1 //===-- ObjectFileMachO.cpp -----------------------------------------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 
9 #include "llvm/ADT/StringRef.h"
10 
11 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h"
12 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h"
13 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h"
14 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h"
15 #include "lldb/Core/Debugger.h"
16 #include "lldb/Core/FileSpecList.h"
17 #include "lldb/Core/Module.h"
18 #include "lldb/Core/ModuleSpec.h"
19 #include "lldb/Core/PluginManager.h"
20 #include "lldb/Core/Progress.h"
21 #include "lldb/Core/Section.h"
22 #include "lldb/Core/StreamFile.h"
23 #include "lldb/Host/Host.h"
24 #include "lldb/Host/SafeMachO.h"
25 #include "lldb/Symbol/DWARFCallFrameInfo.h"
26 #include "lldb/Symbol/LocateSymbolFile.h"
27 #include "lldb/Symbol/ObjectFile.h"
28 #include "lldb/Target/DynamicLoader.h"
29 #include "lldb/Target/MemoryRegionInfo.h"
30 #include "lldb/Target/Platform.h"
31 #include "lldb/Target/Process.h"
32 #include "lldb/Target/SectionLoadList.h"
33 #include "lldb/Target/Target.h"
34 #include "lldb/Target/Thread.h"
35 #include "lldb/Target/ThreadList.h"
36 #include "lldb/Utility/ArchSpec.h"
37 #include "lldb/Utility/DataBuffer.h"
38 #include "lldb/Utility/FileSpec.h"
39 #include "lldb/Utility/Log.h"
40 #include "lldb/Utility/RangeMap.h"
41 #include "lldb/Utility/RegisterValue.h"
42 #include "lldb/Utility/Status.h"
43 #include "lldb/Utility/StreamString.h"
44 #include "lldb/Utility/Timer.h"
45 #include "lldb/Utility/UUID.h"
46 
47 #include "llvm/ADT/DenseSet.h"
48 #include "llvm/Support/FormatVariadic.h"
49 #include "llvm/Support/MemoryBuffer.h"
50 
51 #include "ObjectFileMachO.h"
52 
53 #if defined(__APPLE__)
54 #include <TargetConditionals.h>
55 // GetLLDBSharedCacheUUID() needs to call dlsym()
56 #include <dlfcn.h>
57 #endif
58 
59 #ifndef __APPLE__
60 #include "Utility/UuidCompatibility.h"
61 #else
62 #include <uuid/uuid.h>
63 #endif
64 
65 #include <bitset>
66 #include <memory>
67 
68 #if LLVM_SUPPORT_XCODE_SIGNPOSTS
69 // Unfortunately the signpost header pulls in the system MachO header, too.
70 #undef CPU_TYPE_ARM
71 #undef CPU_TYPE_ARM64
72 #undef CPU_TYPE_ARM64_32
73 #undef CPU_TYPE_I386
74 #undef CPU_TYPE_X86_64
75 #undef MH_BINDATLOAD
76 #undef MH_BUNDLE
77 #undef MH_CIGAM
78 #undef MH_CIGAM_64
79 #undef MH_CORE
80 #undef MH_DSYM
81 #undef MH_DYLDLINK
82 #undef MH_DYLIB
83 #undef MH_DYLIB_STUB
84 #undef MH_DYLINKER
85 #undef MH_DYLINKER
86 #undef MH_EXECUTE
87 #undef MH_FVMLIB
88 #undef MH_INCRLINK
89 #undef MH_KEXT_BUNDLE
90 #undef MH_MAGIC
91 #undef MH_MAGIC_64
92 #undef MH_NOUNDEFS
93 #undef MH_OBJECT
94 #undef MH_OBJECT
95 #undef MH_PRELOAD
96 
97 #undef LC_BUILD_VERSION
98 #undef LC_VERSION_MIN_MACOSX
99 #undef LC_VERSION_MIN_IPHONEOS
100 #undef LC_VERSION_MIN_TVOS
101 #undef LC_VERSION_MIN_WATCHOS
102 
103 #undef PLATFORM_MACOS
104 #undef PLATFORM_MACCATALYST
105 #undef PLATFORM_IOS
106 #undef PLATFORM_IOSSIMULATOR
107 #undef PLATFORM_TVOS
108 #undef PLATFORM_TVOSSIMULATOR
109 #undef PLATFORM_WATCHOS
110 #undef PLATFORM_WATCHOSSIMULATOR
111 #endif
112 
113 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull
114 using namespace lldb;
115 using namespace lldb_private;
116 using namespace llvm::MachO;
117 
118 LLDB_PLUGIN_DEFINE(ObjectFileMachO)
119 
120 // Some structure definitions needed for parsing the dyld shared cache files
121 // found on iOS devices.
122 
123 struct lldb_copy_dyld_cache_header_v1 {
124   char magic[16];         // e.g. "dyld_v0    i386", "dyld_v1   armv7", etc.
125   uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info
126   uint32_t mappingCount;  // number of dyld_cache_mapping_info entries
127   uint32_t imagesOffset;
128   uint32_t imagesCount;
129   uint64_t dyldBaseAddress;
130   uint64_t codeSignatureOffset;
131   uint64_t codeSignatureSize;
132   uint64_t slideInfoOffset;
133   uint64_t slideInfoSize;
134   uint64_t localSymbolsOffset;
135   uint64_t localSymbolsSize;
136   uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13
137                     // and later
138 };
139 
140 struct lldb_copy_dyld_cache_mapping_info {
141   uint64_t address;
142   uint64_t size;
143   uint64_t fileOffset;
144   uint32_t maxProt;
145   uint32_t initProt;
146 };
147 
148 struct lldb_copy_dyld_cache_local_symbols_info {
149   uint32_t nlistOffset;
150   uint32_t nlistCount;
151   uint32_t stringsOffset;
152   uint32_t stringsSize;
153   uint32_t entriesOffset;
154   uint32_t entriesCount;
155 };
156 struct lldb_copy_dyld_cache_local_symbols_entry {
157   uint32_t dylibOffset;
158   uint32_t nlistStartIndex;
159   uint32_t nlistCount;
160 };
161 
PrintRegisterValue(RegisterContext * reg_ctx,const char * name,const char * alt_name,size_t reg_byte_size,Stream & data)162 static void PrintRegisterValue(RegisterContext *reg_ctx, const char *name,
163                                const char *alt_name, size_t reg_byte_size,
164                                Stream &data) {
165   const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name);
166   if (reg_info == nullptr)
167     reg_info = reg_ctx->GetRegisterInfoByName(alt_name);
168   if (reg_info) {
169     lldb_private::RegisterValue reg_value;
170     if (reg_ctx->ReadRegister(reg_info, reg_value)) {
171       if (reg_info->byte_size >= reg_byte_size)
172         data.Write(reg_value.GetBytes(), reg_byte_size);
173       else {
174         data.Write(reg_value.GetBytes(), reg_info->byte_size);
175         for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; ++i)
176           data.PutChar(0);
177       }
178       return;
179     }
180   }
181   // Just write zeros if all else fails
182   for (size_t i = 0; i < reg_byte_size; ++i)
183     data.PutChar(0);
184 }
185 
186 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 {
187 public:
RegisterContextDarwin_x86_64_Mach(lldb_private::Thread & thread,const DataExtractor & data)188   RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread,
189                                     const DataExtractor &data)
190       : RegisterContextDarwin_x86_64(thread, 0) {
191     SetRegisterDataFrom_LC_THREAD(data);
192   }
193 
InvalidateAllRegisters()194   void InvalidateAllRegisters() override {
195     // Do nothing... registers are always valid...
196   }
197 
SetRegisterDataFrom_LC_THREAD(const DataExtractor & data)198   void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) {
199     lldb::offset_t offset = 0;
200     SetError(GPRRegSet, Read, -1);
201     SetError(FPURegSet, Read, -1);
202     SetError(EXCRegSet, Read, -1);
203     bool done = false;
204 
205     while (!done) {
206       int flavor = data.GetU32(&offset);
207       if (flavor == 0)
208         done = true;
209       else {
210         uint32_t i;
211         uint32_t count = data.GetU32(&offset);
212         switch (flavor) {
213         case GPRRegSet:
214           for (i = 0; i < count; ++i)
215             (&gpr.rax)[i] = data.GetU64(&offset);
216           SetError(GPRRegSet, Read, 0);
217           done = true;
218 
219           break;
220         case FPURegSet:
221           // TODO: fill in FPU regs....
222           // SetError (FPURegSet, Read, -1);
223           done = true;
224 
225           break;
226         case EXCRegSet:
227           exc.trapno = data.GetU32(&offset);
228           exc.err = data.GetU32(&offset);
229           exc.faultvaddr = data.GetU64(&offset);
230           SetError(EXCRegSet, Read, 0);
231           done = true;
232           break;
233         case 7:
234         case 8:
235         case 9:
236           // fancy flavors that encapsulate of the above flavors...
237           break;
238 
239         default:
240           done = true;
241           break;
242         }
243       }
244     }
245   }
246 
Create_LC_THREAD(Thread * thread,Stream & data)247   static bool Create_LC_THREAD(Thread *thread, Stream &data) {
248     RegisterContextSP reg_ctx_sp(thread->GetRegisterContext());
249     if (reg_ctx_sp) {
250       RegisterContext *reg_ctx = reg_ctx_sp.get();
251 
252       data.PutHex32(GPRRegSet); // Flavor
253       data.PutHex32(GPRWordCount);
254       PrintRegisterValue(reg_ctx, "rax", nullptr, 8, data);
255       PrintRegisterValue(reg_ctx, "rbx", nullptr, 8, data);
256       PrintRegisterValue(reg_ctx, "rcx", nullptr, 8, data);
257       PrintRegisterValue(reg_ctx, "rdx", nullptr, 8, data);
258       PrintRegisterValue(reg_ctx, "rdi", nullptr, 8, data);
259       PrintRegisterValue(reg_ctx, "rsi", nullptr, 8, data);
260       PrintRegisterValue(reg_ctx, "rbp", nullptr, 8, data);
261       PrintRegisterValue(reg_ctx, "rsp", nullptr, 8, data);
262       PrintRegisterValue(reg_ctx, "r8", nullptr, 8, data);
263       PrintRegisterValue(reg_ctx, "r9", nullptr, 8, data);
264       PrintRegisterValue(reg_ctx, "r10", nullptr, 8, data);
265       PrintRegisterValue(reg_ctx, "r11", nullptr, 8, data);
266       PrintRegisterValue(reg_ctx, "r12", nullptr, 8, data);
267       PrintRegisterValue(reg_ctx, "r13", nullptr, 8, data);
268       PrintRegisterValue(reg_ctx, "r14", nullptr, 8, data);
269       PrintRegisterValue(reg_ctx, "r15", nullptr, 8, data);
270       PrintRegisterValue(reg_ctx, "rip", nullptr, 8, data);
271       PrintRegisterValue(reg_ctx, "rflags", nullptr, 8, data);
272       PrintRegisterValue(reg_ctx, "cs", nullptr, 8, data);
273       PrintRegisterValue(reg_ctx, "fs", nullptr, 8, data);
274       PrintRegisterValue(reg_ctx, "gs", nullptr, 8, data);
275 
276       //            // Write out the FPU registers
277       //            const size_t fpu_byte_size = sizeof(FPU);
278       //            size_t bytes_written = 0;
279       //            data.PutHex32 (FPURegSet);
280       //            data.PutHex32 (fpu_byte_size/sizeof(uint64_t));
281       //            bytes_written += data.PutHex32(0); // uint32_t pad[0]
282       //            bytes_written += data.PutHex32(0); // uint32_t pad[1]
283       //            bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2,
284       //            data);   // uint16_t    fcw;    // "fctrl"
285       //            bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2,
286       //            data);  // uint16_t    fsw;    // "fstat"
287       //            bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1,
288       //            data);   // uint8_t     ftw;    // "ftag"
289       //            bytes_written += data.PutHex8  (0); // uint8_t pad1;
290       //            bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2,
291       //            data);     // uint16_t    fop;    // "fop"
292       //            bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4,
293       //            data);    // uint32_t    ip;     // "fioff"
294       //            bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2,
295       //            data);    // uint16_t    cs;     // "fiseg"
296       //            bytes_written += data.PutHex16 (0); // uint16_t    pad2;
297       //            bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4,
298       //            data);   // uint32_t    dp;     // "fooff"
299       //            bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2,
300       //            data);    // uint16_t    ds;     // "foseg"
301       //            bytes_written += data.PutHex16 (0); // uint16_t    pad3;
302       //            bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4,
303       //            data);    // uint32_t    mxcsr;
304       //            bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL,
305       //            4, data);// uint32_t    mxcsrmask;
306       //            bytes_written += WriteRegister (reg_ctx, "stmm0", NULL,
307       //            sizeof(MMSReg), data);
308       //            bytes_written += WriteRegister (reg_ctx, "stmm1", NULL,
309       //            sizeof(MMSReg), data);
310       //            bytes_written += WriteRegister (reg_ctx, "stmm2", NULL,
311       //            sizeof(MMSReg), data);
312       //            bytes_written += WriteRegister (reg_ctx, "stmm3", NULL,
313       //            sizeof(MMSReg), data);
314       //            bytes_written += WriteRegister (reg_ctx, "stmm4", NULL,
315       //            sizeof(MMSReg), data);
316       //            bytes_written += WriteRegister (reg_ctx, "stmm5", NULL,
317       //            sizeof(MMSReg), data);
318       //            bytes_written += WriteRegister (reg_ctx, "stmm6", NULL,
319       //            sizeof(MMSReg), data);
320       //            bytes_written += WriteRegister (reg_ctx, "stmm7", NULL,
321       //            sizeof(MMSReg), data);
322       //            bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL,
323       //            sizeof(XMMReg), data);
324       //            bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL,
325       //            sizeof(XMMReg), data);
326       //            bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL,
327       //            sizeof(XMMReg), data);
328       //            bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL,
329       //            sizeof(XMMReg), data);
330       //            bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL,
331       //            sizeof(XMMReg), data);
332       //            bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL,
333       //            sizeof(XMMReg), data);
334       //            bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL,
335       //            sizeof(XMMReg), data);
336       //            bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL,
337       //            sizeof(XMMReg), data);
338       //            bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL,
339       //            sizeof(XMMReg), data);
340       //            bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL,
341       //            sizeof(XMMReg), data);
342       //            bytes_written += WriteRegister (reg_ctx, "xmm10", NULL,
343       //            sizeof(XMMReg), data);
344       //            bytes_written += WriteRegister (reg_ctx, "xmm11", NULL,
345       //            sizeof(XMMReg), data);
346       //            bytes_written += WriteRegister (reg_ctx, "xmm12", NULL,
347       //            sizeof(XMMReg), data);
348       //            bytes_written += WriteRegister (reg_ctx, "xmm13", NULL,
349       //            sizeof(XMMReg), data);
350       //            bytes_written += WriteRegister (reg_ctx, "xmm14", NULL,
351       //            sizeof(XMMReg), data);
352       //            bytes_written += WriteRegister (reg_ctx, "xmm15", NULL,
353       //            sizeof(XMMReg), data);
354       //
355       //            // Fill rest with zeros
356       //            for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++
357       //            i)
358       //                data.PutChar(0);
359 
360       // Write out the EXC registers
361       data.PutHex32(EXCRegSet);
362       data.PutHex32(EXCWordCount);
363       PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data);
364       PrintRegisterValue(reg_ctx, "err", nullptr, 4, data);
365       PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 8, data);
366       return true;
367     }
368     return false;
369   }
370 
371 protected:
DoReadGPR(lldb::tid_t tid,int flavor,GPR & gpr)372   int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; }
373 
DoReadFPU(lldb::tid_t tid,int flavor,FPU & fpu)374   int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; }
375 
DoReadEXC(lldb::tid_t tid,int flavor,EXC & exc)376   int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; }
377 
DoWriteGPR(lldb::tid_t tid,int flavor,const GPR & gpr)378   int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override {
379     return 0;
380   }
381 
DoWriteFPU(lldb::tid_t tid,int flavor,const FPU & fpu)382   int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override {
383     return 0;
384   }
385 
DoWriteEXC(lldb::tid_t tid,int flavor,const EXC & exc)386   int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override {
387     return 0;
388   }
389 };
390 
391 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 {
392 public:
RegisterContextDarwin_i386_Mach(lldb_private::Thread & thread,const DataExtractor & data)393   RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread,
394                                   const DataExtractor &data)
395       : RegisterContextDarwin_i386(thread, 0) {
396     SetRegisterDataFrom_LC_THREAD(data);
397   }
398 
InvalidateAllRegisters()399   void InvalidateAllRegisters() override {
400     // Do nothing... registers are always valid...
401   }
402 
SetRegisterDataFrom_LC_THREAD(const DataExtractor & data)403   void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) {
404     lldb::offset_t offset = 0;
405     SetError(GPRRegSet, Read, -1);
406     SetError(FPURegSet, Read, -1);
407     SetError(EXCRegSet, Read, -1);
408     bool done = false;
409 
410     while (!done) {
411       int flavor = data.GetU32(&offset);
412       if (flavor == 0)
413         done = true;
414       else {
415         uint32_t i;
416         uint32_t count = data.GetU32(&offset);
417         switch (flavor) {
418         case GPRRegSet:
419           for (i = 0; i < count; ++i)
420             (&gpr.eax)[i] = data.GetU32(&offset);
421           SetError(GPRRegSet, Read, 0);
422           done = true;
423 
424           break;
425         case FPURegSet:
426           // TODO: fill in FPU regs....
427           // SetError (FPURegSet, Read, -1);
428           done = true;
429 
430           break;
431         case EXCRegSet:
432           exc.trapno = data.GetU32(&offset);
433           exc.err = data.GetU32(&offset);
434           exc.faultvaddr = data.GetU32(&offset);
435           SetError(EXCRegSet, Read, 0);
436           done = true;
437           break;
438         case 7:
439         case 8:
440         case 9:
441           // fancy flavors that encapsulate of the above flavors...
442           break;
443 
444         default:
445           done = true;
446           break;
447         }
448       }
449     }
450   }
451 
Create_LC_THREAD(Thread * thread,Stream & data)452   static bool Create_LC_THREAD(Thread *thread, Stream &data) {
453     RegisterContextSP reg_ctx_sp(thread->GetRegisterContext());
454     if (reg_ctx_sp) {
455       RegisterContext *reg_ctx = reg_ctx_sp.get();
456 
457       data.PutHex32(GPRRegSet); // Flavor
458       data.PutHex32(GPRWordCount);
459       PrintRegisterValue(reg_ctx, "eax", nullptr, 4, data);
460       PrintRegisterValue(reg_ctx, "ebx", nullptr, 4, data);
461       PrintRegisterValue(reg_ctx, "ecx", nullptr, 4, data);
462       PrintRegisterValue(reg_ctx, "edx", nullptr, 4, data);
463       PrintRegisterValue(reg_ctx, "edi", nullptr, 4, data);
464       PrintRegisterValue(reg_ctx, "esi", nullptr, 4, data);
465       PrintRegisterValue(reg_ctx, "ebp", nullptr, 4, data);
466       PrintRegisterValue(reg_ctx, "esp", nullptr, 4, data);
467       PrintRegisterValue(reg_ctx, "ss", nullptr, 4, data);
468       PrintRegisterValue(reg_ctx, "eflags", nullptr, 4, data);
469       PrintRegisterValue(reg_ctx, "eip", nullptr, 4, data);
470       PrintRegisterValue(reg_ctx, "cs", nullptr, 4, data);
471       PrintRegisterValue(reg_ctx, "ds", nullptr, 4, data);
472       PrintRegisterValue(reg_ctx, "es", nullptr, 4, data);
473       PrintRegisterValue(reg_ctx, "fs", nullptr, 4, data);
474       PrintRegisterValue(reg_ctx, "gs", nullptr, 4, data);
475 
476       // Write out the EXC registers
477       data.PutHex32(EXCRegSet);
478       data.PutHex32(EXCWordCount);
479       PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data);
480       PrintRegisterValue(reg_ctx, "err", nullptr, 4, data);
481       PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 4, data);
482       return true;
483     }
484     return false;
485   }
486 
487 protected:
DoReadGPR(lldb::tid_t tid,int flavor,GPR & gpr)488   int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; }
489 
DoReadFPU(lldb::tid_t tid,int flavor,FPU & fpu)490   int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; }
491 
DoReadEXC(lldb::tid_t tid,int flavor,EXC & exc)492   int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; }
493 
DoWriteGPR(lldb::tid_t tid,int flavor,const GPR & gpr)494   int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override {
495     return 0;
496   }
497 
DoWriteFPU(lldb::tid_t tid,int flavor,const FPU & fpu)498   int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override {
499     return 0;
500   }
501 
DoWriteEXC(lldb::tid_t tid,int flavor,const EXC & exc)502   int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override {
503     return 0;
504   }
505 };
506 
507 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm {
508 public:
RegisterContextDarwin_arm_Mach(lldb_private::Thread & thread,const DataExtractor & data)509   RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread,
510                                  const DataExtractor &data)
511       : RegisterContextDarwin_arm(thread, 0) {
512     SetRegisterDataFrom_LC_THREAD(data);
513   }
514 
InvalidateAllRegisters()515   void InvalidateAllRegisters() override {
516     // Do nothing... registers are always valid...
517   }
518 
SetRegisterDataFrom_LC_THREAD(const DataExtractor & data)519   void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) {
520     lldb::offset_t offset = 0;
521     SetError(GPRRegSet, Read, -1);
522     SetError(FPURegSet, Read, -1);
523     SetError(EXCRegSet, Read, -1);
524     bool done = false;
525 
526     while (!done) {
527       int flavor = data.GetU32(&offset);
528       uint32_t count = data.GetU32(&offset);
529       lldb::offset_t next_thread_state = offset + (count * 4);
530       switch (flavor) {
531       case GPRAltRegSet:
532       case GPRRegSet:
533         // On ARM, the CPSR register is also included in the count but it is
534         // not included in gpr.r so loop until (count-1).
535         for (uint32_t i = 0; i < (count - 1); ++i) {
536           gpr.r[i] = data.GetU32(&offset);
537         }
538         // Save cpsr explicitly.
539         gpr.cpsr = data.GetU32(&offset);
540 
541         SetError(GPRRegSet, Read, 0);
542         offset = next_thread_state;
543         break;
544 
545       case FPURegSet: {
546         uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0];
547         const int fpu_reg_buf_size = sizeof(fpu.floats);
548         if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle,
549                               fpu_reg_buf) == fpu_reg_buf_size) {
550           offset += fpu_reg_buf_size;
551           fpu.fpscr = data.GetU32(&offset);
552           SetError(FPURegSet, Read, 0);
553         } else {
554           done = true;
555         }
556       }
557         offset = next_thread_state;
558         break;
559 
560       case EXCRegSet:
561         if (count == 3) {
562           exc.exception = data.GetU32(&offset);
563           exc.fsr = data.GetU32(&offset);
564           exc.far = data.GetU32(&offset);
565           SetError(EXCRegSet, Read, 0);
566         }
567         done = true;
568         offset = next_thread_state;
569         break;
570 
571       // Unknown register set flavor, stop trying to parse.
572       default:
573         done = true;
574       }
575     }
576   }
577 
Create_LC_THREAD(Thread * thread,Stream & data)578   static bool Create_LC_THREAD(Thread *thread, Stream &data) {
579     RegisterContextSP reg_ctx_sp(thread->GetRegisterContext());
580     if (reg_ctx_sp) {
581       RegisterContext *reg_ctx = reg_ctx_sp.get();
582 
583       data.PutHex32(GPRRegSet); // Flavor
584       data.PutHex32(GPRWordCount);
585       PrintRegisterValue(reg_ctx, "r0", nullptr, 4, data);
586       PrintRegisterValue(reg_ctx, "r1", nullptr, 4, data);
587       PrintRegisterValue(reg_ctx, "r2", nullptr, 4, data);
588       PrintRegisterValue(reg_ctx, "r3", nullptr, 4, data);
589       PrintRegisterValue(reg_ctx, "r4", nullptr, 4, data);
590       PrintRegisterValue(reg_ctx, "r5", nullptr, 4, data);
591       PrintRegisterValue(reg_ctx, "r6", nullptr, 4, data);
592       PrintRegisterValue(reg_ctx, "r7", nullptr, 4, data);
593       PrintRegisterValue(reg_ctx, "r8", nullptr, 4, data);
594       PrintRegisterValue(reg_ctx, "r9", nullptr, 4, data);
595       PrintRegisterValue(reg_ctx, "r10", nullptr, 4, data);
596       PrintRegisterValue(reg_ctx, "r11", nullptr, 4, data);
597       PrintRegisterValue(reg_ctx, "r12", nullptr, 4, data);
598       PrintRegisterValue(reg_ctx, "sp", nullptr, 4, data);
599       PrintRegisterValue(reg_ctx, "lr", nullptr, 4, data);
600       PrintRegisterValue(reg_ctx, "pc", nullptr, 4, data);
601       PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data);
602 
603       // Write out the EXC registers
604       //            data.PutHex32 (EXCRegSet);
605       //            data.PutHex32 (EXCWordCount);
606       //            WriteRegister (reg_ctx, "exception", NULL, 4, data);
607       //            WriteRegister (reg_ctx, "fsr", NULL, 4, data);
608       //            WriteRegister (reg_ctx, "far", NULL, 4, data);
609       return true;
610     }
611     return false;
612   }
613 
614 protected:
DoReadGPR(lldb::tid_t tid,int flavor,GPR & gpr)615   int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; }
616 
DoReadFPU(lldb::tid_t tid,int flavor,FPU & fpu)617   int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; }
618 
DoReadEXC(lldb::tid_t tid,int flavor,EXC & exc)619   int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; }
620 
DoReadDBG(lldb::tid_t tid,int flavor,DBG & dbg)621   int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; }
622 
DoWriteGPR(lldb::tid_t tid,int flavor,const GPR & gpr)623   int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override {
624     return 0;
625   }
626 
DoWriteFPU(lldb::tid_t tid,int flavor,const FPU & fpu)627   int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override {
628     return 0;
629   }
630 
DoWriteEXC(lldb::tid_t tid,int flavor,const EXC & exc)631   int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override {
632     return 0;
633   }
634 
DoWriteDBG(lldb::tid_t tid,int flavor,const DBG & dbg)635   int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override {
636     return -1;
637   }
638 };
639 
640 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 {
641 public:
RegisterContextDarwin_arm64_Mach(lldb_private::Thread & thread,const DataExtractor & data)642   RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread,
643                                    const DataExtractor &data)
644       : RegisterContextDarwin_arm64(thread, 0) {
645     SetRegisterDataFrom_LC_THREAD(data);
646   }
647 
InvalidateAllRegisters()648   void InvalidateAllRegisters() override {
649     // Do nothing... registers are always valid...
