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& dataSP, const FileSpec* file, addr_t offset, addr_t length) 67 { 68 if (ObjectFileMachO::MagicBytesMatch(dataSP)) 69 { 70 std::auto_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module, dataSP, 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& dataSP) 101 { 102 DataExtractor data(dataSP, lldb::endian::InlHostByteOrder(), 4); 103 uint32_t offset = 0; 104 uint32_t magic = data.GetU32(&offset); 105 return MachHeaderSizeFromMagic(magic) != 0; 106 } 107 108 109 ObjectFileMachO::ObjectFileMachO(Module* module, DataBufferSP& dataSP, const FileSpec* file, addr_t offset, addr_t length) : 110 ObjectFile(module, file, offset, length, dataSP), 111 m_mutex (Mutex::eMutexTypeRecursive), 112 m_header(), 113 m_sections_ap(), 114 m_symtab_ap(), 115 m_entry_point_address () 116 { 117 ::memset (&m_header, 0, sizeof(m_header)); 118 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 119 } 120 121 122 ObjectFileMachO::~ObjectFileMachO() 123 { 124 } 125 126 127 bool 128 ObjectFileMachO::ParseHeader () 129 { 130 lldb_private::Mutex::Locker locker(m_mutex); 131 bool can_parse = false; 132 uint32_t offset = 0; 133 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 134 // Leave magic in the original byte order 135 m_header.magic = m_data.GetU32(&offset); 136 switch (m_header.magic) 137 { 138 case HeaderMagic32: 139 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 140 m_data.SetAddressByteSize(4); 141 can_parse = true; 142 break; 143 144 case HeaderMagic64: 145 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 146 m_data.SetAddressByteSize(8); 147 can_parse = true; 148 break; 149 150 case HeaderMagic32Swapped: 151 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 152 m_data.SetAddressByteSize(4); 153 can_parse = true; 154 break; 155 156 case HeaderMagic64Swapped: 157 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 158 m_data.SetAddressByteSize(8); 159 can_parse = true; 160 break; 161 162 default: 163 break; 164 } 165 166 if (can_parse) 167 { 168 m_data.GetU32(&offset, &m_header.cputype, 6); 169 170 ArchSpec mach_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 171 172 if (SetModulesArchitecture (mach_arch)) 173 { 174 // Read in all only the load command data 175 DataBufferSP data_sp(m_file.ReadFileContents(m_offset, m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic))); 176 m_data.SetData (data_sp); 177 return true; 178 } 179 } 180 else 181 { 182 memset(&m_header, 0, sizeof(struct mach_header)); 183 } 184 return false; 185 } 186 187 188 ByteOrder 189 ObjectFileMachO::GetByteOrder () const 190 { 191 lldb_private::Mutex::Locker locker(m_mutex); 192 return m_data.GetByteOrder (); 193 } 194 195 bool 196 ObjectFileMachO::IsExecutable() const 197 { 198 return m_header.filetype == HeaderFileTypeExecutable; 199 } 200 201 size_t 202 ObjectFileMachO::GetAddressByteSize () const 203 { 204 lldb_private::Mutex::Locker locker(m_mutex); 205 return m_data.GetAddressByteSize (); 206 } 207 208 AddressClass 209 ObjectFileMachO::GetAddressClass (lldb::addr_t file_addr) 210 { 211 Symtab *symtab = GetSymtab(); 212 if (symtab) 213 { 214 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 215 if (symbol) 216 { 217 const AddressRange *range_ptr = symbol->GetAddressRangePtr(); 218 if (range_ptr) 219 { 220 const Section *section = range_ptr->GetBaseAddress().GetSection(); 221 if (section) 222 { 223 const SectionType section_type = section->GetType(); 224 switch (section_type) 225 { 226 case eSectionTypeInvalid: return eAddressClassUnknown; 227 case eSectionTypeCode: 228 if (m_header.cputype == llvm::MachO::CPUTypeARM) 229 { 230 // For ARM we have a bit in the n_desc field of the symbol 231 // that tells us ARM/Thumb which is bit 0x0008. 232 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 233 return eAddressClassCodeAlternateISA; 234 } 235 return eAddressClassCode; 236 237 case eSectionTypeContainer: return eAddressClassUnknown; 238 case eSectionTypeData: 239 case eSectionTypeDataCString: 240 case eSectionTypeDataCStringPointers: 241 case eSectionTypeDataSymbolAddress: 242 case eSectionTypeData4: 243 case eSectionTypeData8: 244 case eSectionTypeData16: 245 case eSectionTypeDataPointers: 246 case eSectionTypeZeroFill: 247 case eSectionTypeDataObjCMessageRefs: 248 case eSectionTypeDataObjCCFStrings: 249 return eAddressClassData; 250 case eSectionTypeDebug: 251 case eSectionTypeDWARFDebugAbbrev: 252 case eSectionTypeDWARFDebugAranges: 253 case eSectionTypeDWARFDebugFrame: 254 case eSectionTypeDWARFDebugInfo: 255 case eSectionTypeDWARFDebugLine: 256 case eSectionTypeDWARFDebugLoc: 257 case eSectionTypeDWARFDebugMacInfo: 258 case eSectionTypeDWARFDebugPubNames: 259 case eSectionTypeDWARFDebugPubTypes: 260 case eSectionTypeDWARFDebugRanges: 261 case eSectionTypeDWARFDebugStr: 262 case eSectionTypeDWARFAppleNames: 263 case eSectionTypeDWARFAppleTypes: 264 case eSectionTypeDWARFAppleNamespaces: 265 case eSectionTypeDWARFAppleObjC: 266 return eAddressClassDebug; 267 case eSectionTypeEHFrame: return eAddressClassRuntime; 268 case eSectionTypeOther: return eAddressClassUnknown; 269 } 270 } 271 } 272 273 const SymbolType symbol_type = symbol->GetType(); 274 switch (symbol_type) 275 { 276 case eSymbolTypeAny: return eAddressClassUnknown; 277 case eSymbolTypeAbsolute: return eAddressClassUnknown; 278 279 case eSymbolTypeCode: 280 case eSymbolTypeTrampoline: 281 if (m_header.cputype == llvm::MachO::CPUTypeARM) 282 { 283 // For ARM we have a bit in the n_desc field of the symbol 284 // that tells us ARM/Thumb which is bit 0x0008. 285 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 286 return eAddressClassCodeAlternateISA; 287 } 288 return eAddressClassCode; 289 290 case eSymbolTypeData: return eAddressClassData; 291 case eSymbolTypeRuntime: return eAddressClassRuntime; 292 case eSymbolTypeException: return eAddressClassRuntime; 293 case eSymbolTypeSourceFile: return eAddressClassDebug; 294 case eSymbolTypeHeaderFile: return eAddressClassDebug; 295 case eSymbolTypeObjectFile: return eAddressClassDebug; 296 case eSymbolTypeCommonBlock: return eAddressClassDebug; 297 case eSymbolTypeBlock: return eAddressClassDebug; 298 case eSymbolTypeLocal: return eAddressClassData; 299 case eSymbolTypeParam: return eAddressClassData; 300 case eSymbolTypeVariable: return eAddressClassData; 301 case eSymbolTypeVariableType: return eAddressClassDebug; 302 case eSymbolTypeLineEntry: return eAddressClassDebug; 303 case eSymbolTypeLineHeader: return eAddressClassDebug; 304 case eSymbolTypeScopeBegin: return eAddressClassDebug; 305 case eSymbolTypeScopeEnd: return eAddressClassDebug; 306 case eSymbolTypeAdditional: return eAddressClassUnknown; 307 case eSymbolTypeCompiler: return eAddressClassDebug; 308 case eSymbolTypeInstrumentation:return eAddressClassDebug; 309 case eSymbolTypeUndefined: return eAddressClassUnknown; 310 case eSymbolTypeObjCClass: return eAddressClassRuntime; 311 case eSymbolTypeObjCMetaClass: return eAddressClassRuntime; 312 case eSymbolTypeObjCIVar: return eAddressClassRuntime; 313 } 314 } 315 } 316 return eAddressClassUnknown; 317 } 318 319 Symtab * 320 ObjectFileMachO::GetSymtab() 321 { 322 lldb_private::Mutex::Locker symfile_locker(m_mutex); 323 if (m_symtab_ap.get() == NULL) 324 { 325 m_symtab_ap.reset(new Symtab(this)); 326 Mutex::Locker symtab_locker (m_symtab_ap->GetMutex()); 327 ParseSymtab (true); 328 m_symtab_ap->Finalize (); 329 } 330 return m_symtab_ap.get(); 331 } 332 333 334 SectionList * 335 ObjectFileMachO::GetSectionList() 336 { 337 lldb_private::Mutex::Locker locker(m_mutex); 338 if (m_sections_ap.get() == NULL) 339 { 340 m_sections_ap.reset(new SectionList()); 341 ParseSections(); 342 } 343 return m_sections_ap.get(); 344 } 345 346 347 size_t 348 ObjectFileMachO::ParseSections () 349 { 350 lldb::user_id_t segID = 0; 351 lldb::user_id_t sectID = 0; 352 struct segment_command_64 load_cmd; 353 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 354 uint32_t i; 355 //bool dump_sections = false; 356 for (i=0; i<m_header.ncmds; ++i) 357 { 358 const uint32_t load_cmd_offset = offset; 359 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 360 break; 361 362 if (load_cmd.cmd == LoadCommandSegment32 || load_cmd.cmd == LoadCommandSegment64) 363 { 364 if (m_data.GetU8(&offset, (uint8_t*)load_cmd.segname, 16)) 365 { 366 load_cmd.vmaddr = m_data.GetAddress(&offset); 367 load_cmd.vmsize = m_data.GetAddress(&offset); 368 load_cmd.fileoff = m_data.GetAddress(&offset); 369 load_cmd.filesize = m_data.GetAddress(&offset); 370 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 371 { 372 373 const bool segment_is_encrypted = (load_cmd.flags & SegmentCommandFlagBitProtectedVersion1) != 0; 374 375 // Keep a list of mach segments around in case we need to 376 // get at data that isn't stored in the abstracted Sections. 