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 eSectionTypeDWARFDebugNames:       return eAddressClassDebug;
259                     case eSectionTypeDWARFDebugTypes:       return eAddressClassDebug;
260                     case eSectionTypeEHFrame:               return eAddressClassRuntime;
261                     case eSectionTypeOther:                 return eAddressClassUnknown;
262                     }
263                 }
264             }
265 
266             const SymbolType symbol_type = symbol->GetType();
267             switch (symbol_type)
268             {
269             case eSymbolTypeAny:            return eAddressClassUnknown;
270             case eSymbolTypeAbsolute:       return eAddressClassUnknown;
271             case eSymbolTypeExtern:         return eAddressClassUnknown;
272 
273             case eSymbolTypeCode:
274             case eSymbolTypeTrampoline:
275                 if (m_header.cputype == llvm::MachO::CPUTypeARM)
276                 {
277                     // For ARM we have a bit in the n_desc field of the symbol
278                     // that tells us ARM/Thumb which is bit 0x0008.
279                     if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB)
280                         return eAddressClassCodeAlternateISA;
281                 }
282                 return eAddressClassCode;
283 
284             case eSymbolTypeData:           return eAddressClassData;
285             case eSymbolTypeRuntime:        return eAddressClassRuntime;
286             case eSymbolTypeException:      return eAddressClassRuntime;
287             case eSymbolTypeSourceFile:     return eAddressClassDebug;
288             case eSymbolTypeHeaderFile:     return eAddressClassDebug;
289             case eSymbolTypeObjectFile:     return eAddressClassDebug;
290             case eSymbolTypeCommonBlock:    return eAddressClassDebug;
291             case eSymbolTypeBlock:          return eAddressClassDebug;
292             case eSymbolTypeLocal:          return eAddressClassData;
293             case eSymbolTypeParam:          return eAddressClassData;
294             case eSymbolTypeVariable:       return eAddressClassData;
295             case eSymbolTypeVariableType:   return eAddressClassDebug;
296             case eSymbolTypeLineEntry:      return eAddressClassDebug;
297             case eSymbolTypeLineHeader:     return eAddressClassDebug;
298             case eSymbolTypeScopeBegin:     return eAddressClassDebug;
299             case eSymbolTypeScopeEnd:       return eAddressClassDebug;
300             case eSymbolTypeAdditional:     return eAddressClassUnknown;
301             case eSymbolTypeCompiler:       return eAddressClassDebug;
302             case eSymbolTypeInstrumentation:return eAddressClassDebug;
303             case eSymbolTypeUndefined:      return eAddressClassUnknown;
304             }
305         }
306     }
307     return eAddressClassUnknown;
308 }
309 
310 Symtab *
311 ObjectFileMachO::GetSymtab()
312 {
313     lldb_private::Mutex::Locker symfile_locker(m_mutex);
314     if (m_symtab_ap.get() == NULL)
315     {
316         m_symtab_ap.reset(new Symtab(this));
317         Mutex::Locker symtab_locker (m_symtab_ap->GetMutex());
318         ParseSymtab (true);
319     }
320     return m_symtab_ap.get();
321 }
322 
323 
324 SectionList *
325 ObjectFileMachO::GetSectionList()
326 {
327     lldb_private::Mutex::Locker locker(m_mutex);
328     if (m_sections_ap.get() == NULL)
329     {
330         m_sections_ap.reset(new SectionList());
331         ParseSections();
332     }
333     return m_sections_ap.get();
334 }
335 
336 
337 size_t
338 ObjectFileMachO::ParseSections ()
339 {
340     lldb::user_id_t segID = 0;
341     lldb::user_id_t sectID = 0;
342     struct segment_command_64 load_cmd;
343     uint32_t offset = MachHeaderSizeFromMagic(m_header.magic);
344     uint32_t i;
345     //bool dump_sections = false;
346     for (i=0; i<m_header.ncmds; ++i)
347     {
348         const uint32_t load_cmd_offset = offset;
349         if (m_data.GetU32(&offset, &load_cmd, 2) == NULL)
350             break;
351 
352         if (load_cmd.cmd == LoadCommandSegment32 || load_cmd.cmd == LoadCommandSegment64)
353         {
354             if (m_data.GetU8(&offset, (uint8_t*)load_cmd.segname, 16))
355             {
356                 load_cmd.vmaddr = m_data.GetAddress(&offset);
357                 load_cmd.vmsize = m_data.GetAddress(&offset);
358                 load_cmd.fileoff = m_data.GetAddress(&offset);
359                 load_cmd.filesize = m_data.GetAddress(&offset);
360                 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4))
361                 {
362 
363                     const bool segment_is_encrypted = (load_cmd.flags & SegmentCommandFlagBitProtectedVersion1) != 0;
364 
365                     // Keep a list of mach segments around in case we need to
366                     // get at data that isn't stored in the abstracted Sections.
367                     m_mach_segments.push_back (load_cmd);
368 
369                     ConstString segment_name (load_cmd.segname, std::min<int>(strlen(load_cmd.segname), sizeof(load_cmd.segname)));
370                     // Use a segment ID of the segment index shifted left by 8 so they
371                     // never conflict with any of the sections.
372                     SectionSP segment_sp;
373                     if (segment_name)
374                     {
375                         segment_sp.reset(new Section (NULL,
376                                                       GetModule(),            // Module to which this section belongs
377                                                       ++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
378                                                       segment_name,           // Name of this section
379                                                       eSectionTypeContainer,  // This section is a container of other sections.
380                                                       load_cmd.vmaddr,        // File VM address == addresses as they are found in the object file
381                                                       load_cmd.vmsize,        // VM size in bytes of this section
382                                                       load_cmd.fileoff,       // Offset to the data for this section in the file
383                                                       load_cmd.filesize,      // Size in bytes of this section as found in the the file
384                                                       load_cmd.flags));       // Flags for this section
385 
386                         segment_sp->SetIsEncrypted (segment_is_encrypted);
387                         m_sections_ap->AddSection(segment_sp);
388                     }
389 
390                     struct section_64 sect64;
391                     ::memset (&sect64, 0, sizeof(sect64));
392                     // Push a section into our mach sections for the section at
393                     // index zero (NListSectionNoSection) if we don't have any
394                     // mach sections yet...
395                     if (m_mach_sections.empty())
396                         m_mach_sections.push_back(sect64);
397                     uint32_t segment_sect_idx;
398                     const lldb::user_id_t first_segment_sectID = sectID + 1;
399 
400 
401                     const uint32_t num_u32s = load_cmd.cmd == LoadCommandSegment32 ? 7 : 8;
402                     for (segment_sect_idx=0; segment_sect_idx<load_cmd.nsects; ++segment_sect_idx)
403                     {
404                         if (m_data.GetU8(&offset, (uint8_t*)sect64.sectname, sizeof(sect64.sectname)) == NULL)
405                             break;
406                         if (m_data.GetU8(&offset, (uint8_t*)sect64.segname, sizeof(sect64.segname)) == NULL)
407                             break;
408                         sect64.addr = m_data.GetAddress(&offset);
409                         sect64.size = m_data.GetAddress(&offset);
410 
411                         if (m_data.GetU32(&offset, &sect64.offset, num_u32s) == NULL)
412                             break;
413 
414                         // Keep a list of mach sections around in case we need to
415                         // get at data that isn't stored in the abstracted Sections.
416                         m_mach_sections.push_back (sect64);
417 
418                         ConstString section_name (sect64.sectname, std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname)));
419                         if (!segment_name)
420                         {
421                             // We have a segment with no name so we need to conjure up
422                             // segments that correspond to the section's segname if there
423                             // isn't already such a section. If there is such a section,
424                             // we resize the section so that it spans all sections.
425                             // We also mark these sections as fake so address matches don't
426                             // hit if they land in the gaps between the child sections.
427                             segment_name.SetTrimmedCStringWithLength(sect64.segname, sizeof(sect64.segname));
428                             segment_sp = m_sections_ap->FindSectionByName (segment_name);
429                             if (segment_sp.get())
430                             {
431                                 Section *segment = segment_sp.get();
432                                 // Grow the section size as needed.
433                                 const lldb::addr_t sect64_min_addr = sect64.addr;
434                                 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size;
435                                 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize();
436                                 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress();
437                                 const lldb::addr_t curr_seg_max_addr = curr_seg_min_addr + curr_seg_byte_size;
438                                 if (sect64_min_addr >= curr_seg_min_addr)
439                                 {
440                                     const lldb::addr_t new_seg_byte_size = sect64_max_addr - curr_seg_min_addr;
441                                     // Only grow the section size if needed
442                                     if (new_seg_byte_size > curr_seg_byte_size)
443                                         segment->SetByteSize (new_seg_byte_size);
444                                 }
445                                 else
446                                 {
447                                     // We need to change the base address of the segment and
448                                     // adjust the child section offsets for all existing children.