650   }
651 
SetRegisterDataFrom_LC_THREAD(const DataExtractor & data)652   void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) {
653     lldb::offset_t offset = 0;
654     SetError(GPRRegSet, Read, -1);
655     SetError(FPURegSet, Read, -1);
656     SetError(EXCRegSet, Read, -1);
657     bool done = false;
658     while (!done) {
659       int flavor = data.GetU32(&offset);
660       uint32_t count = data.GetU32(&offset);
661       lldb::offset_t next_thread_state = offset + (count * 4);
662       switch (flavor) {
663       case GPRRegSet:
664         // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1
665         // 32-bit register)
666         if (count >= (33 * 2) + 1) {
667           for (uint32_t i = 0; i < 29; ++i)
668             gpr.x[i] = data.GetU64(&offset);
669           gpr.fp = data.GetU64(&offset);
670           gpr.lr = data.GetU64(&offset);
671           gpr.sp = data.GetU64(&offset);
672           gpr.pc = data.GetU64(&offset);
673           gpr.cpsr = data.GetU32(&offset);
674           SetError(GPRRegSet, Read, 0);
675         }
676         offset = next_thread_state;
677         break;
678       case FPURegSet: {
679         uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0];
680         const int fpu_reg_buf_size = sizeof(fpu);
681         if (fpu_reg_buf_size == count * sizeof(uint32_t) &&
682             data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle,
683                               fpu_reg_buf) == fpu_reg_buf_size) {
684           SetError(FPURegSet, Read, 0);
685         } else {
686           done = true;
687         }
688       }
689         offset = next_thread_state;
690         break;
691       case EXCRegSet:
692         if (count == 4) {
693           exc.far = data.GetU64(&offset);
694           exc.esr = data.GetU32(&offset);
695           exc.exception = data.GetU32(&offset);
696           SetError(EXCRegSet, Read, 0);
697         }
698         offset = next_thread_state;
699         break;
700       default:
701         done = true;
702         break;
703       }
704     }
705   }
706 
Create_LC_THREAD(Thread * thread,Stream & data)707   static bool Create_LC_THREAD(Thread *thread, Stream &data) {
708     RegisterContextSP reg_ctx_sp(thread->GetRegisterContext());
709     if (reg_ctx_sp) {
710       RegisterContext *reg_ctx = reg_ctx_sp.get();
711 
712       data.PutHex32(GPRRegSet); // Flavor
713       data.PutHex32(GPRWordCount);
714       PrintRegisterValue(reg_ctx, "x0", nullptr, 8, data);
715       PrintRegisterValue(reg_ctx, "x1", nullptr, 8, data);
716       PrintRegisterValue(reg_ctx, "x2", nullptr, 8, data);
717       PrintRegisterValue(reg_ctx, "x3", nullptr, 8, data);
718       PrintRegisterValue(reg_ctx, "x4", nullptr, 8, data);
719       PrintRegisterValue(reg_ctx, "x5", nullptr, 8, data);
720       PrintRegisterValue(reg_ctx, "x6", nullptr, 8, data);
721       PrintRegisterValue(reg_ctx, "x7", nullptr, 8, data);
722       PrintRegisterValue(reg_ctx, "x8", nullptr, 8, data);
723       PrintRegisterValue(reg_ctx, "x9", nullptr, 8, data);
724       PrintRegisterValue(reg_ctx, "x10", nullptr, 8, data);
725       PrintRegisterValue(reg_ctx, "x11", nullptr, 8, data);
726       PrintRegisterValue(reg_ctx, "x12", nullptr, 8, data);
727       PrintRegisterValue(reg_ctx, "x13", nullptr, 8, data);
728       PrintRegisterValue(reg_ctx, "x14", nullptr, 8, data);
729       PrintRegisterValue(reg_ctx, "x15", nullptr, 8, data);
730       PrintRegisterValue(reg_ctx, "x16", nullptr, 8, data);
731       PrintRegisterValue(reg_ctx, "x17", nullptr, 8, data);
732       PrintRegisterValue(reg_ctx, "x18", nullptr, 8, data);
733       PrintRegisterValue(reg_ctx, "x19", nullptr, 8, data);
734       PrintRegisterValue(reg_ctx, "x20", nullptr, 8, data);
735       PrintRegisterValue(reg_ctx, "x21", nullptr, 8, data);
736       PrintRegisterValue(reg_ctx, "x22", nullptr, 8, data);
737       PrintRegisterValue(reg_ctx, "x23", nullptr, 8, data);
738       PrintRegisterValue(reg_ctx, "x24", nullptr, 8, data);
739       PrintRegisterValue(reg_ctx, "x25", nullptr, 8, data);
740       PrintRegisterValue(reg_ctx, "x26", nullptr, 8, data);
741       PrintRegisterValue(reg_ctx, "x27", nullptr, 8, data);
742       PrintRegisterValue(reg_ctx, "x28", nullptr, 8, data);
743       PrintRegisterValue(reg_ctx, "fp", nullptr, 8, data);
744       PrintRegisterValue(reg_ctx, "lr", nullptr, 8, data);
745       PrintRegisterValue(reg_ctx, "sp", nullptr, 8, data);
746       PrintRegisterValue(reg_ctx, "pc", nullptr, 8, data);
747       PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data);
748 
749       // Write out the EXC registers
750       //            data.PutHex32 (EXCRegSet);
751       //            data.PutHex32 (EXCWordCount);
752       //            WriteRegister (reg_ctx, "far", NULL, 8, data);
753       //            WriteRegister (reg_ctx, "esr", NULL, 4, data);
754       //            WriteRegister (reg_ctx, "exception", NULL, 4, data);
755       return true;
756     }
757     return false;
758   }
759 
760 protected:
DoReadGPR(lldb::tid_t tid,int flavor,GPR & gpr)761   int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; }
762 
DoReadFPU(lldb::tid_t tid,int flavor,FPU & fpu)763   int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; }
764 
DoReadEXC(lldb::tid_t tid,int flavor,EXC & exc)765   int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; }
766 
DoReadDBG(lldb::tid_t tid,int flavor,DBG & dbg)767   int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; }
768 
DoWriteGPR(lldb::tid_t tid,int flavor,const GPR & gpr)769   int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override {
770     return 0;
771   }
772 
DoWriteFPU(lldb::tid_t tid,int flavor,const FPU & fpu)773   int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override {
774     return 0;
775   }
776 
DoWriteEXC(lldb::tid_t tid,int flavor,const EXC & exc)777   int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override {
778     return 0;
779   }
780 
DoWriteDBG(lldb::tid_t tid,int flavor,const DBG & dbg)781   int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override {
782     return -1;
783   }
784 };
785 
MachHeaderSizeFromMagic(uint32_t magic)786 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) {
787   switch (magic) {
788   case MH_MAGIC:
789   case MH_CIGAM:
790     return sizeof(struct llvm::MachO::mach_header);
791 
792   case MH_MAGIC_64:
793   case MH_CIGAM_64:
794     return sizeof(struct llvm::MachO::mach_header_64);
795     break;
796 
797   default:
798     break;
799   }
800   return 0;
801 }
802 
803 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008
804 
805 char ObjectFileMachO::ID;
806 
Initialize()807 void ObjectFileMachO::Initialize() {
808   PluginManager::RegisterPlugin(
809       GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance,
810       CreateMemoryInstance, GetModuleSpecifications, SaveCore);
811 }
812 
Terminate()813 void ObjectFileMachO::Terminate() {
814   PluginManager::UnregisterPlugin(CreateInstance);
815 }
816 
GetPluginNameStatic()817 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() {
818   static ConstString g_name("mach-o");
819   return g_name;
820 }
821 
GetPluginDescriptionStatic()822 const char *ObjectFileMachO::GetPluginDescriptionStatic() {
823   return "Mach-o object file reader (32 and 64 bit)";
824 }
825 
CreateInstance(const lldb::ModuleSP & module_sp,DataBufferSP & data_sp,lldb::offset_t data_offset,const FileSpec * file,lldb::offset_t file_offset,lldb::offset_t length)826 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp,
827                                             DataBufferSP &data_sp,
828                                             lldb::offset_t data_offset,
829                                             const FileSpec *file,
830                                             lldb::offset_t file_offset,
831                                             lldb::offset_t length) {
832   if (!data_sp) {
833     data_sp = MapFileData(*file, length, file_offset);
834     if (!data_sp)
835       return nullptr;
836     data_offset = 0;
837   }
838 
839   if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length))
840     return nullptr;
841 
842   // Update the data to contain the entire file if it doesn't already
843   if (data_sp->GetByteSize() < length) {
844     data_sp = MapFileData(*file, length, file_offset);
845     if (!data_sp)
846       return nullptr;
847     data_offset = 0;
848   }
849   auto objfile_up = std::make_unique<ObjectFileMachO>(
850       module_sp, data_sp, data_offset, file, file_offset, length);
851   if (!objfile_up || !objfile_up->ParseHeader())
852     return nullptr;
853 
854   return objfile_up.release();
855 }
856 
CreateMemoryInstance(const lldb::ModuleSP & module_sp,DataBufferSP & data_sp,const ProcessSP & process_sp,lldb::addr_t header_addr)857 ObjectFile *ObjectFileMachO::CreateMemoryInstance(
858     const lldb::ModuleSP &module_sp, DataBufferSP &data_sp,
859     const ProcessSP &process_sp, lldb::addr_t header_addr) {
860   if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) {
861     std::unique_ptr<ObjectFile> objfile_up(
862         new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr));
863     if (objfile_up.get() && objfile_up->ParseHeader())
864       return objfile_up.release();
865   }
866   return nullptr;
867 }
868 
GetModuleSpecifications(const lldb_private::FileSpec & file,lldb::DataBufferSP & data_sp,lldb::offset_t data_offset,lldb::offset_t file_offset,lldb::offset_t length,lldb_private::ModuleSpecList & specs)869 size_t ObjectFileMachO::GetModuleSpecifications(
870     const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp,
871     lldb::offset_t data_offset, lldb::offset_t file_offset,
872     lldb::offset_t length, lldb_private::ModuleSpecList &specs) {
873   const size_t initial_count = specs.GetSize();
874 
875   if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) {
876     DataExtractor data;
877     data.SetData(data_sp);
878     llvm::MachO::mach_header header;
879     if (ParseHeader(data, &data_offset, header)) {
880       size_t header_and_load_cmds =
881           header.sizeofcmds + MachHeaderSizeFromMagic(header.magic);
882       if (header_and_load_cmds >= data_sp->GetByteSize()) {
883         data_sp = MapFileData(file, header_and_load_cmds, file_offset);
884         data.SetData(data_sp);
885         data_offset = MachHeaderSizeFromMagic(header.magic);
886       }
887       if (data_sp) {
888         ModuleSpec base_spec;
889         base_spec.GetFileSpec() = file;
890         base_spec.SetObjectOffset(file_offset);
891         base_spec.SetObjectSize(length);
892         GetAllArchSpecs(header, data, data_offset, base_spec, specs);
893       }
894     }
895   }
896   return specs.GetSize() - initial_count;
897 }
898 
GetSegmentNameTEXT()899 ConstString ObjectFileMachO::GetSegmentNameTEXT() {
900   static ConstString g_segment_name_TEXT("__TEXT");
901   return g_segment_name_TEXT;
902 }
903 
GetSegmentNameDATA()904 ConstString ObjectFileMachO::GetSegmentNameDATA() {
905   static ConstString g_segment_name_DATA("__DATA");
906   return g_segment_name_DATA;
907 }
908 
GetSegmentNameDATA_DIRTY()909 ConstString ObjectFileMachO::GetSegmentNameDATA_DIRTY() {
910   static ConstString g_segment_name("__DATA_DIRTY");
911   return g_segment_name;
912 }
913 
GetSegmentNameDATA_CONST()914 ConstString ObjectFileMachO::GetSegmentNameDATA_CONST() {
915   static ConstString g_segment_name("__DATA_CONST");
916   return g_segment_name;
917 }
918 
GetSegmentNameOBJC()919 ConstString ObjectFileMachO::GetSegmentNameOBJC() {
920   static ConstString g_segment_name_OBJC("__OBJC");
921   return g_segment_name_OBJC;
922 }
923 
GetSegmentNameLINKEDIT()924 ConstString ObjectFileMachO::GetSegmentNameLINKEDIT() {
925   static ConstString g_section_name_LINKEDIT("__LINKEDIT");
926   return g_section_name_LINKEDIT;
927 }
928 
GetSegmentNameDWARF()929 ConstString ObjectFileMachO::GetSegmentNameDWARF() {
930   static ConstString g_section_name("__DWARF");
931   return g_section_name;
932 }
933 
GetSectionNameEHFrame()934 ConstString ObjectFileMachO::GetSectionNameEHFrame() {
935   static ConstString g_section_name_eh_frame("__eh_frame");
936   return g_section_name_eh_frame;
937 }
938 
MagicBytesMatch(DataBufferSP & data_sp,lldb::addr_t data_offset,lldb::addr_t data_length)939 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp,
940                                       lldb::addr_t data_offset,
941                                       lldb::addr_t data_length) {
942   DataExtractor data;
943   data.SetData(data_sp, data_offset, data_length);
944   lldb::offset_t offset = 0;
945   uint32_t magic = data.GetU32(&offset);
946   return MachHeaderSizeFromMagic(magic) != 0;
947 }
948 
ObjectFileMachO(const lldb::ModuleSP & module_sp,DataBufferSP & data_sp,lldb::offset_t data_offset,const FileSpec * file,lldb::offset_t file_offset,lldb::offset_t length)949 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp,
950                                  DataBufferSP &data_sp,
951                                  lldb::offset_t data_offset,
952                                  const FileSpec *file,
953                                  lldb::offset_t file_offset,
954                                  lldb::offset_t length)
955     : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset),
956       m_mach_segments(), m_mach_sections(), m_entry_point_address(),
957       m_thread_context_offsets(), m_thread_context_offsets_valid(false),
958       m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) {
959   ::memset(&m_header, 0, sizeof(m_header));
960   ::memset(&m_dysymtab, 0, sizeof(m_dysymtab));
961 }
962 
ObjectFileMachO(const lldb::ModuleSP & module_sp,lldb::DataBufferSP & header_data_sp,const lldb::ProcessSP & process_sp,lldb::addr_t header_addr)963 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp,
964                                  lldb::DataBufferSP &header_data_sp,
965                                  const lldb::ProcessSP &process_sp,
966                                  lldb::addr_t header_addr)
967     : ObjectFile(module_sp, process_sp, header_addr, header_data_sp),
968       m_mach_segments(), m_mach_sections(), m_entry_point_address(),
969       m_thread_context_offsets(), m_thread_context_offsets_valid(false),
970       m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) {
971   ::memset(&m_header, 0, sizeof(m_header));
972   ::memset(&m_dysymtab, 0, sizeof(m_dysymtab));
973 }
974 
ParseHeader(DataExtractor & data,lldb::offset_t * data_offset_ptr,llvm::MachO::mach_header & header)975 bool ObjectFileMachO::ParseHeader(DataExtractor &data,
976                                   lldb::offset_t *data_offset_ptr,
977                                   llvm::MachO::mach_header &header) {
978   data.SetByteOrder(endian::InlHostByteOrder());
979   // Leave magic in the original byte order
980   header.magic = data.GetU32(data_offset_ptr);
981   bool can_parse = false;
982   bool is_64_bit = false;
983   switch (header.magic) {
984   case MH_MAGIC:
985     data.SetByteOrder(endian::InlHostByteOrder());
986     data.SetAddressByteSize(4);
987     can_parse = true;
988     break;
989 
990   case MH_MAGIC_64:
991     data.SetByteOrder(endian::InlHostByteOrder());
992     data.SetAddressByteSize(8);
993     can_parse = true;
994     is_64_bit = true;
995     break;
996 
997   case MH_CIGAM:
998     data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig
999                           ? eByteOrderLittle
1000                           : eByteOrderBig);
1001     data.SetAddressByteSize(4);
1002     can_parse = true;
1003     break;
1004 
1005   case MH_CIGAM_64:
1006     data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig
1007                           ? eByteOrderLittle
1008                           : eByteOrderBig);
1009     data.SetAddressByteSize(8);
1010     is_64_bit = true;
1011     can_parse = true;
1012     break;
1013 
1014   default:
1015     break;
1016   }
1017 
1018   if (can_parse) {
1019     data.GetU32(data_offset_ptr, &header.cputype, 6);
1020     if (is_64_bit)
1021       *data_offset_ptr += 4;
1022     return true;
1023   } else {
1024     memset(&header, 0, sizeof(header));
1025   }
1026   return false;
1027 }
1028 
ParseHeader()1029 bool ObjectFileMachO::ParseHeader() {
1030   ModuleSP module_sp(GetModule());
1031   if (!module_sp)
1032     return false;
1033 
1034   std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
1035   bool can_parse = false;
1036   lldb::offset_t offset = 0;
1037   m_data.SetByteOrder(endian::InlHostByteOrder());
1038   // Leave magic in the original byte order
1039   m_header.magic = m_data.GetU32(&offset);
1040   switch (m_header.magic) {
1041   case MH_MAGIC:
1042     m_data.SetByteOrder(endian::InlHostByteOrder());
1043     m_data.SetAddressByteSize(4);
1044     can_parse = true;
1045     break;
1046 
1047   case MH_MAGIC_64:
1048     m_data.SetByteOrder(endian::InlHostByteOrder());
1049     m_data.SetAddressByteSize(8);
1050     can_parse = true;
1051     break;
1052 
1053   case MH_CIGAM:
1054     m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig
1055                             ? eByteOrderLittle
1056                             : eByteOrderBig);
1057     m_data.SetAddressByteSize(4);
1058     can_parse = true;
1059     break;
1060 
1061   case MH_CIGAM_64:
1062     m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig
1063                             ? eByteOrderLittle
1064                             : eByteOrderBig);
1065     m_data.SetAddressByteSize(8);
1066     can_parse = true;
1067     break;
1068 
1069   default:
1070     break;
1071   }
1072 
1073   if (can_parse) {
1074     m_data.GetU32(&offset, &m_header.cputype, 6);
1075 
1076     ModuleSpecList all_specs;
1077     ModuleSpec base_spec;
1078     GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic),
1079                     base_spec, all_specs);
1080 
1081     for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) {
1082       ArchSpec mach_arch =
1083           all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture();
1084 
1085       // Check if the module has a required architecture
1086       const ArchSpec &module_arch = module_sp->GetArchitecture();
1087       if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch))
1088         continue;
1089 
1090       if (SetModulesArchitecture(mach_arch)) {
1091         const size_t header_and_lc_size =
1092             m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic);
1093         if (m_data.GetByteSize() < header_and_lc_size) {
1094           DataBufferSP data_sp;
1095           ProcessSP process_sp(m_process_wp.lock());
1096           if (process_sp) {
1097             data_sp = ReadMemory(process_sp, m_memory_addr, header_and_lc_size);
1098           } else {
1099             // Read in all only the load command data from the file on disk
1100             data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset);
1101             if (data_sp->GetByteSize() != header_and_lc_size)
1102               continue;
1103           }
1104           if (data_sp)
1105             m_data.SetData(data_sp);
1106         }
1107       }
1108       return true;
1109     }
1110     // None found.
1111     return false;
1112   } else {
1113     memset(&m_header, 0, sizeof(struct llvm::MachO::mach_header));
1114   }
1115   return false;
1116 }
1117 
GetByteOrder() const1118 ByteOrder ObjectFileMachO::GetByteOrder() const {
1119   return m_data.GetByteOrder();
1120 }
1121 
IsExecutable() const1122 bool ObjectFileMachO::IsExecutable() const {
1123   return m_header.filetype == MH_EXECUTE;
1124 }
1125 
IsDynamicLoader() const1126 bool ObjectFileMachO::IsDynamicLoader() const {
1127   return m_header.filetype == MH_DYLINKER;
1128 }
1129 
GetAddressByteSize() const1130 uint32_t ObjectFileMachO::GetAddressByteSize() const {
1131   return m_data.GetAddressByteSize();
1132 }
1133 
GetAddressClass(lldb::addr_t file_addr)1134 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) {
1135   Symtab *symtab = GetSymtab();
1136   if (!symtab)
1137     return AddressClass::eUnknown;
1138 
1139   Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr);
1140   if (symbol) {
1141     if (symbol->ValueIsAddress()) {
1142       SectionSP section_sp(symbol->GetAddressRef().GetSection());
1143       if (section_sp) {
1144         const lldb::SectionType section_type = section_sp->GetType();
1145         switch (section_type) {
1146         case eSectionTypeInvalid:
1147           return AddressClass::eUnknown;
1148 
1149         case eSectionTypeCode:
1150           if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) {
1151             // For ARM we have a bit in the n_desc field of the symbol that
1152             // tells us ARM/Thumb which is bit 0x0008.
1153             if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB)
1154               return AddressClass::eCodeAlternateISA;
1155           }
1156           return AddressClass::eCode;
1157 
1158         case eSectionTypeContainer:
1159           return AddressClass::eUnknown;
1160 
1161         case eSectionTypeData:
1162         case eSectionTypeDataCString:
1163         case eSectionTypeDataCStringPointers:
1164         case eSectionTypeDataSymbolAddress:
1165         case eSectionTypeData4:
1166         case eSectionTypeData8:
1167         case eSectionTypeData16:
1168         case eSectionTypeDataPointers:
1169         case eSectionTypeZeroFill:
1170         case eSectionTypeDataObjCMessageRefs:
1171         case eSectionTypeDataObjCCFStrings:
1172         case eSectionTypeGoSymtab:
1173           return AddressClass::eData;
1174 
1175         case eSectionTypeDebug:
1176         case eSectionTypeDWARFDebugAbbrev:
1177         case eSectionTypeDWARFDebugAbbrevDwo:
1178         case eSectionTypeDWARFDebugAddr:
1179         case eSectionTypeDWARFDebugAranges:
1180         case eSectionTypeDWARFDebugCuIndex:
1181         case eSectionTypeDWARFDebugFrame:
1182         case eSectionTypeDWARFDebugInfo:
1183         case eSectionTypeDWARFDebugInfoDwo:
1184         case eSectionTypeDWARFDebugLine:
1185         case eSectionTypeDWARFDebugLineStr:
1186         case eSectionTypeDWARFDebugLoc:
1187         case eSectionTypeDWARFDebugLocDwo:
1188         case eSectionTypeDWARFDebugLocLists:
1189         case eSectionTypeDWARFDebugLocListsDwo:
1190         case eSectionTypeDWARFDebugMacInfo:
1191         case eSectionTypeDWARFDebugMacro:
1192         case eSectionTypeDWARFDebugNames:
1193         case eSectionTypeDWARFDebugPubNames:
1194         case eSectionTypeDWARFDebugPubTypes:
1195         case eSectionTypeDWARFDebugRanges:
1196         case eSectionTypeDWARFDebugRngLists:
1197         case eSectionTypeDWARFDebugRngListsDwo:
1198         case eSectionTypeDWARFDebugStr:
1199         case eSectionTypeDWARFDebugStrDwo:
1200         case eSectionTypeDWARFDebugStrOffsets:
1201         case eSectionTypeDWARFDebugStrOffsetsDwo:
1202         case eSectionTypeDWARFDebugTuIndex:
1203         case eSectionTypeDWARFDebugTypes:
1204         case eSectionTypeDWARFDebugTypesDwo:
1205         case eSectionTypeDWARFAppleNames:
1206         case eSectionTypeDWARFAppleTypes:
1207         case eSectionTypeDWARFAppleNamespaces:
1208         case eSectionTypeDWARFAppleObjC:
1209         case eSectionTypeDWARFGNUDebugAltLink:
1210           return AddressClass::eDebug;
1211 
1212         case eSectionTypeEHFrame:
1213         case eSectionTypeARMexidx:
1214         case eSectionTypeARMextab:
1215         case eSectionTypeCompactUnwind:
1216           return AddressClass::eRuntime;
1217 
1218         case eSectionTypeAbsoluteAddress:
1219         case eSectionTypeELFSymbolTable:
1220         case eSectionTypeELFDynamicSymbols:
1221         case eSectionTypeELFRelocationEntries:
1222         case eSectionTypeELFDynamicLinkInfo:
1223         case eSectionTypeOther:
1224           return AddressClass::eUnknown;
1225         }
1226       }
1227     }
1228 
1229     const SymbolType symbol_type = symbol->GetType();
1230     switch (symbol_type) {
1231     case eSymbolTypeAny:
1232       return AddressClass::eUnknown;
1233     case eSymbolTypeAbsolute:
1234       return AddressClass::eUnknown;
1235 
1236     case eSymbolTypeCode:
1237     case eSymbolTypeTrampoline:
1238     case eSymbolTypeResolver:
1239       if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) {
1240         // For ARM we have a bit in the n_desc field of the symbol that tells
1241         // us ARM/Thumb which is bit 0x0008.
1242         if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB)
1243           return AddressClass::eCodeAlternateISA;
1244       }
1245       return AddressClass::eCode;
1246 
1247     case eSymbolTypeData:
1248       return AddressClass::eData;
1249     case eSymbolTypeRuntime:
1250       return AddressClass::eRuntime;
1251     case eSymbolTypeException:
1252       return AddressClass::eRuntime;
1253     case eSymbolTypeSourceFile:
1254       return AddressClass::eDebug;
1255     case eSymbolTypeHeaderFile:
1256       return AddressClass::eDebug;
1257     case eSymbolTypeObjectFile:
1258       return AddressClass::eDebug;
1259     case eSymbolTypeCommonBlock:
1260       return AddressClass::eDebug;
1261     case eSymbolTypeBlock:
1262       return AddressClass::eDebug;
1263     case eSymbolTypeLocal:
1264       return AddressClass::eData;
1265     case eSymbolTypeParam:
1266       return AddressClass::eData;
1267     case eSymbolTypeVariable:
1268       return AddressClass::eData;
1269     case eSymbolTypeVariableType:
1270       return AddressClass::eDebug;
1271     case eSymbolTypeLineEntry:
1272       return AddressClass::eDebug;
1273     case eSymbolTypeLineHeader:
1274       return AddressClass::eDebug;
1275     case eSymbolTypeScopeBegin:
1276       return AddressClass::eDebug;
1277     case eSymbolTypeScopeEnd:
1278       return AddressClass::eDebug;
1279     case eSymbolTypeAdditional:
1280       return AddressClass::eUnknown;
1281     case eSymbolTypeCompiler:
1282       return AddressClass::eDebug;
1283     case eSymbolTypeInstrumentation:
1284       return AddressClass::eDebug;
1285     case eSymbolTypeUndefined:
1286       return AddressClass::eUnknown;
1287     case eSymbolTypeObjCClass:
1288       return AddressClass::eRuntime;
1289     case eSymbolTypeObjCMetaClass:
1290       return AddressClass::eRuntime;
1291     case eSymbolTypeObjCIVar:
1292       return AddressClass::eRuntime;
1293     case eSymbolTypeReExported:
1294       return AddressClass::eRuntime;
1295     }
1296   }
1297   return AddressClass::eUnknown;
1298 }
1299 
GetSymtab()1300 Symtab *ObjectFileMachO::GetSymtab() {
1301   ModuleSP module_sp(GetModule());
1302   if (module_sp) {
1303     std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
1304     if (m_symtab_up == nullptr) {
1305       m_symtab_up = std::make_unique<Symtab>(this);
1306       std::lock_guard<std::recursive_mutex> symtab_guard(
1307           m_symtab_up->GetMutex());
1308       ParseSymtab();
1309       m_symtab_up->Finalize();
1310     }
1311   }
1312   return m_symtab_up.get();
1313 }
1314 
IsStripped()1315 bool ObjectFileMachO::IsStripped() {
1316   if (m_dysymtab.cmd == 0) {
1317     ModuleSP module_sp(GetModule());
1318     if (module_sp) {
1319       lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
1320       for (uint32_t i = 0; i < m_header.ncmds; ++i) {
1321         const lldb::offset_t load_cmd_offset = offset;
1322 
1323         llvm::MachO::load_command lc;
1324         if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr)
1325           break;
1326         if (lc.cmd == LC_DYSYMTAB) {
1327           m_dysymtab.cmd = lc.cmd;
1328           m_dysymtab.cmdsize = lc.cmdsize;
1329           if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym,
1330                             (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) ==
1331               nullptr) {
1332             // Clear m_dysymtab if we were unable to read all items from the
1333             // load command
1334             ::memset(&m_dysymtab, 0, sizeof(m_dysymtab));
1335           }
1336         }
1337         offset = load_cmd_offset + lc.cmdsize;
1338       }
1339     }
1340   }
1341   if (m_dysymtab.cmd)
1342     return m_dysymtab.nlocalsym <= 1;
1343   return false;
1344 }
1345 
GetEncryptedFileRanges()1346 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() {
1347   EncryptedFileRanges result;
1348   lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
1349 
1350   llvm::MachO::encryption_info_command encryption_cmd;
1351   for (uint32_t i = 0; i < m_header.ncmds; ++i) {
1352     const lldb::offset_t load_cmd_offset = offset;
1353     if (m_data.GetU32(&offset, &encryption_cmd, 2) == nullptr)
1354       break;
1355 
1356     // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the
1357     // 3 fields we care about, so treat them the same.
1358     if (encryption_cmd.cmd == LC_ENCRYPTION_INFO ||
1359         encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) {
1360       if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) {
1361         if (encryption_cmd.cryptid != 0) {
1362           EncryptedFileRanges::Entry entry;
1363           entry.SetRangeBase(encryption_cmd.cryptoff);
1364           entry.SetByteSize(encryption_cmd.cryptsize);
1365           result.Append(entry);
1366         }
1367       }
1368     }
1369     offset = load_cmd_offset + encryption_cmd.cmdsize;
1370   }
1371 
1372   return result;
1373 }
1374 
SanitizeSegmentCommand(llvm::MachO::segment_command_64 & seg_cmd,uint32_t cmd_idx)1375 void ObjectFileMachO::SanitizeSegmentCommand(
1376     llvm::MachO::segment_command_64 &seg_cmd, uint32_t cmd_idx) {
1377   if (m_length == 0 || seg_cmd.filesize == 0)
1378     return;
1379 
1380   if ((m_header.flags & MH_DYLIB_IN_CACHE) && !IsInMemory()) {
1381     // In shared cache images, the load commands are relative to the
1382     // shared cache file, and not the specific image we are
1383     // examining. Let's fix this up so that it looks like a normal
1384     // image.
1385     if (strncmp(seg_cmd.segname, "__TEXT", sizeof(seg_cmd.segname)) == 0)
1386       m_text_address = seg_cmd.vmaddr;
1387     if (strncmp(seg_cmd.segname, "__LINKEDIT", sizeof(seg_cmd.segname)) == 0)
1388       m_linkedit_original_offset = seg_cmd.fileoff;
1389 
1390     seg_cmd.fileoff = seg_cmd.vmaddr - m_text_address;
1391   }
1392 
1393   if (seg_cmd.fileoff > m_length) {
1394     // We have a load command that says it extends past the end of the file.
1395     // This is likely a corrupt file.  We don't have any way to return an error
1396     // condition here (this method was likely invoked from something like
1397     // ObjectFile::GetSectionList()), so we just null out the section contents,
1398     // and dump a message to stdout.  The most common case here is core file
1399     // debugging with a truncated file.
1400     const char *lc_segment_name =
1401         seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT";
1402     GetModule()->ReportWarning(
1403         "load command %u %s has a fileoff (0x%" PRIx64
1404         ") that extends beyond the end of the file (0x%" PRIx64
1405         "), ignoring this section",
1406         cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length);
1407 
1408     seg_cmd.fileoff = 0;
1409     seg_cmd.filesize = 0;
1410   }
1411 
1412   if (seg_cmd.fileoff + seg_cmd.filesize > m_length) {
1413     // We have a load command that says it extends past the end of the file.
1414     // This is likely a corrupt file.  We don't have any way to return an error
1415     // condition here (this method was likely invoked from something like
1416     // ObjectFile::GetSectionList()), so we just null out the section contents,
1417     // and dump a message to stdout.  The most common case here is core file
1418     // debugging with a truncated file.