377 m_mach_segments.push_back (load_cmd); 378 379 ConstString segment_name (load_cmd.segname, std::min<int>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 380 // Use a segment ID of the segment index shifted left by 8 so they 381 // never conflict with any of the sections. 382 SectionSP segment_sp; 383 if (segment_name) 384 { 385 segment_sp.reset(new Section (NULL, 386 GetModule(), // Module to which this section belongs 387 ++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 388 segment_name, // Name of this section 389 eSectionTypeContainer, // This section is a container of other sections. 390 load_cmd.vmaddr, // File VM address == addresses as they are found in the object file 391 load_cmd.vmsize, // VM size in bytes of this section 392 load_cmd.fileoff, // Offset to the data for this section in the file 393 load_cmd.filesize, // Size in bytes of this section as found in the the file 394 load_cmd.flags)); // Flags for this section 395 396 segment_sp->SetIsEncrypted (segment_is_encrypted); 397 m_sections_ap->AddSection(segment_sp); 398 } 399 400 struct section_64 sect64; 401 ::memset (§64, 0, sizeof(sect64)); 402 // Push a section into our mach sections for the section at 403 // index zero (NListSectionNoSection) if we don't have any 404 // mach sections yet... 405 if (m_mach_sections.empty()) 406 m_mach_sections.push_back(sect64); 407 uint32_t segment_sect_idx; 408 const lldb::user_id_t first_segment_sectID = sectID + 1; 409 410 411 const uint32_t num_u32s = load_cmd.cmd == LoadCommandSegment32 ? 7 : 8; 412 for (segment_sect_idx=0; segment_sect_idx<load_cmd.nsects; ++segment_sect_idx) 413 { 414 if (m_data.GetU8(&offset, (uint8_t*)sect64.sectname, sizeof(sect64.sectname)) == NULL) 415 break; 416 if (m_data.GetU8(&offset, (uint8_t*)sect64.segname, sizeof(sect64.segname)) == NULL) 417 break; 418 sect64.addr = m_data.GetAddress(&offset); 419 sect64.size = m_data.GetAddress(&offset); 420 421 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 422 break; 423 424 // Keep a list of mach sections around in case we need to 425 // get at data that isn't stored in the abstracted Sections. 426 m_mach_sections.push_back (sect64); 427 428 ConstString section_name (sect64.sectname, std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 429 if (!segment_name) 430 { 431 // We have a segment with no name so we need to conjure up 432 // segments that correspond to the section's segname if there 433 // isn't already such a section. If there is such a section, 434 // we resize the section so that it spans all sections. 435 // We also mark these sections as fake so address matches don't 436 // hit if they land in the gaps between the child sections. 437 segment_name.SetTrimmedCStringWithLength(sect64.segname, sizeof(sect64.segname)); 438 segment_sp = m_sections_ap->FindSectionByName (segment_name); 439 if (segment_sp.get()) 440 { 441 Section *segment = segment_sp.get(); 442 // Grow the section size as needed. 443 const lldb::addr_t sect64_min_addr = sect64.addr; 444 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 445 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 446 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 447 const lldb::addr_t curr_seg_max_addr = curr_seg_min_addr + curr_seg_byte_size; 448 if (sect64_min_addr >= curr_seg_min_addr) 449 { 450 const lldb::addr_t new_seg_byte_size = sect64_max_addr - curr_seg_min_addr; 451 // Only grow the section size if needed 452 if (new_seg_byte_size > curr_seg_byte_size) 453 segment->SetByteSize (new_seg_byte_size); 454 } 455 else 456 { 457 // We need to change the base address of the segment and 458 // adjust the child section offsets for all existing children. 459 const lldb::addr_t slide_amount = sect64_min_addr - curr_seg_min_addr; 460 segment->Slide(slide_amount, false); 461 segment->GetChildren().Slide (-slide_amount, false); 462 segment->SetByteSize (curr_seg_max_addr - sect64_min_addr); 463 } 464 465 // Grow the section size as needed. 466 if (sect64.offset) 467 { 468 const lldb::addr_t segment_min_file_offset = segment->GetFileOffset(); 469 const lldb::addr_t segment_max_file_offset = segment_min_file_offset + segment->GetFileSize(); 470 471 const lldb::addr_t section_min_file_offset = sect64.offset; 472 const lldb::addr_t section_max_file_offset = section_min_file_offset + sect64.size; 473 const lldb::addr_t new_file_offset = std::min (section_min_file_offset, segment_min_file_offset); 474 const lldb::addr_t new_file_size = std::max (section_max_file_offset, segment_max_file_offset) - new_file_offset; 475 segment->SetFileOffset (new_file_offset); 476 segment->SetFileSize (new_file_size); 477 } 478 } 479 else 480 { 481 // Create a fake section for the section's named segment 482 segment_sp.reset(new Section(segment_sp.get(), // Parent section 483 GetModule(), // Module to which this section belongs 484 ++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 485 segment_name, // Name of this section 486 eSectionTypeContainer, // This section is a container of other sections. 487 sect64.addr, // File VM address == addresses as they are found in the object file 488 sect64.size, // VM size in bytes of this section 489 sect64.offset, // Offset to the data for this section in the file 490 sect64.offset ? sect64.size : 0, // Size in bytes of this section as found in the the file 491 load_cmd.flags)); // Flags for this section 492 segment_sp->SetIsFake(true); 493 m_sections_ap->AddSection(segment_sp); 494 segment_sp->SetIsEncrypted (segment_is_encrypted); 495 } 496 } 497 assert (segment_sp.get()); 498 499 uint32_t mach_sect_type = sect64.flags & SectionFlagMaskSectionType; 500 static ConstString g_sect_name_objc_data ("__objc_data"); 501 static ConstString g_sect_name_objc_msgrefs ("__objc_msgrefs"); 502 static ConstString g_sect_name_objc_selrefs ("__objc_selrefs"); 503 static ConstString g_sect_name_objc_classrefs ("__objc_classrefs"); 504 static ConstString g_sect_name_objc_superrefs ("__objc_superrefs"); 505 static ConstString g_sect_name_objc_const ("__objc_const"); 506 static ConstString g_sect_name_objc_classlist ("__objc_classlist"); 507 static ConstString g_sect_name_cfstring ("__cfstring"); 508 509 static ConstString g_sect_name_dwarf_debug_abbrev ("__debug_abbrev"); 510 static ConstString g_sect_name_dwarf_debug_aranges ("__debug_aranges"); 511 static ConstString g_sect_name_dwarf_debug_frame ("__debug_frame"); 512 static ConstString g_sect_name_dwarf_debug_info ("__debug_info"); 513 static ConstString g_sect_name_dwarf_debug_line ("__debug_line"); 514 static ConstString g_sect_name_dwarf_debug_loc ("__debug_loc"); 515 static ConstString g_sect_name_dwarf_debug_macinfo ("__debug_macinfo"); 516 static ConstString g_sect_name_dwarf_debug_pubnames ("__debug_pubnames"); 517 static ConstString g_sect_name_dwarf_debug_pubtypes ("__debug_pubtypes"); 518 static ConstString g_sect_name_dwarf_debug_ranges ("__debug_ranges"); 519 static ConstString g_sect_name_dwarf_debug_str ("__debug_str"); 520 static ConstString g_sect_name_dwarf_apple_names ("__apple_names"); 521 static ConstString g_sect_name_dwarf_apple_types ("__apple_types"); 522 static ConstString g_sect_name_dwarf_apple_namespaces ("__apple_namespac"); 523 static ConstString g_sect_name_dwarf_apple_objc ("__apple_objc"); 524 static ConstString g_sect_name_eh_frame ("__eh_frame"); 525 static ConstString g_sect_name_DATA ("__DATA"); 526 static ConstString g_sect_name_TEXT ("__TEXT"); 527 528 SectionType sect_type = eSectionTypeOther; 529 530 if (section_name == g_sect_name_dwarf_debug_abbrev) 531 sect_type = eSectionTypeDWARFDebugAbbrev; 532 else if (section_name == g_sect_name_dwarf_debug_aranges) 533 sect_type = eSectionTypeDWARFDebugAranges; 534 else if (section_name == g_sect_name_dwarf_debug_frame) 535 sect_type = eSectionTypeDWARFDebugFrame; 536 else if (section_name == g_sect_name_dwarf_debug_info) 537 sect_type = eSectionTypeDWARFDebugInfo; 538 else if (section_name == g_sect_name_dwarf_debug_line) 539 sect_type = eSectionTypeDWARFDebugLine; 540 else if (section_name == g_sect_name_dwarf_debug_loc) 541 sect_type = eSectionTypeDWARFDebugLoc; 542 else if (section_name == g_sect_name_dwarf_debug_macinfo) 543 sect_type = eSectionTypeDWARFDebugMacInfo; 544 else if (section_name == g_sect_name_dwarf_debug_pubnames) 545 sect_type = eSectionTypeDWARFDebugPubNames; 546 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 547 sect_type = eSectionTypeDWARFDebugPubTypes; 548 else if (section_name == g_sect_name_dwarf_debug_ranges) 549 sect_type = eSectionTypeDWARFDebugRanges; 550 else if (section_name == g_sect_name_dwarf_debug_str) 551 sect_type = eSectionTypeDWARFDebugStr; 552 else if (section_name == g_sect_name_dwarf_apple_names) 553 sect_type = eSectionTypeDWARFAppleNames; 554 else if (section_name == g_sect_name_dwarf_apple_types) 555 sect_type = eSectionTypeDWARFAppleTypes; 556 else if (section_name == g_sect_name_dwarf_apple_namespaces) 557 sect_type = eSectionTypeDWARFAppleNamespaces; 558 else if (section_name == g_sect_name_dwarf_apple_objc) 559 sect_type = eSectionTypeDWARFAppleObjC; 560 else if (section_name == g_sect_name_objc_selrefs) 561 sect_type = eSectionTypeDataCStringPointers; 562 else if (section_name == g_sect_name_objc_msgrefs) 563 sect_type = eSectionTypeDataObjCMessageRefs; 564 else if (section_name == g_sect_name_eh_frame) 565 sect_type = eSectionTypeEHFrame; 566 else if (section_name == g_sect_name_cfstring) 567 sect_type = eSectionTypeDataObjCCFStrings; 568 else if (section_name == g_sect_name_objc_data || 569 section_name == g_sect_name_objc_classrefs || 570 section_name == g_sect_name_objc_superrefs || 571 section_name == g_sect_name_objc_const || 572 section_name == g_sect_name_objc_classlist) 573 { 574 sect_type = eSectionTypeDataPointers; 575 } 576 577 if (sect_type == eSectionTypeOther) 578 { 579 switch (mach_sect_type) 580 { 581 // TODO: categorize sections by other flags for regular sections 582 case SectionTypeRegular: 583 if (segment_sp->GetName() == g_sect_name_TEXT) 584 sect_type = eSectionTypeCode; 585 else if (segment_sp->GetName() == g_sect_name_DATA) 586 sect_type = eSectionTypeData; 587 else 588 sect_type = eSectionTypeOther; 589 break; 590 case SectionTypeZeroFill: sect_type = eSectionTypeZeroFill; break; 591 case SectionTypeCStringLiterals: sect_type = eSectionTypeDataCString; break; // section with only literal C strings 592 case SectionType4ByteLiterals: sect_type = eSectionTypeData4; break; // section with only 4 byte literals 593 case SectionType8ByteLiterals: sect_type = eSectionTypeData8; break; // section with only 8 byte literals 594 case SectionTypeLiteralPointers: sect_type = eSectionTypeDataPointers; break; // section with only pointers to literals 595 case SectionTypeNonLazySymbolPointers: sect_type = eSectionTypeDataPointers; break; // section with only non-lazy symbol pointers 596 case SectionTypeLazySymbolPointers: sect_type = eSectionTypeDataPointers; break; // section with only lazy symbol pointers 597 case SectionTypeSymbolStubs: sect_type = eSectionTypeCode; break; // section with only symbol stubs, byte size of stub in the reserved2 field 598 case SectionTypeModuleInitFunctionPointers: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for initialization 599 case SectionTypeModuleTermFunctionPointers: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for termination 600 case SectionTypeCoalesced: sect_type = eSectionTypeOther; break; 601 case SectionTypeZeroFillLarge: sect_type = eSectionTypeZeroFill; break; 602 case SectionTypeInterposing: sect_type = eSectionTypeCode; break; // section with only pairs of function pointers for interposing 603 case SectionType16ByteLiterals: sect_type = eSectionTypeData16; break; // section with only 16 byte literals 604 case SectionTypeDTraceObjectFormat: sect_type = eSectionTypeDebug; break; 605 case SectionTypeLazyDylibSymbolPointers: sect_type = eSectionTypeDataPointers; break; 606 default: break; 607 } 608 } 609 610 SectionSP section_sp(new Section(segment_sp.get(), 611 GetModule(), 612 ++sectID, 613 section_name, 614 sect_type, 615 sect64.addr - segment_sp->GetFileAddress(), 616 sect64.size, 617 sect64.offset, 618 sect64.offset == 0 ? 0 : sect64.size, 619 sect64.flags)); 620 // Set the section to be encrypted to match the segment 621 section_sp->SetIsEncrypted (segment_is_encrypted); 622 623 segment_sp->GetChildren().AddSection(section_sp); 624 625 if (segment_sp->IsFake()) 626 { 627 segment_sp.reset(); 628 segment_name.Clear(); 629 } 630 } 631 if (segment_sp && m_header.filetype == HeaderFileTypeDSYM) 632 { 633 if (first_segment_sectID <= sectID) 634 { 635 lldb::user_id_t sect_uid; 636 for (sect_uid = first_segment_sectID; sect_uid <= sectID; ++sect_uid) 637 { 638 SectionSP curr_section_sp(segment_sp->GetChildren().FindSectionByID (sect_uid)); 639 SectionSP next_section_sp; 640 if (sect_uid + 1 <= sectID) 641 next_section_sp = segment_sp->GetChildren().FindSectionByID (sect_uid+1); 642 643 if (curr_section_sp.get()) 644 { 645 if (curr_section_sp->GetByteSize() == 0) 646 { 647 if (next_section_sp.get() != NULL) 648 curr_section_sp->SetByteSize ( next_section_sp->GetFileAddress() - curr_section_sp->GetFileAddress() ); 649 else 650 curr_section_sp->SetByteSize ( load_cmd.vmsize ); 651 } 652 } 653 } 654 } 655 } 656 } 657 } 658 } 659 else if (load_cmd.cmd == LoadCommandDynamicSymtabInfo) 660 { 661 m_dysymtab.cmd = load_cmd.cmd; 662 m_dysymtab.cmdsize = load_cmd.cmdsize; 663 m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 664 } 665 666 offset = load_cmd_offset + load_cmd.cmdsize; 667 } 668 // if (dump_sections) 669 // { 670 // StreamFile s(stdout); 671 // m_sections_ap->Dump(&s, true); 672 // } 673 return sectID; // Return the number of sections we registered with the module 674 } 675 676 class MachSymtabSectionInfo 677 { 678 public: 679 680 MachSymtabSectionInfo (SectionList *section_list) : 681 m_section_list (section_list), 682 m_section_infos() 683 { 684 // Get the number of sections down to a depth of 1 to include 685 // all segments and their sections, but no other sections that 686 // may be added for debug map or 687 m_section_infos.resize(section_list->GetNumSections(1)); 688 } 689 690 691 Section * 692 GetSection (uint8_t n_sect, addr_t file_addr) 693 { 694 if (n_sect == 0) 695 return NULL; 696 if (n_sect < m_section_infos.size()) 697 { 698 if (m_section_infos[n_sect].section == NULL) 699 { 700 Section *section = m_section_list->FindSectionByID (n_sect).get(); 701 m_section_infos[n_sect].section = section; 702 if (section != NULL) 703 { 704 m_section_infos[n_sect].vm_range.SetBaseAddress (section->GetFileAddress()); 705 m_section_infos[n_sect].vm_range.SetByteSize (section->GetByteSize()); 706 } 707 else 708 { 709 Host::SystemLog (Host::eSystemLogError, "error: unable to find section for section %u\n", n_sect); 710 } 711 } 712 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) 713 { 714 // Symbol is in section. 715 return m_section_infos[n_sect].section; 716 } 717 else if (m_section_infos[n_sect].vm_range.GetByteSize () == 0 && 718 m_section_infos[n_sect].vm_range.GetBaseAddress() == file_addr) 719 { 720 // Symbol is in section with zero size, but has the same start 721 // address as the section. This can happen with linker symbols 722 // (symbols that start with the letter 'l' or 'L'. 