449                                     const lldb::addr_t slide_amount = sect64_min_addr - curr_seg_min_addr;
450                                     segment->Slide(slide_amount, false);
451                                     segment->GetChildren().Slide (-slide_amount, false);
452                                     segment->SetByteSize (curr_seg_max_addr - sect64_min_addr);
453                                 }
454 
455                                 // Grow the section size as needed.
456                                 if (sect64.offset)
457                                 {
458                                     const lldb::addr_t segment_min_file_offset = segment->GetFileOffset();
459                                     const lldb::addr_t segment_max_file_offset = segment_min_file_offset + segment->GetFileSize();
460 
461                                     const lldb::addr_t section_min_file_offset = sect64.offset;
462                                     const lldb::addr_t section_max_file_offset = section_min_file_offset + sect64.size;
463                                     const lldb::addr_t new_file_offset = std::min (section_min_file_offset, segment_min_file_offset);
464                                     const lldb::addr_t new_file_size = std::max (section_max_file_offset, segment_max_file_offset) - new_file_offset;
465                                     segment->SetFileOffset (new_file_offset);
466                                     segment->SetFileSize (new_file_size);
467                                 }
468                             }
469                             else
470                             {
471                                 // Create a fake section for the section's named segment
472                                 segment_sp.reset(new Section(segment_sp.get(),       // Parent section
473                                                              GetModule(),            // Module to which this section belongs
474                                                              ++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
475                                                              segment_name,           // Name of this section
476                                                              eSectionTypeContainer,  // This section is a container of other sections.
477                                                              sect64.addr,            // File VM address == addresses as they are found in the object file
478                                                              sect64.size,            // VM size in bytes of this section
479                                                              sect64.offset,          // Offset to the data for this section in the file
480                                                              sect64.offset ? sect64.size : 0,        // Size in bytes of this section as found in the the file
481                                                              load_cmd.flags));       // Flags for this section
482                                 segment_sp->SetIsFake(true);
483                                 m_sections_ap->AddSection(segment_sp);
484                                 segment_sp->SetIsEncrypted (segment_is_encrypted);
485                             }
486                         }
487                         assert (segment_sp.get());
488 
489                         uint32_t mach_sect_type = sect64.flags & SectionFlagMaskSectionType;
490                         static ConstString g_sect_name_objc_data ("__objc_data");
491                         static ConstString g_sect_name_objc_msgrefs ("__objc_msgrefs");
492                         static ConstString g_sect_name_objc_selrefs ("__objc_selrefs");
493                         static ConstString g_sect_name_objc_classrefs ("__objc_classrefs");
494                         static ConstString g_sect_name_objc_superrefs ("__objc_superrefs");
495                         static ConstString g_sect_name_objc_const ("__objc_const");
496                         static ConstString g_sect_name_objc_classlist ("__objc_classlist");
497                         static ConstString g_sect_name_cfstring ("__cfstring");
498 
499                         static ConstString g_sect_name_dwarf_debug_abbrev ("__debug_abbrev");
500                         static ConstString g_sect_name_dwarf_debug_aranges ("__debug_aranges");
501                         static ConstString g_sect_name_dwarf_debug_frame ("__debug_frame");
502                         static ConstString g_sect_name_dwarf_debug_info ("__debug_info");
503                         static ConstString g_sect_name_dwarf_debug_line ("__debug_line");
504                         static ConstString g_sect_name_dwarf_debug_loc ("__debug_loc");
505                         static ConstString g_sect_name_dwarf_debug_macinfo ("__debug_macinfo");
506                         static ConstString g_sect_name_dwarf_debug_pubnames ("__debug_pubnames");
507                         static ConstString g_sect_name_dwarf_debug_pubtypes ("__debug_pubtypes");
508                         static ConstString g_sect_name_dwarf_debug_ranges ("__debug_ranges");
509                         static ConstString g_sect_name_dwarf_debug_str ("__debug_str");
510                         static ConstString g_sect_name_dwarf_debug_names ("__debug_names");
511                         static ConstString g_sect_name_dwarf_debug_types ("__debug_types");
512                         static ConstString g_sect_name_eh_frame ("__eh_frame");
513                         static ConstString g_sect_name_DATA ("__DATA");
514                         static ConstString g_sect_name_TEXT ("__TEXT");
515 
516                         SectionType sect_type = eSectionTypeOther;
517 
518                         if (section_name == g_sect_name_dwarf_debug_abbrev)
519                             sect_type = eSectionTypeDWARFDebugAbbrev;
520                         else if (section_name == g_sect_name_dwarf_debug_aranges)
521                             sect_type = eSectionTypeDWARFDebugAranges;
522                         else if (section_name == g_sect_name_dwarf_debug_frame)
523                             sect_type = eSectionTypeDWARFDebugFrame;
524                         else if (section_name == g_sect_name_dwarf_debug_info)
525                             sect_type = eSectionTypeDWARFDebugInfo;
526                         else if (section_name == g_sect_name_dwarf_debug_line)
527                             sect_type = eSectionTypeDWARFDebugLine;
528                         else if (section_name == g_sect_name_dwarf_debug_loc)
529                             sect_type = eSectionTypeDWARFDebugLoc;
530                         else if (section_name == g_sect_name_dwarf_debug_macinfo)
531                             sect_type = eSectionTypeDWARFDebugMacInfo;
532                         else if (section_name == g_sect_name_dwarf_debug_pubnames)
533                             sect_type = eSectionTypeDWARFDebugPubNames;
534                         else if (section_name == g_sect_name_dwarf_debug_pubtypes)
535                             sect_type = eSectionTypeDWARFDebugPubTypes;
536                         else if (section_name == g_sect_name_dwarf_debug_ranges)
537                             sect_type = eSectionTypeDWARFDebugRanges;
538                         else if (section_name == g_sect_name_dwarf_debug_str)
539                             sect_type = eSectionTypeDWARFDebugStr;
540                         else if (section_name == g_sect_name_dwarf_debug_names)
541                             sect_type = eSectionTypeDWARFDebugNames;
542                         else if (section_name == g_sect_name_dwarf_debug_types)
543                             sect_type = eSectionTypeDWARFDebugTypes;
544                         else if (section_name == g_sect_name_objc_selrefs)
545                             sect_type = eSectionTypeDataCStringPointers;
546                         else if (section_name == g_sect_name_objc_msgrefs)
547                             sect_type = eSectionTypeDataObjCMessageRefs;
548                         else if (section_name == g_sect_name_eh_frame)
549                             sect_type = eSectionTypeEHFrame;
550                         else if (section_name == g_sect_name_cfstring)
551                             sect_type = eSectionTypeDataObjCCFStrings;
552                         else if (section_name == g_sect_name_objc_data ||
553                                  section_name == g_sect_name_objc_classrefs ||
554                                  section_name == g_sect_name_objc_superrefs ||
555                                  section_name == g_sect_name_objc_const ||
556                                  section_name == g_sect_name_objc_classlist)
557                         {
558                             sect_type = eSectionTypeDataPointers;
559                         }
560 
561                         if (sect_type == eSectionTypeOther)
562                         {
563                             switch (mach_sect_type)
564                             {
565                             // TODO: categorize sections by other flags for regular sections
566                             case SectionTypeRegular:
567                                 if (segment_sp->GetName() == g_sect_name_TEXT)
568                                     sect_type = eSectionTypeCode;
569                                 else if (segment_sp->GetName() == g_sect_name_DATA)
570                                     sect_type = eSectionTypeData;
571                                 else
572                                     sect_type = eSectionTypeOther;
573                                 break;
574                             case SectionTypeZeroFill:                   sect_type = eSectionTypeZeroFill; break;
575                             case SectionTypeCStringLiterals:            sect_type = eSectionTypeDataCString;    break; // section with only literal C strings
576                             case SectionType4ByteLiterals:              sect_type = eSectionTypeData4;    break; // section with only 4 byte literals
577                             case SectionType8ByteLiterals:              sect_type = eSectionTypeData8;    break; // section with only 8 byte literals
578                             case SectionTypeLiteralPointers:            sect_type = eSectionTypeDataPointers;  break; // section with only pointers to literals
579                             case SectionTypeNonLazySymbolPointers:      sect_type = eSectionTypeDataPointers;  break; // section with only non-lazy symbol pointers
580                             case SectionTypeLazySymbolPointers:         sect_type = eSectionTypeDataPointers;  break; // section with only lazy symbol pointers