1419     const char *lc_segment_name =
1420         seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT";
1421     GetModule()->ReportWarning(
1422         "load command %u %s has a fileoff + filesize (0x%" PRIx64
1423         ") that extends beyond the end of the file (0x%" PRIx64
1424         "), the segment will be truncated to match",
1425         cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length);
1426 
1427     // Truncate the length
1428     seg_cmd.filesize = m_length - seg_cmd.fileoff;
1429   }
1430 }
1431 
1432 static uint32_t
GetSegmentPermissions(const llvm::MachO::segment_command_64 & seg_cmd)1433 GetSegmentPermissions(const llvm::MachO::segment_command_64 &seg_cmd) {
1434   uint32_t result = 0;
1435   if (seg_cmd.initprot & VM_PROT_READ)
1436     result |= ePermissionsReadable;
1437   if (seg_cmd.initprot & VM_PROT_WRITE)
1438     result |= ePermissionsWritable;
1439   if (seg_cmd.initprot & VM_PROT_EXECUTE)
1440     result |= ePermissionsExecutable;
1441   return result;
1442 }
1443 
GetSectionType(uint32_t flags,ConstString section_name)1444 static lldb::SectionType GetSectionType(uint32_t flags,
1445                                         ConstString section_name) {
1446 
1447   if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS))
1448     return eSectionTypeCode;
1449 
1450   uint32_t mach_sect_type = flags & SECTION_TYPE;
1451   static ConstString g_sect_name_objc_data("__objc_data");
1452   static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs");
1453   static ConstString g_sect_name_objc_selrefs("__objc_selrefs");
1454   static ConstString g_sect_name_objc_classrefs("__objc_classrefs");
1455   static ConstString g_sect_name_objc_superrefs("__objc_superrefs");
1456   static ConstString g_sect_name_objc_const("__objc_const");
1457   static ConstString g_sect_name_objc_classlist("__objc_classlist");
1458   static ConstString g_sect_name_cfstring("__cfstring");
1459 
1460   static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev");
1461   static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges");
1462   static ConstString g_sect_name_dwarf_debug_frame("__debug_frame");
1463   static ConstString g_sect_name_dwarf_debug_info("__debug_info");
1464   static ConstString g_sect_name_dwarf_debug_line("__debug_line");
1465   static ConstString g_sect_name_dwarf_debug_loc("__debug_loc");
1466   static ConstString g_sect_name_dwarf_debug_loclists("__debug_loclists");
1467   static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo");
1468   static ConstString g_sect_name_dwarf_debug_names("__debug_names");
1469   static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames");
1470   static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes");
1471   static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges");
1472   static ConstString g_sect_name_dwarf_debug_str("__debug_str");
1473   static ConstString g_sect_name_dwarf_debug_types("__debug_types");
1474   static ConstString g_sect_name_dwarf_apple_names("__apple_names");
1475   static ConstString g_sect_name_dwarf_apple_types("__apple_types");
1476   static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac");
1477   static ConstString g_sect_name_dwarf_apple_objc("__apple_objc");
1478   static ConstString g_sect_name_eh_frame("__eh_frame");
1479   static ConstString g_sect_name_compact_unwind("__unwind_info");
1480   static ConstString g_sect_name_text("__text");
1481   static ConstString g_sect_name_data("__data");
1482   static ConstString g_sect_name_go_symtab("__gosymtab");
1483 
1484   if (section_name == g_sect_name_dwarf_debug_abbrev)
1485     return eSectionTypeDWARFDebugAbbrev;
1486   if (section_name == g_sect_name_dwarf_debug_aranges)
1487     return eSectionTypeDWARFDebugAranges;
1488   if (section_name == g_sect_name_dwarf_debug_frame)
1489     return eSectionTypeDWARFDebugFrame;
1490   if (section_name == g_sect_name_dwarf_debug_info)
1491     return eSectionTypeDWARFDebugInfo;
1492   if (section_name == g_sect_name_dwarf_debug_line)
1493     return eSectionTypeDWARFDebugLine;
1494   if (section_name == g_sect_name_dwarf_debug_loc)
1495     return eSectionTypeDWARFDebugLoc;
1496   if (section_name == g_sect_name_dwarf_debug_loclists)
1497     return eSectionTypeDWARFDebugLocLists;
1498   if (section_name == g_sect_name_dwarf_debug_macinfo)
1499     return eSectionTypeDWARFDebugMacInfo;
1500   if (section_name == g_sect_name_dwarf_debug_names)
1501     return eSectionTypeDWARFDebugNames;
1502   if (section_name == g_sect_name_dwarf_debug_pubnames)
1503     return eSectionTypeDWARFDebugPubNames;
1504   if (section_name == g_sect_name_dwarf_debug_pubtypes)
1505     return eSectionTypeDWARFDebugPubTypes;
1506   if (section_name == g_sect_name_dwarf_debug_ranges)
1507     return eSectionTypeDWARFDebugRanges;
1508   if (section_name == g_sect_name_dwarf_debug_str)
1509     return eSectionTypeDWARFDebugStr;
1510   if (section_name == g_sect_name_dwarf_debug_types)
1511     return eSectionTypeDWARFDebugTypes;
1512   if (section_name == g_sect_name_dwarf_apple_names)
1513     return eSectionTypeDWARFAppleNames;
1514   if (section_name == g_sect_name_dwarf_apple_types)
1515     return eSectionTypeDWARFAppleTypes;
1516   if (section_name == g_sect_name_dwarf_apple_namespaces)
1517     return eSectionTypeDWARFAppleNamespaces;
1518   if (section_name == g_sect_name_dwarf_apple_objc)
1519     return eSectionTypeDWARFAppleObjC;
1520   if (section_name == g_sect_name_objc_selrefs)
1521     return eSectionTypeDataCStringPointers;
1522   if (section_name == g_sect_name_objc_msgrefs)
1523     return eSectionTypeDataObjCMessageRefs;
1524   if (section_name == g_sect_name_eh_frame)
1525     return eSectionTypeEHFrame;
1526   if (section_name == g_sect_name_compact_unwind)
1527     return eSectionTypeCompactUnwind;
1528   if (section_name == g_sect_name_cfstring)
1529     return eSectionTypeDataObjCCFStrings;
1530   if (section_name == g_sect_name_go_symtab)
1531     return eSectionTypeGoSymtab;
1532   if (section_name == g_sect_name_objc_data ||
1533       section_name == g_sect_name_objc_classrefs ||
1534       section_name == g_sect_name_objc_superrefs ||
1535       section_name == g_sect_name_objc_const ||
1536       section_name == g_sect_name_objc_classlist) {
1537     return eSectionTypeDataPointers;
1538   }
1539 
1540   switch (mach_sect_type) {
1541   // TODO: categorize sections by other flags for regular sections
1542   case S_REGULAR:
1543     if (section_name == g_sect_name_text)
1544       return eSectionTypeCode;
1545     if (section_name == g_sect_name_data)
1546       return eSectionTypeData;
1547     return eSectionTypeOther;
1548   case S_ZEROFILL:
1549     return eSectionTypeZeroFill;
1550   case S_CSTRING_LITERALS: // section with only literal C strings
1551     return eSectionTypeDataCString;
1552   case S_4BYTE_LITERALS: // section with only 4 byte literals
1553     return eSectionTypeData4;
1554   case S_8BYTE_LITERALS: // section with only 8 byte literals
1555     return eSectionTypeData8;
1556   case S_LITERAL_POINTERS: // section with only pointers to literals
1557     return eSectionTypeDataPointers;
1558   case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers
1559     return eSectionTypeDataPointers;
1560   case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers
1561     return eSectionTypeDataPointers;
1562   case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in
1563                        // the reserved2 field
1564     return eSectionTypeCode;
1565   case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for
1566                                  // initialization
1567     return eSectionTypeDataPointers;
1568   case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for
1569                                  // termination
1570     return eSectionTypeDataPointers;
1571   case S_COALESCED:
1572     return eSectionTypeOther;
1573   case S_GB_ZEROFILL:
1574     return eSectionTypeZeroFill;
1575   case S_INTERPOSING: // section with only pairs of function pointers for
1576                       // interposing
1577     return eSectionTypeCode;
1578   case S_16BYTE_LITERALS: // section with only 16 byte literals
1579     return eSectionTypeData16;
1580   case S_DTRACE_DOF:
1581     return eSectionTypeDebug;
1582   case S_LAZY_DYLIB_SYMBOL_POINTERS:
1583     return eSectionTypeDataPointers;
1584   default:
1585     return eSectionTypeOther;
1586   }
1587 }
1588 
1589 struct ObjectFileMachO::SegmentParsingContext {
1590   const EncryptedFileRanges EncryptedRanges;
1591   lldb_private::SectionList &UnifiedList;
1592   uint32_t NextSegmentIdx = 0;
1593   uint32_t NextSectionIdx = 0;
1594   bool FileAddressesChanged = false;
1595 
SegmentParsingContextObjectFileMachO::SegmentParsingContext1596   SegmentParsingContext(EncryptedFileRanges EncryptedRanges,
1597                         lldb_private::SectionList &UnifiedList)
1598       : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {}
1599 };
1600 
ProcessSegmentCommand(const llvm::MachO::load_command & load_cmd_,lldb::offset_t offset,uint32_t cmd_idx,SegmentParsingContext & context)1601 void ObjectFileMachO::ProcessSegmentCommand(
1602     const llvm::MachO::load_command &load_cmd_, lldb::offset_t offset,
1603     uint32_t cmd_idx, SegmentParsingContext &context) {
1604   llvm::MachO::segment_command_64 load_cmd;
1605   memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_));
1606 
1607   if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16))
1608     return;
1609 
1610   ModuleSP module_sp = GetModule();
1611   const bool is_core = GetType() == eTypeCoreFile;
1612   const bool is_dsym = (m_header.filetype == MH_DSYM);
1613   bool add_section = true;
1614   bool add_to_unified = true;
1615   ConstString const_segname(
1616       load_cmd.segname, strnlen(load_cmd.segname, sizeof(load_cmd.segname)));
1617 
1618   SectionSP unified_section_sp(
1619       context.UnifiedList.FindSectionByName(const_segname));
1620   if (is_dsym && unified_section_sp) {
1621     if (const_segname == GetSegmentNameLINKEDIT()) {
1622       // We need to keep the __LINKEDIT segment private to this object file
1623       // only
1624       add_to_unified = false;
1625     } else {
1626       // This is the dSYM file and this section has already been created by the
1627       // object file, no need to create it.
1628       add_section = false;
1629     }
1630   }
1631   load_cmd.vmaddr = m_data.GetAddress(&offset);
1632   load_cmd.vmsize = m_data.GetAddress(&offset);
1633   load_cmd.fileoff = m_data.GetAddress(&offset);
1634   load_cmd.filesize = m_data.GetAddress(&offset);
1635   if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4))
1636     return;
1637 
1638   SanitizeSegmentCommand(load_cmd, cmd_idx);
1639 
1640   const uint32_t segment_permissions = GetSegmentPermissions(load_cmd);
1641   const bool segment_is_encrypted =
1642       (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0;
1643 
1644   // Keep a list of mach segments around in case we need to get at data that
1645   // isn't stored in the abstracted Sections.
1646   m_mach_segments.push_back(load_cmd);
1647 
1648   // Use a segment ID of the segment index shifted left by 8 so they never
1649   // conflict with any of the sections.
1650   SectionSP segment_sp;
1651   if (add_section && (const_segname || is_core)) {
1652     segment_sp = std::make_shared<Section>(
1653         module_sp, // Module to which this section belongs
1654         this,      // Object file to which this sections belongs
1655         ++context.NextSegmentIdx
1656             << 8, // Section ID is the 1 based segment index
1657         // shifted right by 8 bits as not to collide with any of the 256
1658         // section IDs that are possible
1659         const_segname,         // Name of this section
1660         eSectionTypeContainer, // This section is a container of other
1661         // sections.
1662         load_cmd.vmaddr, // File VM address == addresses as they are
1663         // found in the object file
1664         load_cmd.vmsize,  // VM size in bytes of this section
1665         load_cmd.fileoff, // Offset to the data for this section in
1666         // the file
1667         load_cmd.filesize, // Size in bytes of this section as found
1668         // in the file
1669         0,               // Segments have no alignment information
1670         load_cmd.flags); // Flags for this section
1671 
1672     segment_sp->SetIsEncrypted(segment_is_encrypted);
1673     m_sections_up->AddSection(segment_sp);
1674     segment_sp->SetPermissions(segment_permissions);
1675     if (add_to_unified)
1676       context.UnifiedList.AddSection(segment_sp);
1677   } else if (unified_section_sp) {
1678     if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) {
1679       // Check to see if the module was read from memory?
1680       if (module_sp->GetObjectFile()->IsInMemory()) {
1681         // We have a module that is in memory and needs to have its file
1682         // address adjusted. We need to do this because when we load a file
1683         // from memory, its addresses will be slid already, yet the addresses
1684         // in the new symbol file will still be unslid.  Since everything is
1685         // stored as section offset, this shouldn't cause any problems.
1686 
1687         // Make sure we've parsed the symbol table from the ObjectFile before
1688         // we go around changing its Sections.
1689         module_sp->GetObjectFile()->GetSymtab();
1690         // eh_frame would present the same problems but we parse that on a per-
1691         // function basis as-needed so it's more difficult to remove its use of
1692         // the Sections.  Realistically, the environments where this code path
1693         // will be taken will not have eh_frame sections.
1694 
1695         unified_section_sp->SetFileAddress(load_cmd.vmaddr);
1696 
1697         // Notify the module that the section addresses have been changed once
1698         // we're done so any file-address caches can be updated.
1699         context.FileAddressesChanged = true;
1700       }
1701     }
1702     m_sections_up->AddSection(unified_section_sp);
1703   }
1704 
1705   llvm::MachO::section_64 sect64;
1706   ::memset(&sect64, 0, sizeof(sect64));
1707   // Push a section into our mach sections for the section at index zero
1708   // (NO_SECT) if we don't have any mach sections yet...
1709   if (m_mach_sections.empty())
1710     m_mach_sections.push_back(sect64);
1711   uint32_t segment_sect_idx;
1712   const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1;
1713 
1714   const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8;
1715   for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects;
1716        ++segment_sect_idx) {
1717     if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname,
1718                      sizeof(sect64.sectname)) == nullptr)
1719       break;
1720     if (m_data.GetU8(&offset, (uint8_t *)sect64.segname,
1721                      sizeof(sect64.segname)) == nullptr)
1722       break;
1723     sect64.addr = m_data.GetAddress(&offset);
1724     sect64.size = m_data.GetAddress(&offset);
1725 
1726     if (m_data.GetU32(&offset, &sect64.offset, num_u32s) == nullptr)
1727       break;
1728 
1729     if ((m_header.flags & MH_DYLIB_IN_CACHE) && !IsInMemory()) {
1730       sect64.offset = sect64.addr - m_text_address;
1731     }
1732 
1733     // Keep a list of mach sections around in case we need to get at data that
1734     // isn't stored in the abstracted Sections.
1735     m_mach_sections.push_back(sect64);
1736 
1737     if (add_section) {
1738       ConstString section_name(
1739           sect64.sectname, strnlen(sect64.sectname, sizeof(sect64.sectname)));
1740       if (!const_segname) {
1741         // We have a segment with no name so we need to conjure up segments
1742         // that correspond to the section's segname if there isn't already such
1743         // a section. If there is such a section, we resize the section so that
1744         // it spans all sections.  We also mark these sections as fake so
1745         // address matches don't hit if they land in the gaps between the child
1746         // sections.
1747         const_segname.SetTrimmedCStringWithLength(sect64.segname,
1748                                                   sizeof(sect64.segname));
1749         segment_sp = context.UnifiedList.FindSectionByName(const_segname);
1750         if (segment_sp.get()) {
1751           Section *segment = segment_sp.get();
1752           // Grow the section size as needed.
1753           const lldb::addr_t sect64_min_addr = sect64.addr;
1754           const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size;
1755           const lldb::addr_t curr_seg_byte_size = segment->GetByteSize();
1756           const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress();
1757           const lldb::addr_t curr_seg_max_addr =
1758               curr_seg_min_addr + curr_seg_byte_size;
1759           if (sect64_min_addr >= curr_seg_min_addr) {
1760             const lldb::addr_t new_seg_byte_size =
1761                 sect64_max_addr - curr_seg_min_addr;
1762             // Only grow the section size if needed
1763             if (new_seg_byte_size > curr_seg_byte_size)
1764               segment->SetByteSize(new_seg_byte_size);
1765           } else {
1766             // We need to change the base address of the segment and adjust the
1767             // child section offsets for all existing children.
1768             const lldb::addr_t slide_amount =
1769                 sect64_min_addr - curr_seg_min_addr;
1770             segment->Slide(slide_amount, false);
1771             segment->GetChildren().Slide(-slide_amount, false);
1772             segment->SetByteSize(curr_seg_max_addr - sect64_min_addr);
1773           }
1774 
1775           // Grow the section size as needed.
1776           if (sect64.offset) {
1777             const lldb::addr_t segment_min_file_offset =
1778                 segment->GetFileOffset();
1779             const lldb::addr_t segment_max_file_offset =
1780                 segment_min_file_offset + segment->GetFileSize();
1781 
1782             const lldb::addr_t section_min_file_offset = sect64.offset;
1783             const lldb::addr_t section_max_file_offset =
1784                 section_min_file_offset + sect64.size;
1785             const lldb::addr_t new_file_offset =
1786                 std::min(section_min_file_offset, segment_min_file_offset);
1787             const lldb::addr_t new_file_size =
1788                 std::max(section_max_file_offset, segment_max_file_offset) -
1789                 new_file_offset;
1790             segment->SetFileOffset(new_file_offset);
1791             segment->SetFileSize(new_file_size);
1792           }
1793         } else {
1794           // Create a fake section for the section's named segment
1795           segment_sp = std::make_shared<Section>(
1796               segment_sp, // Parent section
1797               module_sp,  // Module to which this section belongs
1798               this,       // Object file to which this section belongs
1799               ++context.NextSegmentIdx
1800                   << 8, // Section ID is the 1 based segment index
1801               // shifted right by 8 bits as not to
1802               // collide with any of the 256 section IDs
1803               // that are possible
1804               const_segname,         // Name of this section
1805               eSectionTypeContainer, // This section is a container of
1806               // other sections.
1807               sect64.addr, // File VM address == addresses as they are
1808               // found in the object file
1809               sect64.size,   // VM size in bytes of this section
1810               sect64.offset, // Offset to the data for this section in
1811               // the file
1812               sect64.offset ? sect64.size : 0, // Size in bytes of
1813               // this section as
1814               // found in the file
1815               sect64.align,
1816               load_cmd.flags); // Flags for this section
1817           segment_sp->SetIsFake(true);
1818           segment_sp->SetPermissions(segment_permissions);
1819           m_sections_up->AddSection(segment_sp);
1820           if (add_to_unified)
1821             context.UnifiedList.AddSection(segment_sp);
1822           segment_sp->SetIsEncrypted(segment_is_encrypted);
1823         }
1824       }
1825       assert(segment_sp.get());
1826 
1827       lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name);
1828 
1829       SectionSP section_sp(new Section(
1830           segment_sp, module_sp, this, ++context.NextSectionIdx, section_name,
1831           sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size,
1832           sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align,
1833           sect64.flags));
1834       // Set the section to be encrypted to match the segment
1835 
1836       bool section_is_encrypted = false;
1837       if (!segment_is_encrypted && load_cmd.filesize != 0)
1838         section_is_encrypted = context.EncryptedRanges.FindEntryThatContains(
1839                                    sect64.offset) != nullptr;
1840 
1841       section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted);
1842       section_sp->SetPermissions(segment_permissions);
1843       segment_sp->GetChildren().AddSection(section_sp);
1844 
1845       if (segment_sp->IsFake()) {
1846         segment_sp.reset();
1847         const_segname.Clear();
1848       }
1849     }
1850   }
1851   if (segment_sp && is_dsym) {
1852     if (first_segment_sectID <= context.NextSectionIdx) {
1853       lldb::user_id_t sect_uid;
1854       for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx;
1855            ++sect_uid) {
1856         SectionSP curr_section_sp(
1857             segment_sp->GetChildren().FindSectionByID(sect_uid));
1858         SectionSP next_section_sp;
1859         if (sect_uid + 1 <= context.NextSectionIdx)
1860           next_section_sp =
1861               segment_sp->GetChildren().FindSectionByID(sect_uid + 1);
1862 
1863         if (curr_section_sp.get()) {
1864           if (curr_section_sp->GetByteSize() == 0) {
1865             if (next_section_sp.get() != nullptr)
1866               curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() -
1867                                            curr_section_sp->GetFileAddress());
1868             else
1869               curr_section_sp->SetByteSize(load_cmd.vmsize);
1870           }
1871         }
1872       }
1873     }
1874   }
1875 }
1876 
ProcessDysymtabCommand(const llvm::MachO::load_command & load_cmd,lldb::offset_t offset)1877 void ObjectFileMachO::ProcessDysymtabCommand(
1878     const llvm::MachO::load_command &load_cmd, lldb::offset_t offset) {
1879   m_dysymtab.cmd = load_cmd.cmd;
1880   m_dysymtab.cmdsize = load_cmd.cmdsize;
1881   m_data.GetU32(&offset, &m_dysymtab.ilocalsym,
1882                 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2);
1883 }
1884 
CreateSections(SectionList & unified_section_list)1885 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) {
1886   if (m_sections_up)
1887     return;
1888 
1889   m_sections_up = std::make_unique<SectionList>();
1890 
1891   lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
1892   // bool dump_sections = false;
1893   ModuleSP module_sp(GetModule());
1894 
1895   offset = MachHeaderSizeFromMagic(m_header.magic);
1896 
1897   SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list);
1898   llvm::MachO::load_command load_cmd;
1899   for (uint32_t i = 0; i < m_header.ncmds; ++i) {
1900     const lldb::offset_t load_cmd_offset = offset;
1901     if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr)
1902       break;
1903 
1904     if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64)
1905       ProcessSegmentCommand(load_cmd, offset, i, context);
1906     else if (load_cmd.cmd == LC_DYSYMTAB)
1907       ProcessDysymtabCommand(load_cmd, offset);
1908 
1909     offset = load_cmd_offset + load_cmd.cmdsize;
1910   }
1911 
1912   if (context.FileAddressesChanged && module_sp)
1913     module_sp->SectionFileAddressesChanged();
1914 }
1915 
1916 class MachSymtabSectionInfo {
1917 public:
MachSymtabSectionInfo(SectionList * section_list)1918   MachSymtabSectionInfo(SectionList *section_list)
1919       : m_section_list(section_list), m_section_infos() {
1920     // Get the number of sections down to a depth of 1 to include all segments
1921     // and their sections, but no other sections that may be added for debug
1922     // map or
1923     m_section_infos.resize(section_list->GetNumSections(1));
1924   }
1925 
GetSection(uint8_t n_sect,addr_t file_addr)1926   SectionSP GetSection(uint8_t n_sect, addr_t file_addr) {
1927     if (n_sect == 0)
1928       return SectionSP();
1929     if (n_sect < m_section_infos.size()) {
1930       if (!m_section_infos[n_sect].section_sp) {
1931         SectionSP section_sp(m_section_list->FindSectionByID(n_sect));
1932         m_section_infos[n_sect].section_sp = section_sp;
1933         if (section_sp) {
1934           m_section_infos[n_sect].vm_range.SetBaseAddress(
1935               section_sp->GetFileAddress());
1936           m_section_infos[n_sect].vm_range.SetByteSize(
1937               section_sp->GetByteSize());
1938         } else {
1939           std::string filename = "<unknown>";
1940           SectionSP first_section_sp(m_section_list->GetSectionAtIndex(0));
1941           if (first_section_sp)
1942             filename = first_section_sp->GetObjectFile()->GetFileSpec().GetPath();
1943 
1944           Host::SystemLog(Host::eSystemLogError,
1945                           "error: unable to find section %d for a symbol in "
1946                           "%s, corrupt file?\n",
1947                           n_sect, filename.c_str());
1948         }
1949       }
1950       if (m_section_infos[n_sect].vm_range.Contains(file_addr)) {
1951         // Symbol is in section.
1952         return m_section_infos[n_sect].section_sp;
1953       } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 &&
1954                  m_section_infos[n_sect].vm_range.GetBaseAddress() ==
1955                      file_addr) {
1956         // Symbol is in section with zero size, but has the same start address
1957         // as the section. This can happen with linker symbols (symbols that
1958         // start with the letter 'l' or 'L'.
1959         return m_section_infos[n_sect].section_sp;
1960       }
1961     }
1962     return m_section_list->FindSectionContainingFileAddress(file_addr);
1963   }
1964 
1965 protected:
1966   struct SectionInfo {
SectionInfoMachSymtabSectionInfo::SectionInfo1967     SectionInfo() : vm_range(), section_sp() {}
1968 
1969     VMRange vm_range;
1970     SectionSP section_sp;
1971   };
1972   SectionList *m_section_list;
1973   std::vector<SectionInfo> m_section_infos;
1974 };
1975 
1976 #define TRIE_SYMBOL_IS_THUMB (1ULL << 63)
1977 struct TrieEntry {
DumpTrieEntry1978   void Dump() const {
1979     printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"",
1980            static_cast<unsigned long long>(address),
1981            static_cast<unsigned long long>(flags),
1982            static_cast<unsigned long long>(other), name.GetCString());
1983     if (import_name)
1984       printf(" -> \"%s\"\n", import_name.GetCString());
1985     else
1986       printf("\n");
1987   }
1988   ConstString name;
1989   uint64_t address = LLDB_INVALID_ADDRESS;
1990   uint64_t flags =
1991       0; // EXPORT_SYMBOL_FLAGS_REEXPORT, EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER,
1992          // TRIE_SYMBOL_IS_THUMB
1993   uint64_t other = 0;
1994   ConstString import_name;
1995 };
1996 
1997 struct TrieEntryWithOffset {
1998   lldb::offset_t nodeOffset;
1999   TrieEntry entry;
2000 
TrieEntryWithOffsetTrieEntryWithOffset2001   TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {}
2002 
DumpTrieEntryWithOffset2003   void Dump(uint32_t idx) const {
2004     printf("[%3u] 0x%16.16llx: ", idx,
2005            static_cast<unsigned long long>(nodeOffset));
2006     entry.Dump();
2007   }
2008 
operator <TrieEntryWithOffset2009   bool operator<(const TrieEntryWithOffset &other) const {
2010     return (nodeOffset < other.nodeOffset);
2011   }
2012 };
2013 
ParseTrieEntries(DataExtractor & data,lldb::offset_t offset,const bool is_arm,addr_t text_seg_base_addr,std::vector<llvm::StringRef> & nameSlices,std::set<lldb::addr_t> & resolver_addresses,std::vector<TrieEntryWithOffset> & reexports,std::vector<TrieEntryWithOffset> & ext_symbols)2014 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset,
2015                              const bool is_arm, addr_t text_seg_base_addr,
2016                              std::vector<llvm::StringRef> &nameSlices,
2017                              std::set<lldb::addr_t> &resolver_addresses,
2018                              std::vector<TrieEntryWithOffset> &reexports,
2019                              std::vector<TrieEntryWithOffset> &ext_symbols) {
2020   if (!data.ValidOffset(offset))
2021     return true;
2022 
2023   // Terminal node -- end of a branch, possibly add this to
2024   // the symbol table or resolver table.
2025   const uint64_t terminalSize = data.GetULEB128(&offset);
2026   lldb::offset_t children_offset = offset + terminalSize;
2027   if (terminalSize != 0) {
2028     TrieEntryWithOffset e(offset);
2029     e.entry.flags = data.GetULEB128(&offset);
2030     const char *import_name = nullptr;
2031     if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) {
2032       e.entry.address = 0;
2033       e.entry.other = data.GetULEB128(&offset); // dylib ordinal
2034       import_name = data.GetCStr(&offset);
2035     } else {
2036       e.entry.address = data.GetULEB128(&offset);
2037       if (text_seg_base_addr != LLDB_INVALID_ADDRESS)
2038         e.entry.address += text_seg_base_addr;
2039       if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) {
2040         e.entry.other = data.GetULEB128(&offset);
2041         uint64_t resolver_addr = e.entry.other;
2042         if (text_seg_base_addr != LLDB_INVALID_ADDRESS)
2043           resolver_addr += text_seg_base_addr;
2044         if (is_arm)
2045           resolver_addr &= THUMB_ADDRESS_BIT_MASK;
2046         resolver_addresses.insert(resolver_addr);
2047       } else
2048         e.entry.other = 0;
2049     }
2050     bool add_this_entry = false;
2051     if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT) &&
2052         import_name && import_name[0]) {
2053       // add symbols that are reexport symbols with a valid import name.
2054       add_this_entry = true;
2055     } else if (e.entry.flags == 0 &&
2056                (import_name == nullptr || import_name[0] == '\0')) {
2057       // add externally visible symbols, in case the nlist record has
2058       // been stripped/omitted.
