ObjectFileMachO.cpp revision 08a73204faf6db215f22f5cefaf84b57cfd69e6a
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 "ObjectFileMachO.h" 11 12#include "lldb/Core/ArchSpec.h" 13#include "lldb/Core/DataBuffer.h" 14#include "lldb/Core/FileSpec.h" 15#include "lldb/Core/FileSpecList.h" 16#include "lldb/Core/Module.h" 17#include "lldb/Core/PluginManager.h" 18#include "lldb/Core/Section.h" 19#include "lldb/Core/StreamFile.h" 20#include "lldb/Core/StreamString.h" 21#include "lldb/Core/Timer.h" 22#include "lldb/Core/UUID.h" 23#include "lldb/Symbol/ObjectFile.h" 24 25 26using namespace lldb; 27using namespace lldb_private; 28using namespace llvm::MachO; 29 30 31void 32ObjectFileMachO::Initialize() 33{ 34 PluginManager::RegisterPlugin (GetPluginNameStatic(), 35 GetPluginDescriptionStatic(), 36 CreateInstance); 37} 38 39void 40ObjectFileMachO::Terminate() 41{ 42 PluginManager::UnregisterPlugin (CreateInstance); 43} 44 45 46const char * 47ObjectFileMachO::GetPluginNameStatic() 48{ 49 return "object-file.mach-o"; 50} 51 52const char * 53ObjectFileMachO::GetPluginDescriptionStatic() 54{ 55 return "Mach-o object file reader (32 and 64 bit)"; 56} 57 58 59ObjectFile * 60ObjectFileMachO::CreateInstance (Module* module, DataBufferSP& dataSP, const FileSpec* file, addr_t offset, addr_t length) 61{ 62 if (ObjectFileMachO::MagicBytesMatch(dataSP)) 63 { 64 std::auto_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module, dataSP, file, offset, length)); 65 if (objfile_ap.get() && objfile_ap->ParseHeader()) 66 return objfile_ap.release(); 67 } 68 return NULL; 69} 70 71 72static uint32_t 73MachHeaderSizeFromMagic(uint32_t magic) 74{ 75 switch (magic) 76 { 77 case HeaderMagic32: 78 case HeaderMagic32Swapped: 79 return sizeof(struct mach_header); 80 81 case HeaderMagic64: 82 case HeaderMagic64Swapped: 83 return sizeof(struct mach_header_64); 84 break; 85 86 default: 87 break; 88 } 89 return 0; 90} 91 92 93bool 94ObjectFileMachO::MagicBytesMatch (DataBufferSP& dataSP) 95{ 96 DataExtractor data(dataSP, eByteOrderHost, 4); 97 uint32_t offset = 0; 98 uint32_t magic = data.GetU32(&offset); 99 return MachHeaderSizeFromMagic(magic) != 0; 100} 101 102 103ObjectFileMachO::ObjectFileMachO(Module* module, DataBufferSP& dataSP, const FileSpec* file, addr_t offset, addr_t length) : 104 ObjectFile(module, file, offset, length, dataSP), 105 m_mutex (Mutex::eMutexTypeRecursive), 106 m_header(), 107 m_sections_ap(), 108 m_symtab_ap() 109{ 110 ::bzero (&m_header, sizeof(m_header)); 111 ::bzero (&m_dysymtab, sizeof(m_dysymtab)); 112} 113 114 115ObjectFileMachO::~ObjectFileMachO() 116{ 117} 118 119 120bool 121ObjectFileMachO::ParseHeader () 122{ 123 lldb_private::Mutex::Locker locker(m_mutex); 124 bool can_parse = false; 125 uint32_t offset = 0; 126 m_data.SetByteOrder (eByteOrderHost); 127 // Leave magic in the original byte order 128 m_header.magic = m_data.GetU32(&offset); 129 switch (m_header.magic) 130 { 131 case HeaderMagic32: 132 m_data.SetByteOrder (eByteOrderHost); 133 m_data.SetAddressByteSize(4); 134 can_parse = true; 135 break; 136 137 case HeaderMagic64: 138 m_data.SetByteOrder (eByteOrderHost); 139 m_data.SetAddressByteSize(8); 140 can_parse = true; 141 break; 142 143 case HeaderMagic32Swapped: 144 m_data.SetByteOrder(eByteOrderHost == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 145 m_data.SetAddressByteSize(4); 146 can_parse = true; 147 break; 148 149 case HeaderMagic64Swapped: 150 m_data.SetByteOrder(eByteOrderHost == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 151 m_data.SetAddressByteSize(8); 152 can_parse = true; 153 break; 154 155 default: 156 break; 157 } 158 159 if (can_parse) 160 { 161 m_data.GetU32(&offset, &m_header.cputype, 6); 162 163 ArchSpec mach_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 164 165 if (SetModulesArchitecture (mach_arch)) 166 { 167 // Read in all only the load command data 168 DataBufferSP data_sp(m_file.ReadFileContents(m_offset, m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic))); 169 m_data.SetData (data_sp); 170 return true; 171 } 172 } 173 else 174 { 175 memset(&m_header, 0, sizeof(struct mach_header)); 176 } 177 return false; 178} 179 180 181ByteOrder 182ObjectFileMachO::GetByteOrder () const 183{ 184 lldb_private::Mutex::Locker locker(m_mutex); 185 return m_data.GetByteOrder (); 186} 187 188bool 189ObjectFileMachO::IsExecutable() const 190{ 191 return m_header.filetype == HeaderFileTypeExecutable; 192} 193 194size_t 195ObjectFileMachO::GetAddressByteSize () const 196{ 197 lldb_private::Mutex::Locker locker(m_mutex); 198 return m_data.GetAddressByteSize (); 199} 200 201 202Symtab * 203ObjectFileMachO::GetSymtab() 204{ 205 lldb_private::Mutex::Locker locker(m_mutex); 206 if (m_symtab_ap.get() == NULL) 207 { 208 m_symtab_ap.reset(new Symtab(this)); 209 Mutex::Locker locker (m_symtab_ap->GetMutex()); 210 ParseSymtab (true); 211 } 212 return m_symtab_ap.get(); 213} 214 215 216SectionList * 217ObjectFileMachO::GetSectionList() 218{ 219 lldb_private::Mutex::Locker locker(m_mutex); 220 if (m_sections_ap.get() == NULL) 221 { 222 m_sections_ap.reset(new SectionList()); 223 ParseSections(); 224 } 225 return m_sections_ap.get(); 226} 227 228 229size_t 230ObjectFileMachO::ParseSections () 231{ 232 lldb::user_id_t segID = 0; 233 lldb::user_id_t sectID = 0; 234 struct segment_command_64 load_cmd; 235 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 236 uint32_t i; 237 //bool dump_sections = false; 238 for (i=0; i<m_header.ncmds; ++i) 239 { 240 const uint32_t load_cmd_offset = offset; 241 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 242 break; 243 244 if (load_cmd.cmd == LoadCommandSegment32 || load_cmd.cmd == LoadCommandSegment64) 245 { 246 if (m_data.GetU8(&offset, (uint8_t*)load_cmd.segname, 16)) 247 { 248 load_cmd.vmaddr = m_data.GetAddress(&offset); 249 load_cmd.vmsize = m_data.GetAddress(&offset); 250 load_cmd.fileoff = m_data.GetAddress(&offset); 251 load_cmd.filesize = m_data.GetAddress(&offset); 252 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 253 { 254 // Keep a list of mach segments around in case we need to 255 // get at data that isn't stored in the abstracted Sections. 256 m_mach_segments.push_back (load_cmd); 257 258 ConstString segment_name (load_cmd.segname, std::min<int>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 259 // Use a segment ID of the segment index shifted left by 8 so they 260 // never conflict with any of the sections. 