581                             case SectionTypeSymbolStubs:                sect_type = eSectionTypeCode;  break; // section with only symbol stubs, byte size of stub in the reserved2 field
582                             case SectionTypeModuleInitFunctionPointers: sect_type = eSectionTypeDataPointers;    break; // section with only function pointers for initialization
583                             case SectionTypeModuleTermFunctionPointers: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for termination
584                             case SectionTypeCoalesced:                  sect_type = eSectionTypeOther; break;
585                             case SectionTypeZeroFillLarge:              sect_type = eSectionTypeZeroFill; break;
586                             case SectionTypeInterposing:                sect_type = eSectionTypeCode;  break; // section with only pairs of function pointers for interposing
587                             case SectionType16ByteLiterals:             sect_type = eSectionTypeData16; break; // section with only 16 byte literals
588                             case SectionTypeDTraceObjectFormat:         sect_type = eSectionTypeDebug; break;
589                             case SectionTypeLazyDylibSymbolPointers:    sect_type = eSectionTypeDataPointers;  break;
590                             default: break;
591                             }
592                         }
593 
594                         SectionSP section_sp(new Section(segment_sp.get(),
595                                                          GetModule(),
596                                                          ++sectID,
597                                                          section_name,
598                                                          sect_type,
599                                                          sect64.addr - segment_sp->GetFileAddress(),
600                                                          sect64.size,
601                                                          sect64.offset,
602                                                          sect64.offset == 0 ? 0 : sect64.size,
603                                                          sect64.flags));
604                         // Set the section to be encrypted to match the segment
605                         section_sp->SetIsEncrypted (segment_is_encrypted);
606 
607                         segment_sp->GetChildren().AddSection(section_sp);
608 
609                         if (segment_sp->IsFake())
610                         {
611                             segment_sp.reset();
612                             segment_name.Clear();
613                         }
614                     }
615                     if (segment_sp && m_header.filetype == HeaderFileTypeDSYM)
616                     {
617                         if (first_segment_sectID <= sectID)
618                         {
619                             lldb::user_id_t sect_uid;
620                             for (sect_uid = first_segment_sectID; sect_uid <= sectID; ++sect_uid)
621                             {
622                                 SectionSP curr_section_sp(segment_sp->GetChildren().FindSectionByID (sect_uid));
623                                 SectionSP next_section_sp;
624                                 if (sect_uid + 1 <= sectID)
625                                     next_section_sp = segment_sp->GetChildren().FindSectionByID (sect_uid+1);
626 
627                                 if (curr_section_sp.get())
628                                 {
629                                     if (curr_section_sp->GetByteSize() == 0)
630                                     {
631                                         if (next_section_sp.get() != NULL)
632                                             curr_section_sp->SetByteSize ( next_section_sp->GetFileAddress() - curr_section_sp->GetFileAddress() );
633                                         else
634                                             curr_section_sp->SetByteSize ( load_cmd.vmsize );
635                                     }
636                                 }
637                             }
638                         }
639                     }
640                 }
641             }
642         }
643         else if (load_cmd.cmd == LoadCommandDynamicSymtabInfo)
644         {
645             m_dysymtab.cmd = load_cmd.cmd;
646             m_dysymtab.cmdsize = load_cmd.cmdsize;
647             m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2);
648         }
649 
650         offset = load_cmd_offset + load_cmd.cmdsize;
651     }
652 //    if (dump_sections)
653 //    {
654 //        StreamFile s(stdout);
655 //        m_sections_ap->Dump(&s, true);
656 //    }
657     return sectID;  // Return the number of sections we registered with the module
658 }
659 
660 class MachSymtabSectionInfo
661 {
662 public:
663 
664     MachSymtabSectionInfo (SectionList *section_list) :
665         m_section_list (section_list),
666         m_section_infos()
667     {
668         // Get the number of sections down to a depth of 1 to include
669         // all segments and their sections, but no other sections that
670         // may be added for debug map or
671         m_section_infos.resize(section_list->GetNumSections(1));
672     }
673 
674 
675     Section *
676     GetSection (uint8_t n_sect, addr_t file_addr)
677     {
678         if (n_sect == 0)
679             return NULL;
680         if (n_sect < m_section_infos.size())
681         {
682             if (m_section_infos[n_sect].section == NULL)
683             {
684                 Section *section = m_section_list->FindSectionByID (n_sect).get();
685                 m_section_infos[n_sect].section = section;
686                 if (section != NULL)
687                 {
688                     m_section_infos[n_sect].vm_range.SetBaseAddress (section->GetFileAddress());
689                     m_section_infos[n_sect].vm_range.SetByteSize (section->GetByteSize());
690                 }
691                 else
692                 {
693                     fprintf (stderr, "error: unable to find section for section %u\n", n_sect);
694                 }
695             }
696             if (m_section_infos[n_sect].vm_range.Contains(file_addr))
697             {
698                 // Symbol is in section.
699                 return m_section_infos[n_sect].section;
700             }
701             else if (m_section_infos[n_sect].vm_range.GetByteSize () == 0 &&
702                      m_section_infos[n_sect].vm_range.GetBaseAddress() == file_addr)
703             {
704                 // Symbol is in section with zero size, but has the same start
705                 // address as the section. This can happen with linker symbols
706                 // (symbols that start with the letter 'l' or 'L'.
707                 return m_section_infos[n_sect].section;
708             }
709         }
710         return m_section_list->FindSectionContainingFileAddress(file_addr).get();
711     }
712 
713 protected:
714     struct SectionInfo
715     {
716         SectionInfo () :
717             vm_range(),
718             section (NULL)
719         {
720         }
721 
722         VMRange vm_range;
723         Section *section;
724     };
725     SectionList *m_section_list;
726     std::vector<SectionInfo> m_section_infos;
727 };
728 
729 
730 
731 size_t
732 ObjectFileMachO::ParseSymtab (bool minimize)
733 {
734     Timer scoped_timer(__PRETTY_FUNCTION__,
735                        "ObjectFileMachO::ParseSymtab () module = %s",
736                        m_file.GetFilename().AsCString(""));
737     struct symtab_command symtab_load_command;
738     uint32_t offset = MachHeaderSizeFromMagic(m_header.magic);
739     uint32_t i;
740     for (i=0; i<m_header.ncmds; ++i)
741     {
742         const uint32_t cmd_offset = offset;
743         // Read in the load command and load command size
744         if (m_data.GetU32(&offset, &symtab_load_command, 2) == NULL)
745             break;
746         // Watch for the symbol table load command
747         if (symtab_load_command.cmd == LoadCommandSymtab)
748         {
749             // Read in the rest of the symtab load command
750             if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4)) // fill in symoff, nsyms, stroff, strsize fields
751             {
752                 Symtab *symtab = m_symtab_ap.get();
753                 SectionList *section_list = GetSectionList();
754                 assert(section_list);
755                 const size_t addr_size = m_data.GetAddressByteSize();
756                 const ByteOrder endian = m_data.GetByteOrder();
757                 bool bit_width_32 = addr_size == 4;
758                 const size_t nlist_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64);
759 
760                 DataBufferSP symtab_data_sp(m_file.ReadFileContents(m_offset + symtab_load_command.symoff, symtab_load_command.nsyms * nlist_size));
761                 DataBufferSP strtab_data_sp(m_file.ReadFileContents(m_offset + symtab_load_command.stroff, symtab_load_command.strsize));
762 
763                 const char *strtab_data = (const char *)strtab_data_sp->GetBytes();
764 //                DataExtractor symtab_data(symtab_data_sp, endian, addr_size);
765 //                DataExtractor strtab_data(strtab_data_sp, endian, addr_size);
766 
767                 static ConstString g_segment_name_TEXT ("__TEXT");
768                 static ConstString g_segment_name_DATA ("__DATA");
769                 static ConstString g_segment_name_OBJC ("__OBJC");
770                 static ConstString g_section_name_eh_frame ("__eh_frame");
771                 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT));
772                 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA));
773                 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC));
774                 SectionSP eh_frame_section_sp;
775                 if (text_section_sp.get())
776                     eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame);
777                 else
778                     eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame);
779 
780                 uint8_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() : NListSectionNoSection;
781                 //uint32_t symtab_offset = 0;
782                 const uint8_t* nlist_data = symtab_data_sp->GetBytes();
783                 assert (symtab_data_sp->GetByteSize()/nlist_size >= symtab_load_command.nsyms);
784 
785 
786                 if (endian != lldb::endian::InlHostByteOrder())
787                 {
788                     // ...
