data (m_data, symtab_load_command.stroff, symtab_load_command.strsize); 816 817 if (strtab_data.GetByteSize() == 0) 818 { 819 if (log) 820 GetModule()->LogMessage(log.get(), "failed to read strtab data"); 821 return 0; 822 } 823 824 static ConstString g_segment_name_TEXT ("__TEXT"); 825 static ConstString g_segment_name_DATA ("__DATA"); 826 static ConstString g_segment_name_OBJC ("__OBJC"); 827 static ConstString g_section_name_eh_frame ("__eh_frame"); 828 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT)); 829 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA)); 830 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC)); 831 SectionSP eh_frame_section_sp; 832 if (text_section_sp.get()) 833 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame); 834 else 835 eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame); 836 837 uint8_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() : NListSectionNoSection; 838 839 uint32_t nlist_data_offset = 0; 840 841 uint32_t N_SO_index = UINT32_MAX; 842 843 MachSymtabSectionInfo section_info (section_list); 844 std::vector<uint32_t> N_FUN_indexes; 845 std::vector<uint32_t> N_NSYM_indexes; 846 std::vector<uint32_t> N_INCL_indexes; 847 std::vector<uint32_t> N_BRAC_indexes; 848 std::vector<uint32_t> N_COMM_indexes; 849 typedef std::map <uint64_t, uint32_t> ValueToSymbolIndexMap; 850 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap; 851 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 852 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 853 // Any symbols that get merged into another will get an entry 854 // in this map so we know 855 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 856 uint32_t nlist_idx = 0; 857 Symbol *symbol_ptr = NULL; 858 859 uint32_t sym_idx = 0; 860 Symbol *sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 861 uint32_t num_syms = symtab->GetNumSymbols(); 862 863 //symtab->Reserve (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 864 for (nlist_idx = 0; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) 865 { 866 struct nlist_64 nlist; 867 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 868 break; 869 870 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 871 nlist.n_type = nlist_data.GetU8_unchecked (&nlist_data_offset); 872 nlist.n_sect = nlist_data.GetU8_unchecked (&nlist_data_offset); 873 nlist.n_desc = nlist_data.GetU16_unchecked (&nlist_data_offset); 874 nlist.n_value = nlist_data.GetAddress_unchecked (&nlist_data_offset); 875 876 SymbolType type = eSymbolTypeInvalid; 877 const char *symbol_name = strtab_data.PeekCStr(nlist.n_strx); 878 if (symbol_name == NULL) 879 { 880 // No symbol should be NULL, even the symbols with no 881 // string values should have an offset zero which points 882 // to an empty C-string 883 Host::SystemLog (Host::eSystemLogError, 884 "error: symbol[%u] has invalid string table offset 0x%x in %s/%s, ignoring symbol\n", 885 nlist_idx, 886 nlist.n_strx, 887 m_module->GetFileSpec().GetDirectory().GetCString(), 888 m_module->GetFileSpec().GetFilename().GetCString()); 889 continue; 890 } 891 const char *symbol_name_non_abi_mangled = NULL; 892 893 if (symbol_name[0] == '\0') 894 symbol_name = NULL; 895 Section* symbol_section = NULL; 896 bool add_nlist = true; 897 bool is_debug = ((nlist.n_type & NlistMaskStab) != 0); 898 899 assert (sym_idx < num_syms); 900 901 sym[sym_idx].SetDebug (is_debug); 902 903 if (is_debug) 904 { 905 switch (nlist.n_type) 906 { 907 case StabGlobalSymbol: 908 // N_GSYM -- global symbol: name,,NO_SECT,type,0 909 // Sometimes the N_GSYM value contains the address. 910 911 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 912 // have the same address, but we want to ensure that we always find only the real symbol, 913 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 914 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 915 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 916 // same address. 917 918 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 919 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 920 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 921 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 922 add_nlist = false; 923 else 924 { 925 sym[sym_idx].SetExternal(true); 926 if (nlist.n_value != 0) 927 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 928 type = eSymbolTypeData; 929 } 930 break; 931 932 case StabFunctionName: 933 // N_FNAME -- procedure name (f77 kludge): name,,NO_SECT,0,0 934 type = eSymbolTypeCompiler; 935 break; 936 937 case StabFunction: 938 // N_FUN -- procedure: name,,n_sect,linenumber,address 939 if (symbol_name) 940 { 941 type = eSymbolTypeCode; 942 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 943 944 N_FUN_addr_to_sym_idx[nlist.n_value] = sym_idx; 945 // We use the current number of symbols in the symbol table in lieu of 946 // using nlist_idx in case we