2059       add_this_entry = true;
2060     }
2061     if (add_this_entry) {
2062       std::string name;
2063       if (!nameSlices.empty()) {
2064         for (auto name_slice : nameSlices)
2065           name.append(name_slice.data(), name_slice.size());
2066       }
2067       if (name.size() > 1) {
2068         // Skip the leading '_'
2069         e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1);
2070       }
2071       if (import_name) {
2072         // Skip the leading '_'
2073         e.entry.import_name.SetCString(import_name + 1);
2074       }
2075       if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT)) {
2076         reexports.push_back(e);
2077       } else {
2078         if (is_arm && (e.entry.address & 1)) {
2079           e.entry.flags |= TRIE_SYMBOL_IS_THUMB;
2080           e.entry.address &= THUMB_ADDRESS_BIT_MASK;
2081         }
2082         ext_symbols.push_back(e);
2083       }
2084     }
2085   }
2086 
2087   const uint8_t childrenCount = data.GetU8(&children_offset);
2088   for (uint8_t i = 0; i < childrenCount; ++i) {
2089     const char *cstr = data.GetCStr(&children_offset);
2090     if (cstr)
2091       nameSlices.push_back(llvm::StringRef(cstr));
2092     else
2093       return false; // Corrupt data
2094     lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset);
2095     if (childNodeOffset) {
2096       if (!ParseTrieEntries(data, childNodeOffset, is_arm, text_seg_base_addr,
2097                             nameSlices, resolver_addresses, reexports,
2098                             ext_symbols)) {
2099         return false;
2100       }
2101     }
2102     nameSlices.pop_back();
2103   }
2104   return true;
2105 }
2106 
GetSymbolType(const char * & symbol_name,bool & demangled_is_synthesized,const SectionSP & text_section_sp,const SectionSP & data_section_sp,const SectionSP & data_dirty_section_sp,const SectionSP & data_const_section_sp,const SectionSP & symbol_section)2107 static SymbolType GetSymbolType(const char *&symbol_name,
2108                                 bool &demangled_is_synthesized,
2109                                 const SectionSP &text_section_sp,
2110                                 const SectionSP &data_section_sp,
2111                                 const SectionSP &data_dirty_section_sp,
2112                                 const SectionSP &data_const_section_sp,
2113                                 const SectionSP &symbol_section) {
2114   SymbolType type = eSymbolTypeInvalid;
2115 
2116   const char *symbol_sect_name = symbol_section->GetName().AsCString();
2117   if (symbol_section->IsDescendant(text_section_sp.get())) {
2118     if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS |
2119                                 S_ATTR_SELF_MODIFYING_CODE |
2120                                 S_ATTR_SOME_INSTRUCTIONS))
2121       type = eSymbolTypeData;
2122     else
2123       type = eSymbolTypeCode;
2124   } else if (symbol_section->IsDescendant(data_section_sp.get()) ||
2125              symbol_section->IsDescendant(data_dirty_section_sp.get()) ||
2126              symbol_section->IsDescendant(data_const_section_sp.get())) {
2127     if (symbol_sect_name &&
2128         ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) {
2129       type = eSymbolTypeRuntime;
2130 
2131       if (symbol_name) {
2132         llvm::StringRef symbol_name_ref(symbol_name);
2133         if (symbol_name_ref.startswith("OBJC_")) {
2134           static const llvm::StringRef g_objc_v2_prefix_class("OBJC_CLASS_$_");
2135           static const llvm::StringRef g_objc_v2_prefix_metaclass(
2136               "OBJC_METACLASS_$_");
2137           static const llvm::StringRef g_objc_v2_prefix_ivar("OBJC_IVAR_$_");
2138           if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) {
2139             symbol_name = symbol_name + g_objc_v2_prefix_class.size();
2140             type = eSymbolTypeObjCClass;
2141             demangled_is_synthesized = true;
2142           } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) {
2143             symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size();
2144             type = eSymbolTypeObjCMetaClass;
2145             demangled_is_synthesized = true;
2146           } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) {
2147             symbol_name = symbol_name + g_objc_v2_prefix_ivar.size();
2148             type = eSymbolTypeObjCIVar;
2149             demangled_is_synthesized = true;
2150           }
2151         }
2152       }
2153     } else if (symbol_sect_name &&
2154                ::strstr(symbol_sect_name, "__gcc_except_tab") ==
2155                    symbol_sect_name) {
2156       type = eSymbolTypeException;
2157     } else {
2158       type = eSymbolTypeData;
2159     }
2160   } else if (symbol_sect_name &&
2161              ::strstr(symbol_sect_name, "__IMPORT") == symbol_sect_name) {
2162     type = eSymbolTypeTrampoline;
2163   }
2164   return type;
2165 }
2166 
2167 // Read the UUID out of a dyld_shared_cache file on-disk.
GetSharedCacheUUID(FileSpec dyld_shared_cache,const ByteOrder byte_order,const uint32_t addr_byte_size)2168 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache,
2169                                          const ByteOrder byte_order,
2170                                          const uint32_t addr_byte_size) {
2171   UUID dsc_uuid;
2172   DataBufferSP DscData = MapFileData(
2173       dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0);
2174   if (!DscData)
2175     return dsc_uuid;
2176   DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size);
2177 
2178   char version_str[7];
2179   lldb::offset_t offset = 0;
2180   memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6);
2181   version_str[6] = '\0';
2182   if (strcmp(version_str, "dyld_v") == 0) {
2183     offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid);
2184     dsc_uuid = UUID::fromOptionalData(
2185         dsc_header_data.GetData(&offset, sizeof(uuid_t)), sizeof(uuid_t));
2186   }
2187   Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS));
2188   if (log && dsc_uuid.IsValid()) {
2189     LLDB_LOGF(log, "Shared cache %s has UUID %s",
2190               dyld_shared_cache.GetPath().c_str(),
2191               dsc_uuid.GetAsString().c_str());
2192   }
2193   return dsc_uuid;
2194 }
2195 
2196 static llvm::Optional<struct nlist_64>
ParseNList(DataExtractor & nlist_data,lldb::offset_t & nlist_data_offset,size_t nlist_byte_size)2197 ParseNList(DataExtractor &nlist_data, lldb::offset_t &nlist_data_offset,
2198            size_t nlist_byte_size) {
2199   struct nlist_64 nlist;
2200   if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size))
2201     return {};
2202   nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset);
2203   nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset);
2204   nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset);
2205   nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset);
2206   nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset);
2207   return nlist;
2208 }
2209 
2210 enum { DebugSymbols = true, NonDebugSymbols = false };
2211 
ParseSymtab()2212 size_t ObjectFileMachO::ParseSymtab() {
2213   LLDB_SCOPED_TIMERF("ObjectFileMachO::ParseSymtab () module = %s",
2214                      m_file.GetFilename().AsCString(""));
2215   ModuleSP module_sp(GetModule());
2216   if (!module_sp)
2217     return 0;
2218 
2219   Progress progress(llvm::formatv("Parsing symbol table for {0}",
2220                                   m_file.GetFilename().AsCString("<Unknown>")));
2221 
2222   llvm::MachO::symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0};
2223   llvm::MachO::linkedit_data_command function_starts_load_command = {0, 0, 0, 0};
2224   llvm::MachO::dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
2225   // The data element of type bool indicates that this entry is thumb
2226   // code.
2227   typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts;
2228 
2229   // Record the address of every function/data that we add to the symtab.
2230   // We add symbols to the table in the order of most information (nlist
2231   // records) to least (function starts), and avoid duplicating symbols
2232   // via this set.
2233   llvm::DenseSet<addr_t> symbols_added;
2234 
2235   // We are using a llvm::DenseSet for "symbols_added" so we must be sure we
2236   // do not add the tombstone or empty keys to the set.
2237   auto add_symbol_addr = [&symbols_added](lldb::addr_t file_addr) {
2238     // Don't add the tombstone or empty keys.
2239     if (file_addr == UINT64_MAX || file_addr == UINT64_MAX - 1)
2240       return;
2241     symbols_added.insert(file_addr);
2242   };
2243   FunctionStarts function_starts;
2244   lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
2245   uint32_t i;
2246   FileSpecList dylib_files;
2247   Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS));
2248   llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_");
2249   llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_");
2250   llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_");
2251 
2252   for (i = 0; i < m_header.ncmds; ++i) {
2253     const lldb::offset_t cmd_offset = offset;
2254     // Read in the load command and load command size
2255     llvm::MachO::load_command lc;
2256     if (m_data.GetU32(&offset, &lc, 2) == nullptr)
2257       break;
2258     // Watch for the symbol table load command
2259     switch (lc.cmd) {
2260     case LC_SYMTAB:
2261       symtab_load_command.cmd = lc.cmd;
2262       symtab_load_command.cmdsize = lc.cmdsize;
2263       // Read in the rest of the symtab load command
2264       if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) ==
2265           nullptr) // fill in symoff, nsyms, stroff, strsize fields
2266         return 0;
2267       break;
2268 
2269     case LC_DYLD_INFO:
2270     case LC_DYLD_INFO_ONLY:
2271       if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) {
2272         dyld_info.cmd = lc.cmd;
2273         dyld_info.cmdsize = lc.cmdsize;
2274       } else {
2275         memset(&dyld_info, 0, sizeof(dyld_info));
2276       }
2277       break;
2278 
2279     case LC_LOAD_DYLIB:
2280     case LC_LOAD_WEAK_DYLIB:
2281     case LC_REEXPORT_DYLIB:
2282     case LC_LOADFVMLIB:
2283     case LC_LOAD_UPWARD_DYLIB: {
2284       uint32_t name_offset = cmd_offset + m_data.GetU32(&offset);
2285       const char *path = m_data.PeekCStr(name_offset);
2286       if (path) {
2287         FileSpec file_spec(path);
2288         // Strip the path if there is @rpath, @executable, etc so we just use
2289         // the basename
2290         if (path[0] == '@')
2291           file_spec.GetDirectory().Clear();
2292 
2293         if (lc.cmd == LC_REEXPORT_DYLIB) {
2294           m_reexported_dylibs.AppendIfUnique(file_spec);
2295         }
2296 
2297         dylib_files.Append(file_spec);
2298       }
2299     } break;
2300 
2301     case LC_FUNCTION_STARTS:
2302       function_starts_load_command.cmd = lc.cmd;
2303       function_starts_load_command.cmdsize = lc.cmdsize;
2304       if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) ==
2305           nullptr) // fill in symoff, nsyms, stroff, strsize fields
2306         memset(&function_starts_load_command, 0,
2307                sizeof(function_starts_load_command));
2308       break;
2309 
2310     default:
2311       break;
2312     }
2313     offset = cmd_offset + lc.cmdsize;
2314   }
2315 
2316   if (!symtab_load_command.cmd)
2317     return 0;
2318 
2319   Symtab *symtab = m_symtab_up.get();
2320   SectionList *section_list = GetSectionList();
2321   if (section_list == nullptr)
2322     return 0;
2323 
2324   const uint32_t addr_byte_size = m_data.GetAddressByteSize();
2325   const ByteOrder byte_order = m_data.GetByteOrder();
2326   bool bit_width_32 = addr_byte_size == 4;
2327   const size_t nlist_byte_size =
2328       bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64);
2329 
2330   DataExtractor nlist_data(nullptr, 0, byte_order, addr_byte_size);
2331   DataExtractor strtab_data(nullptr, 0, byte_order, addr_byte_size);
2332   DataExtractor function_starts_data(nullptr, 0, byte_order, addr_byte_size);
2333   DataExtractor indirect_symbol_index_data(nullptr, 0, byte_order,
2334                                            addr_byte_size);
2335   DataExtractor dyld_trie_data(nullptr, 0, byte_order, addr_byte_size);
2336 
2337   const addr_t nlist_data_byte_size =
2338       symtab_load_command.nsyms * nlist_byte_size;
2339   const addr_t strtab_data_byte_size = symtab_load_command.strsize;
2340   addr_t strtab_addr = LLDB_INVALID_ADDRESS;
2341 
2342   ProcessSP process_sp(m_process_wp.lock());
2343   Process *process = process_sp.get();
2344 
2345   uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete;
2346   bool is_shared_cache_image = m_header.flags & MH_DYLIB_IN_CACHE;
2347   bool is_local_shared_cache_image = is_shared_cache_image && !IsInMemory();
2348   SectionSP linkedit_section_sp(
2349       section_list->FindSectionByName(GetSegmentNameLINKEDIT()));
2350 
2351   if (process && m_header.filetype != llvm::MachO::MH_OBJECT &&
2352       !is_local_shared_cache_image) {
2353     Target &target = process->GetTarget();
2354 
2355     memory_module_load_level = target.GetMemoryModuleLoadLevel();
2356 
2357     // Reading mach file from memory in a process or core file...
2358 
2359     if (linkedit_section_sp) {
2360       addr_t linkedit_load_addr =
2361           linkedit_section_sp->GetLoadBaseAddress(&target);
2362       if (linkedit_load_addr == LLDB_INVALID_ADDRESS) {
2363         // We might be trying to access the symbol table before the
2364         // __LINKEDIT's load address has been set in the target. We can't
2365         // fail to read the symbol table, so calculate the right address
2366         // manually
2367         linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage(
2368             m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get());
2369       }
2370 
2371       const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset();
2372       const addr_t symoff_addr = linkedit_load_addr +
2373                                  symtab_load_command.symoff -
2374                                  linkedit_file_offset;
2375       strtab_addr = linkedit_load_addr + symtab_load_command.stroff -
2376                     linkedit_file_offset;
2377 
2378         // Always load dyld - the dynamic linker - from memory if we didn't
2379         // find a binary anywhere else. lldb will not register
2380         // dylib/framework/bundle loads/unloads if we don't have the dyld
2381         // symbols, we force dyld to load from memory despite the user's
2382         // target.memory-module-load-level setting.
2383         if (memory_module_load_level == eMemoryModuleLoadLevelComplete ||
2384             m_header.filetype == llvm::MachO::MH_DYLINKER) {
2385           DataBufferSP nlist_data_sp(
2386               ReadMemory(process_sp, symoff_addr, nlist_data_byte_size));
2387           if (nlist_data_sp)
2388             nlist_data.SetData(nlist_data_sp, 0, nlist_data_sp->GetByteSize());
2389           if (m_dysymtab.nindirectsyms != 0) {
2390             const addr_t indirect_syms_addr = linkedit_load_addr +
2391                                               m_dysymtab.indirectsymoff -
2392                                               linkedit_file_offset;
2393             DataBufferSP indirect_syms_data_sp(ReadMemory(
2394                 process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4));
2395             if (indirect_syms_data_sp)
2396               indirect_symbol_index_data.SetData(
2397                   indirect_syms_data_sp, 0,
2398                   indirect_syms_data_sp->GetByteSize());
2399             // If this binary is outside the shared cache,
2400             // cache the string table.
2401             // Binaries in the shared cache all share a giant string table,
2402             // and we can't share the string tables across multiple
2403             // ObjectFileMachO's, so we'd end up re-reading this mega-strtab
2404             // for every binary in the shared cache - it would be a big perf
2405             // problem. For binaries outside the shared cache, it's faster to
2406             // read the entire strtab at once instead of piece-by-piece as we
2407             // process the nlist records.
2408             if (!is_shared_cache_image) {
2409               DataBufferSP strtab_data_sp(
2410                   ReadMemory(process_sp, strtab_addr, strtab_data_byte_size));
2411               if (strtab_data_sp) {
2412                 strtab_data.SetData(strtab_data_sp, 0,
2413                                     strtab_data_sp->GetByteSize());
2414               }
2415             }
2416           }
2417         if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) {
2418           if (function_starts_load_command.cmd) {
2419             const addr_t func_start_addr =
2420                 linkedit_load_addr + function_starts_load_command.dataoff -
2421                 linkedit_file_offset;
2422             DataBufferSP func_start_data_sp(
2423                 ReadMemory(process_sp, func_start_addr,
2424                            function_starts_load_command.datasize));
2425             if (func_start_data_sp)
2426               function_starts_data.SetData(func_start_data_sp, 0,
2427                                            func_start_data_sp->GetByteSize());
2428           }
2429         }
2430       }
2431     }
2432   } else {
2433     if (is_local_shared_cache_image) {
2434       // The load commands in shared cache images are relative to the
2435       // beginning of the shared cache, not the library image. The
2436       // data we get handed when creating the ObjectFileMachO starts
2437       // at the beginning of a specific library and spans to the end
2438       // of the cache to be able to reach the shared LINKEDIT
2439       // segments. We need to convert the load command offsets to be
2440       // relative to the beginning of our specific image.
2441       lldb::addr_t linkedit_offset = linkedit_section_sp->GetFileOffset();
2442       lldb::offset_t linkedit_slide =
2443           linkedit_offset - m_linkedit_original_offset;
2444       symtab_load_command.symoff += linkedit_slide;
2445       symtab_load_command.stroff += linkedit_slide;
2446       dyld_info.export_off += linkedit_slide;
2447       m_dysymtab.indirectsymoff += linkedit_slide;
2448       function_starts_load_command.dataoff += linkedit_slide;
2449     }
2450 
2451     nlist_data.SetData(m_data, symtab_load_command.symoff,
2452                        nlist_data_byte_size);
2453     strtab_data.SetData(m_data, symtab_load_command.stroff,
2454                         strtab_data_byte_size);
2455 
2456     if (dyld_info.export_size > 0) {
2457       dyld_trie_data.SetData(m_data, dyld_info.export_off,
2458                              dyld_info.export_size);
2459     }
2460 
2461     if (m_dysymtab.nindirectsyms != 0) {
2462       indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff,
2463                                          m_dysymtab.nindirectsyms * 4);
2464     }
2465     if (function_starts_load_command.cmd) {
2466       function_starts_data.SetData(m_data, function_starts_load_command.dataoff,
2467                                    function_starts_load_command.datasize);
2468     }
2469   }
2470 
2471   const bool have_strtab_data = strtab_data.GetByteSize() > 0;
2472 
2473   ConstString g_segment_name_TEXT = GetSegmentNameTEXT();
2474   ConstString g_segment_name_DATA = GetSegmentNameDATA();
2475   ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY();
2476   ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST();
2477   ConstString g_segment_name_OBJC = GetSegmentNameOBJC();
2478   ConstString g_section_name_eh_frame = GetSectionNameEHFrame();
2479   SectionSP text_section_sp(
2480       section_list->FindSectionByName(g_segment_name_TEXT));
2481   SectionSP data_section_sp(
2482       section_list->FindSectionByName(g_segment_name_DATA));
2483   SectionSP data_dirty_section_sp(
2484       section_list->FindSectionByName(g_segment_name_DATA_DIRTY));
2485   SectionSP data_const_section_sp(
2486       section_list->FindSectionByName(g_segment_name_DATA_CONST));
2487   SectionSP objc_section_sp(
2488       section_list->FindSectionByName(g_segment_name_OBJC));
2489   SectionSP eh_frame_section_sp;
2490   if (text_section_sp.get())
2491     eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName(
2492         g_section_name_eh_frame);
2493   else
2494     eh_frame_section_sp =
2495         section_list->FindSectionByName(g_section_name_eh_frame);
2496 
2497   const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM);
2498   const bool always_thumb = GetArchitecture().IsAlwaysThumbInstructions();
2499 
2500   // lldb works best if it knows the start address of all functions in a
2501   // module. Linker symbols or debug info are normally the best source of
2502   // information for start addr / size but they may be stripped in a released
2503   // binary. Two additional sources of information exist in Mach-O binaries:
2504   //    LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each
2505   //    function's start address in the
2506   //                         binary, relative to the text section.
2507   //    eh_frame           - the eh_frame FDEs have the start addr & size of
2508   //    each function
2509   //  LC_FUNCTION_STARTS is the fastest source to read in, and is present on
2510   //  all modern binaries.
2511   //  Binaries built to run on older releases may need to use eh_frame
2512   //  information.
2513 
2514   if (text_section_sp && function_starts_data.GetByteSize()) {
2515     FunctionStarts::Entry function_start_entry;
2516     function_start_entry.data = false;
2517     lldb::offset_t function_start_offset = 0;
2518     function_start_entry.addr = text_section_sp->GetFileAddress();
2519     uint64_t delta;
2520     while ((delta = function_starts_data.GetULEB128(&function_start_offset)) >
2521            0) {
2522       // Now append the current entry
2523       function_start_entry.addr += delta;
2524       if (is_arm) {
2525         if (function_start_entry.addr & 1) {
2526           function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK;
2527           function_start_entry.data = true;
2528         } else if (always_thumb) {
2529           function_start_entry.data = true;
2530         }
2531       }
2532       function_starts.Append(function_start_entry);
2533     }
2534   } else {
2535     // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the
2536     // load command claiming an eh_frame but it doesn't actually have the
2537     // eh_frame content.  And if we have a dSYM, we don't need to do any of
2538     // this fill-in-the-missing-symbols works anyway - the debug info should
2539     // give us all the functions in the module.
2540     if (text_section_sp.get() && eh_frame_section_sp.get() &&
2541         m_type != eTypeDebugInfo) {
2542       DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp,
2543                                   DWARFCallFrameInfo::EH);
2544       DWARFCallFrameInfo::FunctionAddressAndSizeVector functions;
2545       eh_frame.GetFunctionAddressAndSizeVector(functions);
2546       addr_t text_base_addr = text_section_sp->GetFileAddress();
2547       size_t count = functions.GetSize();
2548       for (size_t i = 0; i < count; ++i) {
2549         const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func =
2550             functions.GetEntryAtIndex(i);
2551         if (func) {
2552           FunctionStarts::Entry function_start_entry;
2553           function_start_entry.addr = func->base - text_base_addr;
2554           if (is_arm) {
2555             if (function_start_entry.addr & 1) {
2556               function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK;
2557               function_start_entry.data = true;
2558             } else if (always_thumb) {
2559               function_start_entry.data = true;
2560             }
2561           }
2562           function_starts.Append(function_start_entry);
2563         }
2564       }
2565     }
2566   }
2567 
2568   const size_t function_starts_count = function_starts.GetSize();
2569 
2570   // For user process binaries (executables, dylibs, frameworks, bundles), if
2571   // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're
2572   // going to assume the binary has been stripped.  Don't allow assembly
2573   // language instruction emulation because we don't know proper function
2574   // start boundaries.
2575   //
2576   // For all other types of binaries (kernels, stand-alone bare board
2577   // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame
2578   // sections - we should not make any assumptions about them based on that.
2579   if (function_starts_count == 0 && CalculateStrata() == eStrataUser) {
2580     m_allow_assembly_emulation_unwind_plans = false;
2581     Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet(
2582         LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND));
2583 
2584     if (unwind_or_symbol_log)
2585       module_sp->LogMessage(
2586           unwind_or_symbol_log,
2587           "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds");
2588   }
2589 
2590   const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get()
2591                                              ? eh_frame_section_sp->GetID()
2592                                              : static_cast<user_id_t>(NO_SECT);
2593 
2594   lldb::offset_t nlist_data_offset = 0;
2595 
2596   uint32_t N_SO_index = UINT32_MAX;
2597 
2598   MachSymtabSectionInfo section_info(section_list);
2599   std::vector<uint32_t> N_FUN_indexes;
2600   std::vector<uint32_t> N_NSYM_indexes;
2601   std::vector<uint32_t> N_INCL_indexes;
2602   std::vector<uint32_t> N_BRAC_indexes;
2603   std::vector<uint32_t> N_COMM_indexes;
2604   typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap;
2605   typedef llvm::DenseMap<uint32_t, uint32_t> NListIndexToSymbolIndexMap;
2606   typedef llvm::DenseMap<const char *, uint32_t> ConstNameToSymbolIndexMap;
2607   ValueToSymbolIndexMap N_FUN_addr_to_sym_idx;
2608   ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx;
2609   ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx;
2610   // Any symbols that get merged into another will get an entry in this map
2611   // so we know
2612   NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx;
2613   uint32_t nlist_idx = 0;
2614   Symbol *symbol_ptr = nullptr;
2615 
2616   uint32_t sym_idx = 0;
2617   Symbol *sym = nullptr;
2618   size_t num_syms = 0;
2619   std::string memory_symbol_name;
2620   uint32_t unmapped_local_symbols_found = 0;
2621 
2622   std::vector<TrieEntryWithOffset> reexport_trie_entries;
2623   std::vector<TrieEntryWithOffset> external_sym_trie_entries;
2624   std::set<lldb::addr_t> resolver_addresses;
2625 
2626   if (dyld_trie_data.GetByteSize() > 0) {
2627     ConstString text_segment_name("__TEXT");
2628     SectionSP text_segment_sp =
2629         GetSectionList()->FindSectionByName(text_segment_name);
2630     lldb::addr_t text_segment_file_addr = LLDB_INVALID_ADDRESS;
2631     if (text_segment_sp)
2632       text_segment_file_addr = text_segment_sp->GetFileAddress();
2633     std::vector<llvm::StringRef> nameSlices;
2634     ParseTrieEntries(dyld_trie_data, 0, is_arm, text_segment_file_addr,
2635                      nameSlices, resolver_addresses, reexport_trie_entries,
2636                      external_sym_trie_entries);
2637   }
2638 
2639   typedef std::set<ConstString> IndirectSymbols;
2640   IndirectSymbols indirect_symbol_names;
2641 
2642 #if TARGET_OS_IPHONE
2643 
2644   // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been
2645   // optimized by moving LOCAL symbols out of the memory mapped portion of
2646   // the DSC. The symbol information has all been retained, but it isn't
2647   // available in the normal nlist data. However, there *are* duplicate
2648   // entries of *some*
2649   // LOCAL symbols in the normal nlist data. To handle this situation
2650   // correctly, we must first attempt
2651   // to parse any DSC unmapped symbol information. If we find any, we set a
2652   // flag that tells the normal nlist parser to ignore all LOCAL symbols.
2653 
2654   if (m_header.flags & MH_DYLIB_IN_CACHE) {
2655     // Before we can start mapping the DSC, we need to make certain the
2656     // target process is actually using the cache we can find.
2657 
2658     // Next we need to determine the correct path for the dyld shared cache.
2659 
2660     ArchSpec header_arch = GetArchitecture();
2661     char dsc_path[PATH_MAX];
2662     char dsc_path_development[PATH_MAX];
2663 
2664     snprintf(
2665         dsc_path, sizeof(dsc_path), "%s%s%s",
2666         "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR
2667                                                    */
2668         "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */
2669         header_arch.GetArchitectureName());
2670 
2671     snprintf(
2672         dsc_path_development, sizeof(dsc_path), "%s%s%s%s",
2673         "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR
2674                                                    */
2675         "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */
2676         header_arch.GetArchitectureName(), ".development");
2677 
2678     FileSpec dsc_nondevelopment_filespec(dsc_path);
2679     FileSpec dsc_development_filespec(dsc_path_development);
2680     FileSpec dsc_filespec;
2681 
2682     UUID dsc_uuid;
2683     UUID process_shared_cache_uuid;
2684     addr_t process_shared_cache_base_addr;
2685 
2686     if (process) {
2687       GetProcessSharedCacheUUID(process, process_shared_cache_base_addr,
2688                                 process_shared_cache_uuid);
2689     }
2690 
2691     // First see if we can find an exact match for the inferior process
2692     // shared cache UUID in the development or non-development shared caches
2693     // on disk.
2694     if (process_shared_cache_uuid.IsValid()) {
2695       if (FileSystem::Instance().Exists(dsc_development_filespec)) {
2696         UUID dsc_development_uuid = GetSharedCacheUUID(
2697             dsc_development_filespec, byte_order, addr_byte_size);
2698         if (dsc_development_uuid.IsValid() &&
2699             dsc_development_uuid == process_shared_cache_uuid) {
2700           dsc_filespec = dsc_development_filespec;
2701           dsc_uuid = dsc_development_uuid;
2702         }
2703       }
2704       if (!dsc_uuid.IsValid() &&
2705           FileSystem::Instance().Exists(dsc_nondevelopment_filespec)) {
2706         UUID dsc_nondevelopment_uuid = GetSharedCacheUUID(
2707             dsc_nondevelopment_filespec, byte_order, addr_byte_size);
2708         if (dsc_nondevelopment_uuid.IsValid() &&
2709             dsc_nondevelopment_uuid == process_shared_cache_uuid) {
2710           dsc_filespec = dsc_nondevelopment_filespec;
2711           dsc_uuid = dsc_nondevelopment_uuid;
2712         }
2713       }
2714     }
2715 
2716     // Failing a UUID match, prefer the development dyld_shared cache if both
2717     // are present.
2718     if (!FileSystem::Instance().Exists(dsc_filespec)) {
2719       if (FileSystem::Instance().Exists(dsc_development_filespec)) {
2720         dsc_filespec = dsc_development_filespec;
2721       } else {
2722         dsc_filespec = dsc_nondevelopment_filespec;
2723       }
2724     }
2725 
2726     /* The dyld_cache_header has a pointer to the
2727        dyld_cache_local_symbols_info structure (localSymbolsOffset).
2728        The dyld_cache_local_symbols_info structure gives us three things:
2729          1. The start and count of the nlist records in the dyld_shared_cache
2730        file
2731          2. The start and size of the strings for these nlist records
2732          3. The start and count of dyld_cache_local_symbols_entry entries
2733 
2734        There is one dyld_cache_local_symbols_entry per dylib/framework in the
2735        dyld shared cache.
2736        The "dylibOffset" field is the Mach-O header of this dylib/framework in
2737        the dyld shared cache.
2738        The dyld_cache_local_symbols_entry also lists the start of this
2739        dylib/framework's nlist records
2740        and the count of how many nlist records there are for this
2741        dylib/framework.
2742     */
2743 
2744     // Process the dyld shared cache header to find the unmapped symbols
2745 
2746     DataBufferSP dsc_data_sp = MapFileData(
2747         dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0);
2748     if (!dsc_uuid.IsValid()) {
2749       dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size);
2750     }
2751     if (dsc_data_sp) {
2752       DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size);
2753 
2754       bool uuid_match = true;
2755       if (dsc_uuid.IsValid() && process) {
2756         if (process_shared_cache_uuid.IsValid() &&
2757             dsc_uuid != process_shared_cache_uuid) {
2758           // The on-disk dyld_shared_cache file is not the same as the one in
2759           // this process' memory, don't use it.
2760           uuid_match = false;
2761           ModuleSP module_sp(GetModule());
2762           if (module_sp)
2763             module_sp->ReportWarning("process shared cache does not match "
2764                                      "on-disk dyld_shared_cache file, some "
2765                                      "symbol names will be missing.");
2766         }
2767       }
2768 
2769       offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset);
2770 
2771       uint32_t mappingOffset = dsc_header_data.GetU32(&offset);
2772 
2773       // If the mappingOffset points to a location inside the header, we've
2774       // opened an old dyld shared cache, and should not proceed further.
2775       if (uuid_match &&
2776           mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) {
2777 
2778         DataBufferSP dsc_mapping_info_data_sp = MapFileData(
2779             dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info),
2780             mappingOffset);
2781 
2782         DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp,
2783                                             byte_order, addr_byte_size);
2784         offset = 0;
2785 
2786         // The File addresses (from the in-memory Mach-O load commands) for
2787         // the shared libraries in the shared library cache need to be
2788         // adjusted by an offset to match up with the dylibOffset identifying
2789         // field in the dyld_cache_local_symbol_entry's.  This offset is
2790         // recorded in mapping_offset_value.