789                     assert (!"UNIMPLEMENTED: Swap all nlist entries");
790                 }
791                 uint32_t N_SO_index = UINT32_MAX;
792 
793                 MachSymtabSectionInfo section_info (section_list);
794                 std::vector<uint32_t> N_FUN_indexes;
795                 std::vector<uint32_t> N_NSYM_indexes;
796                 std::vector<uint32_t> N_INCL_indexes;
797                 std::vector<uint32_t> N_BRAC_indexes;
798                 std::vector<uint32_t> N_COMM_indexes;
799                 typedef std::map <uint64_t, uint32_t> ValueToSymbolIndexMap;
800                 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap;
801                 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx;
802                 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx;
803                 // Any symbols that get merged into another will get an entry
804                 // in this map so we know
805                 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx;
806                 uint32_t nlist_idx = 0;
807                 Symbol *symbol_ptr = NULL;
808 
809                 uint32_t sym_idx = 0;
810                 Symbol *sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms);
811                 uint32_t num_syms = symtab->GetNumSymbols();
812 
813                 //symtab->Reserve (symtab_load_command.nsyms + m_dysymtab.nindirectsyms);
814                 for (nlist_idx = 0; nlist_idx < symtab_load_command.nsyms; ++nlist_idx)
815                 {
816                     struct nlist_64 nlist;
817                     if (bit_width_32)
818                     {
819                         struct nlist* nlist32_ptr = (struct nlist*)(nlist_data + (nlist_idx * nlist_size));
820                         nlist.n_strx = nlist32_ptr->n_strx;
821                         nlist.n_type = nlist32_ptr->n_type;
822                         nlist.n_sect = nlist32_ptr->n_sect;
823                         nlist.n_desc = nlist32_ptr->n_desc;
824                         nlist.n_value = nlist32_ptr->n_value;
825                     }
826                     else
827                     {
828                         nlist = *((struct nlist_64*)(nlist_data + (nlist_idx * nlist_size)));
829                     }
830 
831                     SymbolType type = eSymbolTypeInvalid;
832                     const char* symbol_name = &strtab_data[nlist.n_strx];
833                     if (symbol_name[0] == '\0')
834                         symbol_name = NULL;
835                     Section* symbol_section = NULL;
836                     bool add_nlist = true;
837                     bool is_debug = ((nlist.n_type & NlistMaskStab) != 0);
838 
839                     assert (sym_idx < num_syms);
840 
841                     sym[sym_idx].SetDebug (is_debug);
842 
843                     if (is_debug)
844                     {
845                         switch (nlist.n_type)
846                         {
847                         case StabGlobalSymbol:
848                             // N_GSYM -- global symbol: name,,NO_SECT,type,0
849                             // Sometimes the N_GSYM value contains the address.
850                             sym[sym_idx].SetExternal(true);
851                             if (nlist.n_value != 0)
852                                 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value);
853                             type = eSymbolTypeData;
854                             break;
855 
856                         case StabFunctionName:
857                             // N_FNAME -- procedure name (f77 kludge): name,,NO_SECT,0,0
858                             type = eSymbolTypeCompiler;
859                             break;
860 
861                         case StabFunction:
862                             // N_FUN -- procedure: name,,n_sect,linenumber,address
863                             if (symbol_name)
864                             {
865                                 type = eSymbolTypeCode;
866                                 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value);
867 
868                                 N_FUN_addr_to_sym_idx[nlist.n_value] = sym_idx;
869                                 // We use the current number of symbols in the symbol table in lieu of
870                                 // using nlist_idx in case we ever start trimming entries out
871                                 N_FUN_indexes.push_back(sym_idx);
872                             }
873                             else
874                             {
875                                 type = eSymbolTypeCompiler;
876 
877                                 if ( !N_FUN_indexes.empty() )
878                                 {
879                                     // Copy the size of the function into the original STAB entry so we don't have
880                                     // to hunt for it later
881                                     symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value);
882                                     N_FUN_indexes.pop_back();
883                                     // We don't really need the end function STAB as it contains the size which
884                                     // we already placed with the original symbol, so don't add it if we want a
885                                     // minimal symbol table
886                                     if (minimize)
887                                         add_nlist = false;
888                                 }
889                             }
890                             break;
891 
892                         case StabStaticSymbol:
893                             // N_STSYM -- static symbol: name,,n_sect,type,address
894                             N_STSYM_addr_to_sym_idx[nlist.n_value] = sym_idx;
895                             symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value);
896                             type = eSymbolTypeData;
897                             break;
898 
899                         case StabLocalCommon:
900                             // N_LCSYM -- .lcomm symbol: name,,n_sect,type,address
901                             symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value);
902                             type = eSymbolTypeCommonBlock;
903                             break;
904 
905                         case StabBeginSymbol:
906                             // N_BNSYM
907                             // We use the current number of symbols in the symbol table in lieu of
908                             // using nlist_idx in case we ever start trimming entries out
909                             if (minimize)
910                             {
911                                 // Skip these if we want minimal symbol tables
912                                 add_nlist = false;
913                             }
914                             else
915                             {
916                                 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value);
917                                 N_NSYM_indexes.push_back(sym_idx);
918                                 type = eSymbolTypeScopeBegin;
919                             }
920                             break;
921 
922                         case StabEndSymbol:
923                             // N_ENSYM
924                             // Set the size of the N_BNSYM to the terminating index of this N_ENSYM
925                             // so that we can always skip the entire symbol if we need to navigate
926                             // more quickly at the source level when parsing STABS
927                             if (minimize)
928                             {
929                                 // Skip these if we want minimal symbol tables
930                                 add_nlist = false;
931                             }
932                             else
933                             {
934                                 if ( !N_NSYM_indexes.empty() )
935                                 {
936                                     symbol_ptr = symtab->SymbolAtIndex(N_NSYM_indexes.back());
937                                     symbol_ptr->SetByteSize(sym_idx + 1);
938                                     symbol_ptr->SetSizeIsSibling(true);
939                                     N_NSYM_indexes.pop_back();
940                                 }
941                                 type = eSymbolTypeScopeEnd;
942                             }
943                             break;
944 
945 
946                         case StabSourceFileOptions:
947                             // N_OPT - emitted with gcc2_compiled and in gcc source
948                             type = eSymbolTypeCompiler;
949                             break;
950 
951                         case StabRegisterSymbol:
952                             // N_RSYM - register sym: name,,NO_SECT,type,register
953                             type = eSymbolTypeVariable;
954                             break;
955 
956                         case StabSourceLine:
957                             // N_SLINE - src line: 0,,n_sect,linenumber,address
958                             symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value);
959                             type = eSymbolTypeLineEntry;
960                             break;
961 
962                         case StabStructureType:
963                             // N_SSYM - structure elt: name,,NO_SECT,type,struct_offset
964                             type = eSymbolTypeVariableType;
965                             break;
966 
967                         case StabSourceFileName:
968                             // N_SO - source file name
969                             type = eSymbolTypeSourceFile;
970                             if (symbol_name == NULL)
971                             {
972                                 if (minimize)
973                                     add_nlist = false;
974                                 if (N_SO_index != UINT32_MAX)
975                                 {
976                                     // Set the size of the N_SO to the terminating index of this N_SO
977                                     // so that we can always skip the entire N_SO if we need to navigate
978                                     // more quickly at the source level when parsing STABS
979                                     symbol_ptr = symtab->SymbolAtIndex(N_SO_index);
980                                     symbol_ptr->SetByteSize(sym_idx + (minimize ? 0 : 1));
981                                     symbol_ptr->SetSizeIsSibling(true);
982                                 }
983                                 N_NSYM_indexes.clear();
984                                 N_INCL_indexes.clear();
985                                 N_BRAC_indexes.clear();
986                                 N_COMM_indexes.clear();
987                                 N_FUN_indexes.clear();
988                                 N_SO_index = UINT32_MAX;
989                             }
990                             else
991                             {
992                                 // We use the current number of symbols in the symbol table in lieu of
993                                 // using nlist_idx in case we ever start trimming entries out
994                                 if (symbol_name[0] == '/')
995                                     N_SO_index = sym_idx;
996                                 else if (minimize && (N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms))
997                                 {
998                                     const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString();
999                                     if (so_path && so_path[0])
1000                                     {
1001                                         std::string full_so_path (so_path);
1002                                         if (*full_so_path.rbegin() != '/')
1003                                             full_so_path += '/';
1004                                         full_so_path += symbol_name;
1005                                         sym[sym_idx - 1].GetMangled().SetValue(full_so_path.c_str(), false);
1006                                         add_nlist = false;
1007                                         m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1;
1008                                     }
1009                                 }
1010                             }
1011 
1012                             break;
1013 
1014                         case StabObjectFileName:
1015                             // N_OSO - object file name: name,,0,0,st_mtime
1016                             type = eSymbolTypeObjectFile;
1017                             break;
1018 
1019                         case StabLocalSymbol:
1020                             // N_LSYM - local sym: name,,NO_SECT,type,offset
1021                             type = eSymbolTypeLocal;
1022                             break;
1023 
1024                         //----------------------------------------------------------------------
1025                         // INCL scopes
1026                         //----------------------------------------------------------------------
1027                         case StabBeginIncludeFileName:
1028                             // N_BINCL - include file beginning: name,,NO_SECT,0,sum
1029                             // We use the current number of symbols in the symbol table in lieu of
1030                             // using nlist_idx in case we ever start trimming entries out
1031                             N_INCL_indexes.push_back(sym_idx);
1032                             type = eSymbolTypeScopeBegin;
1033                             break;
1034 
1035                         case StabEndIncludeFile:
1036                             // N_EINCL - include file end: name,,NO_SECT,0,0
1037                             // Set the size of the N_BINCL to the terminating index of this N_EINCL
1038                             // so that we can always skip the entire symbol if we need to navigate
1039                             // more quickly at the source level when parsing STABS
1040                             if ( !N_INCL_indexes.empty() )
1041                             {
1042                                 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back());
1043                                 symbol_ptr->SetByteSize(sym_idx + 1);
1044                                 symbol_ptr->SetSizeIsSibling(true);
1045                                 N_INCL_indexes.pop_back();
1046                             }
1047                             type = eSymbolTypeScopeEnd;
1048                             break;
1049 
1050                         case StabIncludeFileName:
1051                             // N_SOL - #included file name: name,,n_sect,0,address