261 SectionSP segment_sp; 262 if (segment_name) 263 { 264 segment_sp.reset(new Section (NULL, 265 GetModule(), // Module to which this section belongs 266 ++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 267 segment_name, // Name of this section 268 eSectionTypeContainer, // This section is a container of other sections. 269 load_cmd.vmaddr, // File VM address == addresses as they are found in the object file 270 load_cmd.vmsize, // VM size in bytes of this section 271 load_cmd.fileoff, // Offset to the data for this section in the file 272 load_cmd.filesize, // Size in bytes of this section as found in the the file 273 load_cmd.flags)); // Flags for this section 274 275 m_sections_ap->AddSection(segment_sp); 276 } 277 278 struct section_64 sect64; 279 ::bzero (§64, sizeof(sect64)); 280 // Push a section into our mach sections for the section at 281 // index zero (NListSectionNoSection) if we don't have any 282 // mach sections yet... 283 if (m_mach_sections.empty()) 284 m_mach_sections.push_back(sect64); 285 uint32_t segment_sect_idx; 286 const lldb::user_id_t first_segment_sectID = sectID + 1; 287 288 289 const uint32_t num_u32s = load_cmd.cmd == LoadCommandSegment32 ? 7 : 8; 290 for (segment_sect_idx=0; segment_sect_idx<load_cmd.nsects; ++segment_sect_idx) 291 { 292 if (m_data.GetU8(&offset, (uint8_t*)sect64.sectname, sizeof(sect64.sectname)) == NULL) 293 break; 294 if (m_data.GetU8(&offset, (uint8_t*)sect64.segname, sizeof(sect64.segname)) == NULL) 295 break; 296 sect64.addr = m_data.GetAddress(&offset); 297 sect64.size = m_data.GetAddress(&offset); 298 299 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 300 break; 301 302 // Keep a list of mach sections around in case we need to 303 // get at data that isn't stored in the abstracted Sections. 304 m_mach_sections.push_back (sect64); 305 306 ConstString section_name (sect64.sectname, std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 307 if (!segment_name) 308 { 309 // We have a segment with no name so we need to conjure up 310 // segments that correspond to the section's segname if there 311 // isn't already such a section. If there is such a section, 312 // we resize the section so that it spans all sections. 313 // We also mark these sections as fake so address matches don't 314 // hit if they land in the gaps between the child sections. 315 segment_name.SetTrimmedCStringWithLength(sect64.segname, sizeof(sect64.segname)); 316 segment_sp = m_sections_ap->FindSectionByName (segment_name); 317 if (segment_sp.get()) 318 { 319 Section *segment = segment_sp.get(); 320 // Grow the section size as needed. 321 const lldb::addr_t sect64_min_addr = sect64.addr; 322 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 323 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 324 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 325 const lldb::addr_t curr_seg_max_addr = curr_seg_min_addr + curr_seg_byte_size; 326 if (sect64_min_addr >= curr_seg_min_addr) 327 { 328 const lldb::addr_t new_seg_byte_size = sect64_max_addr - curr_seg_min_addr; 329 // Only grow the section size if needed 330 if (new_seg_byte_size > curr_seg_byte_size) 331 segment->SetByteSize (new_seg_byte_size); 332 } 333 else 334 { 335 // We need to change the base address of the segment and 336 // adjust the child section offsets for all existing children. 337 const lldb::addr_t slide_amount = sect64_min_addr - curr_seg_min_addr; 338 segment->Slide(slide_amount, false); 339 segment->GetChildren().Slide (-slide_amount, false); 340 segment->SetByteSize (curr_seg_max_addr - sect64_min_addr); 341 } 342 343 // Grow the section size as needed. 344 if (sect64.offset) 345 { 346 const lldb::addr_t segment_min_file_offset = segment->GetFileOffset(); 347 const lldb::addr_t segment_max_file_offset = segment_min_file_offset + segment->GetFileSize(); 348 349 const lldb::addr_t section_min_file_offset = sect64.offset; 350 const lldb::addr_t section_max_file_offset = section_min_file_offset + sect64.size; 351 const lldb::addr_t new_file_offset = std::min (section_min_file_offset, segment_min_file_offset); 352 const lldb::addr_t new_file_size = std::max (section_max_file_offset, segment_max_file_offset) - new_file_offset; 353 segment->SetFileOffset (new_file_offset); 354 segment->SetFileSize (new_file_size); 355 } 356 } 357 else 358 { 359 // Create a fake section for the section's named segment 360 segment_sp.reset(new Section(segment_sp.get(), // Parent section 361 GetModule(), // Module to which this section belongs 362 ++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 363 segment_name, // Name of this section 364 eSectionTypeContainer, // This section is a container of other sections. 365 sect64.addr, // File VM address == addresses as they are found in the object file 366 sect64.size, // VM size in bytes of this section 367 sect64.offset, // Offset to the data for this section in the file 368 sect64.offset ? sect64.size : 0, // Size in bytes of this section as found in the the file 369 load_cmd.flags)); // Flags for this section 370 segment_sp->SetIsFake(true); 371 m_sections_ap->AddSection(segment_sp); 372 } 373 } 374 assert (segment_sp.get()); 375 376 uint32_t mach_sect_type = sect64.flags & SectionFlagMaskSectionType; 377 static ConstString g_sect_name_objc_data ("__objc_data"); 378 static ConstString g_sect_name_objc_msgrefs ("__objc_msgrefs"); 379 static ConstString g_sect_name_objc_selrefs ("__objc_selrefs"); 380 static ConstString g_sect_name_objc_classrefs ("__objc_classrefs"); 381 static ConstString g_sect_name_objc_superrefs ("__objc_superrefs"); 382 static ConstString g_sect_name_objc_const ("__objc_const"); 383 static ConstString g_sect_name_objc_classlist ("__objc_classlist"); 384 static ConstString g_sect_name_cfstring ("__cfstring"); 385 386 static ConstString g_sect_name_dwarf_debug_abbrev ("__debug_abbrev"); 387 static ConstString g_sect_name_dwarf_debug_aranges ("__debug_aranges"); 388 static ConstString g_sect_name_dwarf_debug_frame ("__debug_frame"); 389 static ConstString g_sect_name_dwarf_debug_info ("__debug_info"); 390 static ConstString g_sect_name_dwarf_debug_line ("__debug_line"); 391 static ConstString g_sect_name_dwarf_debug_loc ("__debug_loc"); 392 static ConstString