2791         const uint64_t mapping_offset_value =
2792             dsc_mapping_info_data.GetU64(&offset);
2793 
2794         offset =
2795             offsetof(struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset);
2796         uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset);
2797         uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset);
2798 
2799         if (localSymbolsOffset && localSymbolsSize) {
2800           // Map the local symbols
2801           DataBufferSP dsc_local_symbols_data_sp =
2802               MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset);
2803 
2804           if (dsc_local_symbols_data_sp) {
2805             DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp,
2806                                                  byte_order, addr_byte_size);
2807 
2808             offset = 0;
2809 
2810             typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap;
2811             typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName;
2812             UndefinedNameToDescMap undefined_name_to_desc;
2813             SymbolIndexToName reexport_shlib_needs_fixup;
2814 
2815             // Read the local_symbols_infos struct in one shot
2816             struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info;
2817             dsc_local_symbols_data.GetU32(&offset,
2818                                           &local_symbols_info.nlistOffset, 6);
2819 
2820             SectionSP text_section_sp(
2821                 section_list->FindSectionByName(GetSegmentNameTEXT()));
2822 
2823             uint32_t header_file_offset =
2824                 (text_section_sp->GetFileAddress() - mapping_offset_value);
2825 
2826             offset = local_symbols_info.entriesOffset;
2827             for (uint32_t entry_index = 0;
2828                  entry_index < local_symbols_info.entriesCount; entry_index++) {
2829               struct lldb_copy_dyld_cache_local_symbols_entry
2830                   local_symbols_entry;
2831               local_symbols_entry.dylibOffset =
2832                   dsc_local_symbols_data.GetU32(&offset);
2833               local_symbols_entry.nlistStartIndex =
2834                   dsc_local_symbols_data.GetU32(&offset);
2835               local_symbols_entry.nlistCount =
2836                   dsc_local_symbols_data.GetU32(&offset);
2837 
2838               if (header_file_offset == local_symbols_entry.dylibOffset) {
2839                 unmapped_local_symbols_found = local_symbols_entry.nlistCount;
2840 
2841                 // The normal nlist code cannot correctly size the Symbols
2842                 // array, we need to allocate it here.
2843                 sym = symtab->Resize(
2844                     symtab_load_command.nsyms + m_dysymtab.nindirectsyms +
2845                     unmapped_local_symbols_found - m_dysymtab.nlocalsym);
2846                 num_syms = symtab->GetNumSymbols();
2847 
2848                 nlist_data_offset =
2849                     local_symbols_info.nlistOffset +
2850                     (nlist_byte_size * local_symbols_entry.nlistStartIndex);
2851                 uint32_t string_table_offset = local_symbols_info.stringsOffset;
2852 
2853                 for (uint32_t nlist_index = 0;
2854                      nlist_index < local_symbols_entry.nlistCount;
2855                      nlist_index++) {
2856                   /////////////////////////////
2857                   {
2858                     llvm::Optional<struct nlist_64> nlist_maybe =
2859                         ParseNList(dsc_local_symbols_data, nlist_data_offset,
2860                                    nlist_byte_size);
2861                     if (!nlist_maybe)
2862                       break;
2863                     struct nlist_64 nlist = *nlist_maybe;
2864 
2865                     SymbolType type = eSymbolTypeInvalid;
2866                     const char *symbol_name = dsc_local_symbols_data.PeekCStr(
2867                         string_table_offset + nlist.n_strx);
2868 
2869                     if (symbol_name == NULL) {
2870                       // No symbol should be NULL, even the symbols with no
2871                       // string values should have an offset zero which
2872                       // points to an empty C-string
2873                       Host::SystemLog(
2874                           Host::eSystemLogError,
2875                           "error: DSC unmapped local symbol[%u] has invalid "
2876                           "string table offset 0x%x in %s, ignoring symbol\n",
2877                           entry_index, nlist.n_strx,
2878                           module_sp->GetFileSpec().GetPath().c_str());
2879                       continue;
2880                     }
2881                     if (symbol_name[0] == '\0')
2882                       symbol_name = NULL;
2883 
2884                     const char *symbol_name_non_abi_mangled = NULL;
2885 
2886                     SectionSP symbol_section;
2887                     uint32_t symbol_byte_size = 0;
2888                     bool add_nlist = true;
2889                     bool is_debug = ((nlist.n_type & N_STAB) != 0);
2890                     bool demangled_is_synthesized = false;
2891                     bool is_gsym = false;
2892                     bool set_value = true;
2893 
2894                     assert(sym_idx < num_syms);
2895 
2896                     sym[sym_idx].SetDebug(is_debug);
2897 
2898                     if (is_debug) {
2899                       switch (nlist.n_type) {
2900                       case N_GSYM:
2901                         // global symbol: name,,NO_SECT,type,0
2902                         // Sometimes the N_GSYM value contains the address.
2903 
2904                         // FIXME: In the .o files, we have a GSYM and a debug
2905                         // symbol for all the ObjC data.  They
2906                         // have the same address, but we want to ensure that
2907                         // we always find only the real symbol, 'cause we
2908                         // don't currently correctly attribute the
2909                         // GSYM one to the ObjCClass/Ivar/MetaClass
2910                         // symbol type.  This is a temporary hack to make
2911                         // sure the ObjectiveC symbols get treated correctly.
2912                         // To do this right, we should coalesce all the GSYM
2913                         // & global symbols that have the same address.
2914 
2915                         is_gsym = true;
2916                         sym[sym_idx].SetExternal(true);
2917 
2918                         if (symbol_name && symbol_name[0] == '_' &&
2919                             symbol_name[1] == 'O') {
2920                           llvm::StringRef symbol_name_ref(symbol_name);
2921                           if (symbol_name_ref.startswith(
2922                                   g_objc_v2_prefix_class)) {
2923                             symbol_name_non_abi_mangled = symbol_name + 1;
2924                             symbol_name =
2925                                 symbol_name + g_objc_v2_prefix_class.size();
2926                             type = eSymbolTypeObjCClass;
2927                             demangled_is_synthesized = true;
2928 
2929                           } else if (symbol_name_ref.startswith(
2930                                          g_objc_v2_prefix_metaclass)) {
2931                             symbol_name_non_abi_mangled = symbol_name + 1;
2932                             symbol_name =
2933                                 symbol_name + g_objc_v2_prefix_metaclass.size();
2934                             type = eSymbolTypeObjCMetaClass;
2935                             demangled_is_synthesized = true;
2936                           } else if (symbol_name_ref.startswith(
2937                                          g_objc_v2_prefix_ivar)) {
2938                             symbol_name_non_abi_mangled = symbol_name + 1;
2939                             symbol_name =
2940                                 symbol_name + g_objc_v2_prefix_ivar.size();
2941                             type = eSymbolTypeObjCIVar;
2942                             demangled_is_synthesized = true;
2943                           }
2944                         } else {
2945                           if (nlist.n_value != 0)
2946                             symbol_section = section_info.GetSection(
2947                                 nlist.n_sect, nlist.n_value);
2948                           type = eSymbolTypeData;
2949                         }
2950                         break;
2951 
2952                       case N_FNAME:
2953                         // procedure name (f77 kludge): name,,NO_SECT,0,0
2954                         type = eSymbolTypeCompiler;
2955                         break;
2956 
2957                       case N_FUN:
2958                         // procedure: name,,n_sect,linenumber,address
2959                         if (symbol_name) {
2960                           type = eSymbolTypeCode;
2961                           symbol_section = section_info.GetSection(
2962                               nlist.n_sect, nlist.n_value);
2963 
2964                           N_FUN_addr_to_sym_idx.insert(
2965                               std::make_pair(nlist.n_value, sym_idx));
2966                           // We use the current number of symbols in the
2967                           // symbol table in lieu of using nlist_idx in case
2968                           // we ever start trimming entries out
2969                           N_FUN_indexes.push_back(sym_idx);
2970                         } else {
2971                           type = eSymbolTypeCompiler;
2972 
2973                           if (!N_FUN_indexes.empty()) {
2974                             // Copy the size of the function into the
2975                             // original
2976                             // STAB entry so we don't have
2977                             // to hunt for it later
2978                             symtab->SymbolAtIndex(N_FUN_indexes.back())
2979                                 ->SetByteSize(nlist.n_value);
2980                             N_FUN_indexes.pop_back();
2981                             // We don't really need the end function STAB as
2982                             // it contains the size which we already placed
2983                             // with the original symbol, so don't add it if
2984                             // we want a minimal symbol table
2985                             add_nlist = false;
2986                           }
2987                         }
2988                         break;
2989 
2990                       case N_STSYM:
2991                         // static symbol: name,,n_sect,type,address
2992                         N_STSYM_addr_to_sym_idx.insert(
2993                             std::make_pair(nlist.n_value, sym_idx));
2994                         symbol_section = section_info.GetSection(nlist.n_sect,
2995                                                                  nlist.n_value);
2996                         if (symbol_name && symbol_name[0]) {
2997                           type = ObjectFile::GetSymbolTypeFromName(
2998                               symbol_name + 1, eSymbolTypeData);
2999                         }
3000                         break;
3001 
3002                       case N_LCSYM:
3003                         // .lcomm symbol: name,,n_sect,type,address
3004                         symbol_section = section_info.GetSection(nlist.n_sect,
3005                                                                  nlist.n_value);
3006                         type = eSymbolTypeCommonBlock;
3007                         break;
3008 
3009                       case N_BNSYM:
3010                         // We use the current number of symbols in the symbol
3011                         // table in lieu of using nlist_idx in case we ever
3012                         // start trimming entries out Skip these if we want
3013                         // minimal symbol tables
3014                         add_nlist = false;
3015                         break;
3016 
3017                       case N_ENSYM:
3018                         // Set the size of the N_BNSYM to the terminating
3019                         // index of this N_ENSYM so that we can always skip
3020                         // the entire symbol if we need to navigate more
3021                         // quickly at the source level when parsing STABS
3022                         // Skip these if we want minimal symbol tables
3023                         add_nlist = false;
3024                         break;
3025 
3026                       case N_OPT:
3027                         // emitted with gcc2_compiled and in gcc source
3028                         type = eSymbolTypeCompiler;
3029                         break;
3030 
3031                       case N_RSYM:
3032                         // register sym: name,,NO_SECT,type,register
3033                         type = eSymbolTypeVariable;
3034                         break;
3035 
3036                       case N_SLINE:
3037                         // src line: 0,,n_sect,linenumber,address
3038                         symbol_section = section_info.GetSection(nlist.n_sect,
3039                                                                  nlist.n_value);
3040                         type = eSymbolTypeLineEntry;
3041                         break;
3042 
3043                       case N_SSYM:
3044                         // structure elt: name,,NO_SECT,type,struct_offset
3045                         type = eSymbolTypeVariableType;
3046                         break;
3047 
3048                       case N_SO:
3049                         // source file name
3050                         type = eSymbolTypeSourceFile;
3051                         if (symbol_name == NULL) {
3052                           add_nlist = false;
3053                           if (N_SO_index != UINT32_MAX) {
3054                             // Set the size of the N_SO to the terminating
3055                             // index of this N_SO so that we can always skip
3056                             // the entire N_SO if we need to navigate more
3057                             // quickly at the source level when parsing STABS
3058                             symbol_ptr = symtab->SymbolAtIndex(N_SO_index);
3059                             symbol_ptr->SetByteSize(sym_idx);
3060                             symbol_ptr->SetSizeIsSibling(true);
3061                           }
3062                           N_NSYM_indexes.clear();
3063                           N_INCL_indexes.clear();
3064                           N_BRAC_indexes.clear();
3065                           N_COMM_indexes.clear();
3066                           N_FUN_indexes.clear();
3067                           N_SO_index = UINT32_MAX;
3068                         } else {
3069                           // We use the current number of symbols in the
3070                           // symbol table in lieu of using nlist_idx in case
3071                           // we ever start trimming entries out
3072                           const bool N_SO_has_full_path = symbol_name[0] == '/';
3073                           if (N_SO_has_full_path) {
3074                             if ((N_SO_index == sym_idx - 1) &&
3075                                 ((sym_idx - 1) < num_syms)) {
3076                               // We have two consecutive N_SO entries where
3077                               // the first contains a directory and the
3078                               // second contains a full path.
3079                               sym[sym_idx - 1].GetMangled().SetValue(
3080                                   ConstString(symbol_name), false);
3081                               m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1;
3082                               add_nlist = false;
3083                             } else {
3084                               // This is the first entry in a N_SO that
3085                               // contains a directory or
3086                               // a full path to the source file
3087                               N_SO_index = sym_idx;
3088                             }
3089                           } else if ((N_SO_index == sym_idx - 1) &&
3090                                      ((sym_idx - 1) < num_syms)) {
3091                             // This is usually the second N_SO entry that
3092                             // contains just the filename, so here we combine
3093                             // it with the first one if we are minimizing the
3094                             // symbol table
3095                             const char *so_path = sym[sym_idx - 1]
3096                                                       .GetMangled()
3097                                                       .GetDemangledName()
3098                                                       .AsCString();
3099                             if (so_path && so_path[0]) {
3100                               std::string full_so_path(so_path);
3101                               const size_t double_slash_pos =
3102                                   full_so_path.find("//");
3103                               if (double_slash_pos != std::string::npos) {
3104                                 // The linker has been generating bad N_SO
3105                                 // entries with doubled up paths
3106                                 // in the format "%s%s" where the first
3107                                 // string in the DW_AT_comp_dir, and the
3108                                 // second is the directory for the source
3109                                 // file so you end up with a path that looks
3110                                 // like "/tmp/src//tmp/src/"
3111                                 FileSpec so_dir(so_path);
3112                                 if (!FileSystem::Instance().Exists(so_dir)) {
3113                                   so_dir.SetFile(
3114                                       &full_so_path[double_slash_pos + 1],
3115                                       FileSpec::Style::native);
3116                                   if (FileSystem::Instance().Exists(so_dir)) {
3117                                     // Trim off the incorrect path
3118                                     full_so_path.erase(0, double_slash_pos + 1);
3119                                   }
3120                                 }
3121                               }
3122                               if (*full_so_path.rbegin() != '/')
3123                                 full_so_path += '/';
3124                               full_so_path += symbol_name;
3125                               sym[sym_idx - 1].GetMangled().SetValue(
3126                                   ConstString(full_so_path.c_str()), false);
3127                               add_nlist = false;
3128                               m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1;
3129                             }
3130                           } else {
3131                             // This could be a relative path to a N_SO
3132                             N_SO_index = sym_idx;
3133                           }
3134                         }
3135                         break;
3136 
3137                       case N_OSO:
3138                         // object file name: name,,0,0,st_mtime
3139                         type = eSymbolTypeObjectFile;
3140                         break;
3141 
3142                       case N_LSYM:
3143                         // local sym: name,,NO_SECT,type,offset
3144                         type = eSymbolTypeLocal;
3145                         break;
3146 
3147                       // INCL scopes
3148                       case N_BINCL:
3149                         // include file beginning: name,,NO_SECT,0,sum We use
3150                         // the current number of symbols in the symbol table
3151                         // in lieu of using nlist_idx in case we ever start
3152                         // trimming entries out
3153                         N_INCL_indexes.push_back(sym_idx);
3154                         type = eSymbolTypeScopeBegin;
3155                         break;
3156 
3157                       case N_EINCL:
3158                         // include file end: name,,NO_SECT,0,0
3159                         // Set the size of the N_BINCL to the terminating
3160                         // index of this N_EINCL so that we can always skip
3161                         // the entire symbol if we need to navigate more
3162                         // quickly at the source level when parsing STABS
3163                         if (!N_INCL_indexes.empty()) {
3164                           symbol_ptr =
3165                               symtab->SymbolAtIndex(N_INCL_indexes.back());
3166                           symbol_ptr->SetByteSize(sym_idx + 1);
3167                           symbol_ptr->SetSizeIsSibling(true);
3168                           N_INCL_indexes.pop_back();
3169                         }
3170                         type = eSymbolTypeScopeEnd;
3171                         break;
3172 
3173                       case N_SOL:
3174                         // #included file name: name,,n_sect,0,address
3175                         type = eSymbolTypeHeaderFile;
3176 
3177                         // We currently don't use the header files on darwin
3178                         add_nlist = false;
3179                         break;
3180 
3181                       case N_PARAMS:
3182                         // compiler parameters: name,,NO_SECT,0,0
3183                         type = eSymbolTypeCompiler;
3184                         break;
3185 
3186                       case N_VERSION:
3187                         // compiler version: name,,NO_SECT,0,0
3188                         type = eSymbolTypeCompiler;
3189                         break;
3190 
3191                       case N_OLEVEL:
3192                         // compiler -O level: name,,NO_SECT,0,0
3193                         type = eSymbolTypeCompiler;
3194                         break;
3195 
3196                       case N_PSYM:
3197                         // parameter: name,,NO_SECT,type,offset
3198                         type = eSymbolTypeVariable;
3199                         break;
3200 
3201                       case N_ENTRY:
3202                         // alternate entry: name,,n_sect,linenumber,address
3203                         symbol_section = section_info.GetSection(nlist.n_sect,
3204                                                                  nlist.n_value);
3205                         type = eSymbolTypeLineEntry;
3206                         break;
3207 
3208                       // Left and Right Braces
3209                       case N_LBRAC:
3210                         // left bracket: 0,,NO_SECT,nesting level,address We
3211                         // use the current number of symbols in the symbol
3212                         // table in lieu of using nlist_idx in case we ever
3213                         // start trimming entries out
3214                         symbol_section = section_info.GetSection(nlist.n_sect,
3215                                                                  nlist.n_value);
3216                         N_BRAC_indexes.push_back(sym_idx);
3217                         type = eSymbolTypeScopeBegin;
3218                         break;
3219 
3220                       case N_RBRAC:
3221                         // right bracket: 0,,NO_SECT,nesting level,address
3222                         // Set the size of the N_LBRAC to the terminating
3223                         // index of this N_RBRAC so that we can always skip
3224                         // the entire symbol if we need to navigate more
3225                         // quickly at the source level when parsing STABS
3226                         symbol_section = section_info.GetSection(nlist.n_sect,
3227                                                                  nlist.n_value);
3228                         if (!N_BRAC_indexes.empty()) {
3229                           symbol_ptr =
3230                               symtab->SymbolAtIndex(N_BRAC_indexes.back());
3231                           symbol_ptr->SetByteSize(sym_idx + 1);
3232                           symbol_ptr->SetSizeIsSibling(true);
3233                           N_BRAC_indexes.pop_back();
3234                         }
3235                         type = eSymbolTypeScopeEnd;
3236                         break;
3237 
3238                       case N_EXCL:
3239                         // deleted include file: name,,NO_SECT,0,sum
3240                         type = eSymbolTypeHeaderFile;
3241                         break;
3242 
3243                       // COMM scopes
3244                       case N_BCOMM:
3245                         // begin common: name,,NO_SECT,0,0
3246                         // We use the current number of symbols in the symbol
3247                         // table in lieu of using nlist_idx in case we ever
3248                         // start trimming entries out
3249                         type = eSymbolTypeScopeBegin;
3250                         N_COMM_indexes.push_back(sym_idx);
3251                         break;
3252 
3253                       case N_ECOML:
3254                         // end common (local name): 0,,n_sect,0,address
3255                         symbol_section = section_info.GetSection(nlist.n_sect,
3256                                                                  nlist.n_value);
3257                         // Fall through
3258 
3259                       case N_ECOMM:
3260                         // end common: name,,n_sect,0,0
3261                         // Set the size of the N_BCOMM to the terminating
3262                         // index of this N_ECOMM/N_ECOML so that we can
3263                         // always skip the entire symbol if we need to
3264                         // navigate more quickly at the source level when
3265                         // parsing STABS
3266                         if (!N_COMM_indexes.empty()) {
3267                           symbol_ptr =
3268                               symtab->SymbolAtIndex(N_COMM_indexes.back());
3269                           symbol_ptr->SetByteSize(sym_idx + 1);
3270                           symbol_ptr->SetSizeIsSibling(true);
3271                           N_COMM_indexes.pop_back();
3272                         }
3273                         type = eSymbolTypeScopeEnd;
3274                         break;
3275 
3276                       case N_LENG:
3277                         // second stab entry with length information
3278                         type = eSymbolTypeAdditional;
3279                         break;
3280 
3281                       default:
3282                         break;
3283                       }
3284                     } else {
3285                       // uint8_t n_pext    = N_PEXT & nlist.n_type;
3286                       uint8_t n_type = N_TYPE & nlist.n_type;
3287                       sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0);
3288 
3289                       switch (n_type) {
3290                       case N_INDR: {
3291                         const char *reexport_name_cstr =
3292                             strtab_data.PeekCStr(nlist.n_value);
3293                         if (reexport_name_cstr && reexport_name_cstr[0]) {
3294                           type = eSymbolTypeReExported;
3295                           ConstString reexport_name(
3296                               reexport_name_cstr +
3297                               ((reexport_name_cstr[0] == '_') ? 1 : 0));
3298                           sym[sym_idx].SetReExportedSymbolName(reexport_name);
3299                           set_value = false;
3300                           reexport_shlib_needs_fixup[sym_idx] = reexport_name;
3301                           indirect_symbol_names.insert(ConstString(
3302                               symbol_name + ((symbol_name[0] == '_') ? 1 : 0)));
3303                         } else
3304                           type = eSymbolTypeUndefined;
3305                       } break;
3306 
3307                       case N_UNDF:
3308                         if (symbol_name && symbol_name[0]) {
3309                           ConstString undefined_name(
3310                               symbol_name + ((symbol_name[0] == '_') ? 1 : 0));
3311                           undefined_name_to_desc[undefined_name] = nlist.n_desc;
3312                         }
3313                       // Fall through
3314                       case N_PBUD:
3315                         type = eSymbolTypeUndefined;
3316                         break;
3317 
3318                       case N_ABS:
3319                         type = eSymbolTypeAbsolute;
3320                         break;
3321 
3322                       case N_SECT: {
3323                         symbol_section = section_info.GetSection(nlist.n_sect,
3324                                                                  nlist.n_value);
3325 
3326                         if (symbol_section == NULL) {
3327                           // TODO: warn about this?
3328                           add_nlist = false;
3329                           break;
3330                         }
3331 
3332                         if (TEXT_eh_frame_sectID == nlist.n_sect) {
3333                           type = eSymbolTypeException;
3334                         } else {
3335                           uint32_t section_type =
3336                               symbol_section->Get() & SECTION_TYPE;
3337 
3338                           switch (section_type) {
3339                           case S_CSTRING_LITERALS:
3340                             type = eSymbolTypeData;
3341                             break; // section with only literal C strings
3342                           case S_4BYTE_LITERALS:
3343                             type = eSymbolTypeData;
3344                             break; // section with only 4 byte literals
3345                           case S_8BYTE_LITERALS:
3346                             type = eSymbolTypeData;
3347                             break; // section with only 8 byte literals
3348                           case S_LITERAL_POINTERS:
3349                             type = eSymbolTypeTrampoline;
3350                             break; // section with only pointers to literals
3351                           case S_NON_LAZY_SYMBOL_POINTERS:
3352                             type = eSymbolTypeTrampoline;
3353                             break; // section with only non-lazy symbol
3354                                    // pointers
3355                           case S_LAZY_SYMBOL_POINTERS:
3356                             type = eSymbolTypeTrampoline;
3357                             break; // section with only lazy symbol pointers
3358                           case S_SYMBOL_STUBS:
3359                             type = eSymbolTypeTrampoline;
3360                             break; // section with only symbol stubs, byte
3361                                    // size of stub in the reserved2 field
3362                           case S_MOD_INIT_FUNC_POINTERS:
3363                             type = eSymbolTypeCode;
3364                             break; // section with only function pointers for
3365                                    // initialization
3366                           case S_MOD_TERM_FUNC_POINTERS:
3367                             type = eSymbolTypeCode;
3368                             break; // section with only function pointers for
3369                                    // termination
3370                           case S_INTERPOSING:
3371                             type = eSymbolTypeTrampoline;
3372                             break; // section with only pairs of function
3373                                    // pointers for interposing
3374                           case S_16BYTE_LITERALS:
3375                             type = eSymbolTypeData;
3376                             break; // section with only 16 byte literals
3377                           case S_DTRACE_DOF:
3378                             type = eSymbolTypeInstrumentation;
3379                             break;
3380                           case S_LAZY_DYLIB_SYMBOL_POINTERS:
3381                             type = eSymbolTypeTrampoline;
3382                             break;
3383                           default:
3384                             switch (symbol_section->GetType()) {
3385                             case lldb::eSectionTypeCode:
3386                               type = eSymbolTypeCode;
3387                               break;
3388                             case eSectionTypeData:
3389                             case eSectionTypeDataCString: // Inlined C string
3390                                                           // data
3391                             case eSectionTypeDataCStringPointers: // Pointers
3392                                                                   // to C
3393                                                                   // string
3394                                                                   // data
3395                             case eSectionTypeDataSymbolAddress:   // Address of
3396                                                                   // a symbol in
3397                                                                   // the symbol
3398                                                                   // table
3399                             case eSectionTypeData4:
3400                             case eSectionTypeData8:
3401                             case eSectionTypeData16:
3402                               type = eSymbolTypeData;
3403                               break;
3404                             default:
3405                               break;
3406                             }
3407                             break;
3408                           }
3409 
3410                           if (type == eSymbolTypeInvalid) {
3411                             const char *symbol_sect_name =
3412                                 symbol_section->GetName().AsCString();
3413                             if (symbol_section->IsDescendant(
3414                                     text_section_sp.get())) {
3415                               if (symbol_section->IsClear(
3416                                       S_ATTR_PURE_INSTRUCTIONS |
3417                                       S_ATTR_SELF_MODIFYING_CODE |
3418                                       S_ATTR_SOME_INSTRUCTIONS))
3419                                 type = eSymbolTypeData;
3420                               else
3421                                 type = eSymbolTypeCode;
3422                             } else if (symbol_section->IsDescendant(
3423                                            data_section_sp.get()) ||
3424                                        symbol_section->IsDescendant(
3425                                            data_dirty_section_sp.get()) ||
3426                                        symbol_section->IsDescendant(
3427                                            data_const_section_sp.get())) {
3428                               if (symbol_sect_name &&
3429                                   ::strstr(symbol_sect_name, "__objc") ==
3430                                       symbol_sect_name) {
3431                                 type = eSymbolTypeRuntime;
3432 
3433                                 if (symbol_name) {
3434                                   llvm::StringRef symbol_name_ref(symbol_name);
3435                                   if (symbol_name_ref.startswith("_OBJC_")) {
3436                                     llvm::StringRef
3437                                         g_objc_v2_prefix_class(
3438                                             "_OBJC_CLASS_$_");
3439                                     llvm::StringRef
3440                                         g_objc_v2_prefix_metaclass(
3441                                             "_OBJC_METACLASS_$_");
3442                                     llvm::StringRef
3443                                         g_objc_v2_prefix_ivar("_OBJC_IVAR_$_");
3444                                     if (symbol_name_ref.startswith(
3445                                             g_objc_v2_prefix_class)) {
3446                                       symbol_name_non_abi_mangled =
3447                                           symbol_name + 1;
3448                                       symbol_name =
3449                                           symbol_name +
3450                                           g_objc_v2_prefix_class.size();
3451                                       type = eSymbolTypeObjCClass;
3452                                       demangled_is_synthesized = true;
3453                                     } else if (
3454                                         symbol_name_ref.startswith(
3455                                             g_objc_v2_prefix_metaclass)) {
3456                                       symbol_name_non_abi_mangled =
3457                                           symbol_name + 1;
3458                                       symbol_name =
3459                                           symbol_name +
3460                                           g_objc_v2_prefix_metaclass.size();
3461                                       type = eSymbolTypeObjCMetaClass;
3462                                       demangled_is_synthesized = true;
3463                                     } else if (symbol_name_ref.startswith(
3464                                                    g_objc_v2_prefix_ivar)) {
3465                                       symbol_name_non_abi_mangled =
3466                                           symbol_name + 1;
3467                                       symbol_name =
3468                                           symbol_name +
3469                                           g_objc_v2_prefix_ivar.size();
3470                                       type = eSymbolTypeObjCIVar;
3471                                       demangled_is_synthesized = true;
3472                                     }
3473                                   }
3474                                 }
3475                               } else if (symbol_sect_name &&
3476                                          ::strstr(symbol_sect_name,
3477                                                   "__gcc_except_tab") ==
3478                                              symbol_sect_name) {
3479                                 type = eSymbolTypeException;
3480                               } else {
3481                                 type = eSymbolTypeData;
3482                               }
3483                             } else if (symbol_sect_name &&
3484                                        ::strstr(symbol_sect_name, "__IMPORT") ==
3485                                            symbol_sect_name) {
3486                               type = eSymbolTypeTrampoline;
3487                             } else if (symbol_section->IsDescendant(
3488                                            objc_section_sp.get())) {
3489                               type = eSymbolTypeRuntime;
3490                               if (symbol_name && symbol_name[0] == '.') {
3491                                 llvm::StringRef symbol_name_ref(symbol_name);
3492                                 llvm::StringRef
3493                                     g_objc_v1_prefix_class(".objc_class_name_");
3494                                 if (symbol_name_ref.startswith(
3495                                         g_objc_v1_prefix_class)) {
3496                                   symbol_name_non_abi_mangled = symbol_name;
3497                                   symbol_name = symbol_name +
3498                                                 g_objc_v1_prefix_class.size();
3499                                   type = eSymbolTypeObjCClass;
3500                                   demangled_is_synthesized = true;
3501                                 }
3502                               }
3503                             }
3504                           }
3505                         }
3506                       } break;
3507                       }
3508                     }
3509 
3510                     if (add_nlist) {
3511                       uint64_t symbol_value = nlist.n_value;
3512                       if (symbol_name_non_abi_mangled) {
3513                         sym[sym_idx].GetMangled().SetMangledName(
3514                             ConstString(symbol_name_non_abi_mangled));
3515                         sym[sym_idx].GetMangled().SetDemangledName(
3516                             ConstString(symbol_name));
3517                       } else {
3518                         bool symbol_name_is_mangled = false;
3519 
3520                         if (symbol_name && symbol_name[0] == '_') {
3521                           symbol_name_is_mangled = symbol_name[1] == '_';
3522                           symbol_name++; // Skip the leading underscore
3523                         }
3524 
3525                         if (symbol_name) {
3526                           ConstString const_symbol_name(symbol_name);
3527                           sym[sym_idx].GetMangled().SetValue(
3528                               const_symbol_name, symbol_name_is_mangled);
3529                           if (is_gsym && is_debug) {
3530                             const char *gsym_name =
3531                                 sym[sym_idx]
3532                                     .GetMangled()
3533                                     .GetName(Mangled::ePreferMangled)
3534                                     .GetCString();
3535                             if (gsym_name)
3536                               N_GSYM_name_to_sym_idx[gsym_name] = sym_idx;
3537                           }
3538                         }
3539                       }
3540                       if (symbol_section) {
3541                         const addr_t section_file_addr =
3542                             symbol_section->GetFileAddress();
3543                         if (symbol_byte_size == 0 &&
3544                             function_starts_count > 0) {
3545                           addr_t symbol_lookup_file_addr = nlist.n_value;
3546                           // Do an exact address match for non-ARM addresses,
3547                           // else get the closest since the symbol might be a
3548                           // thumb symbol which has an address with bit zero
3549                           // set
3550                           FunctionStarts::Entry *func_start_entry =
3551                               function_starts.FindEntry(symbol_lookup_file_addr,
3552                                                         !is_arm);
3553                           if (is_arm && func_start_entry) {
3554                             // Verify that the function start address is the
3555                             // symbol address (ARM) or the symbol address + 1
3556                             // (thumb)
3557                             if (func_start_entry->addr !=
3558                                     symbol_lookup_file_addr &&
3559                                 func_start_entry->addr !=
3560                                     (symbol_lookup_file_addr + 1)) {
3561                               // Not the right entry, NULL it out...