1052                             type = eSymbolTypeHeaderFile;
1053 
1054                             // We currently don't use the header files on darwin
1055                             if (minimize)
1056                                 add_nlist = false;
1057                             break;
1058 
1059                         case StabCompilerParameters:
1060                             // N_PARAMS - compiler parameters: name,,NO_SECT,0,0
1061                             type = eSymbolTypeCompiler;
1062                             break;
1063 
1064                         case StabCompilerVersion:
1065                             // N_VERSION - compiler version: name,,NO_SECT,0,0
1066                             type = eSymbolTypeCompiler;
1067                             break;
1068 
1069                         case StabCompilerOptLevel:
1070                             // N_OLEVEL - compiler -O level: name,,NO_SECT,0,0
1071                             type = eSymbolTypeCompiler;
1072                             break;
1073 
1074                         case StabParameter:
1075                             // N_PSYM - parameter: name,,NO_SECT,type,offset
1076                             type = eSymbolTypeVariable;
1077                             break;
1078 
1079                         case StabAlternateEntry:
1080                             // N_ENTRY - alternate entry: name,,n_sect,linenumber,address
1081                             symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value);
1082                             type = eSymbolTypeLineEntry;
1083                             break;
1084 
1085                         //----------------------------------------------------------------------
1086                         // Left and Right Braces
1087                         //----------------------------------------------------------------------
1088                         case StabLeftBracket:
1089                             // N_LBRAC - left bracket: 0,,NO_SECT,nesting level,address
1090                             // We use the current number of symbols in the symbol table in lieu of
1091                             // using nlist_idx in case we ever start trimming entries out
1092                             symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value);
1093                             N_BRAC_indexes.push_back(sym_idx);
1094                             type = eSymbolTypeScopeBegin;
1095                             break;
1096 
1097                         case StabRightBracket:
1098                             // N_RBRAC - right bracket: 0,,NO_SECT,nesting level,address
1099                             // Set the size of the N_LBRAC to the terminating index of this N_RBRAC
1100                             // so that we can always skip the entire symbol if we need to navigate
1101                             // more quickly at the source level when parsing STABS
1102                             symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value);
1103                             if ( !N_BRAC_indexes.empty() )
1104                             {
1105                                 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back());
1106                                 symbol_ptr->SetByteSize(sym_idx + 1);
1107                                 symbol_ptr->SetSizeIsSibling(true);
1108                                 N_BRAC_indexes.pop_back();
1109                             }
1110                             type = eSymbolTypeScopeEnd;
1111                             break;
1112 
1113                         case StabDeletedIncludeFile:
1114                             // N_EXCL - deleted include file: name,,NO_SECT,0,sum
1115                             type = eSymbolTypeHeaderFile;
1116                             break;
1117 
1118                         //----------------------------------------------------------------------
1119                         // COMM scopes
1120                         //----------------------------------------------------------------------
1121                         case StabBeginCommon:
1122                             // N_BCOMM - begin common: name,,NO_SECT,0,0
1123                             // We use the current number of symbols in the symbol table in lieu of
1124                             // using nlist_idx in case we ever start trimming entries out
1125                             type = eSymbolTypeScopeBegin;
1126                             N_COMM_indexes.push_back(sym_idx);
1127                             break;
1128 
1129                         case StabEndCommonLocal:
1130                             // N_ECOML - end common (local name): 0,,n_sect,0,address
1131                             symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value);
1132                             // Fall through
1133 
1134                         case StabEndCommon:
1135                             // N_ECOMM - end common: name,,n_sect,0,0
1136                             // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML
1137                             // so that we can always skip the entire symbol if we need to navigate
1138                             // more quickly at the source level when parsing STABS
1139                             if ( !N_COMM_indexes.empty() )
1140                             {
1141                                 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back());
1142                                 symbol_ptr->SetByteSize(sym_idx + 1);
1143                                 symbol_ptr->SetSizeIsSibling(true);
1144                                 N_COMM_indexes.pop_back();
1145                             }
1146                             type = eSymbolTypeScopeEnd;
1147                             break;
1148 
1149                         case StabLength:
1150                             // N_LENG - second stab entry with length information
1151                             type = eSymbolTypeAdditional;
1152                             break;
1153 
1154                         default: break;
1155                         }
1156                     }
1157                     else
1158                     {
1159                         //uint8_t n_pext    = NlistMaskPrivateExternal & nlist.n_type;
1160                         uint8_t n_type  = NlistMaskType & nlist.n_type;
1161                         sym[sym_idx].SetExternal((NlistMaskExternal & nlist.n_type) != 0);
1162 
1163                         if (symbol_name && ::strstr (symbol_name, ".objc") == symbol_name)
1164                         {
1165                             type = eSymbolTypeRuntime;
1166                         }
1167                         else
1168                         {
1169                             switch (n_type)
1170                             {
1171                             case NListTypeIndirect:         // N_INDR - Fall through
1172                             case NListTypePreboundUndefined:// N_PBUD - Fall through
1173                             case NListTypeUndefined:        // N_UNDF
1174                                 type = eSymbolTypeExtern;
1175                                 break;
1176 
1177                             case NListTypeAbsolute:         // N_ABS
1178                                 type = eSymbolTypeAbsolute;
1179                                 break;
1180 
1181                             case NListTypeSection:          // N_SECT
1182                                 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value);
1183 
1184                                 if (symbol_section == NULL)
1185                                 {
1186                                     // TODO: warn about this?
1187                                     add_nlist = false;
1188                                     break;
1189                                 }
1190 
1191                                 if (TEXT_eh_frame_sectID == nlist.n_sect)
1192                                 {
1193                                     type = eSymbolTypeException;
1194                                 }
1195                                 else
1196                                 {
1197                                     uint32_t section_type = symbol_section->Get() & SectionFlagMaskSectionType;
1198 
1199                                     switch (section_type)
1200                                     {
1201                                     case SectionTypeRegular:                     break; // regular section
1202                                     //case SectionTypeZeroFill:                 type = eSymbolTypeData;    break; // zero fill on demand section
1203                                     case SectionTypeCStringLiterals:            type = eSymbolTypeData;    break; // section with only literal C strings
1204                                     case SectionType4ByteLiterals:              type = eSymbolTypeData;    break; // section with only 4 byte literals
1205                                     case SectionType8ByteLiterals:              type = eSymbolTypeData;    break; // section with only 8 byte literals
1206                                     case SectionTypeLiteralPointers:            type = eSymbolTypeTrampoline; break; // section with only pointers to literals
1207                                     case SectionTypeNonLazySymbolPointers:      type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers
1208                                     case SectionTypeLazySymbolPointers:         type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers
1209                                     case SectionTypeSymbolStubs:                type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field
1210                                     case SectionTypeModuleInitFunctionPointers: type = eSymbolTypeCode;    break; // section with only function pointers for initialization
1211                                     case SectionTypeModuleTermFunctionPointers: type = eSymbolTypeCode;    break; // section with only function pointers for termination
1212                                     //case SectionTypeCoalesced:                type = eSymbolType;    break; // section contains symbols that are to be coalesced
1213                                     //case SectionTypeZeroFillLarge:            type = eSymbolTypeData;    break; // zero fill on demand section (that can be larger than 4 gigabytes)
1214                                     case SectionTypeInterposing:                type = eSymbolTypeTrampoline;  break; // section with only pairs of function pointers for interposing
1215                                     case SectionType16ByteLiterals:             type = eSymbolTypeData;    break; // section with only 16 byte literals
1216                                     case SectionTypeDTraceObjectFormat:         type = eSymbolTypeInstrumentation; break;
1217                                     case SectionTypeLazyDylibSymbolPointers:    type = eSymbolTypeTrampoline; break;
1218                                     default: break;
1219                                     }
1220 
1221                                     if (type == eSymbolTypeInvalid)
1222                                     {
1223                                         const char *symbol_sect_name = symbol_section->GetName().AsCString();
1224                                         if (symbol_section->IsDescendant (text_section_sp.get()))
1225                                         {
1226                                             if (symbol_section->IsClear(SectionAttrUserPureInstructions |
1227                                                                         SectionAttrUserSelfModifyingCode |
1228                                                                         SectionAttrSytemSomeInstructions))
1229                                                 type = eSymbolTypeData;
1230                                             else
1231                                                 type = eSymbolTypeCode;
1232                                         }
1233                                         else
1234                                         if (symbol_section->IsDescendant(data_section_sp.get()))
1235                                         {
1236                                             if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name)
1237                                             {
1238                                                 type = eSymbolTypeRuntime;
1239                                             }
1240                                             else
1241                                             if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name)
1242                                             {
1243                                                 type = eSymbolTypeException;
1244                                             }
1245                                             else
1246                                             {
1247                                                 type = eSymbolTypeData;
1248                                             }
1249                                         }
1250                                         else
1251                                         if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name)
1252                                         {
1253                                             type = eSymbolTypeTrampoline;
1254                                         }
1255                                         else
1256                                         if (symbol_section->IsDescendant(objc_section_sp.get()))
1257                                         {
1258                                             type = eSymbolTypeRuntime;
1259                                         }
1260                                     }
1261                                 }
1262                                 break;
1263                             }
1264                         }
1265                     }
1266                     if (add_nlist)
1267                     {
1268                         bool symbol_name_is_mangled = false;
1269                         if (symbol_name && symbol_name[0] == '_')
1270                         {
1271                             symbol_name_is_mangled = symbol_name[1] == '_';
1272                             symbol_name++;  // Skip the leading underscore
1273                         }
1274                         uint64_t symbol_value = nlist.n_value;
1275 
1276                         if (symbol_name)
1277                             sym[sym_idx].GetMangled().SetValue(symbol_name, symbol_name_is_mangled);
1278                         if (is_debug == false)
1279                         {
1280                             if (type == eSymbolTypeCode)
1281                             {
1282                                 // See if we can find a N_FUN entry for any code symbols.