723 return m_section_infos[n_sect].section; 724 } 725 } 726 return m_section_list->FindSectionContainingFileAddress(file_addr).get(); 727 } 728 729 protected: 730 struct SectionInfo 731 { 732 SectionInfo () : 733 vm_range(), 734 section (NULL) 735 { 736 } 737 738 VMRange vm_range; 739 Section *section; 740 }; 741 SectionList *m_section_list; 742 std::vector<SectionInfo> m_section_infos; 743 }; 744 745 746 747 size_t 748 ObjectFileMachO::ParseSymtab (bool minimize) 749 { 750 Timer scoped_timer(__PRETTY_FUNCTION__, 751 "ObjectFileMachO::ParseSymtab () module = %s", 752 m_file.GetFilename().AsCString("")); 753 struct symtab_command symtab_load_command; 754 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 755 uint32_t i; 756 757 LogSP log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_SYMBOLS)); 758 759 for (i=0; i<m_header.ncmds; ++i) 760 { 761 const uint32_t cmd_offset = offset; 762 // Read in the load command and load command size 763 if (m_data.GetU32(&offset, &symtab_load_command, 2) == NULL) 764 break; 765 // Watch for the symbol table load command 766 if (symtab_load_command.cmd == LoadCommandSymtab) 767 { 768 // Read in the rest of the symtab load command 769 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4)) // fill in symoff, nsyms, stroff, strsize fields 770 { 771 if (symtab_load_command.symoff == 0) 772 { 773 if (log) 774 GetModule()->LogMessage(log.get(), "LC_SYMTAB.symoff == 0"); 775 return 0; 776 } 777 778 if (symtab_load_command.stroff == 0) 779 { 780 if (log) 781 GetModule()->LogMessage(log.get(), "LC_SYMTAB.stroff == 0"); 782 return 0; 783 } 784 785 if (symtab_load_command.nsyms == 0) 786 { 787 if (log) 788 GetModule()->LogMessage(log.get(), "LC_SYMTAB.nsyms == 0"); 789 return 0; 790 } 791 792 if (symtab_load_command.strsize == 0) 793 { 794 if (log) 795 GetModule()->LogMessage(log.get(), "LC_SYMTAB.strsize == 0"); 796 return 0; 797 } 798 799 Symtab *symtab = m_symtab_ap.get(); 800 SectionList *section_list = GetSectionList(); 801 if (section_list == NULL) 802 return 0; 803 804 const size_t addr_byte_size = m_data.GetAddressByteSize(); 805 const ByteOrder byte_order = m_data.GetByteOrder(); 806 bool bit_width_32 = addr_byte_size == 4; 807 const size_t nlist_byte_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 808 809 DataBufferSP symtab_data_sp(m_file.ReadFileContents (m_offset + symtab_load_command.symoff, 810 symtab_load_command.nsyms * nlist_byte_size)); 811 812 if (symtab_data_sp.get() == NULL || 813 symtab_data_sp->GetBytes() == NULL || 814 symtab_data_sp->GetByteSize() == 0) 815 { 816 if (log) 817 GetModule()->LogMessage(log.get(), "failed to read nlist data"); 818 return 0; 819 } 820 821 DataBufferSP strtab_data_sp(m_file.ReadFileContents (m_offset + symtab_load_command.stroff, 822 symtab_load_command.strsize)); 823 824 if (strtab_data_sp.get() == NULL || 825 strtab_data_sp->GetBytes() == NULL || 826 strtab_data_sp->GetByteSize() == 0) 827 { 828 if (log) 829 GetModule()->LogMessage(log.get(), "failed to read strtab data"); 830 return 0; 831 } 832 833 const char *strtab_data = (const char *)strtab_data_sp->GetBytes(); 834 const size_t strtab_data_len = strtab_data_sp->GetByteSize(); 835 836 static ConstString g_segment_name_TEXT ("__TEXT"); 837 static ConstString g_segment_name_DATA ("__DATA"); 838 static ConstString g_segment_name_OBJC ("__OBJC"); 839 static ConstString g_section_name_eh_frame ("__eh_frame"); 840 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT)); 841 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA)); 842 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC)); 843 SectionSP eh_frame_section_sp; 844 if (text_section_sp.get()) 845 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame); 846 else 847 eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame); 848 849 uint8_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() : NListSectionNoSection; 850 //uint32_t symtab_offset = 0; 851 assert (symtab_data_sp->GetByteSize()/nlist_byte_size >= symtab_load_command.nsyms); 852 853 uint32_t nlist_data_offset = 0; 854 DataExtractor nlist_data (symtab_data_sp, byte_order, addr_byte_size); 855 856 uint32_t N_SO_index = UINT32_MAX; 857 858 MachSymtabSectionInfo section_info (section_list); 859 std::vector<uint32_t> N_FUN_indexes; 860 std::vector<uint32_t> N_NSYM_indexes; 861 std::vector<uint32_t> N_INCL_indexes; 862 std::vector<uint32_t> N_BRAC_indexes; 863 std::vector<uint32_t> N_COMM_indexes; 864 typedef std::map <uint64_t, uint32_t> ValueToSymbolIndexMap; 865 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap; 866 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 867 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 868 // Any symbols that get merged into another will get an entry 869 // in this map so we know 870 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 871 uint32_t nlist_idx = 0; 872 Symbol *symbol_ptr = NULL; 873 874 uint32_t sym_idx = 0; 875 Symbol *sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 876 uint32_t num_syms = symtab->GetNumSymbols(); 877 878 //symtab->Reserve (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 879 for (nlist_idx = 0; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) 880 { 881 struct nlist_64 nlist; 882 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 883 break; 884 885 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 886 nlist.n_type = nlist_data.GetU8_unchecked (&nlist_data_offset); 887 nlist.n_sect = nlist_data.GetU8_unchecked (&nlist_data_offset); 888 nlist.n_desc = nlist_data.GetU16_unchecked (&nlist_data_offset); 889 nlist.n_value = nlist_data.GetAddress_unchecked (&nlist_data_offset); 890 891 SymbolType type = eSymbolTypeInvalid; 892 if (nlist.n_strx >= strtab_data_len) 893 { 894 // No symbol should be NULL, even the symbols with no 895 // string values should have an offset zero which points 896 // to an empty C-string 897 Host::SystemLog (Host::eSystemLogError, 898 "error: symbol[%u] has invalid string table offset 0x%x in %s/%s, ignoring symbol\n", 899 nlist_idx, 900 nlist.n_strx, 901 m_module->GetFileSpec().GetDirectory().GetCString(), 902 m_module->GetFileSpec().GetFilename().GetCString()); 903 continue; 904 } 905 const char *symbol_name = &strtab_data[nlist.n_strx]; 906 const char *symbol_name_non_abi_mangled = NULL; 907 908 if (symbol_name[0] == '\0') 909 symbol_name = NULL; 910 Section* symbol_section = NULL; 911 bool add_nlist = true; 912 bool is_debug = ((nlist.n_type & NlistMaskStab) != 0); 913 914 assert (sym_idx < num_syms); 915 916 sym[sym_idx].SetDebug (is_debug); 917 918 if (is_debug) 919 { 920 switch (nlist.n_type) 921 { 922 case StabGlobalSymbol: 923 // N_GSYM -- global symbol: name,,NO_SECT,type,0 924 // Sometimes the N_GSYM value contains the address. 925 926 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 927 // have the same address, but we want to ensure that we always find only the real symbol, 928 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 929 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 930 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 931 // same address. 932 933 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 934 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 935 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 936 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 937 add_nlist = false; 938 else 939 { 940 sym[sym_idx].SetExternal(true); 941 if (nlist.n_value != 0) 942 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 943 type = eSymbolTypeData; 944 } 945 break; 946 947 case StabFunctionName: 948 // N_FNAME -- procedure name (f77 kludge): name,,NO_SECT,0,0 949 type = eSymbolTypeCompiler; 950 break; 951 952 case StabFunction: 953 // N_FUN -- procedure: name,,n_sect,linenumber,address 954 if (symbol_name) 955 { 956 type = eSymbolTypeCode; 957 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 958 959 N_FUN_addr_to_sym_idx[nlist.n_value] = sym_idx; 960 // We use the current number of symbols in the symbol table in lieu of 961 // using nlist_idx in case we ever start trimming entries out 962 N_FUN_indexes.push_back(sym_idx); 963 } 964 else 965 { 966 type = eSymbolTypeCompiler; 967 968 if ( !N_FUN_indexes.empty() ) 969 { 970 // Copy the size of the function into the original STAB entry so we don't have 971 // to hunt for it later 972 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 973 N_FUN_indexes.pop_back(); 974 // We don't really need the end function STAB as it contains the size which 975 // we already placed with the original symbol, so don't add it if we want a 976 // minimal symbol table 977 if (minimize) 978 add_nlist = false; 979 } 980 } 981 break; 982 983 case StabStaticSymbol: 984 // N_STSYM -- static symbol: name,,n_sect,type,address 985 N_STSYM_addr_to_sym_idx[nlist.n_value] = sym_idx; 986 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 987 type = eSymbolTypeData; 988 break; 989 990 case StabLocalCommon: 991 // N_LCSYM -- .lcomm symbol: name,,n_sect,type,address 992 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 993 type = eSymbolTypeCommonBlock; 994 break; 995 996 case StabBeginSymbol: 997 // N_BNSYM 998 // We use the current