g_sect_name_dwarf_debug_macinfo ("__debug_macinfo"); 393 static ConstString g_sect_name_dwarf_debug_pubnames ("__debug_pubnames"); 394 static ConstString g_sect_name_dwarf_debug_pubtypes ("__debug_pubtypes"); 395 static ConstString g_sect_name_dwarf_debug_ranges ("__debug_ranges"); 396 static ConstString g_sect_name_dwarf_debug_str ("__debug_str"); 397 static ConstString g_sect_name_eh_frame ("__eh_frame"); 398 static ConstString g_sect_name_DATA ("__DATA"); 399 static ConstString g_sect_name_TEXT ("__TEXT"); 400 401 SectionType sect_type = eSectionTypeOther; 402 403 if (section_name == g_sect_name_dwarf_debug_abbrev) 404 sect_type = eSectionTypeDWARFDebugAbbrev; 405 else if (section_name == g_sect_name_dwarf_debug_aranges) 406 sect_type = eSectionTypeDWARFDebugAranges; 407 else if (section_name == g_sect_name_dwarf_debug_frame) 408 sect_type = eSectionTypeDWARFDebugFrame; 409 else if (section_name == g_sect_name_dwarf_debug_info) 410 sect_type = eSectionTypeDWARFDebugInfo; 411 else if (section_name == g_sect_name_dwarf_debug_line) 412 sect_type = eSectionTypeDWARFDebugLine; 413 else if (section_name == g_sect_name_dwarf_debug_loc) 414 sect_type = eSectionTypeDWARFDebugLoc; 415 else if (section_name == g_sect_name_dwarf_debug_macinfo) 416 sect_type = eSectionTypeDWARFDebugMacInfo; 417 else if (section_name == g_sect_name_dwarf_debug_pubnames) 418 sect_type = eSectionTypeDWARFDebugPubNames; 419 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 420 sect_type = eSectionTypeDWARFDebugPubTypes; 421 else if (section_name == g_sect_name_dwarf_debug_ranges) 422 sect_type = eSectionTypeDWARFDebugRanges; 423 else if (section_name == g_sect_name_dwarf_debug_str) 424 sect_type = eSectionTypeDWARFDebugStr; 425 else if (section_name == g_sect_name_objc_selrefs) 426 sect_type = eSectionTypeDataCStringPointers; 427 else if (section_name == g_sect_name_objc_msgrefs) 428 sect_type = eSectionTypeDataObjCMessageRefs; 429 else if (section_name == g_sect_name_eh_frame) 430 sect_type = eSectionTypeEHFrame; 431 else if (section_name == g_sect_name_cfstring) 432 sect_type = eSectionTypeDataObjCCFStrings; 433 else if (section_name == g_sect_name_objc_data || 434 section_name == g_sect_name_objc_classrefs || 435 section_name == g_sect_name_objc_superrefs || 436 section_name == g_sect_name_objc_const || 437 section_name == g_sect_name_objc_classlist) 438 { 439 sect_type = eSectionTypeDataPointers; 440 } 441 442 if (sect_type == eSectionTypeOther) 443 { 444 switch (mach_sect_type) 445 { 446 // TODO: categorize sections by other flags for regular sections 447 case SectionTypeRegular: 448 if (segment_sp->GetName() == g_sect_name_TEXT) 449 sect_type = eSectionTypeCode; 450 else if (segment_sp->GetName() == g_sect_name_DATA) 451 sect_type = eSectionTypeData; 452 else 453 sect_type = eSectionTypeOther; 454 break; 455 case SectionTypeZeroFill: sect_type = eSectionTypeZeroFill; break; 456 case SectionTypeCStringLiterals: sect_type = eSectionTypeDataCString; break; // section with only literal C strings 457 case SectionType4ByteLiterals: sect_type = eSectionTypeData4; break; // section with only 4 byte literals 458 case SectionType8ByteLiterals: sect_type = eSectionTypeData8; break; // section with only 8 byte literals 459 case SectionTypeLiteralPointers: sect_type = eSectionTypeDataPointers; break; // section with only pointers to literals 460 case SectionTypeNonLazySymbolPointers: sect_type = eSectionTypeDataPointers; break; // section with only non-lazy symbol pointers 461 case SectionTypeLazySymbolPointers: sect_type = eSectionTypeDataPointers; break; // section with only lazy symbol pointers 462 case SectionTypeSymbolStubs: sect_type = eSectionTypeCode; break; // section with only symbol stubs, byte size of stub in the reserved2 field 463 case SectionTypeModuleInitFunctionPointers: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for initialization 464 case SectionTypeModuleTermFunctionPointers: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for termination 465 case SectionTypeCoalesced: sect_type = eSectionTypeOther; break; 466 case SectionTypeZeroFillLarge: sect_type = eSectionTypeZeroFill; break; 467 case SectionTypeInterposing: sect_type = eSectionTypeCode; break; // section with only pairs of function pointers for interposing 468 case SectionType16ByteLiterals: sect_type = eSectionTypeData16; break; // section with only 16 byte literals 469 case SectionTypeDTraceObjectFormat: sect_type = eSectionTypeDebug; break; 470 case SectionTypeLazyDylibSymbolPointers: sect_type = eSectionTypeDataPointers; break; 471 default: break; 472 } 473 } 474 475 SectionSP section_sp(new Section(segment_sp.get(), 476 GetModule(), 477 ++sectID, 478 section_name, 479 sect_type, 480 sect64.addr - segment_sp->GetFileAddress(), 481 sect64.size, 482 sect64.offset, 483 sect64.offset == 0 ? 0 : sect64.size, 484 sect64.flags)); 485 segment_sp->GetChildren().AddSection(section_sp); 486 487 if (segment_sp->IsFake()) 488 { 489 segment_sp.reset(); 490 segment_name.Clear(); 491 } 492 } 493 if (m_header.filetype == HeaderFileTypeDSYM) 494 { 495 if (first_segment_sectID <= sectID) 496 { 497 lldb::user_id_t sect_uid; 498 for (sect_uid = first_segment_sectID; sect_uid <= sectID; ++sect_uid) 499 { 500 SectionSP curr_section_sp(segment_sp->GetChildren().FindSectionByID (sect_uid)); 501 SectionSP next_section_sp; 502 if (sect_uid + 1 <= sectID) 503 next_section_sp = segment_sp->GetChildren().FindSectionByID (sect_uid+1); 504 505 if (curr_section_sp.get()) 506 { 507 if (curr_section_sp->GetByteSize() == 0) 508 { 509 if (next_section_sp.get() != NULL) 510 curr_section_sp->SetByteSize ( next_section_sp->GetFileAddress() - curr_section_sp->GetFileAddress() ); 511 else 512 curr_section_sp->SetByteSize ( load_cmd.vmsize ); 513 } 514 } 515 } 516 } 517 } 518 } 519 } 520 } 521 else if (load_cmd.cmd == LoadCommandDynamicSymtabInfo) 522 { 523 m_dysymtab.cmd = load_cmd.cmd; 524 m_dysymtab.cmdsize = load_cmd.cmdsize; 525 m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 526 } 527 528 offset = load_cmd_offset + load_cmd.cmdsize; 529 } 530// if (dump_sections) 531// { 532// StreamFile s(stdout); 533// m_sections_ap->Dump(&s, true); 534// } 535 return sectID; // Return the number of sections we registered with the module 536} 537 538class MachSymtabSectionInfo 539{ 540public: 541 542 MachSymtabSectionInfo (SectionList *section_list) : 543 m_section_list (section_list), 544 m_section_infos() 545 { 546 // Get the number of sections down to a depth of 1 to include 547 // all segments and their sections, but no other sections that 548 // may be added for debug map or 549 m_section_infos.resize(section_list->GetNumSections(1)); 550 } 551 552 553 Section * 554 GetSection (uint8_t n_sect, addr_t file_addr) 555 { 556 if (n_sect == 0) 557 return NULL; 558 if (n_sect < m_section_infos.size()) 559 { 560 if (m_section_infos[n_sect].section == NULL) 561 { 562 Section *section = m_section_list->FindSectionByID (n_sect).get(); 563 m_section_infos[n_sect].section = section; 564 assert (section != NULL); 565 m_section_infos[n_sect].vm_range.SetBaseAddress (section->GetFileAddress()); 566 m_section_infos[n_sect].vm_range.SetByteSize (section->GetByteSize()); 567 } 568 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) 569 return m_section_infos[n_sect].section; 570 } 571 return m_section_list->FindSectionContainingFileAddress(file_addr).get(); 572 } 573 574protected: 575 struct SectionInfo 576 { 577 SectionInfo () : 578 vm_range(), 579 section (NULL) 580 { 581 } 582 583 VMRange vm_range; 584 Section *section; 585 }; 586 SectionList *m_section_list; 587 std::vector<SectionInfo> m_section_infos; 588}; 589 590 591 592size_t 593ObjectFileMachO::ParseSymtab (bool minimize) 594{ 595 Timer scoped_timer(__PRETTY_FUNCTION__, 596 "ObjectFileMachO::ParseSymtab () module = %s", 597 m_file.GetFilename().AsCString("")); 598 struct symtab_command symtab_load_command; 599 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 600 uint32_t i; 601 for (i=0; i<m_header.ncmds; ++i) 602 { 603 const uint32_t cmd_offset = offset; 604 // Read in the load command and load command size 605 if (m_data.GetU32(&offset, &symtab_load_command, 2) == NULL) 606 break; 607 // Watch for the symbol table load command 608 if (symtab_load_command.cmd == LoadCommandSymtab) 609 { 610 // Read in the rest of the symtab load command 611 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4)) // fill in symoff, nsyms, stroff, strsize fields 612 { 613 Symtab *symtab = m_symtab_ap.get(); 614 SectionList *section_list = GetSectionList(); 615 assert(section_list); 616 const size_t addr_size = m_data.GetAddressByteSize(); 617 const ByteOrder endian = m_data.GetByteOrder(); 618 bool bit_width_32 = addr_size == 4; 619 const size_t nlist_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 620 621 DataBufferSP symtab_data_sp(m_file.ReadFileContents(m_offset + symtab_load_command.symoff, symtab_load_command.nsyms * nlist_size)); 622 DataBufferSP strtab_data_sp(m_file.ReadFileContents(m_offset + symtab_load_command.stroff, symtab_load_command.strsize)); 623 624 const char *strtab_data = (const char *)strtab_data_sp->GetBytes(); 625// DataExtractor symtab_data(symtab_data_sp, endian, addr_size); 626// DataExtractor strtab_data(strtab_data_sp, endian, addr_size); 627 628 static ConstString g_segment_name_TEXT ("__TEXT"); 629 static ConstString g_segment_name_DATA ("__DATA"); 630 static ConstString g_segment_name_OBJC ("__OBJC"); 631 static ConstString g_section_name_eh_frame ("__eh_frame"); 632 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT)); 633 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA)); 634 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC)); 635 SectionSP eh_frame_section_sp; 636 if (text_section_sp.get()) 637 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame); 638 else 639 eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame); 640 641 uint8_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() : NListSectionNoSection; 642 //uint32_t symtab_offset = 0; 643 const uint8_t* nlist_data = symtab_data_sp->GetBytes(); 644 assert (symtab_data_sp->GetByteSize()/nlist_size >= symtab_load_command.nsyms); 645 646 647 if (endian != eByteOrderHost) 648 { 649 // ... 650 assert (!"UNIMPLEMENTED: Swap all nlist entries"); 651 } 652 uint32_t N_SO_index = UINT32_MAX; 653 654 MachSymtabSectionInfo section_info (section_list); 655 std::vector<uint32_t> N_FUN_indexes; 656 std::vector<uint32_t> N_NSYM_indexes; 657 std::vector<uint32_t> N_INCL_indexes; 658 std::vector<uint32_t> N_BRAC_indexes; 659 std::vector<uint32_t> N_COMM_indexes; 660 typedef std::map <uint64_t, uint32_t> ValueToSymbolIndexMap; 661 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap; 662 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 663 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 664 // Any symbols that get merged into another will get an entry 665 // in this map so we know 666 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 667 uint32_t nlist_idx = 0; 668 Symbol *symbol_ptr = NULL; 669 670 uint32_t sym_idx = 0; 671 Symbol *sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 672 uint32_t num_syms = symtab->GetNumSymbols(); 673 674 //symtab->Reserve (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 675 for (nlist_idx = 0; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) 676 { 677 struct nlist_64 nlist; 678 if (bit_width_32) 679 { 680 struct nlist* nlist32_ptr = (struct nlist*)(nlist_data + (nlist_idx * nlist_size)); 681 nlist.n_strx = nlist32_ptr->n_strx; 682 nlist.n_type = nlist32_ptr->n_type; 683 nlist.n_sect = nlist32_ptr->n_sect; 684 nlist.n_desc = nlist32_ptr->n_desc; 685 nlist.n_value = nlist32_ptr->n_value; 686 } 687 else 688 { 689 nlist = *((struct nlist_64*)(nlist_data + (nlist_idx * nlist_size))); 690 } 691 692 SymbolType type = eSymbolTypeInvalid; 693 const char* symbol_name = &strtab_data[nlist.n_strx]; 694 if (symbol_name[0] == '\0') 695 symbol_name = NULL; 696 Section* symbol_section = NULL; 697 bool add_nlist = true; 698 bool is_debug = ((nlist.n_type & NlistMaskStab) != 0); 699 700 assert (sym_idx < num_syms); 701 702 sym[sym_idx].SetDebug (is_debug); 703 704 if (is_debug) 705 { 706 switch (nlist.n_type) 707 { 708 case StabGlobalSymbol: 709 // N_GSYM -- global symbol: name,,NO_SECT,type,0 710 // Sometimes the N_GSYM value contains the address. 