1283                                 // If we do find a match, and the name matches, then we
1284                                 // can merge the two into just the function symbol to avoid
1285                                 // duplicate entries in the symbol table
1286                                 ValueToSymbolIndexMap::const_iterator pos = N_FUN_addr_to_sym_idx.find (nlist.n_value);
1287                                 if (pos != N_FUN_addr_to_sym_idx.end())
1288                                 {
1289                                     if ((symbol_name_is_mangled == true && sym[sym_idx].GetMangled().GetMangledName() == sym[pos->second].GetMangled().GetMangledName()) ||
1290                                         (symbol_name_is_mangled == false && sym[sym_idx].GetMangled().GetDemangledName() == sym[pos->second].GetMangled().GetDemangledName()))
1291                                     {
1292                                         m_nlist_idx_to_sym_idx[nlist_idx] = pos->second;
1293                                         // We just need the flags from the linker symbol, so put these flags
1294                                         // into the N_FUN flags to avoid duplicate symbols in the symbol table
1295                                         sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc);
1296                                         sym[sym_idx].Clear();
1297                                         continue;
1298                                     }
1299                                 }
1300                             }
1301                             else if (type == eSymbolTypeData)
1302                             {
1303                                 // See if we can find a N_STSYM entry for any data symbols.
1304                                 // If we do find a match, and the name matches, then we
1305                                 // can merge the two into just the Static symbol to avoid
1306                                 // duplicate entries in the symbol table
1307                                 ValueToSymbolIndexMap::const_iterator pos = N_STSYM_addr_to_sym_idx.find (nlist.n_value);
1308                                 if (pos != N_STSYM_addr_to_sym_idx.end())
1309                                 {
1310                                     if ((symbol_name_is_mangled == true && sym[sym_idx].GetMangled().GetMangledName() == sym[pos->second].GetMangled().GetMangledName()) ||
1311                                         (symbol_name_is_mangled == false && sym[sym_idx].GetMangled().GetDemangledName() == sym[pos->second].GetMangled().GetDemangledName()))
1312                                     {
1313                                         m_nlist_idx_to_sym_idx[nlist_idx] = pos->second;
1314                                         // We just need the flags from the linker symbol, so put these flags
1315                                         // into the N_STSYM flags to avoid duplicate symbols in the symbol table
1316                                         sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc);
1317                                         sym[sym_idx].Clear();
1318                                         continue;
1319                                     }
1320                                 }
1321                             }
1322                         }
1323                         if (symbol_section != NULL)
1324                             symbol_value -= symbol_section->GetFileAddress();
1325 
1326                         sym[sym_idx].SetID (nlist_idx);
1327                         sym[sym_idx].SetType (type);
1328                         sym[sym_idx].GetAddressRangeRef().GetBaseAddress().SetSection (symbol_section);
1329                         sym[sym_idx].GetAddressRangeRef().GetBaseAddress().SetOffset (symbol_value);
1330                         sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc);
1331 
1332                         ++sym_idx;
1333                     }
1334                     else
1335                     {
1336                         sym[sym_idx].Clear();
1337                     }
1338 
1339                 }
1340 
1341                 // STAB N_GSYM entries end up having a symbol type eSymbolTypeGlobal and when the symbol value
1342                 // is zero, the address of the global ends up being in a non-STAB entry. Try and fix up all
1343                 // such entries by figuring out what the address for the global is by looking up this non-STAB
1344                 // entry and copying the value into the debug symbol's value to save us the hassle in the
1345                 // debug symbol parser.
1346 
1347                 Symbol *global_symbol = NULL;
1348                 for (nlist_idx = 0;
1349                      nlist_idx < symtab_load_command.nsyms && (global_symbol = symtab->FindSymbolWithType (eSymbolTypeData, Symtab::eDebugYes, Symtab::eVisibilityAny, nlist_idx)) != NULL;
1350                      nlist_idx++)
1351                 {
1352                     if (global_symbol->GetValue().GetFileAddress() == 0)
1353                     {
1354                         std::vector<uint32_t> indexes;
1355                         if (symtab->AppendSymbolIndexesWithName (global_symbol->GetMangled().GetName(), indexes) > 0)
1356                         {
1357                             std::vector<uint32_t>::const_iterator pos;
1358                             std::vector<uint32_t>::const_iterator end = indexes.end();
1359                             for (pos = indexes.begin(); pos != end; ++pos)
1360                             {
1361                                 symbol_ptr = symtab->SymbolAtIndex(*pos);
1362                                 if (symbol_ptr != global_symbol && symbol_ptr->IsDebug() == false)
1363                                 {
1364                                     global_symbol->SetValue(symbol_ptr->GetValue());
1365                                     break;
1366                                 }
1367                             }
1368                         }
1369                     }
1370                 }
1371 
1372                 // Trim our symbols down to just what we ended up with after
1373                 // removing any symbols.