ever start trimming entries out 947 N_FUN_indexes.push_back(sym_idx); 948 } 949 else 950 { 951 type = eSymbolTypeCompiler; 952 953 if ( !N_FUN_indexes.empty() ) 954 { 955 // Copy the size of the function into the original STAB entry so we don't have 956 // to hunt for it later 957 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 958 N_FUN_indexes.pop_back(); 959 // We don't really need the end function STAB as it contains the size which 960 // we already placed with the original symbol, so don't add it if we want a 961 // minimal symbol table 962 if (minimize) 963 add_nlist = false; 964 } 965 } 966 break; 967 968 case StabStaticSymbol: 969 // N_STSYM -- static symbol: name,,n_sect,type,address 970 N_STSYM_addr_to_sym_idx[nlist.n_value] = sym_idx; 971 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 972 type = eSymbolTypeData; 973 break; 974 975 case StabLocalCommon: 976 // N_LCSYM -- .lcomm symbol: name,,n_sect,type,address 977 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 978 type = eSymbolTypeCommonBlock; 979 break; 980 981 case StabBeginSymbol: 982 // N_BNSYM 983 // We use the current number of symbols in the symbol table in lieu of 984 // using nlist_idx in case we ever start trimming entries out 985 if (minimize) 986 { 987 // Skip these if we want minimal symbol tables 988 add_nlist = false; 989 } 990 else 991 { 992 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 993 N_NSYM_indexes.push_back(sym_idx); 994 type = eSymbolTypeScopeBegin; 995 } 996 break; 997 998 case StabEndSymbol: 999 // N_ENSYM 1000 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 1001 // so that we can always skip the entire symbol if we need to navigate 1002 // more quickly at the source level when parsing STABS 1003 if (minimize) 1004 { 1005 // Skip these if we want minimal symbol tables 1006 add_nlist = false; 1007 } 1008 else 1009 { 1010 if ( !N_NSYM_indexes.empty() ) 1011 { 1012 symbol_ptr = symtab->SymbolAtIndex(N_NSYM_indexes.back()); 1013 symbol_ptr->SetByteSize(sym_idx + 1); 1014 symbol_ptr->SetSizeIsSibling(true); 1015 N_NSYM_indexes.pop_back(); 1016 } 1017 type = eSymbolTypeScopeEnd; 1018 } 1019 break; 1020 1021 1022 case StabSourceFileOptions: 1023 // N_OPT - emitted with gcc2_compiled and in gcc source 1024 type = eSymbolTypeCompiler; 1025 break; 1026 1027 case StabRegisterSymbol: 1028 // N_RSYM - register sym: name,,NO_SECT,type,register 1029 type = eSymbolTypeVariable; 1030 break; 1031 1032 case StabSourceLine: 1033 // N_SLINE - src line: 0,,n_sect,linenumber,address 1034 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1035 type = eSymbolTypeLineEntry; 1036 break; 1037 1038 case StabStructureType: 1039 // N_SSYM - structure elt: name,,NO_SECT,type,struct_offset 1040 type = eSymbolTypeVariableType; 1041 break; 1042 1043 case StabSourceFileName: 1044 // N_SO - source file name 1045 type = eSymbolTypeSourceFile; 1046 if (symbol_name == NULL) 1047 { 1048 if (minimize) 1049 add_nlist = false; 1050 if (N_SO_index != UINT32_MAX) 1051 { 1052 // Set the size of the N_SO to the terminating index of this N_SO 1053 // so that we can always skip the entire N_SO if we need to navigate 1054 // more quickly at the source level when parsing STABS 1055 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 1056 symbol_ptr->SetByteSize(sym_idx + (minimize ? 0 : 1)); 1057 symbol_ptr->SetSizeIsSibling(true); 1058 } 1059 N_NSYM_indexes.clear(); 1060 N_INCL_indexes.clear(); 1061 N_BRAC_indexes.clear(); 1062 N_COMM_indexes.clear(); 1063 N_FUN_indexes.clear(); 1064 N_SO_index = UINT32_MAX; 1065 } 1066 else 1067 { 1068 // We use the current number of symbols in the symbol table in lieu of 1069 // using nlist_idx in case we ever start trimming entries out 1070 if (symbol_name[0] == '/') 1071 N_SO_index = sym_idx; 1072 else if (minimize && (N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 1073 { 1074 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 1075 if (so_path && so_path[0]) 1076 { 1077 std::string full_so_path (so_path); 1078 if (*full_so_path.rbegin() != '/') 1079 full_so_path += '/'; 1080 full_so_path += symbol_name; 1081 sym[sym_idx - 1].GetMangled().SetValue(full_so_path.c_str(), false); 1082 add_nlist = false; 1083 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 1084 } 1085 } 1086 } 1087 1088 break; 1089 1090 case StabObjectFileName: 1091 // N_OSO - object file name: name,,0,0,st_mtime 1092 type = eSymbolTypeObjectFile; 1093 break; 1094 1095 case StabLocalSymbol: 1096 // N_LSYM - local sym: name,,NO_SECT,type,offset 1097 type = eSymbolTypeLocal; 1098 break; 1099 1100 //---------------------------------------------------------------------- 1101 // INCL scopes 1102 //---------------------------------------------------------------------- 1103 case StabBeginIncludeFileName: 1104 // N_BINCL - include file beginning: name,,NO_SECT,0,sum 1105 // We use the current number of symbols in the