711 sym[sym_idx].SetExternal(true); 712 if (nlist.n_value != 0) 713 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 714 type = eSymbolTypeData; 715 break; 716 717 case StabFunctionName: 718 // N_FNAME -- procedure name (f77 kludge): name,,NO_SECT,0,0 719 type = eSymbolTypeCompiler; 720 break; 721 722 case StabFunction: 723 // N_FUN -- procedure: name,,n_sect,linenumber,address 724 if (symbol_name) 725 { 726 type = eSymbolTypeCode; 727 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 728 729 N_FUN_addr_to_sym_idx[nlist.n_value] = sym_idx; 730 // We use the current number of symbols in the symbol table in lieu of 731 // using nlist_idx in case we ever start trimming entries out 732 N_FUN_indexes.push_back(sym_idx); 733 } 734 else 735 { 736 type = eSymbolTypeCompiler; 737 738 if ( !N_FUN_indexes.empty() ) 739 { 740 // Copy the size of the function into the original STAB entry so we don't have 741 // to hunt for it later 742 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 743 N_FUN_indexes.pop_back(); 744 // We don't really need the end function STAB as it contains the size which 745 // we already placed with the original symbol, so don't add it if we want a 746 // minimal symbol table 747 if (minimize) 748 add_nlist = false; 749 } 750 } 751 break; 752 753 case StabStaticSymbol: 754 // N_STSYM -- static symbol: name,,n_sect,type,address 755 N_STSYM_addr_to_sym_idx[nlist.n_value] = sym_idx; 756 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 757 type = eSymbolTypeData; 758 break; 759 760 case StabLocalCommon: 761 // N_LCSYM -- .lcomm symbol: name,,n_sect,type,address 762 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 763 type = eSymbolTypeCommonBlock; 764 break; 765 766 case StabBeginSymbol: 767 // N_BNSYM 768 // We use the current number of symbols in the symbol table in lieu of 769 // using nlist_idx in case we ever start trimming entries out 770 if (minimize) 771 { 772 // Skip these if we want minimal symbol tables 773 add_nlist = false; 774 } 775 else 776 { 777 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 778 N_NSYM_indexes.push_back(sym_idx); 779 type = eSymbolTypeScopeBegin; 780 } 781 break; 782 783 case StabEndSymbol: 784 // N_ENSYM 785 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 786 // so that we can always skip the entire symbol if we need to navigate 787 // more quickly at the source level when parsing STABS 788 if (minimize) 789 { 790 // Skip these if we want minimal symbol tables 791 add_nlist = false; 792 } 793 else 794 { 795 if ( !N_NSYM_indexes.empty() ) 796 { 797 symbol_ptr = symtab->SymbolAtIndex(N_NSYM_indexes.back()); 798 symbol_ptr->SetByteSize(sym_idx + 1); 799 symbol_ptr->SetSizeIsSibling(true); 800 N_NSYM_indexes.pop_back(); 801 } 802 type = eSymbolTypeScopeEnd; 803 } 804 break; 805 806 807 case StabSourceFileOptions: 808 // N_OPT - emitted with gcc2_compiled and in gcc source 809 type = eSymbolTypeCompiler; 810 break; 811 812 case StabRegisterSymbol: 813 // N_RSYM - register sym: name,,NO_SECT,type,register 814 type = eSymbolTypeVariable; 815 break; 816 817 case StabSourceLine: 818 // N_SLINE - src line: 0,,n_sect,linenumber,address 819 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 820 type = eSymbolTypeLineEntry; 821 break; 822 823 case StabStructureType: 824 // N_SSYM - structure elt: name,,NO_SECT,type,struct_offset 825 type = eSymbolTypeVariableType; 826 break; 827 828 case StabSourceFileName: 829 // N_SO - source file name 830 type = eSymbolTypeSourceFile; 831 if (symbol_name == NULL) 832 { 833 if (minimize) 834 add_nlist = false; 835 if (N_SO_index != UINT32_MAX) 836 { 837 // Set the size of the N_SO to the terminating index of this N_SO 838 // so that we can always skip the entire N_SO if we need to navigate 839 // more quickly at the source level when parsing STABS 840 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 841 symbol_ptr->SetByteSize(sym_idx + (minimize ? 0 : 1)); 842 symbol_ptr->SetSizeIsSibling(true); 843 } 844 N_NSYM_indexes.clear(); 845 N_INCL_indexes.clear(); 846 N_BRAC_indexes.clear(); 847 N_COMM_indexes.clear(); 848 N_FUN_indexes.clear(); 849 N_SO_index = UINT32_MAX; 850 } 851 else 852 { 853 // We use the current number of symbols in the symbol table in lieu of 854 // using nlist_idx in case we ever start trimming entries out 855 if (symbol_name[0] == '/') 856 N_SO_index = sym_idx; 857 else if (minimize && (N_SO_index == sym_idx - 1)) 858 { 859 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 860 if (so_path && so_path[0]) 861 { 862 std::string full_so_path (so_path); 863 if (*full_so_path.rbegin() != '/') 864 full_so_path += '/'; 865 full_so_path += symbol_name; 866 sym[sym_idx - 1].GetMangled().SetValue(full_so_path.c_str(), false); 867 add_nlist = false; 868 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 869 } 870 } 871 } 872 873 break; 874 875 case StabObjectFileName: 876 // N_OSO - object file name: name,,0,0,st_mtime 877 type = eSymbolTypeObjectFile; 878 break; 879 880 case StabLocalSymbol: 881 // N_LSYM - local sym: name,,NO_SECT,type,offset 882 type = eSymbolTypeLocal; 883 break; 884 885 //---------------------------------------------------------------------- 886 // INCL scopes 887 //---------------------------------------------------------------------- 888 case StabBeginIncludeFileName: 889 // N_BINCL - include file beginning: name,,NO_SECT,0,sum 890 // We use the current number of symbols in the symbol table in lieu of 891 // using nlist_idx in case we ever start trimming entries out 892 N_INCL_indexes.push_back(sym_idx); 893 type = eSymbolTypeScopeBegin; 894 break; 895 896 case StabEndIncludeFile: 897 // N_EINCL - include file end: name,,NO_SECT,0,0 898 // Set the size of the N_BINCL to the terminating index of this N_EINCL 899 // so that we can always skip the entire symbol if we need to navigate 900 // more quickly at the source level when parsing STABS 901 if ( !N_INCL_indexes.empty() ) 902 { 903 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 904 symbol_ptr->SetByteSize(sym_idx + 1); 905 symbol_ptr->SetSizeIsSibling(true); 906 N_INCL_indexes.pop_back(); 907 } 908 type = eSymbolTypeScopeEnd; 909 break; 910 911 case StabIncludeFileName: 912 // N_SOL - #included file name: name,,n_sect,0,address 913 type = eSymbolTypeHeaderFile; 914 915 // We currently don't use the header files on darwin 916 if (minimize) 917 add_nlist = false; 918 break; 919 920 case StabCompilerParameters: 921 // N_PARAMS - compiler parameters: name,,NO_SECT,0,0 922 type = eSymbolTypeCompiler; 923 break; 924 925 case StabCompilerVersion: 926 // N_VERSION - compiler version: name,,NO_SECT,0,0 927 type = eSymbolTypeCompiler; 