1374                 if (sym_idx < num_syms)
1375                 {
1376                     num_syms = sym_idx;
1377                     sym = symtab->Resize (num_syms);
1378                 }
1379 
1380                 // Now synthesize indirect symbols
1381                 if (m_dysymtab.nindirectsyms != 0)
1382                 {
1383                     DataBufferSP indirect_symbol_indexes_sp(m_file.ReadFileContents(m_offset + m_dysymtab.indirectsymoff, m_dysymtab.nindirectsyms * 4));
1384 
1385                     if (indirect_symbol_indexes_sp && indirect_symbol_indexes_sp->GetByteSize())
1386                     {
1387                         NListIndexToSymbolIndexMap::const_iterator end_index_pos = m_nlist_idx_to_sym_idx.end();
1388                         DataExtractor indirect_symbol_index_data (indirect_symbol_indexes_sp, m_data.GetByteOrder(), m_data.GetAddressByteSize());
1389 
1390                         for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); ++sect_idx)
1391                         {
1392                             if ((m_mach_sections[sect_idx].flags & SectionFlagMaskSectionType) == SectionTypeSymbolStubs)
1393                             {
1394                                 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2;
1395                                 if (symbol_stub_byte_size == 0)
1396                                     continue;
1397 
1398                                 const uint32_t num_symbol_stubs = m_mach_sections[sect_idx].size / symbol_stub_byte_size;
1399 
1400                                 if (num_symbol_stubs == 0)
1401                                     continue;
1402 
1403                                 const uint32_t symbol_stub_index_offset = m_mach_sections[sect_idx].reserved1;
1404                                 uint32_t synthetic_stub_sym_id = symtab_load_command.nsyms;
1405                                 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx)
1406                                 {
1407                                     const uint32_t symbol_stub_index = symbol_stub_index_offset + stub_idx;
1408                                     const lldb::addr_t symbol_stub_addr = m_mach_sections[sect_idx].addr + (stub_idx * symbol_stub_byte_size);
1409                                     uint32_t symbol_stub_offset = symbol_stub_index * 4;
1410                                     if (indirect_symbol_index_data.ValidOffsetForDataOfSize(symbol_stub_offset, 4))
1411                                     {
1412                                         const uint32_t stub_sym_id = indirect_symbol_index_data.GetU32 (&symbol_stub_offset);
1413                                         if (stub_sym_id & (IndirectSymbolAbsolute | IndirectSymbolLocal))
1414                                             continue;
1415 
1416                                         NListIndexToSymbolIndexMap::const_iterator index_pos = m_nlist_idx_to_sym_idx.find (stub_sym_id);
1417                                         Symbol *stub_symbol = NULL;
1418                                         if (index_pos != end_index_pos)
1419                                         {
1420                                             // We have a remapping from the original nlist index to
1421                                             // a current symbol index, so just look this up by index
1422                                             stub_symbol = symtab->SymbolAtIndex (index_pos->second);
1423                                         }
1424                                         else
1425                                         {
1426                                             // We need to lookup a symbol using the original nlist
1427                                             // symbol index since this index is coming from the
1428                                             // S_SYMBOL_STUBS
1429                                             stub_symbol = symtab->FindSymbolByID (stub_sym_id);
1430                                         }
1431 
1432                                         assert (stub_symbol);
1433                                         if (stub_symbol)
1434                                         {
1435                                             Address so_addr(symbol_stub_addr, section_list);
1436 
1437                                             if (stub_symbol->GetType() == eSymbolTypeExtern)
1438                                             {
1439                                                 // Change the external symbol into a trampoline that makes sense
1440                                                 // These symbols were N_UNDF N_EXT, and are useless to us, so we
1441                                                 // can re-use them so we don't have to make up a synthetic symbol
1442                                                 // for no good reason.
1443                                                 stub_symbol->SetType (eSymbolTypeTrampoline);
1444                                                 stub_symbol->SetExternal (false);
1445                                                 stub_symbol->GetAddressRangeRef().GetBaseAddress() = so_addr;
1446                                                 stub_symbol->GetAddressRangeRef().SetByteSize (symbol_stub_byte_size);
1447                                             }
1448                                             else
1449                                             {
1450                                                 // Make a synthetic symbol to describe the trampoline stub
1451                                                 if (sym_idx >= num_syms)
1452                                                     sym = symtab->Resize (++num_syms);
1453                                                 sym[sym_idx].SetID (synthetic_stub_sym_id++);
1454                                                 sym[sym_idx].GetMangled() = stub_symbol->GetMangled();
1455                                                 sym[sym_idx].SetType (eSymbolTypeTrampoline);
1456                                                 sym[sym_idx].SetIsSynthetic (true);
1457                                                 sym[sym_idx].GetAddressRangeRef().GetBaseAddress() = so_addr;
1458                                                 sym[sym_idx].GetAddressRangeRef().SetByteSize (symbol_stub_byte_size);
1459                                                 ++sym_idx;
1460                                             }
1461                                         }
1462                                     }
1463                                 }
1464                             }
1465                         }
1466                     }
1467                 }
1468 
1469                 return symtab->GetNumSymbols();
1470             }
1471         }
1472         offset = cmd_offset + symtab_load_command.cmdsize;
1473     }
1474     return 0;
1475 }
1476 
1477 
1478 void
1479 ObjectFileMachO::Dump (Stream *s)
1480 {
1481     lldb_private::Mutex::Locker locker(m_mutex);
1482     s->Printf("%.*p: ", (int)sizeof(void*) * 2, this);
1483     s->Indent();
1484     if (m_header.magic == HeaderMagic64 || m_header.magic == HeaderMagic64Swapped)
1485         s->PutCString("ObjectFileMachO64");
1486     else
1487         s->PutCString("ObjectFileMachO32");
1488 
1489     ArchSpec header_arch(eArchTypeMachO, m_header.cputype, m_header.cpusubtype);
1490 
1491     *s << ", file = '" << m_file << "', arch = " << header_arch.GetArchitectureName() << "\n";
1492 
1493     if (m_sections_ap.get())
1494         m_sections_ap->Dump(s, NULL, true, UINT32_MAX);
1495 
1496     if (m_symtab_ap.get())
1497         m_symtab_ap->Dump(s, NULL, eSortOrderNone);
1498 }
1499 
1500 
1501 bool
1502 ObjectFileMachO::GetUUID (lldb_private::UUID* uuid)
1503 {
1504     lldb_private::Mutex::Locker locker(m_mutex);
1505     struct uuid_command load_cmd;
1506     uint32_t offset = MachHeaderSizeFromMagic(m_header.magic);
1507     uint32_t i;
1508     for (i=0; i<m_header.ncmds; ++i)
1509     {
1510         const uint32_t cmd_offset = offset;
1511         if (m_data.GetU32(&offset, &load_cmd, 2) == NULL)
1512             break;
1513 
1514         if (load_cmd.cmd == LoadCommandUUID)
1515         {
1516             const uint8_t *uuid_bytes = m_data.PeekData(offset, 16);
1517             if (uuid_bytes)
1518             {
1519                 uuid->SetBytes (uuid_bytes);
1520                 return true;
1521             }
1522             return false;
1523         }
1524         offset = cmd_offset + load_cmd.cmdsize;
1525     }
1526     return false;
1527 }
1528 
1529 
1530 uint32_t
1531 ObjectFileMachO::GetDependentModules (FileSpecList& files)
1532 {
1533     lldb_private::Mutex::Locker locker(m_mutex);
1534     struct load_command load_cmd;
1535     uint32_t offset = MachHeaderSizeFromMagic(m_header.magic);
1536     uint32_t count = 0;
1537     const bool resolve_path = false; // Don't resolve the dependend file paths since they may not reside on this system
1538     uint32_t i;
1539     for (i=0; i<m_header.ncmds; ++i)
1540     {
1541         const uint32_t cmd_offset = offset;
1542         if (m_data.GetU32(&offset, &load_cmd, 2) == NULL)
1543             break;
1544 
1545         switch (load_cmd.cmd)
1546         {
1547         case LoadCommandDylibLoad:
1548         case LoadCommandDylibLoadWeak:
1549         case LoadCommandDylibReexport:
1550         case LoadCommandDynamicLinkerLoad:
1551         case LoadCommandFixedVMShlibLoad:
1552         case LoadCommandDylibLoadUpward:
1553             {
1554                 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset);
1555                 const char *path = m_data.PeekCStr(name_offset);
1556                 // Skip any path that starts with '@' since these are usually:
1557                 // @executable_path/.../file
1558                 // @rpath/.../file
1559                 if (path && path[0] != '@')
1560                 {
1561                     FileSpec file_spec(path, resolve_path);
1562                     if (files.AppendIfUnique(file_spec))
1563                         count++;
1564                 }
1565             }
1566             break;
1567 
1568         default:
1569             break;
1570         }
1571         offset = cmd_offset + load_cmd.cmdsize;
1572     }
1573     return count;
1574 }
1575 
1576 lldb_private::Address
1577 ObjectFileMachO::GetEntryPointAddress ()
1578 {
1579     // If the object file is not an executable it can't hold the entry point.  m_entry_point_address
1580     // is initialized to an invalid address, so we can just return that.
1581     // If m_entry_point_address is valid it means we've found it already, so return the cached value.
1582 
1583     if (!IsExecutable() || m_entry_point_address.IsValid())
1584         return m_entry_point_address;
1585 
1586     // Otherwise, look for the UnixThread or Thread command.  The data for the Thread command is given in
1587     // /usr/include/mach-o.h, but it is basically:
1588     //
1589     //  uint32_t flavor  - this is the flavor argument you would pass to thread_get_state
1590     //  uint32_t count   - this is the count of longs in the thread state data
1591     //  struct XXX_thread_state state - this is the structure from <machine/thread_status.h> corresponding to the flavor.
1592     //  <repeat this trio>
1593     //
1594     // So we just keep reading the various register flavors till we find the GPR one, then read the PC out of there.
1595     // FIXME: We will need to have a "RegisterContext data provider" class at some point that can get all the registers
1596     // out of data in this form & attach them to a given thread.  That should underlie the MacOS X User process plugin,
1597     // and we'll also need it for the MacOS X Core File process plugin.  When we have that we can also use it here.