symbol table in lieu of 1106 // using nlist_idx in case we ever start trimming entries out 1107 N_INCL_indexes.push_back(sym_idx); 1108 type = eSymbolTypeScopeBegin; 1109 break; 1110 1111 case StabEndIncludeFile: 1112 // N_EINCL - include file end: name,,NO_SECT,0,0 1113 // Set the size of the N_BINCL to the terminating index of this N_EINCL 1114 // so that we can always skip the entire symbol if we need to navigate 1115 // more quickly at the source level when parsing STABS 1116 if ( !N_INCL_indexes.empty() ) 1117 { 1118 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 1119 symbol_ptr->SetByteSize(sym_idx + 1); 1120 symbol_ptr->SetSizeIsSibling(true); 1121 N_INCL_indexes.pop_back(); 1122 } 1123 type = eSymbolTypeScopeEnd; 1124 break; 1125 1126 case StabIncludeFileName: 1127 // N_SOL - #included file name: name,,n_sect,0,address 1128 type = eSymbolTypeHeaderFile; 1129 1130 // We currently don't use the header files on darwin 1131 if (minimize) 1132 add_nlist = false; 1133 break; 1134 1135 case StabCompilerParameters: 1136 // N_PARAMS - compiler parameters: name,,NO_SECT,0,0 1137 type = eSymbolTypeCompiler; 1138 break; 1139 1140 case StabCompilerVersion: 1141 // N_VERSION - compiler version: name,,NO_SECT,0,0 1142 type = eSymbolTypeCompiler; 1143 break; 1144 1145 case StabCompilerOptLevel: 1146 // N_OLEVEL - compiler -O level: name,,NO_SECT,0,0 1147 type = eSymbolTypeCompiler; 1148 break; 1149 1150 case StabParameter: 1151 // N_PSYM - parameter: name,,NO_SECT,type,offset 1152 type = eSymbolTypeVariable; 1153 break; 1154 1155 case StabAlternateEntry: 1156 // N_ENTRY - alternate entry: name,,n_sect,linenumber,address 1157 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1158 type = eSymbolTypeLineEntry; 1159 break; 1160 1161 //---------------------------------------------------------------------- 1162 // Left and Right Braces 1163 //---------------------------------------------------------------------- 1164 case StabLeftBracket: 1165 // N_LBRAC - left bracket: 0,,NO_SECT,nesting level,address 1166 // We use the current number of symbols in the symbol table in lieu of 1167 // using nlist_idx in case we ever start trimming entries out 1168 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1169 N_BRAC_indexes.push_back(sym_idx); 1170 type = eSymbolTypeScopeBegin; 1171 break; 1172 1173 case StabRightBracket: 1174 // N_RBRAC - right bracket: 0,,NO_SECT,nesting level,address 1175 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 1176 // so that we can always skip the entire symbol if we need to navigate 1177 // more quickly at the source level when parsing STABS 1178 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1179 if ( !N_BRAC_indexes.empty() ) 1180 { 1181 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 1182 symbol_ptr->SetByteSize(sym_idx + 1); 1183 symbol_ptr->SetSizeIsSibling(true); 1184 N_BRAC_indexes.pop_back(); 1185 } 1186 type = eSymbolTypeScopeEnd; 1187 break; 1188 1189 case StabDeletedIncludeFile: 1190 // N_EXCL - deleted include file: name,,NO_SECT,0,sum 1191 type = eSymbolTypeHeaderFile; 1192 break; 1193 1194 //---------------------------------------------------------------------- 1195 // COMM scopes 1196 //---------------------------------------------------------------------- 1197 case StabBeginCommon: 1198 // N_BCOMM - begin common: name,,NO_SECT,0,0 1199 // We use the current number of symbols in the symbol table in lieu of 1200 // using nlist_idx in case we ever start trimming entries out 1201 type = eSymbolTypeScopeBegin; 1202 N_COMM_indexes.push_back(sym_idx); 1203 break; 1204 1205 case StabEndCommonLocal: 1206 // N_ECOML - end common (local name): 0,,n_sect,0,address 1207 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1208 // Fall through 1209 1210 case StabEndCommon: 1211 // N_ECOMM - end common: name,,n_sect,0,0 1212 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 1213 // so that we can always skip the entire symbol if we need to navigate 1214 // more quickly at the source level when parsing STABS 1215 if ( !N_COMM_indexes.empty() ) 1216 { 1217 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 1218 symbol_ptr->SetByteSize(sym_idx + 1); 1219 symbol_ptr->SetSizeIsSibling(true); 1220 N_COMM_indexes.pop_back(); 1221 } 1222 type = eSymbolTypeScopeEnd; 1223 break; 1224 1225 case StabLength: 1226 // N_LENG - second stab entry with length information 1227 type = eSymbolTypeAdditional; 1228 break; 1229 1230 default: break; 1231 } 1232 } 1233 else 1234 { 1235 //uint8_t n_pext = NlistMaskPrivateExternal & nlist.n_type; 1236 uint8_t n_type = NlistMaskType & nlist.n_type; 1237 sym[sym_idx].SetExternal((NlistMaskExternal & nlist.n_type) != 0); 1238 1239 switch (n_type) 1240 { 1241 case NListTypeIndirect: // N_INDR - Fall through 1242 case NListTypePreboundUndefined:// N_PBUD - Fall through 1243 case NListTypeUndefined: // N_UNDF 1244 type = eSymbolTypeUndefined; 