928 break; 929 930 case StabCompilerOptLevel: 931 // N_OLEVEL - compiler -O level: name,,NO_SECT,0,0 932 type = eSymbolTypeCompiler; 933 break; 934 935 case StabParameter: 936 // N_PSYM - parameter: name,,NO_SECT,type,offset 937 type = eSymbolTypeVariable; 938 break; 939 940 case StabAlternateEntry: 941 // N_ENTRY - alternate entry: name,,n_sect,linenumber,address 942 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 943 type = eSymbolTypeLineEntry; 944 break; 945 946 //---------------------------------------------------------------------- 947 // Left and Right Braces 948 //---------------------------------------------------------------------- 949 case StabLeftBracket: 950 // N_LBRAC - left bracket: 0,,NO_SECT,nesting level,address 951 // We use the current number of symbols in the symbol table in lieu of 952 // using nlist_idx in case we ever start trimming entries out 953 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 954 N_BRAC_indexes.push_back(sym_idx); 955 type = eSymbolTypeScopeBegin; 956 break; 957 958 case StabRightBracket: 959 // N_RBRAC - right bracket: 0,,NO_SECT,nesting level,address 960 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 961 // so that we can always skip the entire symbol if we need to navigate 962 // more quickly at the source level when parsing STABS 963 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 964 if ( !N_BRAC_indexes.empty() ) 965 { 966 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 967 symbol_ptr->SetByteSize(sym_idx + 1); 968 symbol_ptr->SetSizeIsSibling(true); 969 N_BRAC_indexes.pop_back(); 970 } 971 type = eSymbolTypeScopeEnd; 972 break; 973 974 case StabDeletedIncludeFile: 975 // N_EXCL - deleted include file: name,,NO_SECT,0,sum 976 type = eSymbolTypeHeaderFile; 977 break; 978 979 //---------------------------------------------------------------------- 980 // COMM scopes 981 //---------------------------------------------------------------------- 982 case StabBeginCommon: 983 // N_BCOMM - begin common: name,,NO_SECT,0,0 984 // We use the current number of symbols in the symbol table in lieu of 985 // using nlist_idx in case we ever start trimming entries out 986 type = eSymbolTypeScopeBegin; 987 N_COMM_indexes.push_back(sym_idx); 988 break; 989 990 case StabEndCommonLocal: 991 // N_ECOML - end common (local name): 0,,n_sect,0,address 992 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 993 // Fall through 994 995 case StabEndCommon: 996 // N_ECOMM - end common: name,,n_sect,0,0 997 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 998 // so that we can always skip the entire symbol if we need to navigate 999 // more quickly at the source level when parsing STABS 1000 if ( !N_COMM_indexes.empty() ) 1001 { 1002 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 1003 symbol_ptr->SetByteSize(sym_idx + 1); 1004 symbol_ptr->SetSizeIsSibling(true); 1005 N_COMM_indexes.pop_back(); 1006 } 1007 type = eSymbolTypeScopeEnd; 1008 break; 1009 1010 case StabLength: 1011 // N_LENG - second stab entry with length information 1012 type = eSymbolTypeAdditional; 1013 break; 1014 1015 default: break; 1016 } 1017 } 1018 else 1019 { 1020 //uint8_t n_pext = NlistMaskPrivateExternal & nlist.n_type; 1021 uint8_t n_type = NlistMaskType & nlist.n_type; 1022 sym[sym_idx].SetExternal((NlistMaskExternal & nlist.n_type) != 0); 1023 1024 if (symbol_name && ::strstr (symbol_name, ".objc") == symbol_name) 1025 { 1026 type = eSymbolTypeRuntime; 1027 } 1028 else 1029 { 1030 switch (n_type) 1031 { 1032 case NListTypeIndirect: // N_INDR - Fall through 1033 case NListTypePreboundUndefined:// N_PBUD - Fall through 1034 case NListTypeUndefined: // N_UNDF 1035 type = eSymbolTypeExtern; 1036 break; 1037 1038 case NListTypeAbsolute: // N_ABS 1039 type = eSymbolTypeAbsolute; 1040 break; 1041 1042 case NListTypeSection: // N_SECT 1043 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 1044 1045 assert(symbol_section != NULL); 1046 if (TEXT_eh_frame_sectID == nlist.n_sect) 1047 { 1048 type = eSymbolTypeException; 1049 } 1050 else 1051 { 1052 uint32_t section_type = symbol_section->GetAllFlagBits() & SectionFlagMaskSectionType; 1053 1054 switch (section_type) 1055 { 1056 case SectionTypeRegular: break; // regular section 1057 //case SectionTypeZeroFill: type = eSymbolTypeData; break; // zero fill on demand section 1058 case SectionTypeCStringLiterals: type = eSymbolTypeData; break; // section with only literal C strings 1059 case SectionType4ByteLiterals: type = eSymbolTypeData; break; // section with only 4 byte literals 1060 case SectionType8ByteLiterals: type = eSymbolTypeData; break; // section with only 8 byte literals 1061 case SectionTypeLiteralPointers: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 1062 case SectionTypeNonLazySymbolPointers: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 1063 case SectionTypeLazySymbolPointers: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 1064 case SectionTypeSymbolStubs: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 1065 case SectionTypeModuleInitFunctionPointers: type = eSymbolTypeCode; break; // section with only function pointers for initialization 1066 case SectionTypeModuleTermFunctionPointers: type = eSymbolTypeCode; break; // section with only function pointers for termination 1067 //case SectionTypeCoalesced: type = eSymbolType; break; // section contains symbols that are to be coalesced 1068 //case SectionTypeZeroFillLarge: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 1069 case SectionTypeInterposing: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 1070 case SectionType16ByteLiterals: type = eSymbolTypeData; break; // section with only 16 byte literals 1071 case SectionTypeDTraceObjectFormat: type = eSymbolTypeInstrumentation; break; 1072 case SectionTypeLazyDylibSymbolPointers: type = eSymbolTypeTrampoline; break; 1073 default: break; 1074 } 1075 1076 if (type == eSymbolTypeInvalid) 1077 { 1078 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 1079 if (symbol_section->IsDescendant (text_section_sp.get())) 1080 { 1081 if (symbol_section->IsClear(SectionAttrUserPureInstructions | 1082 SectionAttrUserSelfModifyingCode | 1083 SectionAttrSytemSomeInstructions)) 1084 type = eSymbolTypeData; 1085 else 1086 type = eSymbolTypeCode; 1087 } 1088 else 1089 if (symbol_section->IsDescendant(data_section_sp.get())) 