1598     //
1599     // For now we hard-code the offsets and flavors we need:
1600     //
1601     //
1602 
1603     lldb_private::Mutex::Locker locker(m_mutex);
1604     struct load_command load_cmd;
1605     uint32_t offset = MachHeaderSizeFromMagic(m_header.magic);
1606     uint32_t i;
1607     lldb::addr_t start_address = LLDB_INVALID_ADDRESS;
1608     bool done = false;
1609 
1610     for (i=0; i<m_header.ncmds; ++i)
1611     {
1612         const uint32_t cmd_offset = offset;
1613         if (m_data.GetU32(&offset, &load_cmd, 2) == NULL)
1614             break;
1615 
1616         switch (load_cmd.cmd)
1617         {
1618         case LoadCommandUnixThread:
1619         case LoadCommandThread:
1620             {
1621                 while (offset < cmd_offset + load_cmd.cmdsize)
1622                 {
1623                     uint32_t flavor = m_data.GetU32(&offset);
1624                     uint32_t count = m_data.GetU32(&offset);
1625                     if (count == 0)
1626                     {
1627                         // We've gotten off somehow, log and exit;
1628                         return m_entry_point_address;
1629                     }
1630 
1631                     switch (m_header.cputype)
1632                     {
1633                     case llvm::MachO::CPUTypeARM:
1634                        if (flavor == 1) // ARM_THREAD_STATE from mach/arm/thread_status.h
1635                        {
1636                            offset += 60;  // This is the offset of pc in the GPR thread state data structure.
1637                            start_address = m_data.GetU32(&offset);
1638                            done = true;
1639                         }
1640                     break;
1641                     case llvm::MachO::CPUTypeI386:
1642                        if (flavor == 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h
1643                        {
1644                            offset += 40;  // This is the offset of eip in the GPR thread state data structure.
1645                            start_address = m_data.GetU32(&offset);
1646                            done = true;
1647                         }
1648                     break;
1649                     case llvm::MachO::CPUTypeX86_64:
1650                        if (flavor == 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h
1651                        {
1652                            offset += 16 * 8;  // This is the offset of rip in the GPR thread state data structure.
1653                            start_address = m_data.GetU64(&offset);
1654                            done = true;
1655                         }
1656                     break;
1657                     default:
1658                         return m_entry_point_address;
1659                     }
1660                     // Haven't found the GPR flavor yet, skip over the data for this flavor:
1661                     if (done)
1662                         break;
1663                     offset += count * 4;
1664                 }
1665             }
1666             break;
1667 
1668         default:
1669             break;
1670         }
1671         if (done)
1672             break;
1673 
1674         // Go to the next load command:
1675         offset = cmd_offset + load_cmd.cmdsize;
1676     }
1677 
1678     if (start_address != LLDB_INVALID_ADDRESS)
1679     {
1680         // We got the start address from the load commands, so now resolve that address in the sections
1681         // of this ObjectFile:
1682         if (!m_entry_point_address.ResolveAddressUsingFileSections (start_address, GetSectionList()))
1683         {
1684             m_entry_point_address.Clear();
1685         }
1686     }
1687     else
1688     {
1689         // We couldn't read the UnixThread load command - maybe it wasn't there.  As a fallback look for the
1690         // "start" symbol in the main executable.
1691 
1692         SymbolContextList contexts;
1693         SymbolContext context;
1694         if (!m_module->FindSymbolsWithNameAndType(ConstString ("start"), eSymbolTypeCode, contexts))
1695             return m_entry_point_address;
1696 
1697         contexts.GetContextAtIndex(0, context);
1698 
1699         m_entry_point_address = context.symbol->GetValue();
1700     }
1701 
1702     return m_entry_point_address;
1703 
1704 }
1705 
1706 ObjectFile::Type
1707 ObjectFileMachO::CalculateType()
1708 {
1709     switch (m_header.filetype)
1710     {
1711         case HeaderFileTypeObject:                                          // 0x1u MH_OBJECT
1712             if (GetAddressByteSize () == 4)
1713             {
1714                 // 32 bit kexts are just object files, but they do have a valid
1715                 // UUID load command.
1716                 UUID uuid;
1717                 if (GetUUID(&uuid))
1718                 {
1719                     // this checking for the UUID load command is not enough
1720                     // we could eventually look for the symbol named
1721                     // "OSKextGetCurrentIdentifier" as this is required of kexts
1722                     if (m_strata == eStrataInvalid)
1723                         m_strata = eStrataKernel;
1724                     return eTypeSharedLibrary;
1725                 }
1726             }
1727             return eTypeObjectFile;
1728 
1729         case HeaderFileTypeExecutable:          return eTypeExecutable;     // 0x2u MH_EXECUTE
1730         case HeaderFileTypeFixedVMShlib:        return eTypeSharedLibrary;  // 0x3u MH_FVMLIB
1731         case HeaderFileTypeCore:                return eTypeCoreFile;       // 0x4u MH_CORE
1732         case HeaderFileTypePreloadedExecutable: return eTypeSharedLibrary;  // 0x5u MH_PRELOAD
1733         case HeaderFileTypeDynamicShlib:        return eTypeSharedLibrary;  // 0x6u MH_DYLIB
1734         case HeaderFileTypeDynamicLinkEditor:   return eTypeDynamicLinker;  // 0x7u MH_DYLINKER
1735         case HeaderFileTypeBundle:              return eTypeSharedLibrary;  // 0x8u MH_BUNDLE
1736         case HeaderFileTypeDynamicShlibStub:    return eTypeStubLibrary;    // 0x9u MH_DYLIB_STUB
1737         case HeaderFileTypeDSYM:                return eTypeDebugInfo;      // 0xAu MH_DSYM
1738         case HeaderFileTypeKextBundle:          return eTypeSharedLibrary;  // 0xBu MH_KEXT_BUNDLE
1739         default:
1740             break;
1741     }
1742     return eTypeUnknown;
1743 }
1744 
1745 ObjectFile::Strata
1746 ObjectFileMachO::CalculateStrata()
1747 {
1748     switch (m_header.filetype)
1749     {
1750         case HeaderFileTypeObject:      // 0x1u MH_OBJECT
1751             {
1752                 // 32 bit kexts are just object files, but they do have a valid
1753                 // UUID load command.
1754                 UUID uuid;
1755                 if (GetUUID(&uuid))
1756                 {
1757                     // this checking for the UUID load command is not enough
1758                     // we could eventually look for the symbol named
1759                     // "OSKextGetCurrentIdentifier" as this is required of kexts
1760                     if (m_type == eTypeInvalid)
1761                         m_type = eTypeSharedLibrary;
1762 
1763                     return eStrataKernel;
1764                 }
1765             }
1766             return eStrataUnknown;
1767 
1768         case HeaderFileTypeExecutable:                                     // 0x2u MH_EXECUTE
1769             // Check for the MH_DYLDLINK bit in the flags
1770             if (m_header.flags & HeaderFlagBitIsDynamicLinkObject)
1771                 return eStrataUser;
1772             return eStrataKernel;
1773 
1774         case HeaderFileTypeFixedVMShlib:        return eStrataUser;         // 0x3u MH_FVMLIB
1775         case HeaderFileTypeCore:                return eStrataUnknown;      // 0x4u MH_CORE
1776         case HeaderFileTypePreloadedExecutable: return eStrataUser;         // 0x5u MH_PRELOAD
1777         case HeaderFileTypeDynamicShlib:        return eStrataUser;         // 0x6u MH_DYLIB
1778         case HeaderFileTypeDynamicLinkEditor:   return eStrataUser;         // 0x7u MH_DYLINKER
1779         case HeaderFileTypeBundle:              return eStrataUser;         // 0x8u MH_BUNDLE
1780         case HeaderFileTypeDynamicShlibStub:    return eStrataUser;         // 0x9u MH_DYLIB_STUB
1781         case HeaderFileTypeDSYM:                return eStrataUnknown;      // 0xAu MH_DSYM
1782         case HeaderFileTypeKextBundle:          return eStrataKernel;       // 0xBu MH_KEXT_BUNDLE
1783         default:
1784             break;
1785     }
1786     return eStrataUnknown;
1787 }
1788 
1789 
1790 bool
1791 ObjectFileMachO::GetArchitecture (ArchSpec &arch)
1792 {
1793     lldb_private::Mutex::Locker locker(m_mutex);
1794     arch.SetArchitecture (eArchTypeMachO, m_header.cputype, m_header.cpusubtype);
1795     return true;
1796 }
1797 
1798 
1799 //------------------------------------------------------------------
1800 // PluginInterface protocol
1801 //------------------------------------------------------------------
1802 const char *
1803 ObjectFileMachO::GetPluginName()
1804 {
1805     return "ObjectFileMachO";
1806 }
1807 
1808 const char *
1809 ObjectFileMachO::GetShortPluginName()
1810 {
1811     return GetPluginNameStatic();
1812 }
1813 
1814 uint32_t
1815 ObjectFileMachO::GetPluginVersion()
1816 {
1817     return 1;
1818 }
1819 
1820