1245 break; 1246 1247 case NListTypeAbsolute: // N_ABS 1248 type = eSymbolTypeAbsolute; 1249 break; 1250 1251 case NListTypeSection: // N_SECT 1252 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1253 1254 if (symbol_section == NULL) 1255 { 1256 // TODO: warn about this? 1257 add_nlist = false; 1258 break; 1259 } 1260 1261 if (TEXT_eh_frame_sectID == nlist.n_sect) 1262 { 1263 type = eSymbolTypeException; 1264 } 1265 else 1266 { 1267 uint32_t section_type = symbol_section->Get() & SectionFlagMaskSectionType; 1268 1269 switch (section_type) 1270 { 1271 case SectionTypeRegular: break; // regular section 1272 //case SectionTypeZeroFill: type = eSymbolTypeData; break; // zero fill on demand section 1273 case SectionTypeCStringLiterals: type = eSymbolTypeData; break; // section with only literal C strings 1274 case SectionType4ByteLiterals: type = eSymbolTypeData; break; // section with only 4 byte literals 1275 case SectionType8ByteLiterals: type = eSymbolTypeData; break; // section with only 8 byte literals 1276 case SectionTypeLiteralPointers: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 1277 case SectionTypeNonLazySymbolPointers: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 1278 case SectionTypeLazySymbolPointers: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 1279 case SectionTypeSymbolStubs: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 1280 case SectionTypeModuleInitFunctionPointers: type = eSymbolTypeCode; break; // section with only function pointers for initialization 1281 case SectionTypeModuleTermFunctionPointers: type = eSymbolTypeCode; break; // section with only function pointers for termination 1282 //case SectionTypeCoalesced: type = eSymbolType; break; // section contains symbols that are to be coalesced 1283 //case SectionTypeZeroFillLarge: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 1284 case SectionTypeInterposing: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 1285 case SectionType16ByteLiterals: type = eSymbolTypeData; break; // section with only 16 byte literals 1286 case SectionTypeDTraceObjectFormat: type = eSymbolTypeInstrumentation; break; 1287 case SectionTypeLazyDylibSymbolPointers: type = eSymbolTypeTrampoline; break; 1288 default: break; 1289 } 1290 1291 if (type == eSymbolTypeInvalid) 1292 { 1293 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 1294 if (symbol_section->IsDescendant (text_section_sp.get())) 1295 { 1296 if (symbol_section->IsClear(SectionAttrUserPureInstructions | 1297 SectionAttrUserSelfModifyingCode | 1298 SectionAttrSytemSomeInstructions)) 1299 type = eSymbolTypeData; 1300 else 1301 type = eSymbolTypeCode; 1302 } 1303 else 1304 if (symbol_section->IsDescendant(data_section_sp.get())) 1305 { 1306 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 1307 { 1308 type = eSymbolTypeRuntime; 1309 1310 if (symbol_name && 1311 symbol_name[0] == '_' && 1312 symbol_name[1] == 'O' && 1313 symbol_name[2] == 'B') 1314 { 1315 llvm::StringRef symbol_name_ref(symbol_name); 1316 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 1317 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 1318 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 1319 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 1320 { 1321 symbol_name_non_abi_mangled = symbol_name + 1; 1322 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 1323 type = eSymbolTypeObjCClass; 1324 } 1325 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 1326 { 1327 symbol_name_non_abi_mangled = symbol_name + 1; 1328 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 1329 type = eSymbolTypeObjCMetaClass; 1330 } 1331 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 1332 { 1333 symbol_name_non_abi_mangled = symbol_name + 1; 1334 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 1335 type = eSymbolTypeObjCIVar; 1336 } 1337 } 1338 } 1339 else 1340 if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 1341 { 1342 type = eSymbolTypeException; 1343 } 1344 else 1345 { 1346 type = eSymbolTypeData; 1347 } 1348 } 1349 else 1350 if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 1351 { 1352 type = eSymbolTypeTrampoline; 1353 } 1354 else 1355 if (symbol_section->IsDescendant(objc_section_sp.get())) 1356 { 1357 type = eSymbolTypeRuntime; 1358 if (symbol_name && symbol_name[0] == '.') 