3562                               func_start_entry = NULL;
3563                             }
3564                           }
3565                           if (func_start_entry) {
3566                             func_start_entry->data = true;
3567 
3568                             addr_t symbol_file_addr = func_start_entry->addr;
3569                             uint32_t symbol_flags = 0;
3570                             if (is_arm) {
3571                               if (symbol_file_addr & 1)
3572                                 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB;
3573                               symbol_file_addr &= THUMB_ADDRESS_BIT_MASK;
3574                             }
3575 
3576                             const FunctionStarts::Entry *next_func_start_entry =
3577                                 function_starts.FindNextEntry(func_start_entry);
3578                             const addr_t section_end_file_addr =
3579                                 section_file_addr +
3580                                 symbol_section->GetByteSize();
3581                             if (next_func_start_entry) {
3582                               addr_t next_symbol_file_addr =
3583                                   next_func_start_entry->addr;
3584                               // Be sure the clear the Thumb address bit when
3585                               // we calculate the size from the current and
3586                               // next address
3587                               if (is_arm)
3588                                 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK;
3589                               symbol_byte_size = std::min<lldb::addr_t>(
3590                                   next_symbol_file_addr - symbol_file_addr,
3591                                   section_end_file_addr - symbol_file_addr);
3592                             } else {
3593                               symbol_byte_size =
3594                                   section_end_file_addr - symbol_file_addr;
3595                             }
3596                           }
3597                         }
3598                         symbol_value -= section_file_addr;
3599                       }
3600 
3601                       if (is_debug == false) {
3602                         if (type == eSymbolTypeCode) {
3603                           // See if we can find a N_FUN entry for any code
3604                           // symbols. If we do find a match, and the name
3605                           // matches, then we can merge the two into just the
3606                           // function symbol to avoid duplicate entries in
3607                           // the symbol table
3608                           auto range =
3609                               N_FUN_addr_to_sym_idx.equal_range(nlist.n_value);
3610                           if (range.first != range.second) {
3611                             bool found_it = false;
3612                             for (auto pos = range.first; pos != range.second;
3613                                  ++pos) {
3614                               if (sym[sym_idx].GetMangled().GetName(
3615                                       Mangled::ePreferMangled) ==
3616                                   sym[pos->second].GetMangled().GetName(
3617                                       Mangled::ePreferMangled)) {
3618                                 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second;
3619                                 // We just need the flags from the linker
3620                                 // symbol, so put these flags
3621                                 // into the N_FUN flags to avoid duplicate
3622                                 // symbols in the symbol table
3623                                 sym[pos->second].SetExternal(
3624                                     sym[sym_idx].IsExternal());
3625                                 sym[pos->second].SetFlags(nlist.n_type << 16 |
3626                                                           nlist.n_desc);
3627                                 if (resolver_addresses.find(nlist.n_value) !=
3628                                     resolver_addresses.end())
3629                                   sym[pos->second].SetType(eSymbolTypeResolver);
3630                                 sym[sym_idx].Clear();
3631                                 found_it = true;
3632                                 break;
3633                               }
3634                             }
3635                             if (found_it)
3636                               continue;
3637                           } else {
3638                             if (resolver_addresses.find(nlist.n_value) !=
3639                                 resolver_addresses.end())
3640                               type = eSymbolTypeResolver;
3641                           }
3642                         } else if (type == eSymbolTypeData ||
3643                                    type == eSymbolTypeObjCClass ||
3644                                    type == eSymbolTypeObjCMetaClass ||
3645                                    type == eSymbolTypeObjCIVar) {
3646                           // See if we can find a N_STSYM entry for any data
3647                           // symbols. If we do find a match, and the name
3648                           // matches, then we can merge the two into just the
3649                           // Static symbol to avoid duplicate entries in the
3650                           // symbol table
3651                           auto range = N_STSYM_addr_to_sym_idx.equal_range(
3652                               nlist.n_value);
3653                           if (range.first != range.second) {
3654                             bool found_it = false;
3655                             for (auto pos = range.first; pos != range.second;
3656                                  ++pos) {
3657                               if (sym[sym_idx].GetMangled().GetName(
3658                                       Mangled::ePreferMangled) ==
3659                                   sym[pos->second].GetMangled().GetName(
3660                                       Mangled::ePreferMangled)) {
3661                                 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second;
3662                                 // We just need the flags from the linker
3663                                 // symbol, so put these flags
3664                                 // into the N_STSYM flags to avoid duplicate
3665                                 // symbols in the symbol table
3666                                 sym[pos->second].SetExternal(
3667                                     sym[sym_idx].IsExternal());
3668                                 sym[pos->second].SetFlags(nlist.n_type << 16 |
3669                                                           nlist.n_desc);
3670                                 sym[sym_idx].Clear();
3671                                 found_it = true;
3672                                 break;
3673                               }
3674                             }
3675                             if (found_it)
3676                               continue;
3677                           } else {
3678                             const char *gsym_name =
3679                                 sym[sym_idx]
3680                                     .GetMangled()
3681                                     .GetName(Mangled::ePreferMangled)
3682                                     .GetCString();
3683                             if (gsym_name) {
3684                               // Combine N_GSYM stab entries with the non
3685                               // stab symbol
3686                               ConstNameToSymbolIndexMap::const_iterator pos =
3687                                   N_GSYM_name_to_sym_idx.find(gsym_name);
3688                               if (pos != N_GSYM_name_to_sym_idx.end()) {
3689                                 const uint32_t GSYM_sym_idx = pos->second;
3690                                 m_nlist_idx_to_sym_idx[nlist_idx] =
3691                                     GSYM_sym_idx;
3692                                 // Copy the address, because often the N_GSYM
3693                                 // address has an invalid address of zero
3694                                 // when the global is a common symbol
3695                                 sym[GSYM_sym_idx].GetAddressRef().SetSection(
3696                                     symbol_section);
3697                                 sym[GSYM_sym_idx].GetAddressRef().SetOffset(
3698                                     symbol_value);
3699                                 add_symbol_addr(sym[GSYM_sym_idx]
3700                                                     .GetAddress()
3701                                                     .GetFileAddress());
3702                                 // We just need the flags from the linker
3703                                 // symbol, so put these flags
3704                                 // into the N_GSYM flags to avoid duplicate
3705                                 // symbols in the symbol table
3706                                 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 |
3707                                                            nlist.n_desc);
3708                                 sym[sym_idx].Clear();
3709                                 continue;
3710                               }
3711                             }
3712                           }
3713                         }
3714                       }
3715 
3716                       sym[sym_idx].SetID(nlist_idx);
3717                       sym[sym_idx].SetType(type);
3718                       if (set_value) {
3719                         sym[sym_idx].GetAddressRef().SetSection(symbol_section);
3720                         sym[sym_idx].GetAddressRef().SetOffset(symbol_value);
3721                         add_symbol_addr(
3722                             sym[sym_idx].GetAddress().GetFileAddress());
3723                       }
3724                       sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc);
3725 
3726                       if (symbol_byte_size > 0)
3727                         sym[sym_idx].SetByteSize(symbol_byte_size);
3728 
3729                       if (demangled_is_synthesized)
3730                         sym[sym_idx].SetDemangledNameIsSynthesized(true);
3731                       ++sym_idx;
3732                     } else {
3733                       sym[sym_idx].Clear();
3734                     }
3735                   }
3736                   /////////////////////////////
3737                 }
3738                 break; // No more entries to consider
3739               }
3740             }
3741 
3742             for (const auto &pos : reexport_shlib_needs_fixup) {
3743               const auto undef_pos = undefined_name_to_desc.find(pos.second);
3744               if (undef_pos != undefined_name_to_desc.end()) {
3745                 const uint8_t dylib_ordinal =
3746                     llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second);
3747                 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize())
3748                   sym[pos.first].SetReExportedSymbolSharedLibrary(
3749                       dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1));
3750               }
3751             }
3752           }
3753         }
3754       }
3755     }
3756   }
3757 
3758   // Must reset this in case it was mutated above!
3759   nlist_data_offset = 0;
3760 #endif
3761 
3762   if (nlist_data.GetByteSize() > 0) {
3763 
3764     // If the sym array was not created while parsing the DSC unmapped
3765     // symbols, create it now.
3766     if (sym == nullptr) {
3767       sym =
3768           symtab->Resize(symtab_load_command.nsyms + m_dysymtab.nindirectsyms);
3769       num_syms = symtab->GetNumSymbols();
3770     }
3771 
3772     if (unmapped_local_symbols_found) {
3773       assert(m_dysymtab.ilocalsym == 0);
3774       nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size);
3775       nlist_idx = m_dysymtab.nlocalsym;
3776     } else {
3777       nlist_idx = 0;
3778     }
3779 
3780     typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap;
3781     typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName;
3782     UndefinedNameToDescMap undefined_name_to_desc;
3783     SymbolIndexToName reexport_shlib_needs_fixup;
3784 
3785     // Symtab parsing is a huge mess. Everything is entangled and the code
3786     // requires access to a ridiculous amount of variables. LLDB depends
3787     // heavily on the proper merging of symbols and to get that right we need
3788     // to make sure we have parsed all the debug symbols first. Therefore we
3789     // invoke the lambda twice, once to parse only the debug symbols and then
3790     // once more to parse the remaining symbols.
3791     auto ParseSymbolLambda = [&](struct nlist_64 &nlist, uint32_t nlist_idx,
3792                                  bool debug_only) {
3793       const bool is_debug = ((nlist.n_type & N_STAB) != 0);
3794       if (is_debug != debug_only)
3795         return true;
3796 
3797       const char *symbol_name_non_abi_mangled = nullptr;
3798       const char *symbol_name = nullptr;
3799 
3800       if (have_strtab_data) {
3801         symbol_name = strtab_data.PeekCStr(nlist.n_strx);
3802 
3803         if (symbol_name == nullptr) {
3804           // No symbol should be NULL, even the symbols with no string values
3805           // should have an offset zero which points to an empty C-string
3806           Host::SystemLog(Host::eSystemLogError,
3807                           "error: symbol[%u] has invalid string table offset "
3808                           "0x%x in %s, ignoring symbol\n",
3809                           nlist_idx, nlist.n_strx,
3810                           module_sp->GetFileSpec().GetPath().c_str());
3811           return true;
3812         }
3813         if (symbol_name[0] == '\0')
3814           symbol_name = nullptr;
3815       } else {
3816         const addr_t str_addr = strtab_addr + nlist.n_strx;
3817         Status str_error;
3818         if (process->ReadCStringFromMemory(str_addr, memory_symbol_name,
3819                                            str_error))
3820           symbol_name = memory_symbol_name.c_str();
3821       }
3822 
3823       SymbolType type = eSymbolTypeInvalid;
3824       SectionSP symbol_section;
3825       lldb::addr_t symbol_byte_size = 0;
3826       bool add_nlist = true;
3827       bool is_gsym = false;
3828       bool demangled_is_synthesized = false;
3829       bool set_value = true;
3830 
3831       assert(sym_idx < num_syms);
3832       sym[sym_idx].SetDebug(is_debug);
3833 
3834       if (is_debug) {
3835         switch (nlist.n_type) {
3836         case N_GSYM:
3837           // global symbol: name,,NO_SECT,type,0
3838           // Sometimes the N_GSYM value contains the address.
3839 
3840           // FIXME: In the .o files, we have a GSYM and a debug symbol for all
3841           // the ObjC data.  They
3842           // have the same address, but we want to ensure that we always find
3843           // only the real symbol, 'cause we don't currently correctly
3844           // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol
3845           // type.  This is a temporary hack to make sure the ObjectiveC
3846           // symbols get treated correctly.  To do this right, we should
3847           // coalesce all the GSYM & global symbols that have the same
3848           // address.
3849           is_gsym = true;
3850           sym[sym_idx].SetExternal(true);
3851 
3852           if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') {
3853             llvm::StringRef symbol_name_ref(symbol_name);
3854             if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) {
3855               symbol_name_non_abi_mangled = symbol_name + 1;
3856               symbol_name = symbol_name + g_objc_v2_prefix_class.size();
3857               type = eSymbolTypeObjCClass;
3858               demangled_is_synthesized = true;
3859 
3860             } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) {
3861               symbol_name_non_abi_mangled = symbol_name + 1;
3862               symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size();
3863               type = eSymbolTypeObjCMetaClass;
3864               demangled_is_synthesized = true;
3865             } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) {
3866               symbol_name_non_abi_mangled = symbol_name + 1;
3867               symbol_name = symbol_name + g_objc_v2_prefix_ivar.size();
3868               type = eSymbolTypeObjCIVar;
3869               demangled_is_synthesized = true;
3870             }
3871           } else {
3872             if (nlist.n_value != 0)
3873               symbol_section =
3874                   section_info.GetSection(nlist.n_sect, nlist.n_value);
3875             type = eSymbolTypeData;
3876           }
3877           break;
3878 
3879         case N_FNAME:
3880           // procedure name (f77 kludge): name,,NO_SECT,0,0
3881           type = eSymbolTypeCompiler;
3882           break;
3883 
3884         case N_FUN:
3885           // procedure: name,,n_sect,linenumber,address
3886           if (symbol_name) {
3887             type = eSymbolTypeCode;
3888             symbol_section =
3889                 section_info.GetSection(nlist.n_sect, nlist.n_value);
3890 
3891             N_FUN_addr_to_sym_idx.insert(
3892                 std::make_pair(nlist.n_value, sym_idx));
3893             // We use the current number of symbols in the symbol table in
3894             // lieu of using nlist_idx in case we ever start trimming entries
3895             // out
3896             N_FUN_indexes.push_back(sym_idx);
3897           } else {
3898             type = eSymbolTypeCompiler;
3899 
3900             if (!N_FUN_indexes.empty()) {
3901               // Copy the size of the function into the original STAB entry
3902               // so we don't have to hunt for it later
3903               symtab->SymbolAtIndex(N_FUN_indexes.back())
3904                   ->SetByteSize(nlist.n_value);
3905               N_FUN_indexes.pop_back();
3906               // We don't really need the end function STAB as it contains
3907               // the size which we already placed with the original symbol,
3908               // so don't add it if we want a minimal symbol table
3909               add_nlist = false;
3910             }
3911           }
3912           break;
3913 
3914         case N_STSYM:
3915           // static symbol: name,,n_sect,type,address
3916           N_STSYM_addr_to_sym_idx.insert(
3917               std::make_pair(nlist.n_value, sym_idx));
3918           symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value);
3919           if (symbol_name && symbol_name[0]) {
3920             type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1,
3921                                                      eSymbolTypeData);
3922           }
3923           break;
3924 
3925         case N_LCSYM:
3926           // .lcomm symbol: name,,n_sect,type,address
3927           symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value);
3928           type = eSymbolTypeCommonBlock;
3929           break;
3930 
3931         case N_BNSYM:
3932           // We use the current number of symbols in the symbol table in lieu
3933           // of using nlist_idx in case we ever start trimming entries out
3934           // Skip these if we want minimal symbol tables
3935           add_nlist = false;
3936           break;
3937 
3938         case N_ENSYM:
3939           // Set the size of the N_BNSYM to the terminating index of this
3940           // N_ENSYM so that we can always skip the entire symbol if we need
3941           // to navigate more quickly at the source level when parsing STABS
3942           // Skip these if we want minimal symbol tables
3943           add_nlist = false;
3944           break;
3945 
3946         case N_OPT:
3947           // emitted with gcc2_compiled and in gcc source
3948           type = eSymbolTypeCompiler;
3949           break;
3950 
3951         case N_RSYM:
3952           // register sym: name,,NO_SECT,type,register
3953           type = eSymbolTypeVariable;
3954           break;
3955 
3956         case N_SLINE:
3957           // src line: 0,,n_sect,linenumber,address
3958           symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value);
3959           type = eSymbolTypeLineEntry;
3960           break;
3961 
3962         case N_SSYM:
3963           // structure elt: name,,NO_SECT,type,struct_offset
3964           type = eSymbolTypeVariableType;
3965           break;
3966 
3967         case N_SO:
3968           // source file name
3969           type = eSymbolTypeSourceFile;
3970           if (symbol_name == nullptr) {
3971             add_nlist = false;
3972             if (N_SO_index != UINT32_MAX) {
3973               // Set the size of the N_SO to the terminating index of this
3974               // N_SO so that we can always skip the entire N_SO if we need
3975               // to navigate more quickly at the source level when parsing
3976               // STABS
3977               symbol_ptr = symtab->SymbolAtIndex(N_SO_index);
3978               symbol_ptr->SetByteSize(sym_idx);
3979               symbol_ptr->SetSizeIsSibling(true);
3980             }
3981             N_NSYM_indexes.clear();
3982             N_INCL_indexes.clear();
3983             N_BRAC_indexes.clear();
3984             N_COMM_indexes.clear();
3985             N_FUN_indexes.clear();
3986             N_SO_index = UINT32_MAX;
3987           } else {
3988             // We use the current number of symbols in the symbol table in
3989             // lieu of using nlist_idx in case we ever start trimming entries
3990             // out
3991             const bool N_SO_has_full_path = symbol_name[0] == '/';
3992             if (N_SO_has_full_path) {
3993               if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) {
3994                 // We have two consecutive N_SO entries where the first
3995                 // contains a directory and the second contains a full path.
3996                 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name),
3997                                                        false);
3998                 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1;
3999                 add_nlist = false;
4000               } else {
4001                 // This is the first entry in a N_SO that contains a
4002                 // directory or a full path to the source file
4003                 N_SO_index = sym_idx;
4004               }
4005             } else if ((N_SO_index == sym_idx - 1) &&
4006                        ((sym_idx - 1) < num_syms)) {
4007               // This is usually the second N_SO entry that contains just the
4008               // filename, so here we combine it with the first one if we are
4009               // minimizing the symbol table
4010               const char *so_path =
4011                   sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString();
4012               if (so_path && so_path[0]) {
4013                 std::string full_so_path(so_path);
4014                 const size_t double_slash_pos = full_so_path.find("//");
4015                 if (double_slash_pos != std::string::npos) {
4016                   // The linker has been generating bad N_SO entries with
4017                   // doubled up paths in the format "%s%s" where the first
4018                   // string in the DW_AT_comp_dir, and the second is the
4019                   // directory for the source file so you end up with a path
4020                   // that looks like "/tmp/src//tmp/src/"
4021                   FileSpec so_dir(so_path);
4022                   if (!FileSystem::Instance().Exists(so_dir)) {
4023                     so_dir.SetFile(&full_so_path[double_slash_pos + 1],
4024                                    FileSpec::Style::native);
4025                     if (FileSystem::Instance().Exists(so_dir)) {
4026                       // Trim off the incorrect path
4027                       full_so_path.erase(0, double_slash_pos + 1);
4028                     }
4029                   }
4030                 }
4031                 if (*full_so_path.rbegin() != '/')
4032                   full_so_path += '/';
4033                 full_so_path += symbol_name;
4034                 sym[sym_idx - 1].GetMangled().SetValue(
4035                     ConstString(full_so_path.c_str()), false);
4036                 add_nlist = false;
4037                 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1;
4038               }
4039             } else {
4040               // This could be a relative path to a N_SO
4041               N_SO_index = sym_idx;
4042             }
4043           }
4044           break;
4045 
4046         case N_OSO:
4047           // object file name: name,,0,0,st_mtime
4048           type = eSymbolTypeObjectFile;
4049           break;
4050 
4051         case N_LSYM:
4052           // local sym: name,,NO_SECT,type,offset
4053           type = eSymbolTypeLocal;
4054           break;
4055 
4056         // INCL scopes
4057         case N_BINCL:
4058           // include file beginning: name,,NO_SECT,0,sum We use the current
4059           // number of symbols in the symbol table in lieu of using nlist_idx
4060           // in case we ever start trimming entries out
4061           N_INCL_indexes.push_back(sym_idx);
4062           type = eSymbolTypeScopeBegin;
4063           break;
4064 
4065         case N_EINCL:
4066           // include file end: name,,NO_SECT,0,0
4067           // Set the size of the N_BINCL to the terminating index of this
4068           // N_EINCL so that we can always skip the entire symbol if we need
4069           // to navigate more quickly at the source level when parsing STABS
4070           if (!N_INCL_indexes.empty()) {
4071             symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back());
4072             symbol_ptr->SetByteSize(sym_idx + 1);
4073             symbol_ptr->SetSizeIsSibling(true);
4074             N_INCL_indexes.pop_back();
4075           }
4076           type = eSymbolTypeScopeEnd;
4077           break;
4078 
4079         case N_SOL:
4080           // #included file name: name,,n_sect,0,address
4081           type = eSymbolTypeHeaderFile;
4082 
4083           // We currently don't use the header files on darwin
4084           add_nlist = false;
4085           break;
4086 
4087         case N_PARAMS:
4088           // compiler parameters: name,,NO_SECT,0,0
4089           type = eSymbolTypeCompiler;
4090           break;
4091 
4092         case N_VERSION:
4093           // compiler version: name,,NO_SECT,0,0
4094           type = eSymbolTypeCompiler;
4095           break;
4096 
4097         case N_OLEVEL:
4098           // compiler -O level: name,,NO_SECT,0,0
4099           type = eSymbolTypeCompiler;
4100           break;
4101 
4102         case N_PSYM:
4103           // parameter: name,,NO_SECT,type,offset
4104           type = eSymbolTypeVariable;
4105           break;
4106 
4107         case N_ENTRY:
4108           // alternate entry: name,,n_sect,linenumber,address
4109           symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value);
4110           type = eSymbolTypeLineEntry;
4111           break;
4112 
4113         // Left and Right Braces
4114         case N_LBRAC:
4115           // left bracket: 0,,NO_SECT,nesting level,address We use the
4116           // current number of symbols in the symbol table in lieu of using
4117           // nlist_idx in case we ever start trimming entries out
4118           symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value);
4119           N_BRAC_indexes.push_back(sym_idx);
4120           type = eSymbolTypeScopeBegin;
4121           break;
4122 
4123         case N_RBRAC:
4124           // right bracket: 0,,NO_SECT,nesting level,address Set the size of
4125           // the N_LBRAC to the terminating index of this N_RBRAC so that we
4126           // can always skip the entire symbol if we need to navigate more
4127           // quickly at the source level when parsing STABS
4128           symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value);
4129           if (!N_BRAC_indexes.empty()) {
4130             symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back());
4131             symbol_ptr->SetByteSize(sym_idx + 1);
4132             symbol_ptr->SetSizeIsSibling(true);
4133             N_BRAC_indexes.pop_back();
4134           }
4135           type = eSymbolTypeScopeEnd;
4136           break;
4137 
4138         case N_EXCL:
4139           // deleted include file: name,,NO_SECT,0,sum
4140           type = eSymbolTypeHeaderFile;
4141           break;
4142 
4143         // COMM scopes
4144         case N_BCOMM:
4145           // begin common: name,,NO_SECT,0,0
4146           // We use the current number of symbols in the symbol table in lieu
4147           // of using nlist_idx in case we ever start trimming entries out
4148           type = eSymbolTypeScopeBegin;
4149           N_COMM_indexes.push_back(sym_idx);
4150           break;
4151 
4152         case N_ECOML:
4153           // end common (local name): 0,,n_sect,0,address
4154           symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value);
4155           LLVM_FALLTHROUGH;
4156 
4157         case N_ECOMM:
4158           // end common: name,,n_sect,0,0
4159           // Set the size of the N_BCOMM to the terminating index of this
4160           // N_ECOMM/N_ECOML so that we can always skip the entire symbol if
4161           // we need to navigate more quickly at the source level when
4162           // parsing STABS
4163           if (!N_COMM_indexes.empty()) {
4164             symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back());
4165             symbol_ptr->SetByteSize(sym_idx + 1);
4166             symbol_ptr->SetSizeIsSibling(true);
4167             N_COMM_indexes.pop_back();
4168           }
4169           type = eSymbolTypeScopeEnd;
4170           break;
4171 
4172         case N_LENG:
4173           // second stab entry with length information
4174           type = eSymbolTypeAdditional;
4175           break;
4176 
4177         default:
4178           break;
4179         }
4180       } else {
4181         uint8_t n_type = N_TYPE & nlist.n_type;
4182         sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0);
4183 
4184         switch (n_type) {
4185         case N_INDR: {
4186           const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value);
4187           if (reexport_name_cstr && reexport_name_cstr[0]) {
4188             type = eSymbolTypeReExported;
4189             ConstString reexport_name(reexport_name_cstr +
4190                                       ((reexport_name_cstr[0] == '_') ? 1 : 0));
4191             sym[sym_idx].SetReExportedSymbolName(reexport_name);
4192             set_value = false;
4193             reexport_shlib_needs_fixup[sym_idx] = reexport_name;
4194             indirect_symbol_names.insert(
4195                 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0)));
4196           } else
4197             type = eSymbolTypeUndefined;
4198         } break;
4199 
4200         case N_UNDF:
4201           if (symbol_name && symbol_name[0]) {
4202             ConstString undefined_name(symbol_name +
4203                                        ((symbol_name[0] == '_') ? 1 : 0));
4204             undefined_name_to_desc[undefined_name] = nlist.n_desc;
4205           }
4206           LLVM_FALLTHROUGH;
4207 
4208         case N_PBUD:
4209           type = eSymbolTypeUndefined;
4210           break;
4211 
4212         case N_ABS:
4213           type = eSymbolTypeAbsolute;
4214           break;
4215 
4216         case N_SECT: {
4217           symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value);
4218 
4219           if (!symbol_section) {
4220             // TODO: warn about this?
