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