1359 { 1360 llvm::StringRef symbol_name_ref(symbol_name); 1361 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 1362 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 1363 { 1364 symbol_name_non_abi_mangled = symbol_name; 1365 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 1366 type = eSymbolTypeObjCClass; 1367 } 1368 } 1369 } 1370 } 1371 } 1372 break; 1373 } 1374 } 1375 1376 if (add_nlist) 1377 { 1378 uint64_t symbol_value = nlist.n_value; 1379 bool symbol_name_is_mangled = false; 1380 1381 if (symbol_name_non_abi_mangled) 1382 { 1383 sym[sym_idx].GetMangled().SetMangledName (symbol_name_non_abi_mangled); 1384 sym[sym_idx].GetMangled().SetDemangledName (symbol_name); 1385 } 1386 else 1387 { 1388 if (symbol_name && symbol_name[0] == '_') 1389 { 1390 symbol_name_is_mangled = symbol_name[1] == '_'; 1391 symbol_name++; // Skip the leading underscore 1392 } 1393 1394 if (symbol_name) 1395 { 1396 sym[sym_idx].GetMangled().SetValue(symbol_name, symbol_name_is_mangled); 1397 } 1398 } 1399 1400 if (is_debug == false) 1401 { 1402 if (type == eSymbolTypeCode) 1403 { 1404 // See if we can find a N_FUN entry for any code symbols. 1405 // If we do find a match, and the name matches, then we 1406 // can merge the two into just the function symbol to avoid 1407 // duplicate entries in the symbol table 1408 ValueToSymbolIndexMap::const_iterator pos = N_FUN_addr_to_sym_idx.find (nlist.n_value); 1409 if (pos != N_FUN_addr_to_sym_idx.end()) 1410 { 1411 if ((symbol_name_is_mangled == true && sym[sym_idx].GetMangled().GetMangledName() == sym[pos->second].GetMangled().GetMangledName()) || 1412 (symbol_name_is_mangled == false && sym[sym_idx].GetMangled().GetDemangledName() == sym[pos->second].GetMangled().GetDemangledName())) 1413 { 1414 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 1415 // We just need the flags from the linker symbol, so put these flags 1416 // into the N_FUN flags to avoid duplicate symbols in the symbol table 1417 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 1418 sym[sym_idx].Clear(); 1419 continue; 1420 } 1421 } 1422 } 1423 else if (type == eSymbolTypeData) 1424 { 1425 // See if we can find a N_STSYM entry for any data symbols. 1426 // If we do find a match, and the name matches, then we 1427 // can merge the two into just the Static symbol to avoid 1428 // duplicate entries in the symbol table 1429 ValueToSymbolIndexMap::const_iterator pos = N_STSYM_addr_to_sym_idx.find (nlist.n_value); 1430 if (pos != N_STSYM_addr_to_sym_idx.end()) 1431 { 1432 if ((symbol_name_is_mangled == true && sym[sym_idx].GetMangled().GetMangledName() == sym[pos->second].GetMangled().GetMangledName()) || 1433 (symbol_name_is_mangled == false && sym[sym_idx].GetMangled().GetDemangledName() == sym[pos->second].GetMangled().GetDemangledName())) 1434 { 1435 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 1436 // We just need the flags from the linker symbol, so put these flags 1437 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 1438 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 1439 sym[sym_idx].Clear(); 1440 continue; 1441 } 1442 } 1443 } 1444 } 1445 if (symbol_section != NULL) 1446 symbol_value -= symbol_section->GetFileAddress(); 1447 1448 sym[sym_idx].SetID (nlist_idx); 1449 sym[sym_idx].SetType (type); 1450 sym[sym_idx].GetAddressRangeRef().GetBaseAddress().SetSection (symbol_section); 1451 sym[sym_idx].GetAddressRangeRef().GetBaseAddress().SetOffset (symbol_value); 1452 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 1453 1454 ++sym_idx; 1455 } 1456 else 1457 { 1458 sym[sym_idx].Clear(); 1459 } 1460 1461 } 1462 1463 // STAB N_GSYM entries end up having a symbol type eSymbolTypeGlobal and when the symbol value 1464 // is zero, the address of the global ends up being in a non-STAB entry. Try and fix up all 1465 // such entries by figuring out what the address for the global is by looking up this non-STAB 1466 // entry and copying the value into the debug symbol's value to save us the hassle in the 1467 // debug symbol parser. 1468 1469 Symbol *global_symbol = NULL; 1470 for (nlist_idx = 0; 1471 nlist_idx < symtab_load_command.nsyms && (global_symbol = symtab->FindSymbolWithType (eSymbolTypeData, Symtab::eDebugYes, Symtab::eVisibilityAny, nlist_idx)) != NULL; 1472 nlist_idx++) 1473 { 1474 if (global_symbol->GetValue().GetFileAddress() == 0) 1475 { 1476 std::vector<uint32_t> indexes; 1477 if (symtab->AppendSymbolIndexesWithName (global_symbol->GetMangled().GetName(), indexes) > 0) 1478 { 1479 std::vector<uint32_t>::const_iterator pos; 1480 std::vector<uint32_t>::const_iterator end = indexes.end(); 1481 for (pos = indexes.begin(); pos != end; ++pos) 1482 { 1483 symbol_ptr = symtab->SymbolAtIndex(*pos); 1484 if (symbol_ptr != global_symbol && symbol_ptr->IsDebug() == false) 1485 { 1486 global_symbol->SetValue(symbol_ptr->GetValue()); 1487 break; 1488 } 1489 } 1490 } 1491 } 1492 } 1493 1494 // Trim our symbols down to just what we ended up with after 1495 // removing any symbols. 