4221             add_nlist = false;
4222             break;
4223           }
4224 
4225           if (TEXT_eh_frame_sectID == nlist.n_sect) {
4226             type = eSymbolTypeException;
4227           } else {
4228             uint32_t section_type = symbol_section->Get() & SECTION_TYPE;
4229 
4230             switch (section_type) {
4231             case S_CSTRING_LITERALS:
4232               type = eSymbolTypeData;
4233               break; // section with only literal C strings
4234             case S_4BYTE_LITERALS:
4235               type = eSymbolTypeData;
4236               break; // section with only 4 byte literals
4237             case S_8BYTE_LITERALS:
4238               type = eSymbolTypeData;
4239               break; // section with only 8 byte literals
4240             case S_LITERAL_POINTERS:
4241               type = eSymbolTypeTrampoline;
4242               break; // section with only pointers to literals
4243             case S_NON_LAZY_SYMBOL_POINTERS:
4244               type = eSymbolTypeTrampoline;
4245               break; // section with only non-lazy symbol pointers
4246             case S_LAZY_SYMBOL_POINTERS:
4247               type = eSymbolTypeTrampoline;
4248               break; // section with only lazy symbol pointers
4249             case S_SYMBOL_STUBS:
4250               type = eSymbolTypeTrampoline;
4251               break; // section with only symbol stubs, byte size of stub in
4252                      // the reserved2 field
4253             case S_MOD_INIT_FUNC_POINTERS:
4254               type = eSymbolTypeCode;
4255               break; // section with only function pointers for initialization
4256             case S_MOD_TERM_FUNC_POINTERS:
4257               type = eSymbolTypeCode;
4258               break; // section with only function pointers for termination
4259             case S_INTERPOSING:
4260               type = eSymbolTypeTrampoline;
4261               break; // section with only pairs of function pointers for
4262                      // interposing
4263             case S_16BYTE_LITERALS:
4264               type = eSymbolTypeData;
4265               break; // section with only 16 byte literals
4266             case S_DTRACE_DOF:
4267               type = eSymbolTypeInstrumentation;
4268               break;
4269             case S_LAZY_DYLIB_SYMBOL_POINTERS:
4270               type = eSymbolTypeTrampoline;
4271               break;
4272             default:
4273               switch (symbol_section->GetType()) {
4274               case lldb::eSectionTypeCode:
4275                 type = eSymbolTypeCode;
4276                 break;
4277               case eSectionTypeData:
4278               case eSectionTypeDataCString:         // Inlined C string data
4279               case eSectionTypeDataCStringPointers: // Pointers to C string
4280                                                     // data
4281               case eSectionTypeDataSymbolAddress:   // Address of a symbol in
4282                                                     // the symbol table
4283               case eSectionTypeData4:
4284               case eSectionTypeData8:
4285               case eSectionTypeData16:
4286                 type = eSymbolTypeData;
4287                 break;
4288               default:
4289                 break;
4290               }
4291               break;
4292             }
4293 
4294             if (type == eSymbolTypeInvalid) {
4295               const char *symbol_sect_name =
4296                   symbol_section->GetName().AsCString();
4297               if (symbol_section->IsDescendant(text_section_sp.get())) {
4298                 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS |
4299                                             S_ATTR_SELF_MODIFYING_CODE |
4300                                             S_ATTR_SOME_INSTRUCTIONS))
4301                   type = eSymbolTypeData;
4302                 else
4303                   type = eSymbolTypeCode;
4304               } else if (symbol_section->IsDescendant(data_section_sp.get()) ||
4305                          symbol_section->IsDescendant(
4306                              data_dirty_section_sp.get()) ||
4307                          symbol_section->IsDescendant(
4308                              data_const_section_sp.get())) {
4309                 if (symbol_sect_name &&
4310                     ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) {
4311                   type = eSymbolTypeRuntime;
4312 
4313                   if (symbol_name) {
4314                     llvm::StringRef symbol_name_ref(symbol_name);
4315                     if (symbol_name_ref.startswith("_OBJC_")) {
4316                       llvm::StringRef g_objc_v2_prefix_class(
4317                           "_OBJC_CLASS_$_");
4318                       llvm::StringRef g_objc_v2_prefix_metaclass(
4319                           "_OBJC_METACLASS_$_");
4320                       llvm::StringRef g_objc_v2_prefix_ivar(
4321                           "_OBJC_IVAR_$_");
4322                       if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) {
4323                         symbol_name_non_abi_mangled = symbol_name + 1;
4324                         symbol_name =
4325                             symbol_name + g_objc_v2_prefix_class.size();
4326                         type = eSymbolTypeObjCClass;
4327                         demangled_is_synthesized = true;
4328                       } else if (symbol_name_ref.startswith(
4329                                      g_objc_v2_prefix_metaclass)) {
4330                         symbol_name_non_abi_mangled = symbol_name + 1;
4331                         symbol_name =
4332                             symbol_name + g_objc_v2_prefix_metaclass.size();
4333                         type = eSymbolTypeObjCMetaClass;
4334                         demangled_is_synthesized = true;
4335                       } else if (symbol_name_ref.startswith(
4336                                      g_objc_v2_prefix_ivar)) {
4337                         symbol_name_non_abi_mangled = symbol_name + 1;
4338                         symbol_name =
4339                             symbol_name + g_objc_v2_prefix_ivar.size();
4340                         type = eSymbolTypeObjCIVar;
4341                         demangled_is_synthesized = true;
4342                       }
4343                     }
4344                   }
4345                 } else if (symbol_sect_name &&
4346                            ::strstr(symbol_sect_name, "__gcc_except_tab") ==
4347                                symbol_sect_name) {
4348                   type = eSymbolTypeException;
4349                 } else {
4350                   type = eSymbolTypeData;
4351                 }
4352               } else if (symbol_sect_name &&
4353                          ::strstr(symbol_sect_name, "__IMPORT") ==
4354                              symbol_sect_name) {
4355                 type = eSymbolTypeTrampoline;
4356               } else if (symbol_section->IsDescendant(objc_section_sp.get())) {
4357                 type = eSymbolTypeRuntime;
4358                 if (symbol_name && symbol_name[0] == '.') {
4359                   llvm::StringRef symbol_name_ref(symbol_name);
4360                   llvm::StringRef g_objc_v1_prefix_class(
4361                       ".objc_class_name_");
4362                   if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) {
4363                     symbol_name_non_abi_mangled = symbol_name;
4364                     symbol_name = symbol_name + g_objc_v1_prefix_class.size();
4365                     type = eSymbolTypeObjCClass;
4366                     demangled_is_synthesized = true;
4367                   }
4368                 }
4369               }
4370             }
4371           }
4372         } break;
4373         }
4374       }
4375 
4376       if (!add_nlist) {
4377         sym[sym_idx].Clear();
4378         return true;
4379       }
4380 
4381       uint64_t symbol_value = nlist.n_value;
4382 
4383       if (symbol_name_non_abi_mangled) {
4384         sym[sym_idx].GetMangled().SetMangledName(
4385             ConstString(symbol_name_non_abi_mangled));
4386         sym[sym_idx].GetMangled().SetDemangledName(ConstString(symbol_name));
4387       } else {
4388         bool symbol_name_is_mangled = false;
4389 
4390         if (symbol_name && symbol_name[0] == '_') {
4391           symbol_name_is_mangled = symbol_name[1] == '_';
4392           symbol_name++; // Skip the leading underscore
4393         }
4394 
4395         if (symbol_name) {
4396           ConstString const_symbol_name(symbol_name);
4397           sym[sym_idx].GetMangled().SetValue(const_symbol_name,
4398                                              symbol_name_is_mangled);
4399         }
4400       }
4401 
4402       if (is_gsym) {
4403         const char *gsym_name = sym[sym_idx]
4404                                     .GetMangled()
4405                                     .GetName(Mangled::ePreferMangled)
4406                                     .GetCString();
4407         if (gsym_name)
4408           N_GSYM_name_to_sym_idx[gsym_name] = sym_idx;
4409       }
4410 
4411       if (symbol_section) {
4412         const addr_t section_file_addr = symbol_section->GetFileAddress();
4413         if (symbol_byte_size == 0 && function_starts_count > 0) {
4414           addr_t symbol_lookup_file_addr = nlist.n_value;
4415           // Do an exact address match for non-ARM addresses, else get the
4416           // closest since the symbol might be a thumb symbol which has an
4417           // address with bit zero set.
4418           FunctionStarts::Entry *func_start_entry =
4419               function_starts.FindEntry(symbol_lookup_file_addr, !is_arm);
4420           if (is_arm && func_start_entry) {
4421             // Verify that the function start address is the symbol address
4422             // (ARM) or the symbol address + 1 (thumb).
4423             if (func_start_entry->addr != symbol_lookup_file_addr &&
4424                 func_start_entry->addr != (symbol_lookup_file_addr + 1)) {
4425               // Not the right entry, NULL it out...
4426               func_start_entry = nullptr;
4427             }
4428           }
4429           if (func_start_entry) {
4430             func_start_entry->data = true;
4431 
4432             addr_t symbol_file_addr = func_start_entry->addr;
4433             if (is_arm)
4434               symbol_file_addr &= THUMB_ADDRESS_BIT_MASK;
4435 
4436             const FunctionStarts::Entry *next_func_start_entry =
4437                 function_starts.FindNextEntry(func_start_entry);
4438             const addr_t section_end_file_addr =
4439                 section_file_addr + symbol_section->GetByteSize();
4440             if (next_func_start_entry) {
4441               addr_t next_symbol_file_addr = next_func_start_entry->addr;
4442               // Be sure the clear the Thumb address bit when we calculate the
4443               // size from the current and next address
4444               if (is_arm)
4445                 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK;
4446               symbol_byte_size = std::min<lldb::addr_t>(
4447                   next_symbol_file_addr - symbol_file_addr,
4448                   section_end_file_addr - symbol_file_addr);
4449             } else {
4450               symbol_byte_size = section_end_file_addr - symbol_file_addr;
4451             }
4452           }
4453         }
4454         symbol_value -= section_file_addr;
4455       }
4456 
4457       if (!is_debug) {
4458         if (type == eSymbolTypeCode) {
4459           // See if we can find a N_FUN entry for any code symbols. If we do
4460           // find a match, and the name matches, then we can merge the two into
4461           // just the function symbol to avoid duplicate entries in the symbol
4462           // table.
4463           std::pair<ValueToSymbolIndexMap::const_iterator,
4464                     ValueToSymbolIndexMap::const_iterator>
4465               range;
4466           range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value);
4467           if (range.first != range.second) {
4468             for (ValueToSymbolIndexMap::const_iterator pos = range.first;
4469                  pos != range.second; ++pos) {
4470               if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) ==
4471                   sym[pos->second].GetMangled().GetName(
4472                       Mangled::ePreferMangled)) {
4473                 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second;
4474                 // We just need the flags from the linker symbol, so put these
4475                 // flags into the N_FUN flags to avoid duplicate symbols in the
4476                 // symbol table.
4477                 sym[pos->second].SetExternal(sym[sym_idx].IsExternal());
4478                 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc);
4479                 if (resolver_addresses.find(nlist.n_value) !=
4480                     resolver_addresses.end())
4481                   sym[pos->second].SetType(eSymbolTypeResolver);
4482                 sym[sym_idx].Clear();
4483                 return true;
4484               }
4485             }
4486           } else {
4487             if (resolver_addresses.find(nlist.n_value) !=
4488                 resolver_addresses.end())
4489               type = eSymbolTypeResolver;
4490           }
4491         } else if (type == eSymbolTypeData || type == eSymbolTypeObjCClass ||
4492                    type == eSymbolTypeObjCMetaClass ||
4493                    type == eSymbolTypeObjCIVar) {
4494           // See if we can find a N_STSYM entry for any data symbols. If we do
4495           // find a match, and the name matches, then we can merge the two into
4496           // just the Static symbol to avoid duplicate entries in the symbol
4497           // table.
4498           std::pair<ValueToSymbolIndexMap::const_iterator,
4499                     ValueToSymbolIndexMap::const_iterator>
4500               range;
4501           range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value);
4502           if (range.first != range.second) {
4503             for (ValueToSymbolIndexMap::const_iterator pos = range.first;
4504                  pos != range.second; ++pos) {
4505               if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) ==
4506                   sym[pos->second].GetMangled().GetName(
4507                       Mangled::ePreferMangled)) {
4508                 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second;
4509                 // We just need the flags from the linker symbol, so put these
4510                 // flags into the N_STSYM flags to avoid duplicate symbols in
4511                 // the symbol table.
4512                 sym[pos->second].SetExternal(sym[sym_idx].IsExternal());
4513                 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc);
4514                 sym[sym_idx].Clear();
4515                 return true;
4516               }
4517             }
4518           } else {
4519             // Combine N_GSYM stab entries with the non stab symbol.
4520             const char *gsym_name = sym[sym_idx]
4521                                         .GetMangled()
4522                                         .GetName(Mangled::ePreferMangled)
4523                                         .GetCString();
4524             if (gsym_name) {
4525               ConstNameToSymbolIndexMap::const_iterator pos =
4526                   N_GSYM_name_to_sym_idx.find(gsym_name);
4527               if (pos != N_GSYM_name_to_sym_idx.end()) {
4528                 const uint32_t GSYM_sym_idx = pos->second;
4529                 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx;
4530                 // Copy the address, because often the N_GSYM address has an
4531                 // invalid address of zero when the global is a common symbol.
4532                 sym[GSYM_sym_idx].GetAddressRef().SetSection(symbol_section);
4533                 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value);
4534                 add_symbol_addr(
4535                     sym[GSYM_sym_idx].GetAddress().GetFileAddress());
4536                 // We just need the flags from the linker symbol, so put these
4537                 // flags into the N_GSYM flags to avoid duplicate symbols in
4538                 // the symbol table.
4539                 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc);
4540                 sym[sym_idx].Clear();
4541                 return true;
4542               }
4543             }
4544           }
4545         }
4546       }
4547 
4548       sym[sym_idx].SetID(nlist_idx);
4549       sym[sym_idx].SetType(type);
4550       if (set_value) {
4551         sym[sym_idx].GetAddressRef().SetSection(symbol_section);
4552         sym[sym_idx].GetAddressRef().SetOffset(symbol_value);
4553         if (symbol_section)
4554           add_symbol_addr(sym[sym_idx].GetAddress().GetFileAddress());
4555       }
4556       sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc);
4557       if (nlist.n_desc & N_WEAK_REF)
4558         sym[sym_idx].SetIsWeak(true);
4559 
4560       if (symbol_byte_size > 0)
4561         sym[sym_idx].SetByteSize(symbol_byte_size);
4562 
4563       if (demangled_is_synthesized)
4564         sym[sym_idx].SetDemangledNameIsSynthesized(true);
4565 
4566       ++sym_idx;
4567       return true;
4568     };
4569 
4570     // First parse all the nlists but don't process them yet. See the next
4571     // comment for an explanation why.
4572     std::vector<struct nlist_64> nlists;
4573     nlists.reserve(symtab_load_command.nsyms);
4574     for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) {
4575       if (auto nlist =
4576               ParseNList(nlist_data, nlist_data_offset, nlist_byte_size))
4577         nlists.push_back(*nlist);
4578       else
4579         break;
4580     }
4581 
4582     // Now parse all the debug symbols. This is needed to merge non-debug
4583     // symbols in the next step. Non-debug symbols are always coalesced into
4584     // the debug symbol. Doing this in one step would mean that some symbols
4585     // won't be merged.
4586     nlist_idx = 0;
4587     for (auto &nlist : nlists) {
4588       if (!ParseSymbolLambda(nlist, nlist_idx++, DebugSymbols))
4589         break;
4590     }
4591 
4592     // Finally parse all the non debug symbols.
4593     nlist_idx = 0;
4594     for (auto &nlist : nlists) {
4595       if (!ParseSymbolLambda(nlist, nlist_idx++, NonDebugSymbols))
4596         break;
4597     }
4598 
4599     for (const auto &pos : reexport_shlib_needs_fixup) {
4600       const auto undef_pos = undefined_name_to_desc.find(pos.second);
4601       if (undef_pos != undefined_name_to_desc.end()) {
4602         const uint8_t dylib_ordinal =
4603             llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second);
4604         if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize())
4605           sym[pos.first].SetReExportedSymbolSharedLibrary(
4606               dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1));
4607       }
4608     }
4609   }
4610 
4611   // Count how many trie symbols we'll add to the symbol table
4612   int trie_symbol_table_augment_count = 0;
4613   for (auto &e : external_sym_trie_entries) {
4614     if (symbols_added.find(e.entry.address) == symbols_added.end())
4615       trie_symbol_table_augment_count++;
4616   }
4617 
4618   if (num_syms < sym_idx + trie_symbol_table_augment_count) {
4619     num_syms = sym_idx + trie_symbol_table_augment_count;
4620     sym = symtab->Resize(num_syms);
4621   }
4622   uint32_t synthetic_sym_id = symtab_load_command.nsyms;
4623 
4624   // Add symbols from the trie to the symbol table.
4625   for (auto &e : external_sym_trie_entries) {
4626     if (symbols_added.find(e.entry.address) != symbols_added.end())
4627       continue;
4628 
4629     // Find the section that this trie address is in, use that to annotate
4630     // symbol type as we add the trie address and name to the symbol table.
4631     Address symbol_addr;
4632     if (module_sp->ResolveFileAddress(e.entry.address, symbol_addr)) {
4633       SectionSP symbol_section(symbol_addr.GetSection());
4634       const char *symbol_name = e.entry.name.GetCString();
4635       bool demangled_is_synthesized = false;
4636       SymbolType type =
4637           GetSymbolType(symbol_name, demangled_is_synthesized, text_section_sp,
4638                         data_section_sp, data_dirty_section_sp,
4639                         data_const_section_sp, symbol_section);
4640 
4641       sym[sym_idx].SetType(type);
4642       if (symbol_section) {
4643         sym[sym_idx].SetID(synthetic_sym_id++);
4644         sym[sym_idx].GetMangled().SetMangledName(ConstString(symbol_name));
4645         if (demangled_is_synthesized)
4646           sym[sym_idx].SetDemangledNameIsSynthesized(true);
4647         sym[sym_idx].SetIsSynthetic(true);
4648         sym[sym_idx].SetExternal(true);
4649         sym[sym_idx].GetAddressRef() = symbol_addr;
4650         add_symbol_addr(symbol_addr.GetFileAddress());
4651         if (e.entry.flags & TRIE_SYMBOL_IS_THUMB)
4652           sym[sym_idx].SetFlags(MACHO_NLIST_ARM_SYMBOL_IS_THUMB);
4653         ++sym_idx;
4654       }
4655     }
4656   }
4657 
4658   if (function_starts_count > 0) {
4659     uint32_t num_synthetic_function_symbols = 0;
4660     for (i = 0; i < function_starts_count; ++i) {
4661       if (symbols_added.find(function_starts.GetEntryRef(i).addr) ==
4662           symbols_added.end())
4663         ++num_synthetic_function_symbols;
4664     }
4665 
4666     if (num_synthetic_function_symbols > 0) {
4667       if (num_syms < sym_idx + num_synthetic_function_symbols) {
4668         num_syms = sym_idx + num_synthetic_function_symbols;
4669         sym = symtab->Resize(num_syms);
4670       }
4671       for (i = 0; i < function_starts_count; ++i) {
4672         const FunctionStarts::Entry *func_start_entry =
4673             function_starts.GetEntryAtIndex(i);
4674         if (symbols_added.find(func_start_entry->addr) == symbols_added.end()) {
4675           addr_t symbol_file_addr = func_start_entry->addr;
4676           uint32_t symbol_flags = 0;
4677           if (func_start_entry->data)
4678             symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB;
4679           Address symbol_addr;
4680           if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) {
4681             SectionSP symbol_section(symbol_addr.GetSection());
4682             uint32_t symbol_byte_size = 0;
4683             if (symbol_section) {
4684               const addr_t section_file_addr = symbol_section->GetFileAddress();
4685               const FunctionStarts::Entry *next_func_start_entry =
4686                   function_starts.FindNextEntry(func_start_entry);
4687               const addr_t section_end_file_addr =
4688                   section_file_addr + symbol_section->GetByteSize();
4689               if (next_func_start_entry) {
4690                 addr_t next_symbol_file_addr = next_func_start_entry->addr;
4691                 if (is_arm)
4692                   next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK;
4693                 symbol_byte_size = std::min<lldb::addr_t>(
4694                     next_symbol_file_addr - symbol_file_addr,
4695                     section_end_file_addr - symbol_file_addr);
4696               } else {
4697                 symbol_byte_size = section_end_file_addr - symbol_file_addr;
4698               }
4699               sym[sym_idx].SetID(synthetic_sym_id++);
4700               // Don't set the name for any synthetic symbols, the Symbol
4701               // object will generate one if needed when the name is accessed
4702               // via accessors.
4703               sym[sym_idx].GetMangled().SetDemangledName(ConstString());
4704               sym[sym_idx].SetType(eSymbolTypeCode);
4705               sym[sym_idx].SetIsSynthetic(true);
4706               sym[sym_idx].GetAddressRef() = symbol_addr;
4707               add_symbol_addr(symbol_addr.GetFileAddress());
4708               if (symbol_flags)
4709                 sym[sym_idx].SetFlags(symbol_flags);
4710               if (symbol_byte_size)
4711                 sym[sym_idx].SetByteSize(symbol_byte_size);
4712               ++sym_idx;
4713             }
4714           }
4715         }
4716       }
4717     }
4718   }
4719 
4720   // Trim our symbols down to just what we ended up with after removing any
4721   // symbols.
4722   if (sym_idx < num_syms) {
4723     num_syms = sym_idx;
4724     sym = symtab->Resize(num_syms);
4725   }
4726 
4727   // Now synthesize indirect symbols
4728   if (m_dysymtab.nindirectsyms != 0) {
4729     if (indirect_symbol_index_data.GetByteSize()) {
4730       NListIndexToSymbolIndexMap::const_iterator end_index_pos =
4731           m_nlist_idx_to_sym_idx.end();
4732 
4733       for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size();
4734            ++sect_idx) {
4735         if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) ==
4736             S_SYMBOL_STUBS) {
4737           uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2;
4738           if (symbol_stub_byte_size == 0)
4739             continue;
4740 
4741           const uint32_t num_symbol_stubs =
4742               m_mach_sections[sect_idx].size / symbol_stub_byte_size;
4743 
4744           if (num_symbol_stubs == 0)
4745             continue;
4746 
4747           const uint32_t symbol_stub_index_offset =
4748               m_mach_sections[sect_idx].reserved1;
4749           for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) {
4750             const uint32_t symbol_stub_index =
4751                 symbol_stub_index_offset + stub_idx;
4752             const lldb::addr_t symbol_stub_addr =
4753                 m_mach_sections[sect_idx].addr +
4754                 (stub_idx * symbol_stub_byte_size);
4755             lldb::offset_t symbol_stub_offset = symbol_stub_index * 4;
4756             if (indirect_symbol_index_data.ValidOffsetForDataOfSize(
4757                     symbol_stub_offset, 4)) {
4758               const uint32_t stub_sym_id =
4759                   indirect_symbol_index_data.GetU32(&symbol_stub_offset);
4760               if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL))
4761                 continue;
4762 
4763               NListIndexToSymbolIndexMap::const_iterator index_pos =
4764                   m_nlist_idx_to_sym_idx.find(stub_sym_id);
4765               Symbol *stub_symbol = nullptr;
4766               if (index_pos != end_index_pos) {
4767                 // We have a remapping from the original nlist index to a
4768                 // current symbol index, so just look this up by index
4769                 stub_symbol = symtab->SymbolAtIndex(index_pos->second);
4770               } else {
4771                 // We need to lookup a symbol using the original nlist symbol
4772                 // index since this index is coming from the S_SYMBOL_STUBS
4773                 stub_symbol = symtab->FindSymbolByID(stub_sym_id);
4774               }
4775 
4776               if (stub_symbol) {
4777                 Address so_addr(symbol_stub_addr, section_list);
4778 
4779                 if (stub_symbol->GetType() == eSymbolTypeUndefined) {
4780                   // Change the external symbol into a trampoline that makes
4781                   // sense These symbols were N_UNDF N_EXT, and are useless
4782                   // to us, so we can re-use them so we don't have to make up
4783                   // a synthetic symbol for no good reason.
4784                   if (resolver_addresses.find(symbol_stub_addr) ==
4785                       resolver_addresses.end())
4786                     stub_symbol->SetType(eSymbolTypeTrampoline);
4787                   else
4788                     stub_symbol->SetType(eSymbolTypeResolver);
4789                   stub_symbol->SetExternal(false);
4790                   stub_symbol->GetAddressRef() = so_addr;
4791                   stub_symbol->SetByteSize(symbol_stub_byte_size);
4792                 } else {
4793                   // Make a synthetic symbol to describe the trampoline stub
4794                   Mangled stub_symbol_mangled_name(stub_symbol->GetMangled());
4795                   if (sym_idx >= num_syms) {
4796                     sym = symtab->Resize(++num_syms);
4797                     stub_symbol = nullptr; // this pointer no longer valid
4798                   }
4799                   sym[sym_idx].SetID(synthetic_sym_id++);
4800                   sym[sym_idx].GetMangled() = stub_symbol_mangled_name;
4801                   if (resolver_addresses.find(symbol_stub_addr) ==
4802                       resolver_addresses.end())
4803                     sym[sym_idx].SetType(eSymbolTypeTrampoline);
4804                   else
4805                     sym[sym_idx].SetType(eSymbolTypeResolver);
4806                   sym[sym_idx].SetIsSynthetic(true);
4807                   sym[sym_idx].GetAddressRef() = so_addr;
4808                   add_symbol_addr(so_addr.GetFileAddress());
4809                   sym[sym_idx].SetByteSize(symbol_stub_byte_size);
4810                   ++sym_idx;
4811                 }
4812               } else {
4813                 if (log)
4814                   log->Warning("symbol stub referencing symbol table symbol "
4815                                "%u that isn't in our minimal symbol table, "
4816                                "fix this!!!",
4817                                stub_sym_id);
4818               }
4819             }
4820           }
4821         }
4822       }
4823     }
4824   }
4825 
4826   if (!reexport_trie_entries.empty()) {
4827     for (const auto &e : reexport_trie_entries) {
4828       if (e.entry.import_name) {
4829         // Only add indirect symbols from the Trie entries if we didn't have
4830         // a N_INDR nlist entry for this already
4831         if (indirect_symbol_names.find(e.entry.name) ==
4832             indirect_symbol_names.end()) {
4833           // Make a synthetic symbol to describe re-exported symbol.
4834           if (sym_idx >= num_syms)
4835             sym = symtab->Resize(++num_syms);
4836           sym[sym_idx].SetID(synthetic_sym_id++);
4837           sym[sym_idx].GetMangled() = Mangled(e.entry.name);
4838           sym[sym_idx].SetType(eSymbolTypeReExported);
4839           sym[sym_idx].SetIsSynthetic(true);
4840           sym[sym_idx].SetReExportedSymbolName(e.entry.import_name);
4841           if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) {
4842             sym[sym_idx].SetReExportedSymbolSharedLibrary(
4843                 dylib_files.GetFileSpecAtIndex(e.entry.other - 1));
4844           }
4845           ++sym_idx;
4846         }
4847       }
4848     }
4849   }
4850 
4851   //        StreamFile s(stdout, false);
4852   //        s.Printf ("Symbol table before CalculateSymbolSizes():\n");
4853   //        symtab->Dump(&s, NULL, eSortOrderNone);
4854   // Set symbol byte sizes correctly since mach-o nlist entries don't have
4855   // sizes
4856   symtab->CalculateSymbolSizes();
4857 
4858   //        s.Printf ("Symbol table after CalculateSymbolSizes():\n");
4859   //        symtab->Dump(&s, NULL, eSortOrderNone);
4860 
4861   return symtab->GetNumSymbols();
4862 }
4863 
Dump(Stream * s)4864 void ObjectFileMachO::Dump(Stream *s) {
4865   ModuleSP module_sp(GetModule());
4866   if (module_sp) {
4867     std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
4868     s->Printf("%p: ", static_cast<void *>(this));
4869     s->Indent();
4870     if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64)
4871       s->PutCString("ObjectFileMachO64");
4872     else
4873       s->PutCString("ObjectFileMachO32");
4874 
4875     *s << ", file = '" << m_file;
4876     ModuleSpecList all_specs;
4877     ModuleSpec base_spec;
4878     GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic),
4879                     base_spec, all_specs);
4880     for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) {
4881       *s << "', triple";
4882       if (e)
4883         s->Printf("[%d]", i);
4884       *s << " = ";
4885       *s << all_specs.GetModuleSpecRefAtIndex(i)
4886                 .GetArchitecture()
4887                 .GetTriple()
4888                 .getTriple();
4889     }
4890     *s << "\n";
4891     SectionList *sections = GetSectionList();
4892     if (sections)
4893       sections->Dump(s->AsRawOstream(), s->GetIndentLevel(), nullptr, true,
4894                      UINT32_MAX);
4895 
4896     if (m_symtab_up)
4897       m_symtab_up->Dump(s, nullptr, eSortOrderNone);
4898   }
4899 }
4900 
GetUUID(const llvm::MachO::mach_header & header,const lldb_private::DataExtractor & data,lldb::offset_t lc_offset)4901 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header,
4902                               const lldb_private::DataExtractor &data,
4903                               lldb::offset_t lc_offset) {
4904   uint32_t i;
4905   llvm::MachO::uuid_command load_cmd;
4906 
4907   lldb::offset_t offset = lc_offset;
4908   for (i = 0; i < header.ncmds; ++i) {
4909     const lldb::offset_t cmd_offset = offset;
4910     if (data.GetU32(&offset, &load_cmd, 2) == nullptr)
4911       break;
4912 
4913     if (load_cmd.cmd == LC_UUID) {
4914       const uint8_t *uuid_bytes = data.PeekData(offset, 16);
4915 
4916       if (uuid_bytes) {
4917         // OpenCL on Mac OS X uses the same UUID for each of its object files.
4918         // We pretend these object files have no UUID to prevent crashing.