number of symbols in the symbol table in lieu of 999 // using nlist_idx in case we ever start trimming entries out 1000 if (minimize) 1001 { 1002 // Skip these if we want minimal symbol tables 1003 add_nlist = false; 1004 } 1005 else 1006 { 1007 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1008 N_NSYM_indexes.push_back(sym_idx); 1009 type = eSymbolTypeScopeBegin; 1010 } 1011 break; 1012 1013 case StabEndSymbol: 1014 // N_ENSYM 1015 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 1016 // so that we can always skip the entire symbol if we need to navigate 1017 // more quickly at the source level when parsing STABS 1018 if (minimize) 1019 { 1020 // Skip these if we want minimal symbol tables 1021 add_nlist = false; 1022 } 1023 else 1024 { 1025 if ( !N_NSYM_indexes.empty() ) 1026 { 1027 symbol_ptr = symtab->SymbolAtIndex(N_NSYM_indexes.back()); 1028 symbol_ptr->SetByteSize(sym_idx + 1); 1029 symbol_ptr->SetSizeIsSibling(true); 1030 N_NSYM_indexes.pop_back(); 1031 } 1032 type = eSymbolTypeScopeEnd; 1033 } 1034 break; 1035 1036 1037 case StabSourceFileOptions: 1038 // N_OPT - emitted with gcc2_compiled and in gcc source 1039 type = eSymbolTypeCompiler; 1040 break; 1041 1042 case StabRegisterSymbol: 1043 // N_RSYM - register sym: name,,NO_SECT,type,register 1044 type = eSymbolTypeVariable; 1045 break; 1046 1047 case StabSourceLine: 1048 // N_SLINE - src line: 0,,n_sect,linenumber,address 1049 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1050 type = eSymbolTypeLineEntry; 1051 break; 1052 1053 case StabStructureType: 1054 // N_SSYM - structure elt: name,,NO_SECT,type,struct_offset 1055 type = eSymbolTypeVariableType; 1056 break; 1057 1058 case StabSourceFileName: 1059 // N_SO - source file name 1060 type = eSymbolTypeSourceFile; 1061 if (symbol_name == NULL) 1062 { 1063 if (minimize) 1064 add_nlist = false; 1065 if (N_SO_index != UINT32_MAX) 1066 { 1067 // Set the size of the N_SO to the terminating index of this N_SO 1068 // so that we can always skip the entire N_SO if we need to navigate 1069 // more quickly at the source level when parsing STABS 1070 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 1071 symbol_ptr->SetByteSize(sym_idx + (minimize ? 0 : 1)); 1072 symbol_ptr->SetSizeIsSibling(true); 1073 } 1074 N_NSYM_indexes.clear(); 1075 N_INCL_indexes.clear(); 1076 N_BRAC_indexes.clear(); 1077 N_COMM_indexes.clear(); 1078 N_FUN_indexes.clear(); 1079 N_SO_index = UINT32_MAX; 1080 } 1081 else 1082 { 1083 // We use the current number of symbols in the symbol table in lieu of 1084 // using nlist_idx in case we ever start trimming entries out 1085 if (symbol_name[0] == '/') 1086 N_SO_index = sym_idx; 1087 else if (minimize && (N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 1088 { 1089 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 1090 if (so_path && so_path[0]) 1091 { 1092 std::string full_so_path (so_path); 1093 if (*full_so_path.rbegin() != '/') 1094 full_so_path += '/'; 1095 full_so_path += symbol_name; 1096 sym[sym_idx - 1].GetMangled().SetValue(full_so_path.c_str(), false); 1097 add_nlist = false; 1098 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 1099 } 1100 } 1101 } 1102 1103 break; 1104 1105 case StabObjectFileName: 1106 // N_OSO - object file name: name,,0,0,st_mtime 1107 type = eSymbolTypeObjectFile; 1108 break; 1109 1110 case StabLocalSymbol: 1111 // N_LSYM - local sym: name,,NO_SECT,type,offset 1112 type = eSymbolTypeLocal; 1113 break; 1114 1115 //---------------------------------------------------------------------- 1116 // INCL scopes 1117 //---------------------------------------------------------------------- 1118 case StabBeginIncludeFileName: 1119 // N_BINCL - include file beginning: name,,NO_SECT,0,sum 1120 // We use the current number of symbols in the symbol table in lieu of 1121 // using nlist_idx in case we ever start trimming entries out 1122 N_INCL_indexes.push_back(sym_idx); 1123 type = eSymbolTypeScopeBegin; 1124 break; 1125 1126 case StabEndIncludeFile: 1127 // N_EINCL - include file end: name,,NO_SECT,0,0 1128 // Set the size of the N_BINCL to the terminating index of this N_EINCL 1129 // so that we can always skip the entire symbol if we need to navigate 1130 // more quickly at the source level when parsing STABS 1131 if ( !N_INCL_indexes.empty() ) 1132 { 1133 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 1134 symbol_ptr->SetByteSize(sym_idx + 1); 1135 symbol_ptr->SetSizeIsSibling(true); 1136 N_INCL_indexes.pop_back(); 1137 } 1138 type = eSymbolTypeScopeEnd; 1139 break; 1140 1141 case StabIncludeFileName: 1142 // N_SOL - #included file name: name,,n_sect,0,address 1143 type = eSymbolTypeHeaderFile; 1144 1145 // We currently don't use the header files on darwin 1146 if (minimize) 1147 add_nlist = false; 1148 break; 1149 1150 case StabCompilerParameters: 1151 // N_PARAMS - compiler parameters: name,,NO_SECT,0,0 1152 type = eSymbolTypeCompiler; 1153 break; 1154 1155 case StabCompilerVersion: 1156 // N_VERSION - compiler version: name,,NO_SECT,0,0 1157 type = eSymbolTypeCompiler; 1158 break; 1159 1160 case StabCompilerOptLevel: 1161 // N_OLEVEL - compiler -O level: name,,NO_SECT,0,0 1162 type = eSymbolTypeCompiler; 1163 break; 1164 1165 case StabParameter: 1166 // N_PSYM - parameter: name,,NO_SECT,type,offset 1167 type = eSymbolTypeVariable; 1168 break; 1169 1170 case StabAlternateEntry: 1171 // N_ENTRY - alternate entry: name,,n_sect,linenumber,address 1172 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1173 type = eSymbolTypeLineEntry; 1174 break; 1175 1176 //---------------------------------------------------------------------- 1177 // Left and Right Braces 1178 //---------------------------------------------------------------------- 1179 case StabLeftBracket: 1180 // N_LBRAC - left bracket: 0,,NO_SECT,nesting level,address 1181 // We use the current number of symbols in the symbol table in lieu of 1182 // using nlist_idx in case we ever start trimming entries out 1183 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1184 N_BRAC_indexes.push_back(sym_idx); 1185 type = eSymbolTypeScopeBegin; 1186 break; 1187 1188 case StabRightBracket: 1189 // N_RBRAC - right bracket: 0,,NO_SECT,nesting level,address 1190 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 1191 // so that we can always skip the entire symbol if we need to navigate 1192 // more quickly at the source level when parsing STABS 1193 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1194 if ( !N_BRAC_indexes.empty() ) 1195 { 1196 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 1197 symbol_ptr->SetByteSize(sym_idx + 1); 1198 symbol_ptr->SetSizeIsSibling(true); 1199 N_BRAC_indexes.pop_back(); 1200 } 1201 type = eSymbolTypeScopeEnd; 1202 break; 1203 1204 case StabDeletedIncludeFile: 1205 // N_EXCL - deleted include file: name,,NO_SECT,0,sum 1206 type = eSymbolTypeHeaderFile; 1207 break; 1208 1209 //---------------------------------------------------------------------- 1210 // COMM scopes 1211 //---------------------------------------------------------------------- 1212 case StabBeginCommon: 1213 // N_BCOMM - begin common: name,,NO_SECT,0,0 1214 // We use the current number of symbols in the symbol table in lieu of 1215 // using nlist_idx in case we ever start trimming entries out 1216 type = eSymbolTypeScopeBegin; 1217 N_COMM_indexes.push_back(sym_idx); 1218 break; 1219 1220 case StabEndCommonLocal: 1221 // N_ECOML - end common (local name): 0,,n_sect,0,address 1222 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1223 // Fall through 1224 1225 case StabEndCommon: 1226 // N_ECOMM - end common: name,,n_sect,0,0 1227 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 1228 // so that we can always skip the entire symbol if we need to navigate 1229 // more quickly at the source level when parsing STABS 1230 if ( !N_COMM_indexes.empty() ) 1231 { 1232 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 1233 symbol_ptr->SetByteSize(sym_idx + 1); 1234 symbol_ptr->SetSizeIsSibling(true); 1235 N_COMM_indexes.pop_back(); 1236 } 1237 type = eSymbolTypeScopeEnd; 1238 break; 1239 1240 case StabLength: 1241 // N_LENG - second stab entry with length information 1242 type = eSymbolTypeAdditional; 1243 break; 1244 1245 default: break; 1246 } 1247 } 1248 else 1249 { 1250 //uint8_t n_pext = NlistMaskPrivateExternal & nlist.n_type; 