1496 if (sym_idx < num_syms) 1497 { 1498 num_syms = sym_idx; 1499 sym = symtab->Resize (num_syms); 1500 } 1501 1502 // Now synthesize indirect symbols 1503 if (m_dysymtab.nindirectsyms != 0) 1504 { 1505 DataExtractor indirect_symbol_index_data (m_data, m_dysymtab.indirectsymoff, m_dysymtab.nindirectsyms * 4); 1506 1507 if (indirect_symbol_index_data.GetByteSize()) 1508 { 1509 NListIndexToSymbolIndexMap::const_iterator end_index_pos = m_nlist_idx_to_sym_idx.end(); 1510 1511 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); ++sect_idx) 1512 { 1513 if ((m_mach_sections[sect_idx].flags & SectionFlagMaskSectionType) == SectionTypeSymbolStubs) 1514 { 1515 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 1516 if (symbol_stub_byte_size == 0) 1517 continue; 1518 1519 const uint32_t num_symbol_stubs = m_mach_sections[sect_idx].size / symbol_stub_byte_size; 1520 1521 if (num_symbol_stubs == 0) 1522 continue; 1523 1524 const uint32_t symbol_stub_index_offset = m_mach_sections[sect_idx].reserved1; 1525 uint32_t synthetic_stub_sym_id = symtab_load_command.nsyms; 1526 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) 1527 { 1528 const uint32_t symbol_stub_index = symbol_stub_index_offset + stub_idx; 1529 const lldb::addr_t symbol_stub_addr = m_mach_sections[sect_idx].addr + (stub_idx * symbol_stub_byte_size); 1530 uint32_t symbol_stub_offset = symbol_stub_index * 4; 1531 if (indirect_symbol_index_data.ValidOffsetForDataOfSize(symbol_stub_offset, 4)) 1532 { 1533 const uint32_t stub_sym_id = indirect_symbol_index_data.GetU32 (&symbol_stub_offset); 1534 if (stub_sym_id & (IndirectSymbolAbsolute | IndirectSymbolLocal)) 1535 continue; 1536 1537 NListIndexToSymbolIndexMap::const_iterator index_pos = m_nlist_idx_to_sym_idx.find (stub_sym_id); 1538 Symbol *stub_symbol = NULL; 1539 if (index_pos != end_index_pos) 1540 { 1541 // We have a remapping from the original nlist index to 1542 // a current symbol index, so just look this up by index 1543 stub_symbol = symtab->SymbolAtIndex (index_pos->second); 1544 } 1545 else 1546 { 1547 // We need to lookup a symbol using the original nlist 1548 // symbol index since this index is coming from the 1549 // S_SYMBOL_STUBS 1550 stub_symbol = symtab->FindSymbolByID (stub_sym_id); 1551 } 1552 1553 assert (stub_symbol); 1554 if (stub_symbol) 1555 { 1556 Address so_addr(symbol_stub_addr, section_list); 1557 1558 if (stub_symbol->GetType() == eSymbolTypeUndefined) 1559 { 1560 // Change the external symbol into a trampoline that makes sense 1561 // These symbols were N_UNDF N_EXT, and are useless to us, so we 1562 // can re-use them so we don't have to make up a synthetic symbol 1563 // for no good reason. 1564 stub_symbol->SetType (eSymbolTypeTrampoline); 1565 stub_symbol->SetExternal (false); 1566 stub_symbol->GetAddressRangeRef().GetBaseAddress() = so_addr; 1567 stub_symbol->GetAddressRangeRef().SetByteSize (symbol_stub_byte_size); 1568 } 1569 else 1570 { 1571 // Make a synthetic symbol to describe the trampoline stub 1572 if (sym_idx >= num_syms) 1573 sym = symtab->Resize (++num_syms); 1574 sym[sym_idx].SetID (synthetic_stub_sym_id++); 1575 sym[sym_idx].GetMangled() = stub_symbol->GetMangled(); 1576 sym[sym_idx].SetType (eSymbolTypeTrampoline); 1577 sym[sym_idx].SetIsSynthetic (true); 1578 sym[sym_idx].GetAddressRangeRef().GetBaseAddress() = so_addr; 1579 sym[sym_idx].GetAddressRangeRef().SetByteSize (symbol_stub_byte_size); 1580 ++sym_idx; 1581 } 1582 } 1583 } 1584 } 1585 } 1586 } 1587 } 1588 } 1589 return symtab->GetNumSymbols(); 1590 } 1591 } 1592 offset = cmd_offset + symtab_load_command.cmdsize; 1593 } 1594 return 0; 1595 } 1596 1597 1598 void 1599 ObjectFileMachO::Dump (Stream *s) 1600 { 1601 lldb_private::Mutex::Locker locker(m_mutex); 1602 s->Printf("%p: ", this); 1603 s->Indent(); 1604 if (m_header.magic == HeaderMagic64 || m_header.magic == HeaderMagic64Swapped) 1605 s->PutCString("ObjectFileMachO64"); 1606 else 1607 s->PutCString("ObjectFileMachO32"); 1608 1609 ArchSpec header_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 1610 1611 *s << ", file = '" << m_file << "', arch = " << header_arch.GetArchitectureName() << "\n"; 1612 1613 if (m_sections_ap.get()) 1614 m_sections_ap->Dump(s, NULL, true, UINT32_MAX); 1615 1616 if (m_symtab_ap.get()) 1617 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 1618 } 1619 1620 1621 bool 1622 ObjectFileMachO::GetUUID (lldb_private::UUID* uuid) 1623 { 1624 lldb_private::Mutex::Locker locker(m_mutex); 1625 struct uuid_command load_cmd; 1626 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 1627 uint32_t i; 1628 for (i=0; i<m_header.ncmds; ++i) 1629 { 1630 const uint32_t cmd_offset = offset; 1631 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1632 break; 1633 1634 if (load_cmd.cmd == LoadCommandUUID) 1635 { 1636 const uint8_t *uuid_bytes = m_data.PeekData(offset, 16); 1637 if (uuid_bytes) 1638 { 1639 uuid->SetBytes (uuid_bytes); 1640 return true; 1641 } 1642 return false; 1643 } 1644 offset = cmd_offset + load_cmd.cmdsize; 1645 } 1646 return false; 1647 } 1648 1649 1650 uint32_t 1651 ObjectFileMachO::GetDependentModules (FileSpecList& files) 1652 { 1653 lldb_private::Mutex::Locker locker(m_mutex); 1654 struct load_command load_cmd; 1655 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 1656 uint32_t count = 0; 1657 const bool resolve_path = false; // Don't resolve