1090 { 1091 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 1092 { 1093 type = eSymbolTypeRuntime; 1094 } 1095 else 1096 if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 1097 { 1098 type = eSymbolTypeException; 1099 } 1100 else 1101 { 1102 type = eSymbolTypeData; 1103 } 1104 } 1105 else 1106 if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 1107 { 1108 type = eSymbolTypeTrampoline; 1109 } 1110 else 1111 if (symbol_section->IsDescendant(objc_section_sp.get())) 1112 { 1113 type = eSymbolTypeRuntime; 1114 } 1115 } 1116 } 1117 break; 1118 } 1119 } 1120 } 1121 if (add_nlist) 1122 { 1123 bool symbol_name_is_mangled = false; 1124 if (symbol_name && symbol_name[0] == '_') 1125 { 1126 symbol_name_is_mangled = symbol_name[1] == '_'; 1127 symbol_name++; // Skip the leading underscore 1128 } 1129 uint64_t symbol_value = nlist.n_value; 1130 1131 if (symbol_name) 1132 sym[sym_idx].GetMangled().SetValue(symbol_name, symbol_name_is_mangled); 1133 if (is_debug == false) 1134 { 1135 if (type == eSymbolTypeCode) 1136 { 1137 // See if we can find a N_FUN entry for any code symbols. 1138 // If we do find a match, and the name matches, then we 1139 // can merge the two into just the function symbol to avoid 1140 // duplicate entries in the symbol table 1141 ValueToSymbolIndexMap::const_iterator pos = N_FUN_addr_to_sym_idx.find (nlist.n_value); 1142 if (pos != N_FUN_addr_to_sym_idx.end()) 1143 { 1144 if ((symbol_name_is_mangled == true && sym[sym_idx].GetMangled().GetMangledName() == sym[pos->second].GetMangled().GetMangledName()) || 1145 (symbol_name_is_mangled == false && sym[sym_idx].GetMangled().GetDemangledName() == sym[pos->second].GetMangled().GetDemangledName())) 1146 { 1147 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 1148 // We just need the flags from the linker symbol, so put these flags 1149 // into the N_FUN flags to avoid duplicate symbols in the symbol table 1150 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 1151 sym[sym_idx].Clear(); 1152 continue; 1153 } 1154 } 1155 } 1156 else if (type == eSymbolTypeData) 1157 { 1158 // See if we can find a N_STSYM entry for any data symbols. 1159 // If we do find a match, and the name matches, then we 1160 // can merge the two into just the Static symbol to avoid 1161 // duplicate entries in the symbol table 1162 ValueToSymbolIndexMap::const_iterator pos = N_STSYM_addr_to_sym_idx.find (nlist.n_value); 1163 if (pos != N_STSYM_addr_to_sym_idx.end()) 1164 { 1165 if ((symbol_name_is_mangled == true && sym[sym_idx].GetMangled().GetMangledName() == sym[pos->second].GetMangled().GetMangledName()) || 1166 (symbol_name_is_mangled == false && sym[sym_idx].GetMangled().GetDemangledName() == sym[pos->second].GetMangled().GetDemangledName())) 1167 { 1168 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 1169 // We just need the flags from the linker symbol, so put these flags 1170 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 1171 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 1172 sym[sym_idx].Clear(); 1173 continue; 1174 } 1175 } 1176 } 1177 } 1178 if (symbol_section != NULL) 1179 symbol_value -= symbol_section->GetFileAddress(); 1180 1181 sym[sym_idx].SetID (nlist_idx); 1182 sym[sym_idx].SetType (type); 1183 sym[sym_idx].GetAddressRangeRef().GetBaseAddress().SetSection (symbol_section); 1184 sym[sym_idx].GetAddressRangeRef().GetBaseAddress().SetOffset (symbol_value); 1185 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 1186 1187 ++sym_idx; 1188 } 1189 else 1190 { 1191 sym[sym_idx].Clear(); 1192 } 1193 1194 } 1195 1196 // STAB N_GSYM entries end up having a symbol type eSymbolTypeGlobal and when the symbol value 1197 // is zero, the address of the global ends up being in a non-STAB entry. Try and fix up all 1198 // such entries by figuring out what the address for the global is by looking up this non-STAB 1199 // entry and copying the value into the debug symbol's value to save us the hassle in the 1200 // debug symbol parser. 1201 1202 Symbol *global_symbol = NULL; 1203 for (nlist_idx = 0; 1204 nlist_idx < symtab_load_command.nsyms && (global_symbol = symtab->FindSymbolWithType (eSymbolTypeData, Symtab::eDebugYes, Symtab::eVisibilityAny, nlist_idx)) != NULL; 1205 nlist_idx++) 1206 { 1207 if (global_symbol->GetValue().GetFileAddress() == 0) 1208 { 1209 std::vector<uint32_t> indexes; 1210 if (symtab->AppendSymbolIndexesWithName (global_symbol->GetMangled().GetName(), indexes) > 0) 1211 { 1212 std::vector<uint32_t>::const_iterator pos; 1213 std::vector<uint32_t>::const_iterator end = indexes.end(); 1214 for (pos = indexes.begin(); pos != end; ++pos) 1215 { 1216 symbol_ptr = symtab->SymbolAtIndex(*pos); 1217 if (symbol_ptr != global_symbol && symbol_ptr->IsDebug() == false) 1218 { 1219 global_symbol->SetValue(symbol_ptr->GetValue()); 1220 break; 1221 } 1222 } 1223 } 1224 } 1225 } 1226 1227 // Trim our symbols down to just what we ended up with after 1228 // removing any symbols. 1229 if (sym_idx < num_syms) 1230 { 1231 num_syms = sym_idx; 1232 sym = symtab->Resize (num_syms); 1233 } 1234 1235 // Now synthesize indirect symbols 1236 if (m_dysymtab.nindirectsyms != 0) 1237 { 1238 DataBufferSP indirect_symbol_indexes_sp(m_file.ReadFileContents(m_offset + m_dysymtab.indirectsymoff, m_dysymtab.nindirectsyms * 4)); 1239 1240 if (indirect_symbol_indexes_sp && indirect_symbol_indexes_sp->GetByteSize()) 1241 { 1242 NListIndexToSymbolIndexMap::const_iterator end_index_pos = m_nlist_idx_to_sym_idx.end(); 1243 DataExtractor indirect_symbol_index_data (indirect_symbol_indexes_sp, m_data.GetByteOrder(), m_data.GetAddressByteSize()); 1244 1245 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); ++sect_idx) 1246 { 1247 if ((m_mach_sections[sect_idx].flags & SectionFlagMaskSectionType) == SectionTypeSymbolStubs) 1248 { 1249 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 1250 if (symbol_stub_byte_size == 0) 1251 continue; 1252 1253 const uint32_t num_symbol_stubs = m_mach_sections[sect_idx].size / symbol_stub_byte_size; 1254 1255 if (num_symbol_stubs == 0) 1256 continue; 1257 1258 const uint32_t symbol_stub_index_offset = m_mach_sections[sect_idx].reserved1; 1259 uint32_t synthetic_stub_sym_id = symtab_load_command.nsyms; 1260 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) 1261 { 1262 const uint32_t symbol_stub_index = symbol_stub_index_offset + stub_idx; 1263 const lldb::addr_t symbol_stub_addr = m_mach_sections[sect_idx].addr + (stub_idx * symbol_stub_byte_size); 1264 uint32_t symbol_stub_offset = symbol_stub_index * 4; 1265 if (indirect_symbol_index_data.ValidOffsetForDataOfSize(symbol_stub_offset, 4)) 1266 { 1267 const uint32_t stub_sym_id = indirect_symbol_index_data.GetU32 (&symbol_stub_offset); 1268 if (stub_sym_id & (IndirectSymbolAbsolute | IndirectSymbolLocal)) 1269 continue; 1270 1271 NListIndexToSymbolIndexMap::const_iterator index_pos = m_nlist_idx_to_sym_idx.find (stub_sym_id); 1272 Symbol *stub_symbol = NULL; 1273 if (index_pos != end_index_pos) 1274 { 1275 // We have a remapping from the original nlist index to 1276 // a current symbol index, so just look this up by index 1277 stub_symbol = symtab->SymbolAtIndex (index_pos->second); 1278 } 1279 else 1280 { 1281 // We need to lookup a symbol using the original nlist 1282 // symbol index since this index is coming from the 1283 // S_SYMBOL_STUBS 1284 stub_symbol = symtab->FindSymbolByID (stub_sym_id); 1285 } 1286 1287 assert (stub_symbol); 1288 if (stub_symbol) 1289 { 1290 Address so_addr(symbol_stub_addr, section_list); 1291 1292 if (stub_symbol->GetType() == eSymbolTypeExtern) 1293 { 1294 // Change the external symbol into a trampoline that makes sense 1295 // These symbols were N_UNDF N_EXT, and are useless to us, so we 1296 // can re-use them so we don't have to make up a synthetic symbol 1297 // for no good reason. 1298 stub_symbol->SetType (eSymbolTypeTrampoline); 1299 stub_symbol->SetExternal (false); 1300 stub_symbol->GetAddressRangeRef().GetBaseAddress() = so_addr; 1301 stub_symbol->GetAddressRangeRef().SetByteSize (symbol_stub_byte_size); 1302 } 1303 else 1304 { 1305 // Make a synthetic symbol to describe the trampoline stub 1306 if (sym_idx >= num_syms) 1307 sym = symtab->Resize (++num_syms); 1308 sym[sym_idx].SetID (synthetic_stub_sym_id++); 1309 sym[sym_idx].GetMangled() = stub_symbol->GetMangled(); 1310 sym[sym_idx].SetType (eSymbolTypeTrampoline); 1311 sym[sym_idx].SetIsSynthetic (true); 1312 sym[sym_idx].GetAddressRangeRef().GetBaseAddress() = so_addr; 1313 sym[sym_idx].GetAddressRangeRef().SetByteSize (symbol_stub_byte_size); 1314 ++sym_idx; 1315 } 1316 } 1317 } 1318 } 1319 } 1320 } 1321 } 1322 } 1323 1324 return symtab->GetNumSymbols(); 1325 } 1326 } 1327 offset = cmd_offset + symtab_load_command.cmdsize; 1328 } 1329 return 0; 1330} 1331 1332 1333void 1334ObjectFileMachO::Dump (Stream *s) 1335{ 1336 lldb_private::Mutex::Locker locker(m_mutex); 1337 s->Printf("%.*p: ", (int)sizeof(void*) * 2, this); 1338 s->Indent(); 1339 if (m_header.magic == HeaderMagic64 || m_header.magic == HeaderMagic64Swapped) 1340 s->PutCString("ObjectFileMachO64"); 1341 else 1342 s->PutCString("ObjectFileMachO32"); 1343 1344 ArchSpec header_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype); 1345 1346 *s << ", file = '" << m_file << "', arch = " << header_arch.AsCString() << "\n"; 1347 1348 if (m_sections_ap.get()) 1349 m_sections_ap->Dump(s, NULL, true); 1350 1351 if (m_symtab_ap.get()) 1352 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 1353} 1354 1355 1356bool 1357ObjectFileMachO::GetUUID (UUID* uuid) 1358{ 1359 lldb_private::Mutex::Locker locker(m_mutex); 1360 struct uuid_command load_cmd; 1361 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 1362 uint32_t i; 1363 for (i=0; i<m_header.ncmds; ++i) 1364 { 1365 const uint32_t cmd_offset = offset; 1366 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1367 break; 1368 1369 if (load_cmd.cmd == LoadCommandUUID) 1370 { 1371 const uint8_t *uuid_bytes = m_data.PeekData(offset, 16); 1372 if (uuid_bytes) 1373 { 1374 uuid->SetBytes (uuid_bytes); 1375 return true; 1376 } 1377 return false; 1378 } 1379 offset = cmd_offset + load_cmd.cmdsize; 1380 } 1381 return false; 1382} 1383 1384 1385uint32_t 1386ObjectFileMachO::GetDependentModules (FileSpecList& files) 1387{ 1388 lldb_private::Mutex::Locker locker(m_mutex); 1389 struct load_command load_cmd; 1390 uint32_t offset = MachHeaderSizeFromMagic(m_header.magic); 1391 uint32_t count = 0; 1392 uint32_t i; 1393 for (i=0; i<m_header.ncmds; ++i) 1394 { 1395 const uint32_t cmd_offset = offset; 1396 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1397 break; 1398 1399 switch (load_cmd.cmd) 1400 { 1401 case LoadCommandDylibLoad: 1402 case LoadCommandDylibLoadWeak: 1403 case LoadCommandDylibReexport: 1404 case LoadCommandDynamicLinkerLoad: 1405 case LoadCommandFixedVMShlibLoad: 1406 case LoadCommandDylibLoadUpward: 1407 { 1408 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 1409 const char *path = m_data.PeekCStr(name_offset); 1410 // Skip any path that starts with '@' since these are usually: 1411 // @executable_path/.../file 1412 // @rpath/.../file 1413 if (path && path[0] != '@') 1414 { 1415 FileSpec file_spec(path); 1416 if (files.AppendIfUnique(file_spec)) 1417 count++; 1418 } 1419 } 1420 break; 1421 1422 default: 1423 break; 1424 } 1425 offset = cmd_offset + load_cmd.cmdsize; 1426 } 1427 return count; 1428} 1429 1430bool 1431ObjectFileMachO::GetTargetTriple (ConstString &target_triple) 1432{ 1433 lldb_private::Mutex::Locker locker(m_mutex); 1434 std::string triple(GetModule()->GetArchitecture().AsCString()); 1435 triple += "-apple-darwin"; 1436 target_triple.SetCString(triple.c_str()); 1437 if (target_triple) 1438 return true; 1439 return false; 1440} 1441 1442 1443//------------------------------------------------------------------ 1444// PluginInterface protocol 1445//------------------------------------------------------------------ 1446const char * 1447ObjectFileMachO::GetPluginName() 1448{ 1449 return "ObjectFileMachO"; 1450} 1451 1452const char * 1453ObjectFileMachO::GetShortPluginName() 1454{ 1455 return GetPluginNameStatic(); 1456} 1457 1458uint32_t 1459ObjectFileMachO::GetPluginVersion() 1460{ 1461 return 1; 1462} 1463 1464void 1465ObjectFileMachO::GetPluginCommandHelp (const char *command, Stream *strm) 1466{ 1467} 1468 1469Error 1470ObjectFileMachO::ExecutePluginCommand (Args &command, Stream *strm) 1471{ 1472 Error error; 1473 error.SetErrorString("No plug-in command are currently supported."); 1474 return error; 1475} 1476 1477Log * 1478ObjectFileMachO::EnablePluginLogging (Stream *strm, Args &command) 1479{ 1480 return NULL; 1481} 1482 1483 1484 1485 1486