1251 uint8_t n_type = NlistMaskType & nlist.n_type; 1252 sym[sym_idx].SetExternal((NlistMaskExternal & nlist.n_type) != 0); 1253 1254 switch (n_type) 1255 { 1256 case NListTypeIndirect: // N_INDR - Fall through 1257 case NListTypePreboundUndefined:// N_PBUD - Fall through 1258 case NListTypeUndefined: // N_UNDF 1259 type = eSymbolTypeUndefined; 1260 break; 1261 1262 case NListTypeAbsolute: // N_ABS 1263 type = eSymbolTypeAbsolute; 1264 break; 1265 1266 case NListTypeSection: // N_SECT 1267 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1268 1269 if (symbol_section == NULL) 1270 { 1271 // TODO: warn about this? 1272 add_nlist = false; 1273 break; 1274 } 1275 1276 if (TEXT_eh_frame_sectID == nlist.n_sect) 1277 { 1278 type = eSymbolTypeException; 1279 } 1280 else 1281 { 1282 uint32_t section_type = symbol_section->Get() & SectionFlagMaskSectionType; 1283 1284 switch (section_type) 1285 { 1286 case SectionTypeRegular: break; // regular section 1287 //case SectionTypeZeroFill: type = eSymbolTypeData; break; // zero fill on demand section 1288 case SectionTypeCStringLiterals: type = eSymbolTypeData; break; // section with only literal C strings 1289 case SectionType4ByteLiterals: type = eSymbolTypeData; break; // section with only 4 byte literals 1290 case SectionType8ByteLiterals: type = eSymbolTypeData; break; // section with only 8 byte literals 1291 case SectionTypeLiteralPointers: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 1292 case SectionTypeNonLazySymbolPointers: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 1293 case SectionTypeLazySymbolPointers: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 1294 case SectionTypeSymbolStubs: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 1295 case SectionTypeModuleInitFunctionPointers: type = eSymbolTypeCode; break; // section with only function pointers for initialization 1296 case SectionTypeModuleTermFunctionPointers: type = eSymbolTypeCode; break; // section with only function pointers for termination 1297 //case SectionTypeCoalesced: type = eSymbolType; break; // section contains symbols that are to be coalesced 1298 //case SectionTypeZeroFillLarge: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 1299 case SectionTypeInterposing: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 1300 case SectionType16ByteLiterals: type = eSymbolTypeData; break; // section with only 16 byte literals 1301 case SectionTypeDTraceObjectFormat: type = eSymbolTypeInstrumentation; break; 1302 case SectionTypeLazyDylibSymbolPointers: type = eSymbolTypeTrampoline; break; 1303 default: break; 1304 } 1305 1306 if (type == eSymbolTypeInvalid) 1307 { 1308 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 1309 if (symbol_section->IsDescendant (text_section_sp.get())) 1310 { 1311 if (symbol_section->IsClear(SectionAttrUserPureInstructions | 1312 SectionAttrUserSelfModifyingCode | 1313 SectionAttrSytemSomeInstructions)) 1314 type = eSymbolTypeData; 1315 else 1316 type = eSymbolTypeCode; 1317 } 1318 else 1319 if (symbol_section->IsDescendant(data_section_sp.get())) 1320 { 1321 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 1322 { 1323 type = eSymbolTypeRuntime; 1324 1325 if (symbol_name && 1326 symbol_name[0] == '_' && 1327 symbol_name[1] == 'O' && 1328 symbol_name[2] == 'B') 1329 { 1330 llvm::StringRef symbol_name_ref(symbol_name); 1331 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 1332 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 1333 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 1334 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 1335 { 1336 symbol_name_non_abi_mangled = symbol_name + 1; 1337 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 1338 type = eSymbolTypeObjCClass; 1339 } 1340 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 1341 { 1342 symbol_name_non_abi_mangled = symbol_name + 1; 1343 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 1344 type = eSymbolTypeObjCMetaClass; 1345 } 1346 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 1347 { 1348 symbol_name_non_abi_mangled = symbol_name + 1; 1349 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 1350 type = eSymbolTypeObjCIVar; 1351 } 1352 } 1353 } 1354 else 1355 if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 1356 { 1357 type = eSymbolTypeException; 1358 } 1359 else 1360 { 1361 type = eSymbolTypeData; 1362 } 1363 } 1364 else 1365 if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 1366 { 1367 type = eSymbolTypeTrampoline; 1368 } 1369 else 1370 if (symbol_section->IsDescendant(objc_section_sp.get())) 1371 { 1372 type = eSymbolTypeRuntime; 1373 if (symbol_name && symbol_name[0] == '.') 1374 { 1375 llvm::StringRef symbol_name_ref(symbol_name); 1376 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 1377 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 1378 { 1379 symbol_name_non_abi_mangled = symbol_name; 1380 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 1381 type = eSymbolTypeObjCClass; 1382 } 1383 } 1384 } 1385 } 1386 } 1387 break; 1388 } 1389 } 1390 1391 if (add_nlist) 1392 { 1393 uint64_t symbol_value = nlist.n_value; 1394 bool symbol_name_is_mangled = false; 1395 1396 if (symbol_name_non_abi_mangled) 1397 { 1398 sym[sym_idx].GetMangled().SetMangledName (symbol_name_non_abi_mangled); 1399 sym[sym_idx].GetMangled().SetDemangledName (symbol_name); 1400 } 1401 else 1402 { 1403 if (symbol_name && symbol_name[0] == '_') 1404 { 1405 symbol_name_is_mangled = symbol_name[1] == '_'; 1406 symbol_name++; // Skip the leading underscore 1407 } 1408 1409 if (symbol_name) 1410 { 1411 sym[sym_idx].GetMangled().SetValue(symbol_name, symbol_name_is_mangled); 1412 } 1413 } 1414 1415 if (is_debug == false) 1416 { 1417 if (type == eSymbolTypeCode) 1418 { 1419 // See if we can find a N_FUN entry for any code symbols. 1420 // If we do find a match, and the name matches, then we 1421 // can merge the two into just the function symbol to avoid 1422 // duplicate entries in the symbol table 1423 ValueToSymbolIndexMap::const_iterator pos = N_FUN_addr_to_sym_idx.find (nlist.n_value); 1424 if (pos != N_FUN_addr_to_sym_idx.end()) 1425 { 1426 if ((symbol_name_is_mangled == true && sym[sym_idx].GetMangled().GetMangledName() == sym[pos->second].GetMangled().GetMangledName()) || 1427 (symbol_name_is_mangled == false && sym[sym_idx].GetMangled().GetDemangledName() == sym[pos->second].GetMangled().GetDemangledName())) 1428 { 1429 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 1430 // We just need the flags from the linker symbol, so put these flags 1431 // into the N_FUN flags to avoid duplicate symbols in the symbol table 1432 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 1433 sym[sym_idx].Clear(); 1434 continue; 1435 } 1436 } 1437 } 1438 else if (type == eSymbolTypeData) 1439 { 1440 // See if we can find a N_STSYM entry for any data symbols. 1441 // If we do find a match, and the name matches, then we 1442 // can merge the two into just the Static symbol to avoid 1443 // duplicate entries in the symbol table 1444 ValueToSymbolIndexMap::const_iterator pos = N_STSYM_addr_to_sym_idx.find (nlist.n_value); 1445 if (pos != N_STSYM_addr_to_sym_idx.end()) 1446 { 1447 if ((symbol_name_is_mangled == true && sym[sym_idx].GetMangled().GetMangledName() == sym[pos->second].GetMangled().GetMangledName()) || 1448 (symbol_name_is_mangled == false && sym[sym_idx].GetMangled().GetDemangledName() == sym[pos->second].GetMangled().GetDemangledName())) 1449 { 1450 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 1451 // We just need the flags from the linker symbol, so put these flags 1452 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 1453 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 1454 sym[sym_idx].Clear(); 1455 continue; 1456 } 1457 } 1458 } 1459 } 1460 if (symbol_section != NULL) 1461 symbol_value -= symbol_section->GetFileAddress(); 1462 1463 sym[sym_idx].SetID (nlist_idx); 1464 sym[sym_idx].SetType (type); 1465 sym[sym_idx].GetAddressRangeRef().GetBaseAddress().SetSection (symbol_section); 1466 sym[sym_idx].GetAddressRangeRef().GetBaseAddress().SetOffset (symbol_value); 1467 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 1468 1469 ++sym_idx; 1470 } 1471 else 1472 { 1473 sym[sym_idx].Clear(); 1474 } 1475 1476 } 1477 1478 // STAB N_GSYM entries end up having a symbol type eSymbolTypeGlobal and when the symbol value 1479 // is zero, the address of the global ends up being in a non-STAB entry. Try and fix up all 1480 // such entries by figuring out what the address for the global is by looking up this non-STAB 1481 // entry and copying the value into the debug symbol's value to save us the hassle in the 1482 // debug symbol parser. 