the dependend file paths since they may not reside on this system 1658 uint32_t i; 1659 for (i=0; i<m_header.ncmds; ++i) 1660 { 1661 const uint32_t cmd_offset = offset; 1662 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1663 break; 1664 1665 switch (load_cmd.cmd) 1666 { 1667 case LoadCommandDylibLoad: 1668 case LoadCommandDylibLoadWeak: 1669 case LoadCommandDylibReexport: 1670 case LoadCommandDynamicLinkerLoad: 1671 case LoadCommandFixedVMShlibLoad: 1672 case LoadCommandDylibLoadUpward: 1673 { 1674 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 1675 const char *path = m_data.PeekCStr(name_offset); 1676 // Skip any path that starts with '@' since these are usually: 1677 // @executable_path/.../file 1678 // @rpath/.../file 1679 if (path && path[0] != '@') 1680 { 1681 FileSpec file_spec(path, resolve_path); 1682 if (files.AppendIfUnique(file_spec)) 1683 count++; 1684 } 1685 } 1686 break; 1687 1688 default: 1689 break; 1690 } 1691 offset = cmd_offset + load_cmd.cmdsize; 1692 } 1693 return count; 1694 } 1695 1696 lldb_private::Address 1697 ObjectFileMachO::GetEntryPointAddress () 1698 { 1699 // If the object file is not an executable it can't hold the entry point. m_entry_point_address 1700 // is initialized to an invalid address, so we can just return that. 1701 // If m_entry_point_address is valid it means we've found it already, so return the cached value. 1702 1703 if (!IsExecutable() || m_entry_point_address.IsValid()) 1704 return m_entry_point_address; 1705 1706 // Otherwise, look for the UnixThread or Thread command. The data for the Thread command is given in 1707 // /usr/include/mach-o.h, but it is basically: 1708 // 1709 // uint32_t flavor - this is the flavor argument you would pass to thread_get_state 1710 // uint32_t count - this is the count of longs in the thread state data 1711 // struct XXX_thread_state state - this is the structure from <machine/thread_status.h> corresponding to the flavor. 1712 // <repeat this trio> 1713 // 1714 // So we just keep reading the various register flavors till we find the GPR one, then read the PC out of there. 1715 // FIXME: We will need to have a "RegisterContext data provider" class at some point that can get all the registers 1716 // out of data in this form & attach them to a given thread. That should underlie the MacOS X User process plugin, 1717 // and we'll also need it for the MacOS X Core File process plugin. When we have that we can also use it here. 1718 // 1719 // For now we hard-code the offsets and flavors we need: 1720 // 1721 // 1722 1723 lldb_private::Mutex::Locker locker(m_mutex); 1724 struct load_command load_cmd; 1725 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 1726 uint32_t i; 1727 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 1728 bool done = false; 1729 1730 for (i=0; i<m_header.ncmds; ++i) 1731 { 1732 const uint32_t cmd_offset = offset; 1733 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1734 break; 1735 1736 switch (load_cmd.cmd) 1737 { 1738 case LoadCommandUnixThread: 1739 case LoadCommandThread: 1740 { 1741 while (offset < cmd_offset + load_cmd.cmdsize) 1742 { 1743 uint32_t flavor = m_data.GetU32(&offset); 1744 uint32_t count = m_data.GetU32(&offset); 1745 if (count == 0) 1746 { 1747 // We've gotten off somehow, log and exit; 1748 return m_entry_point_address; 1749 } 1750 1751 switch (m_header.cputype) 1752 { 1753 case llvm::MachO::CPUTypeARM: 1754 if (flavor == 1) // ARM_THREAD_STATE from mach/arm/thread_status.h 1755 { 1756 offset += 60; // This is the offset of pc in the GPR thread state data structure. 1757 start_address = m_data.GetU32(&offset); 1758 done = true; 1759 } 1760 break; 1761 case llvm::MachO::CPUTypeI386: 1762 if (flavor == 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 1763 { 1764 offset += 40; // This is the offset of eip in the GPR thread state data structure. 1765 start_address = m_data.GetU32(&offset); 1766 done = true; 1767 } 1768 break; 1769 case llvm::MachO::CPUTypeX86_64: 1770 if (flavor == 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 1771 { 1772 offset += 16 * 8; // This is the offset of rip in the GPR thread state data structure. 1773 start_address = m_data.GetU64(&offset); 1774 done = true; 1775 } 1776 break; 1777 default: 1778 return m_entry_point_address; 1779 } 1780 // Haven't found the GPR flavor yet, skip over the data for this flavor: 1781 if (done) 1782 break; 1783 offset += count * 4; 1784 } 1785 } 1786 break; 1787 1788 default: 1789 break; 1790 } 1791 if (done) 1792 break; 1793 1794 // Go to the next load command: 1795 offset = cmd_offset + load_cmd.cmdsize; 1796 } 1797 1798 if (start_address != LLDB_INVALID_ADDRESS) 1799 { 1800 // We got the start address from the load commands, so now resolve that address in the sections 1801 // of this ObjectFile: 1802 if (!m_entry_point_address.ResolveAddressUsingFileSections (start_address, GetSectionList())) 1803 { 1804 m_entry_point_address.Clear(); 1805 } 1806 } 1807 else 1808 { 1809 // We couldn't read the UnixThread load command - maybe it wasn't there. As a fallback look for the 1810 // "start" symbol in the main executable. 1811 1812 SymbolContextList contexts; 1813 SymbolContext context; 1814 if (!m_module->FindSymbolsWithNameAndType(ConstString ("start"), eSymbolTypeCode, contexts)) 1815 return m_entry_point_address; 1816 1817 contexts.GetContextAtIndex(0, context); 1818 1819 m_entry_point_address = context.symbol->GetValue(); 1820 } 1821 1822 return m_entry_point_address; 1823 1824 } 1825 1826 ObjectFile::Type 1827 ObjectFileMachO::CalculateType() 1828 { 1829 switch (m_header.filetype) 1830 { 1831 case HeaderFileTypeObject: // 0x1u MH_OBJECT 1832 if (GetAddressByteSize () == 4) 1833 { 1834 // 32 bit kexts are just object files, but they do have a valid 1835 // UUID load command. 1836 UUID uuid; 1837 if (GetUUID(&uuid)) 1838 { 1839 // this checking for the UUID load command is not enough 1840 // we could eventually look for the symbol named 1841 // "OSKextGetCurrentIdentifier" as this is required of kexts 1842 if (m_strata == eStrataInvalid) 1843 m_strata = eStrataKernel; 1844 return eTypeSharedLibrary; 1845 } 1846 } 1847 return eTypeObjectFile; 1848 1849 case HeaderFileTypeExecutable: return eTypeExecutable; // 0x2u MH_EXECUTE 1850 case HeaderFileTypeFixedVMShlib: return eTypeSharedLibrary; // 0x3u MH_FVMLIB 1851 case HeaderFileTypeCore: return eTypeCoreFile; // 0x4u MH_CORE 1852 case HeaderFileTypePreloadedExecutable: return eTypeSharedLibrary; // 0x5u MH_PRELOAD 1853 case HeaderFileTypeDynamicShlib: return eTypeSharedLibrary; // 0x6u MH_DYLIB 1854 case HeaderFileTypeDynamicLinkEditor: return eTypeDynamicLinker; // 0x7u MH_DYLINKER 1855 case HeaderFileTypeBundle: return eTypeSharedLibrary; // 0x8u MH_BUNDLE 1856 case HeaderFileTypeDynamicShlibStub: return eTypeStubLibrary; // 0x9u MH_DYLIB_STUB 1857 case HeaderFileTypeDSYM: return eTypeDebugInfo; // 0xAu MH_DSYM 1858 case HeaderFileTypeKextBundle: return eTypeSharedLibrary; // 0xBu MH_KEXT_BUNDLE 1859 default: 1860 break; 1861 } 1862 return eTypeUnknown; 1863 } 1864 1865 ObjectFile::Strata 1866 ObjectFileMachO::CalculateStrata() 1867 { 1868 switch (m_header.filetype) 1869 { 1870 case HeaderFileTypeObject: // 0x1u MH_OBJECT 1871 { 1872 // 32 bit kexts are just object files, but they do have a valid 1873 // UUID load command. 1874 UUID uuid; 1875 if (GetUUID(&uuid)) 1876 { 1877 // this checking for the UUID load command is not enough 1878 // we could eventually look for the symbol named 1879 // "OSKextGetCurrentIdentifier" as this is required of kexts 1880 if (m_type == eTypeInvalid) 1881 m_type = eTypeSharedLibrary; 1882 1883 return eStrataKernel; 1884 } 1885 } 1886 return eStrataUnknown; 1887 1888 case HeaderFileTypeExecutable: // 0x2u MH_EXECUTE 1889 // Check for the MH_DYLDLINK bit in the flags 1890 if (m_header.flags & HeaderFlagBitIsDynamicLinkObject) 1891 return eStrataUser; 1892 return eStrataKernel; 1893 1894 case HeaderFileTypeFixedVMShlib: return eStrataUser; // 0x3u MH_FVMLIB 1895 case HeaderFileTypeCore: return eStrataUnknown; // 0x4u MH_CORE 1896 case HeaderFileTypePreloadedExecutable: return eStrataUser; // 0x5u MH_PRELOAD 1897 case HeaderFileTypeDynamicShlib: return eStrataUser; // 0x6u MH_DYLIB 1898 case HeaderFileTypeDynamicLinkEditor: return eStrataUser; // 0x7u MH_DYLINKER 1899 case HeaderFileTypeBundle: return eStrataUser; // 0x8u MH_BUNDLE 1900 case HeaderFileTypeDynamicShlibStub: return eStrataUser; // 0x9u MH_DYLIB_STUB 1901 case HeaderFileTypeDSYM: return eStrataUnknown; // 0xAu MH_DSYM 1902 case HeaderFileTypeKextBundle: return eStrataKernel; // 0xBu MH_KEXT_BUNDLE 1903 default: 1904 break; 1905 } 1906 return eStrataUnknown; 1907 } 1908 1909 1910 bool 1911 ObjectFileMachO::GetArchitecture (ArchSpec &arch) 1912 { 1913 lldb_private::Mutex::Locker locker(m_mutex); 1914 arch.SetArchitecture (eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 1915 1916 // Files with type MH_PRELOAD are currently used in cases where the image 1917 // debugs at the addresses in the file itself. Below we set the OS to 1918 // unknown to make sure we use the DynamicLoaderStatic()... 1919 if (m_header.filetype == HeaderFileTypePreloadedExecutable) 1920 { 1921 arch.GetTriple().setOS (llvm::Triple::UnknownOS); 1922 } 1923 1924 return true; 1925 } 1926 1927 1928 //------------------------------------------------------------------ 1929 // PluginInterface protocol 1930 //------------------------------------------------------------------ 1931 const char * 1932 ObjectFileMachO::GetPluginName() 1933 { 1934 return "ObjectFileMachO"; 1935 } 1936 1937 const char * 1938 ObjectFileMachO::GetShortPluginName() 1939 { 1940 return GetPluginNameStatic(); 1941 } 1942 1943 uint32_t 1944 ObjectFileMachO::GetPluginVersion() 1945 { 1946 return 1; 1947 } 1948 1949