1483 1484 Symbol *global_symbol = NULL; 1485 for (nlist_idx = 0; 1486 nlist_idx < symtab_load_command.nsyms && (global_symbol = symtab->FindSymbolWithType (eSymbolTypeData, Symtab::eDebugYes, Symtab::eVisibilityAny, nlist_idx)) != NULL; 1487 nlist_idx++) 1488 { 1489 if (global_symbol->GetValue().GetFileAddress() == 0) 1490 { 1491 std::vector<uint32_t> indexes; 1492 if (symtab->AppendSymbolIndexesWithName (global_symbol->GetMangled().GetName(), indexes) > 0) 1493 { 1494 std::vector<uint32_t>::const_iterator pos; 1495 std::vector<uint32_t>::const_iterator end = indexes.end(); 1496 for (pos = indexes.begin(); pos != end; ++pos) 1497 { 1498 symbol_ptr = symtab->SymbolAtIndex(*pos); 1499 if (symbol_ptr != global_symbol && symbol_ptr->IsDebug() == false) 1500 { 1501 global_symbol->SetValue(symbol_ptr->GetValue()); 1502 break; 1503 } 1504 } 1505 } 1506 } 1507 } 1508 1509 // Trim our symbols down to just what we ended up with after 1510 // removing any symbols. 1511 if (sym_idx < num_syms) 1512 { 1513 num_syms = sym_idx; 1514 sym = symtab->Resize (num_syms); 1515 } 1516 1517 // Now synthesize indirect symbols 1518 if (m_dysymtab.nindirectsyms != 0) 1519 { 1520 DataBufferSP indirect_symbol_indexes_sp(m_file.ReadFileContents(m_offset + m_dysymtab.indirectsymoff, m_dysymtab.nindirectsyms * 4)); 1521 1522 if (indirect_symbol_indexes_sp && indirect_symbol_indexes_sp->GetByteSize()) 1523 { 1524 NListIndexToSymbolIndexMap::const_iterator end_index_pos = m_nlist_idx_to_sym_idx.end(); 1525 DataExtractor indirect_symbol_index_data (indirect_symbol_indexes_sp, m_data.GetByteOrder(), m_data.GetAddressByteSize()); 1526 1527 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); ++sect_idx) 1528 { 1529 if ((m_mach_sections[sect_idx].flags & SectionFlagMaskSectionType) == SectionTypeSymbolStubs) 1530 { 1531 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 1532 if (symbol_stub_byte_size == 0) 1533 continue; 1534 1535 const uint32_t num_symbol_stubs = m_mach_sections[sect_idx].size / symbol_stub_byte_size; 1536 1537 if (num_symbol_stubs == 0) 1538 continue; 1539 1540 const uint32_t symbol_stub_index_offset = m_mach_sections[sect_idx].reserved1; 1541 uint32_t synthetic_stub_sym_id = symtab_load_command.nsyms; 1542 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) 1543 { 1544 const uint32_t symbol_stub_index = symbol_stub_index_offset + stub_idx; 1545 const lldb::addr_t symbol_stub_addr = m_mach_sections[sect_idx].addr + (stub_idx * symbol_stub_byte_size); 1546 uint32_t symbol_stub_offset = symbol_stub_index * 4; 1547 if (indirect_symbol_index_data.ValidOffsetForDataOfSize(symbol_stub_offset, 4)) 1548 { 1549 const uint32_t stub_sym_id = indirect_symbol_index_data.GetU32 (&symbol_stub_offset); 1550 if (stub_sym_id & (IndirectSymbolAbsolute | IndirectSymbolLocal)) 1551 continue; 1552 1553 NListIndexToSymbolIndexMap::const_iterator index_pos = m_nlist_idx_to_sym_idx.find (stub_sym_id); 1554 Symbol *stub_symbol = NULL; 1555 if (index_pos != end_index_pos) 1556 { 1557 // We have a remapping from the original nlist index to 1558 // a current symbol index, so just look this up by index 1559 stub_symbol = symtab->SymbolAtIndex (index_pos->second); 1560 } 1561 else 1562 { 1563 // We need to lookup a symbol using the original nlist 1564 // symbol index since this index is coming from the 1565 // S_SYMBOL_STUBS 1566 stub_symbol = symtab->FindSymbolByID (stub_sym_id); 1567 } 1568 1569 assert (stub_symbol); 1570 if (stub_symbol) 1571 { 1572 Address so_addr(symbol_stub_addr, section_list); 1573 1574 if (stub_symbol->GetType() == eSymbolTypeUndefined) 1575 { 1576 // Change the external symbol into a trampoline that makes sense 1577 // These symbols were N_UNDF N_EXT, and are useless to us, so we 1578 // can re-use them so we don't have to make up a synthetic symbol 1579 // for no good reason. 1580 stub_symbol->SetType (eSymbolTypeTrampoline); 1581 stub_symbol->SetExternal (false); 1582 stub_symbol->GetAddressRangeRef().GetBaseAddress() = so_addr; 1583 stub_symbol->GetAddressRangeRef().SetByteSize (symbol_stub_byte_size); 1584 } 1585 else 1586 { 1587 // Make a synthetic symbol to describe the trampoline stub 1588 if (sym_idx >= num_syms) 1589 sym = symtab->Resize (++num_syms); 1590 sym[sym_idx].SetID (synthetic_stub_sym_id++); 1591 sym[sym_idx].GetMangled() = stub_symbol->GetMangled(); 1592 sym[sym_idx].SetType (eSymbolTypeTrampoline); 1593 sym[sym_idx].SetIsSynthetic (true); 1594 sym[sym_idx].GetAddressRangeRef().GetBaseAddress() = so_addr; 1595 sym[sym_idx].GetAddressRangeRef().SetByteSize (symbol_stub_byte_size); 1596 ++sym_idx; 1597 } 1598 } 1599 } 1600 } 1601 } 1602 } 1603 } 1604 } 1605 return symtab->GetNumSymbols(); 1606 } 1607 } 1608 offset = cmd_offset + symtab_load_command.cmdsize; 1609 } 1610 return 0; 1611 } 1612 1613 1614 void 1615 ObjectFileMachO::Dump (Stream *s) 1616 { 1617 lldb_private::Mutex::Locker locker(m_mutex); 1618 s->Printf("%p: ", this); 1619 s->Indent(); 1620 if (m_header.magic == HeaderMagic64 || m_header.magic == HeaderMagic64Swapped) 1621 s->PutCString("ObjectFileMachO64"); 1622 else 1623 s->PutCString("ObjectFileMachO32"); 1624 1625 ArchSpec header_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 1626 1627 *s << ", file = '" << m_file << "', arch = " << header_arch.GetArchitectureName() << "\n"; 1628 1629 if (m_sections_ap.get()) 1630 m_sections_ap->Dump(s, NULL, true, UINT32_MAX); 1631 1632 if (m_symtab_ap.get()) 1633 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 1634 } 1635 1636 1637 bool 1638 ObjectFileMachO::GetUUID (lldb_private::UUID* uuid) 1639 { 1640 lldb_private::Mutex::Locker locker(m_mutex); 1641 struct uuid_command load_cmd; 1642 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 1643 uint32_t i; 1644 for (i=0; i<m_header.ncmds; ++i) 1645 { 1646 const uint32_t cmd_offset = offset; 1647 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1648 break; 1649 1650 if (load_cmd.cmd == LoadCommandUUID) 1651 { 1652 const uint8_t *uuid_bytes = m_data.PeekData(offset, 16); 1653 if (uuid_bytes) 1654 { 1655 uuid->SetBytes (uuid_bytes); 1656 return true; 1657 } 1658 return false; 1659 } 1660 offset = cmd_offset + load_cmd.cmdsize; 1661 } 1662 return false; 1663 } 1664 1665 1666 uint32_t 1667 ObjectFileMachO::GetDependentModules (FileSpecList& files) 1668 { 1669 lldb_private::Mutex::Locker locker(m_mutex); 1670 struct load_command load_cmd; 1671 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 1672 uint32_t count = 0; 1673 const bool resolve_path = false; // Don't resolve the dependend file paths since they may not reside on this system 1674 uint32_t i; 1675 for (i=0; i<m_header.ncmds; ++i) 1676 { 1677 const uint32_t cmd_offset = offset; 1678 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1679 break; 1680 1681 switch (load_cmd.cmd) 1682 { 1683 case LoadCommandDylibLoad: 1684 case LoadCommandDylibLoadWeak: 1685 case LoadCommandDylibReexport: 1686 case LoadCommandDynamicLinkerLoad: 1687 case LoadCommandFixedVMShlibLoad: 1688 case LoadCommandDylibLoadUpward: 1689 { 1690 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 1691 const char *path = m_data.PeekCStr(name_offset); 1692 // Skip any path that starts with '@' since these are usually: 1693 // @executable_path/.../file 1694 // @rpath/.../file 1695 if (path && path[0] != '@') 1696 { 1697 FileSpec file_spec(path, resolve_path); 1698 if (files.AppendIfUnique(file_spec)) 1699 count++; 1700 } 1701 } 1702 break; 1703 1704 default: 1705 break; 1706 } 1707 offset = cmd_offset + load_cmd.cmdsize; 1708 } 1709 return count; 1710 } 1711 1712 lldb_private::Address 1713 ObjectFileMachO::GetEntryPointAddress () 1714 { 1715 // If the object file is not an executable it can't hold the entry point. m_entry_point_address 1716 // is initialized to an invalid address, so we can just return that. 1717 // If m_entry_point_address is valid it means we've found it already, so return the cached value. 1718 1719 if (!IsExecutable() || m_entry_point_address.IsValid()) 1720 return m_entry_point_address; 1721 1722 // Otherwise, look for the UnixThread or Thread command. The data for the Thread command is given in 1723 // /usr/include/mach-o.h, but it is basically: 1724 // 1725 // uint32_t flavor - this is the flavor argument you would pass to thread_get_state 1726 // uint32_t count - this is the count of longs in the thread state data 1727 // struct XXX_thread_state state - this is the structure from <machine/thread_status.h> corresponding to the flavor. 1728 // <repeat this trio> 1729 // 1730 // So we just keep reading the various register flavors till we find the GPR one, then read the PC out of there. 1731 // FIXME: We will need to have a "RegisterContext data provider" class at some point that can get all the registers 1732 // out of data in this form & attach them to a given thread. That should underlie the MacOS X User process plugin, 1733 // and we'll also need it for the MacOS X Core File process plugin. When we have that we can also use it here. 1734 // 1735 // For now we hard-code the offsets and flavors we need: 1736 // 1737 // 1738 1739 lldb_private::Mutex::Locker locker(m_mutex); 1740 struct load_command load_cmd; 1741 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 1742 uint32_t i; 1743 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 1744 bool done = false; 1745 1746 for (i=0; i<m_header.ncmds; ++i) 1747 { 1748 const uint32_t cmd_offset = offset; 1749 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1750 break; 1751 1752 switch (load_cmd.cmd) 1753 { 1754 case LoadCommandUnixThread: 1755 case LoadCommandThread: 1756 { 1757 while (offset < cmd_offset + load_cmd.cmdsize) 1758 { 1759 uint32_t flavor = m_data.GetU32(&offset); 1760 uint32_t count = m_data.GetU32(&offset); 1761 if (count == 0) 1762 { 1763 // We've gotten off somehow, log and exit; 1764 return m_entry_point_address; 1765 } 1766 1767 switch (m_header.cputype) 1768 { 1769 case llvm::MachO::CPUTypeARM: 1770 if (flavor == 1) // ARM_THREAD_STATE from mach/arm/thread_status.h 1771 { 1772 offset += 60; // This is the offset of pc in the GPR thread state data structure. 1773 start_address = m_data.GetU32(&offset); 1774 done = true; 1775 } 1776 break; 1777 case llvm::MachO::CPUTypeI386: 1778 if (flavor == 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 1779 { 1780 offset += 40; // This is the offset of eip in the GPR thread state data structure. 1781 start_address = m_data.GetU32(&offset); 1782 done = true; 1783 } 1784 break; 1785 case llvm::MachO::CPUTypeX86_64: 1786 if (flavor == 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 1787 { 1788 offset += 16 * 8; // This is the offset of rip in the GPR thread state data structure. 1789 start_address = m_data.GetU64(&offset); 1790 done = true; 1791 } 1792 break; 1793 default: 1794 return m_entry_point_address; 1795 } 1796 // Haven't found the GPR flavor yet, skip over the data for this flavor: 1797 if (done) 1798 break; 1799 offset += count * 4; 1800 } 1801 } 1802 break; 1803 1804 default: 1805 break; 1806 } 1807 if (done) 1808 break; 1809 1810 // Go to the next load command: 1811 offset = cmd_offset + load_cmd.cmdsize; 1812 } 1813 1814 if (start_address != LLDB_INVALID_ADDRESS) 1815 { 1816 // We got the start address from the load commands, so now resolve that address in the sections 1817 // of this ObjectFile: 1818 if (!m_entry_point_address.ResolveAddressUsingFileSections (start_address, GetSectionList())) 1819 { 1820 m_entry_point_address.Clear(); 1821 } 1822 } 1823 else 1824 { 1825 // We couldn't read the UnixThread load command - maybe it wasn't there. As a fallback look for the 1826 // "start" symbol in the main executable. 1827 1828 SymbolContextList contexts; 1829 SymbolContext context; 1830 if (!m_module->FindSymbolsWithNameAndType(ConstString ("start"), eSymbolTypeCode, contexts)) 1831 return m_entry_point_address; 1832 1833 contexts.GetContextAtIndex(0, context); 1834 1835 m_entry_point_address = context.symbol->GetValue(); 1836 } 1837 1838 return m_entry_point_address; 1839 1840 } 1841 1842 ObjectFile::Type 1843 ObjectFileMachO::CalculateType() 1844 { 1845 switch (m_header.filetype) 1846 { 1847 case HeaderFileTypeObject: // 0x1u MH_OBJECT 1848 if (GetAddressByteSize () == 4) 1849 { 1850 // 32 bit kexts are just object files, but they do have a valid 1851 // UUID load command. 1852 UUID uuid; 1853 if (GetUUID(&uuid)) 1854 { 1855 // this checking for the UUID load command is not enough 1856 // we could eventually look for the symbol named 1857 // "OSKextGetCurrentIdentifier" as this is required of kexts 1858 if (m_strata == eStrataInvalid) 1859 m_strata = eStrataKernel; 1860 return eTypeSharedLibrary; 1861 } 1862 } 1863 return eTypeObjectFile; 1864 1865 case HeaderFileTypeExecutable: return eTypeExecutable; // 0x2u MH_EXECUTE 1866 case HeaderFileTypeFixedVMShlib: return eTypeSharedLibrary; // 0x3u MH_FVMLIB 1867 case HeaderFileTypeCore: return eTypeCoreFile; // 0x4u MH_CORE 1868 case HeaderFileTypePreloadedExecutable: return eTypeSharedLibrary; // 0x5u MH_PRELOAD 1869 case HeaderFileTypeDynamicShlib: return eTypeSharedLibrary; // 0x6u MH_DYLIB 1870 case HeaderFileTypeDynamicLinkEditor: return eTypeDynamicLinker; // 0x7u MH_DYLINKER 1871 case HeaderFileTypeBundle: return eTypeSharedLibrary; // 0x8u MH_BUNDLE 1872 case HeaderFileTypeDynamicShlibStub: return eTypeStubLibrary; // 0x9u MH_DYLIB_STUB 1873 case HeaderFileTypeDSYM: return eTypeDebugInfo; // 0xAu MH_DSYM 1874 case HeaderFileTypeKextBundle: return eTypeSharedLibrary; // 0xBu MH_KEXT_BUNDLE 1875 default: 1876 break; 1877 } 1878 return eTypeUnknown; 1879 } 1880 1881 ObjectFile::Strata 1882 ObjectFileMachO::CalculateStrata() 1883 { 1884 switch (m_header.filetype) 1885 { 1886 case HeaderFileTypeObject: // 0x1u MH_OBJECT 1887 { 1888 // 32 bit kexts are just object files, but they do have a valid 1889 // UUID load command. 1890 UUID uuid; 1891 if (GetUUID(&uuid)) 1892 { 1893 // this checking for the UUID load command is not enough 1894 // we could eventually look for the symbol named 1895 // "OSKextGetCurrentIdentifier" as this is required of kexts 1896 if (m_type == eTypeInvalid) 1897 m_type = eTypeSharedLibrary; 1898 1899 return eStrataKernel; 1900 } 1901 } 1902 return eStrataUnknown; 1903 1904 case HeaderFileTypeExecutable: // 0x2u MH_EXECUTE 1905 // Check for the MH_DYLDLINK bit in the flags 1906 if (m_header.flags & HeaderFlagBitIsDynamicLinkObject) 1907 return eStrataUser; 1908 return eStrataKernel; 1909 1910 case HeaderFileTypeFixedVMShlib: return eStrataUser; // 0x3u MH_FVMLIB 1911 case HeaderFileTypeCore: return eStrataUnknown; // 0x4u MH_CORE 1912 case HeaderFileTypePreloadedExecutable: return eStrataUser; // 0x5u MH_PRELOAD 1913 case HeaderFileTypeDynamicShlib: return eStrataUser; // 0x6u MH_DYLIB 1914 case HeaderFileTypeDynamicLinkEditor: return eStrataUser; // 0x7u MH_DYLINKER 1915 case HeaderFileTypeBundle: return eStrataUser; // 0x8u MH_BUNDLE 1916 case HeaderFileTypeDynamicShlibStub: return eStrataUser; // 0x9u MH_DYLIB_STUB 1917 case HeaderFileTypeDSYM: return eStrataUnknown; // 0xAu MH_DSYM 1918 case HeaderFileTypeKextBundle: return eStrataKernel; // 0xBu MH_KEXT_BUNDLE 1919 default: 1920 break; 1921 } 1922 return eStrataUnknown; 1923 } 1924 1925 1926 bool 1927 ObjectFileMachO::GetArchitecture (ArchSpec &arch) 1928 { 1929 lldb_private::Mutex::Locker locker(m_mutex); 1930 arch.SetArchitecture (eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 1931 1932 // Files with type MH_PRELOAD are currently used in cases where the image 1933 // debugs at the addresses in the file itself. Below we set the OS to 1934 // unknown to make sure we use the DynamicLoaderStatic()... 1935 if (m_header.filetype == HeaderFileTypePreloadedExecutable) 1936 { 1937 arch.GetTriple().setOS (llvm::Triple::UnknownOS); 1938 } 1939 1940 return true; 1941 } 1942 1943 1944 //------------------------------------------------------------------ 1945 // PluginInterface protocol 1946 //------------------------------------------------------------------ 1947 const char * 1948 ObjectFileMachO::GetPluginName() 1949 { 1950 return "ObjectFileMachO"; 1951 } 1952 1953 const char * 1954 ObjectFileMachO::GetShortPluginName() 1955 { 1956 return GetPluginNameStatic(); 1957 } 1958 1959 uint32_t 1960 ObjectFileMachO::GetPluginVersion() 1961 { 1962 return 1; 1963 } 1964 1965