4919 
4920         const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8,
4921                                        0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63,
4922                                        0xbb, 0x14, 0xf0, 0x0d};
4923 
4924         if (!memcmp(uuid_bytes, opencl_uuid, 16))
4925           return UUID();
4926 
4927         return UUID::fromOptionalData(uuid_bytes, 16);
4928       }
4929       return UUID();
4930     }
4931     offset = cmd_offset + load_cmd.cmdsize;
4932   }
4933   return UUID();
4934 }
4935 
GetOSName(uint32_t cmd)4936 static llvm::StringRef GetOSName(uint32_t cmd) {
4937   switch (cmd) {
4938   case llvm::MachO::LC_VERSION_MIN_IPHONEOS:
4939     return llvm::Triple::getOSTypeName(llvm::Triple::IOS);
4940   case llvm::MachO::LC_VERSION_MIN_MACOSX:
4941     return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX);
4942   case llvm::MachO::LC_VERSION_MIN_TVOS:
4943     return llvm::Triple::getOSTypeName(llvm::Triple::TvOS);
4944   case llvm::MachO::LC_VERSION_MIN_WATCHOS:
4945     return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS);
4946   default:
4947     llvm_unreachable("unexpected LC_VERSION load command");
4948   }
4949 }
4950 
4951 namespace {
4952 struct OSEnv {
4953   llvm::StringRef os_type;
4954   llvm::StringRef environment;
OSEnv__anon14812a410411::OSEnv4955   OSEnv(uint32_t cmd) {
4956     switch (cmd) {
4957     case llvm::MachO::PLATFORM_MACOS:
4958       os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX);
4959       return;
4960     case llvm::MachO::PLATFORM_IOS:
4961       os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS);
4962       return;
4963     case llvm::MachO::PLATFORM_TVOS:
4964       os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS);
4965       return;
4966     case llvm::MachO::PLATFORM_WATCHOS:
4967       os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS);
4968       return;
4969       // NEED_BRIDGEOS_TRIPLE      case llvm::MachO::PLATFORM_BRIDGEOS:
4970       // NEED_BRIDGEOS_TRIPLE        os_type =
4971       // llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS);
4972       // NEED_BRIDGEOS_TRIPLE        return;
4973     case llvm::MachO::PLATFORM_MACCATALYST:
4974       os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS);
4975       environment = llvm::Triple::getEnvironmentTypeName(llvm::Triple::MacABI);
4976       return;
4977     case llvm::MachO::PLATFORM_IOSSIMULATOR:
4978       os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS);
4979       environment =
4980           llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator);
4981       return;
4982     case llvm::MachO::PLATFORM_TVOSSIMULATOR:
4983       os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS);
4984       environment =
4985           llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator);
4986       return;
4987     case llvm::MachO::PLATFORM_WATCHOSSIMULATOR:
4988       os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS);
4989       environment =
4990           llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator);
4991       return;
4992     default: {
4993       Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS |
4994                                                       LIBLLDB_LOG_PROCESS));
4995       LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION");
4996     }
4997     }
4998   }
4999 };
5000 
5001 struct MinOS {
5002   uint32_t major_version, minor_version, patch_version;
MinOS__anon14812a410411::MinOS5003   MinOS(uint32_t version)
5004       : major_version(version >> 16), minor_version((version >> 8) & 0xffu),
5005         patch_version(version & 0xffu) {}
5006 };
5007 } // namespace
5008 
GetAllArchSpecs(const llvm::MachO::mach_header & header,const lldb_private::DataExtractor & data,lldb::offset_t lc_offset,ModuleSpec & base_spec,lldb_private::ModuleSpecList & all_specs)5009 void ObjectFileMachO::GetAllArchSpecs(const llvm::MachO::mach_header &header,
5010                                       const lldb_private::DataExtractor &data,
5011                                       lldb::offset_t lc_offset,
5012                                       ModuleSpec &base_spec,
5013                                       lldb_private::ModuleSpecList &all_specs) {
5014   auto &base_arch = base_spec.GetArchitecture();
5015   base_arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype);
5016   if (!base_arch.IsValid())
5017     return;
5018 
5019   bool found_any = false;
5020   auto add_triple = [&](const llvm::Triple &triple) {
5021     auto spec = base_spec;
5022     spec.GetArchitecture().GetTriple() = triple;
5023     if (spec.GetArchitecture().IsValid()) {
5024       spec.GetUUID() = ObjectFileMachO::GetUUID(header, data, lc_offset);
5025       all_specs.Append(spec);
5026       found_any = true;
5027     }
5028   };
5029 
5030   // Set OS to an unspecified unknown or a "*" so it can match any OS
5031   llvm::Triple base_triple = base_arch.GetTriple();
5032   base_triple.setOS(llvm::Triple::UnknownOS);
5033   base_triple.setOSName(llvm::StringRef());
5034 
5035   if (header.filetype == MH_PRELOAD) {
5036     if (header.cputype == CPU_TYPE_ARM) {
5037       // If this is a 32-bit arm binary, and it's a standalone binary, force
5038       // the Vendor to Apple so we don't accidentally pick up the generic
5039       // armv7 ABI at runtime.  Apple's armv7 ABI always uses r7 for the
5040       // frame pointer register; most other armv7 ABIs use a combination of
5041       // r7 and r11.
5042       base_triple.setVendor(llvm::Triple::Apple);
5043     } else {
5044       // Set vendor to an unspecified unknown or a "*" so it can match any
5045       // vendor This is required for correct behavior of EFI debugging on
5046       // x86_64
5047       base_triple.setVendor(llvm::Triple::UnknownVendor);
5048       base_triple.setVendorName(llvm::StringRef());
5049     }
5050     return add_triple(base_triple);
5051   }
5052 
5053   llvm::MachO::load_command load_cmd;
5054 
5055   // See if there is an LC_VERSION_MIN_* load command that can give
5056   // us the OS type.
5057   lldb::offset_t offset = lc_offset;
5058   for (uint32_t i = 0; i < header.ncmds; ++i) {
5059     const lldb::offset_t cmd_offset = offset;
5060     if (data.GetU32(&offset, &load_cmd, 2) == NULL)
5061       break;
5062 
5063     llvm::MachO::version_min_command version_min;
5064     switch (load_cmd.cmd) {
5065     case llvm::MachO::LC_VERSION_MIN_MACOSX:
5066     case llvm::MachO::LC_VERSION_MIN_IPHONEOS:
5067     case llvm::MachO::LC_VERSION_MIN_TVOS:
5068     case llvm::MachO::LC_VERSION_MIN_WATCHOS: {
5069       if (load_cmd.cmdsize != sizeof(version_min))
5070         break;
5071       if (data.ExtractBytes(cmd_offset, sizeof(version_min),
5072                             data.GetByteOrder(), &version_min) == 0)
5073         break;
5074       MinOS min_os(version_min.version);
5075       llvm::SmallString<32> os_name;
5076       llvm::raw_svector_ostream os(os_name);
5077       os << GetOSName(load_cmd.cmd) << min_os.major_version << '.'
5078          << min_os.minor_version << '.' << min_os.patch_version;
5079 
5080       auto triple = base_triple;
5081       triple.setOSName(os.str());
5082 
5083       // Disambiguate legacy simulator platforms.
5084       if (load_cmd.cmd != llvm::MachO::LC_VERSION_MIN_MACOSX &&
5085           (base_triple.getArch() == llvm::Triple::x86_64 ||
5086            base_triple.getArch() == llvm::Triple::x86)) {
5087         // The combination of legacy LC_VERSION_MIN load command and
5088         // x86 architecture always indicates a simulator environment.
5089         // The combination of LC_VERSION_MIN and arm architecture only
5090         // appears for native binaries. Back-deploying simulator
5091         // binaries on Apple Silicon Macs use the modern unambigous
5092         // LC_BUILD_VERSION load commands; no special handling required.
5093         triple.setEnvironment(llvm::Triple::Simulator);
5094       }
5095       add_triple(triple);
5096       break;
5097     }
5098     default:
5099       break;
5100     }
5101 
5102     offset = cmd_offset + load_cmd.cmdsize;
5103   }
5104 
5105   // See if there are LC_BUILD_VERSION load commands that can give
5106   // us the OS type.
5107   offset = lc_offset;
5108   for (uint32_t i = 0; i < header.ncmds; ++i) {
5109     const lldb::offset_t cmd_offset = offset;
5110     if (data.GetU32(&offset, &load_cmd, 2) == NULL)
5111       break;
5112 
5113     do {
5114       if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) {
5115         llvm::MachO::build_version_command build_version;
5116         if (load_cmd.cmdsize < sizeof(build_version)) {
5117           // Malformed load command.
5118           break;
5119         }
5120         if (data.ExtractBytes(cmd_offset, sizeof(build_version),
5121                               data.GetByteOrder(), &build_version) == 0)
5122           break;
5123         MinOS min_os(build_version.minos);
5124         OSEnv os_env(build_version.platform);
5125         llvm::SmallString<16> os_name;
5126         llvm::raw_svector_ostream os(os_name);
5127         os << os_env.os_type << min_os.major_version << '.'
5128            << min_os.minor_version << '.' << min_os.patch_version;
5129         auto triple = base_triple;
5130         triple.setOSName(os.str());
5131         os_name.clear();
5132         if (!os_env.environment.empty())
5133           triple.setEnvironmentName(os_env.environment);
5134         add_triple(triple);
5135       }
5136     } while (0);
5137     offset = cmd_offset + load_cmd.cmdsize;
5138   }
5139 
5140   if (!found_any) {
5141     if (header.filetype == MH_KEXT_BUNDLE) {
5142       base_triple.setVendor(llvm::Triple::Apple);
5143       add_triple(base_triple);
5144     } else {
5145       // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT
5146       // so lets not say our Vendor is Apple, leave it as an unspecified
5147       // unknown.
5148       base_triple.setVendor(llvm::Triple::UnknownVendor);
5149       base_triple.setVendorName(llvm::StringRef());
5150       add_triple(base_triple);
5151     }
5152   }
5153 }
5154 
GetArchitecture(ModuleSP module_sp,const llvm::MachO::mach_header & header,const lldb_private::DataExtractor & data,lldb::offset_t lc_offset)5155 ArchSpec ObjectFileMachO::GetArchitecture(
5156     ModuleSP module_sp, const llvm::MachO::mach_header &header,
5157     const lldb_private::DataExtractor &data, lldb::offset_t lc_offset) {
5158   ModuleSpecList all_specs;
5159   ModuleSpec base_spec;
5160   GetAllArchSpecs(header, data, MachHeaderSizeFromMagic(header.magic),
5161                   base_spec, all_specs);
5162 
5163   // If the object file offers multiple alternative load commands,
5164   // pick the one that matches the module.
5165   if (module_sp) {
5166     const ArchSpec &module_arch = module_sp->GetArchitecture();
5167     for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) {
5168       ArchSpec mach_arch =
5169           all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture();
5170       if (module_arch.IsCompatibleMatch(mach_arch))
5171         return mach_arch;
5172     }
5173   }
5174 
5175   // Return the first arch we found.
5176   if (all_specs.GetSize() == 0)
5177     return {};
5178   return all_specs.GetModuleSpecRefAtIndex(0).GetArchitecture();
5179 }
5180 
GetUUID()5181 UUID ObjectFileMachO::GetUUID() {
5182   ModuleSP module_sp(GetModule());
5183   if (module_sp) {
5184     std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5185     lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
5186     return GetUUID(m_header, m_data, offset);
5187   }
5188   return UUID();
5189 }
5190 
GetDependentModules(FileSpecList & files)5191 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) {
5192   uint32_t count = 0;
5193   ModuleSP module_sp(GetModule());
5194   if (module_sp) {
5195     std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5196     llvm::MachO::load_command load_cmd;
5197     lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
5198     std::vector<std::string> rpath_paths;
5199     std::vector<std::string> rpath_relative_paths;
5200     std::vector<std::string> at_exec_relative_paths;
5201     uint32_t i;
5202     for (i = 0; i < m_header.ncmds; ++i) {
5203       const uint32_t cmd_offset = offset;
5204       if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr)
5205         break;
5206 
5207       switch (load_cmd.cmd) {
5208       case LC_RPATH:
5209       case LC_LOAD_DYLIB:
5210       case LC_LOAD_WEAK_DYLIB:
5211       case LC_REEXPORT_DYLIB:
5212       case LC_LOAD_DYLINKER:
5213       case LC_LOADFVMLIB:
5214       case LC_LOAD_UPWARD_DYLIB: {
5215         uint32_t name_offset = cmd_offset + m_data.GetU32(&offset);
5216         const char *path = m_data.PeekCStr(name_offset);
5217         if (path) {
5218           if (load_cmd.cmd == LC_RPATH)
5219             rpath_paths.push_back(path);
5220           else {
5221             if (path[0] == '@') {
5222               if (strncmp(path, "@rpath", strlen("@rpath")) == 0)
5223                 rpath_relative_paths.push_back(path + strlen("@rpath"));
5224               else if (strncmp(path, "@executable_path",
5225                                strlen("@executable_path")) == 0)
5226                 at_exec_relative_paths.push_back(path +
5227                                                  strlen("@executable_path"));
5228             } else {
5229               FileSpec file_spec(path);
5230               if (files.AppendIfUnique(file_spec))
5231                 count++;
5232             }
5233           }
5234         }
5235       } break;
5236 
5237       default:
5238         break;
5239       }
5240       offset = cmd_offset + load_cmd.cmdsize;
5241     }
5242 
5243     FileSpec this_file_spec(m_file);
5244     FileSystem::Instance().Resolve(this_file_spec);
5245 
5246     if (!rpath_paths.empty()) {
5247       // Fixup all LC_RPATH values to be absolute paths
5248       std::string loader_path("@loader_path");
5249       std::string executable_path("@executable_path");
5250       for (auto &rpath : rpath_paths) {
5251         if (llvm::StringRef(rpath).startswith(loader_path)) {
5252           rpath.erase(0, loader_path.size());
5253           rpath.insert(0, this_file_spec.GetDirectory().GetCString());
5254         } else if (llvm::StringRef(rpath).startswith(executable_path)) {
5255           rpath.erase(0, executable_path.size());
5256           rpath.insert(0, this_file_spec.GetDirectory().GetCString());
5257         }
5258       }
5259 
5260       for (const auto &rpath_relative_path : rpath_relative_paths) {
5261         for (const auto &rpath : rpath_paths) {
5262           std::string path = rpath;
5263           path += rpath_relative_path;
5264           // It is OK to resolve this path because we must find a file on disk
5265           // for us to accept it anyway if it is rpath relative.
5266           FileSpec file_spec(path);
5267           FileSystem::Instance().Resolve(file_spec);
5268           if (FileSystem::Instance().Exists(file_spec) &&
5269               files.AppendIfUnique(file_spec)) {
5270             count++;
5271             break;
5272           }
5273         }
5274       }
5275     }
5276 
5277     // We may have @executable_paths but no RPATHS.  Figure those out here.
5278     // Only do this if this object file is the executable.  We have no way to
5279     // get back to the actual executable otherwise, so we won't get the right
5280     // path.
5281     if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) {
5282       FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent();
5283       for (const auto &at_exec_relative_path : at_exec_relative_paths) {
5284         FileSpec file_spec =
5285             exec_dir.CopyByAppendingPathComponent(at_exec_relative_path);
5286         if (FileSystem::Instance().Exists(file_spec) &&
5287             files.AppendIfUnique(file_spec))
5288           count++;
5289       }
5290     }
5291   }
5292   return count;
5293 }
5294 
GetEntryPointAddress()5295 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() {
5296   // If the object file is not an executable it can't hold the entry point.
5297   // m_entry_point_address is initialized to an invalid address, so we can just
5298   // return that. If m_entry_point_address is valid it means we've found it
5299   // already, so return the cached value.
5300 
5301   if ((!IsExecutable() && !IsDynamicLoader()) ||
5302       m_entry_point_address.IsValid()) {
5303     return m_entry_point_address;
5304   }
5305 
5306   // Otherwise, look for the UnixThread or Thread command.  The data for the
5307   // Thread command is given in /usr/include/mach-o.h, but it is basically:
5308   //
5309   //  uint32_t flavor  - this is the flavor argument you would pass to
5310   //  thread_get_state
5311   //  uint32_t count   - this is the count of longs in the thread state data
5312   //  struct XXX_thread_state state - this is the structure from
5313   //  <machine/thread_status.h> corresponding to the flavor.
5314   //  <repeat this trio>
5315   //
5316   // So we just keep reading the various register flavors till we find the GPR
5317   // one, then read the PC out of there.
5318   // FIXME: We will need to have a "RegisterContext data provider" class at some
5319   // point that can get all the registers
5320   // out of data in this form & attach them to a given thread.  That should
5321   // underlie the MacOS X User process plugin, and we'll also need it for the
5322   // MacOS X Core File process plugin.  When we have that we can also use it
5323   // here.
5324   //
5325   // For now we hard-code the offsets and flavors we need:
5326   //
5327   //
5328 
5329   ModuleSP module_sp(GetModule());
5330   if (module_sp) {
5331     std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5332     llvm::MachO::load_command load_cmd;
5333     lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
5334     uint32_t i;
5335     lldb::addr_t start_address = LLDB_INVALID_ADDRESS;
5336     bool done = false;
5337 
5338     for (i = 0; i < m_header.ncmds; ++i) {
5339       const lldb::offset_t cmd_offset = offset;
5340       if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr)
5341         break;
5342 
5343       switch (load_cmd.cmd) {
5344       case LC_UNIXTHREAD:
5345       case LC_THREAD: {
5346         while (offset < cmd_offset + load_cmd.cmdsize) {
5347           uint32_t flavor = m_data.GetU32(&offset);
5348           uint32_t count = m_data.GetU32(&offset);
5349           if (count == 0) {
5350             // We've gotten off somehow, log and exit;
5351             return m_entry_point_address;
5352           }
5353 
5354           switch (m_header.cputype) {
5355           case llvm::MachO::CPU_TYPE_ARM:
5356             if (flavor == 1 ||
5357                 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32
5358                              // from mach/arm/thread_status.h
5359             {
5360               offset += 60; // This is the offset of pc in the GPR thread state
5361                             // data structure.
5362               start_address = m_data.GetU32(&offset);
5363               done = true;
5364             }
5365             break;
5366           case llvm::MachO::CPU_TYPE_ARM64:
5367           case llvm::MachO::CPU_TYPE_ARM64_32:
5368             if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h
5369             {
5370               offset += 256; // This is the offset of pc in the GPR thread state
5371                              // data structure.
5372               start_address = m_data.GetU64(&offset);
5373               done = true;
5374             }
5375             break;
5376           case llvm::MachO::CPU_TYPE_I386:
5377             if (flavor ==
5378                 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h
5379             {
5380               offset += 40; // This is the offset of eip in the GPR thread state
5381                             // data structure.
5382               start_address = m_data.GetU32(&offset);
5383               done = true;
5384             }
5385             break;
5386           case llvm::MachO::CPU_TYPE_X86_64:
5387             if (flavor ==
5388                 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h
5389             {
5390               offset += 16 * 8; // This is the offset of rip in the GPR thread
5391                                 // state data structure.
5392               start_address = m_data.GetU64(&offset);
5393               done = true;
5394             }
5395             break;
5396           default:
5397             return m_entry_point_address;
5398           }
5399           // Haven't found the GPR flavor yet, skip over the data for this
5400           // flavor:
5401           if (done)
5402             break;
5403           offset += count * 4;
5404         }
5405       } break;
5406       case LC_MAIN: {
5407         ConstString text_segment_name("__TEXT");
5408         uint64_t entryoffset = m_data.GetU64(&offset);
5409         SectionSP text_segment_sp =
5410             GetSectionList()->FindSectionByName(text_segment_name);
5411         if (text_segment_sp) {
5412           done = true;
5413           start_address = text_segment_sp->GetFileAddress() + entryoffset;
5414         }
5415       } break;
5416 
5417       default:
5418         break;
5419       }
5420       if (done)
5421         break;
5422 
5423       // Go to the next load command:
5424       offset = cmd_offset + load_cmd.cmdsize;
5425     }
5426 
5427     if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) {
5428       if (GetSymtab()) {
5429         Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType(
5430             ConstString("_dyld_start"), SymbolType::eSymbolTypeCode,
5431             Symtab::eDebugAny, Symtab::eVisibilityAny);
5432         if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) {
5433           start_address = dyld_start_sym->GetAddress().GetFileAddress();
5434         }
5435       }
5436     }
5437 
5438     if (start_address != LLDB_INVALID_ADDRESS) {
5439       // We got the start address from the load commands, so now resolve that
5440       // address in the sections of this ObjectFile:
5441       if (!m_entry_point_address.ResolveAddressUsingFileSections(
5442               start_address, GetSectionList())) {
5443         m_entry_point_address.Clear();
5444       }
5445     } else {
5446       // We couldn't read the UnixThread load command - maybe it wasn't there.
5447       // As a fallback look for the "start" symbol in the main executable.
5448 
5449       ModuleSP module_sp(GetModule());
5450 
5451       if (module_sp) {
5452         SymbolContextList contexts;
5453         SymbolContext context;
5454         module_sp->FindSymbolsWithNameAndType(ConstString("start"),
5455                                               eSymbolTypeCode, contexts);
5456         if (contexts.GetSize()) {
5457           if (contexts.GetContextAtIndex(0, context))
5458             m_entry_point_address = context.symbol->GetAddress();
5459         }
5460       }
5461     }
5462   }
5463 
5464   return m_entry_point_address;
5465 }
5466 
GetBaseAddress()5467 lldb_private::Address ObjectFileMachO::GetBaseAddress() {
5468   lldb_private::Address header_addr;
5469   SectionList *section_list = GetSectionList();
5470   if (section_list) {
5471     SectionSP text_segment_sp(
5472         section_list->FindSectionByName(GetSegmentNameTEXT()));
5473     if (text_segment_sp) {
5474       header_addr.SetSection(text_segment_sp);
5475       header_addr.SetOffset(0);
5476     }
5477   }
5478   return header_addr;
5479 }
5480 
GetNumThreadContexts()5481 uint32_t ObjectFileMachO::GetNumThreadContexts() {
5482   ModuleSP module_sp(GetModule());
5483   if (module_sp) {
5484     std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5485     if (!m_thread_context_offsets_valid) {
5486       m_thread_context_offsets_valid = true;
5487       lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
5488       FileRangeArray::Entry file_range;
5489       llvm::MachO::thread_command thread_cmd;
5490       for (uint32_t i = 0; i < m_header.ncmds; ++i) {
5491         const uint32_t cmd_offset = offset;
5492         if (m_data.GetU32(&offset, &thread_cmd, 2) == nullptr)
5493           break;
5494 
5495         if (thread_cmd.cmd == LC_THREAD) {
5496           file_range.SetRangeBase(offset);
5497           file_range.SetByteSize(thread_cmd.cmdsize - 8);
5498           m_thread_context_offsets.Append(file_range);
5499         }
5500         offset = cmd_offset + thread_cmd.cmdsize;
5501       }
5502     }
5503   }
5504   return m_thread_context_offsets.GetSize();
5505 }
5506 
GetIdentifierString()5507 std::string ObjectFileMachO::GetIdentifierString() {
5508   std::string result;
5509   ModuleSP module_sp(GetModule());
5510   if (module_sp) {
5511     std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5512 
5513     // First, look over the load commands for an LC_NOTE load command with
5514     // data_owner string "kern ver str" & use that if found.
5515     lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
5516     for (uint32_t i = 0; i < m_header.ncmds; ++i) {
5517       const uint32_t cmd_offset = offset;
5518       llvm::MachO::load_command lc;
5519       if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr)
5520         break;
5521       if (lc.cmd == LC_NOTE) {
5522         char data_owner[17];
5523         m_data.CopyData(offset, 16, data_owner);
5524         data_owner[16] = '\0';
5525         offset += 16;
5526         uint64_t fileoff = m_data.GetU64_unchecked(&offset);
5527         uint64_t size = m_data.GetU64_unchecked(&offset);
5528 
5529         // "kern ver str" has a uint32_t version and then a nul terminated
5530         // c-string.
5531         if (strcmp("kern ver str", data_owner) == 0) {
5532           offset = fileoff;
5533           uint32_t version;
5534           if (m_data.GetU32(&offset, &version, 1) != nullptr) {
5535             if (version == 1) {
5536               uint32_t strsize = size - sizeof(uint32_t);
5537               char *buf = (char *)malloc(strsize);
5538               if (buf) {
5539                 m_data.CopyData(offset, strsize, buf);
5540                 buf[strsize - 1] = '\0';
5541                 result = buf;
5542                 if (buf)
5543                   free(buf);
5544                 return result;
5545               }
5546             }
5547           }
5548         }
5549       }
5550       offset = cmd_offset + lc.cmdsize;
5551     }
5552 
5553     // Second, make a pass over the load commands looking for an obsolete
5554     // LC_IDENT load command.
5555     offset = MachHeaderSizeFromMagic(m_header.magic);
5556     for (uint32_t i = 0; i < m_header.ncmds; ++i) {
5557       const uint32_t cmd_offset = offset;
5558       llvm::MachO::ident_command ident_command;
5559       if (m_data.GetU32(&offset, &ident_command, 2) == nullptr)
5560         break;
5561       if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) {
5562         char *buf = (char *)malloc(ident_command.cmdsize);
5563         if (buf != nullptr && m_data.CopyData(offset, ident_command.cmdsize,
5564                                               buf) == ident_command.cmdsize) {
5565           buf[ident_command.cmdsize - 1] = '\0';
5566           result = buf;
5567         }
5568         if (buf)
5569           free(buf);
5570       }
5571       offset = cmd_offset + ident_command.cmdsize;
5572     }
5573   }
5574   return result;
5575 }
5576 
GetAddressMask()5577 addr_t ObjectFileMachO::GetAddressMask() {
5578   addr_t mask = 0;
5579   ModuleSP module_sp(GetModule());
5580   if (module_sp) {
5581     std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5582     lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
5583     for (uint32_t i = 0; i < m_header.ncmds; ++i) {
5584       const uint32_t cmd_offset = offset;
5585       llvm::MachO::load_command lc;
5586       if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr)
5587         break;
5588       if (lc.cmd == LC_NOTE) {
5589         char data_owner[17];
5590         m_data.CopyData(offset, 16, data_owner);
5591         data_owner[16] = '\0';
5592         offset += 16;
5593         uint64_t fileoff = m_data.GetU64_unchecked(&offset);
5594 
5595         // "addrable bits" has a uint32_t version and a uint32_t
5596         // number of bits used in addressing.
5597         if (strcmp("addrable bits", data_owner) == 0) {
5598           offset = fileoff;
5599           uint32_t version;
5600           if (m_data.GetU32(&offset, &version, 1) != nullptr) {
5601             if (version == 3) {
5602               uint32_t num_addr_bits = m_data.GetU32_unchecked(&offset);
5603               if (num_addr_bits != 0) {
5604                 mask = ~((1ULL << num_addr_bits) - 1);
5605               }
5606               break;
5607             }
5608           }
5609         }
5610       }
5611       offset = cmd_offset + lc.cmdsize;
5612     }
5613   }
5614   return mask;
5615 }
5616 
GetCorefileMainBinaryInfo(addr_t & address,UUID & uuid,ObjectFile::BinaryType & type)5617 bool ObjectFileMachO::GetCorefileMainBinaryInfo(addr_t &address, UUID &uuid,
5618                                                 ObjectFile::BinaryType &type) {
5619   address = LLDB_INVALID_ADDRESS;
5620   uuid.Clear();
5621   ModuleSP module_sp(GetModule());
5622   if (module_sp) {
5623     std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5624     lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
5625     for (uint32_t i = 0; i < m_header.ncmds; ++i) {
5626       const uint32_t cmd_offset = offset;
5627       llvm::MachO::load_command lc;
5628       if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr)
5629         break;
5630       if (lc.cmd == LC_NOTE) {
5631         char data_owner[17];
5632         memset(data_owner, 0, sizeof(data_owner));
5633         m_data.CopyData(offset, 16, data_owner);
5634         offset += 16;
5635         uint64_t fileoff = m_data.GetU64_unchecked(&offset);
5636         uint64_t size = m_data.GetU64_unchecked(&offset);
5637 
5638         // "main bin spec" (main binary specification) data payload is
5639         // formatted:
5640         //    uint32_t version       [currently 1]
5641         //    uint32_t type          [0 == unspecified, 1 == kernel,
5642         //                            2 == user process, 3 == firmware ]
5643         //    uint64_t address       [ UINT64_MAX if address not specified ]
5644         //    uuid_t   uuid          [ all zero's if uuid not specified ]
5645         //    uint32_t log2_pagesize [ process page size in log base
5646         //                             2, e.g. 4k pages are 12.
5647         //                             0 for unspecified ]
5648         //    uint32_t unused        [ for alignment ]
5649 
5650         if (strcmp("main bin spec", data_owner) == 0 && size >= 32) {
5651           offset = fileoff;
5652           uint32_t version;
5653           if (m_data.GetU32(&offset, &version, 1) != nullptr && version == 1) {
5654             uint32_t binspec_type = 0;
5655             uuid_t raw_uuid;
5656             memset(raw_uuid, 0, sizeof(uuid_t));
5657 
5658             if (m_data.GetU32(&offset, &binspec_type, 1) &&
5659                 m_data.GetU64(&offset, &address, 1) &&
5660                 m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) {
5661               uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t));
5662               // convert the "main bin spec" type into our
5663               // ObjectFile::BinaryType enum
5664               switch (binspec_type) {
5665               case 0:
5666                 type = eBinaryTypeUnknown;
5667                 break;
5668               case 1:
5669                 type = eBinaryTypeKernel;
5670                 break;
5671               case 2:
5672                 type = eBinaryTypeUser;
5673                 break;
5674               case 3:
5675                 type = eBinaryTypeStandalone;
5676                 break;
5677               }
5678               return true;
5679             }
5680           }
5681         }
5682       }
5683       offset = cmd_offset + lc.cmdsize;
5684     }
5685   }
5686   return false;
5687 }
5688 
5689 lldb::RegisterContextSP
GetThreadContextAtIndex(uint32_t idx,lldb_private::Thread & thread)5690 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx,
5691                                          lldb_private::Thread &thread) {
5692   lldb::RegisterContextSP reg_ctx_sp;
5693 
5694   ModuleSP module_sp(GetModule());
5695   if (module_sp) {
5696     std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5697     if (!m_thread_context_offsets_valid)
5698       GetNumThreadContexts();
5699 
5700     const FileRangeArray::Entry *thread_context_file_range =
5701         m_thread_context_offsets.GetEntryAtIndex(idx);
5702     if (thread_context_file_range) {
5703 
5704       DataExtractor data(m_data, thread_context_file_range->GetRangeBase(),
5705                          thread_context_file_range->GetByteSize());
5706 
5707       switch (m_header.cputype) {
5708       case llvm::MachO::CPU_TYPE_ARM64:
5709       case llvm::MachO::CPU_TYPE_ARM64_32:
5710         reg_ctx_sp =
5711             std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data);
5712         break;
5713 
5714       case llvm::MachO::CPU_TYPE_ARM:
5715         reg_ctx_sp =
5716             std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data);
5717         break;
5718 
5719       case llvm::MachO::CPU_TYPE_I386:
5720         reg_ctx_sp =
5721             std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data);
5722         break;
5723 
5724       case llvm::MachO::CPU_TYPE_X86_64:
5725         reg_ctx_sp =
5726             std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data);
5727         break;
5728       }
5729     }
5730   }
5731   return reg_ctx_sp;
5732 }
5733 
CalculateType()5734 ObjectFile::Type ObjectFileMachO::CalculateType() {
5735   switch (m_header.filetype) {
5736   case MH_OBJECT: // 0x1u
5737     if (GetAddressByteSize() == 4) {
5738       // 32 bit kexts are just object files, but they do have a valid
5739       // UUID load command.
5740       if (GetUUID()) {
5741         // this checking for the UUID load command is not enough we could
5742         // eventually look for the symbol named "OSKextGetCurrentIdentifier" as
5743         // this is required of kexts
5744         if (m_strata == eStrataInvalid)
5745           m_strata = eStrataKernel;
5746         return eTypeSharedLibrary;
5747       }
5748     }
5749     return eTypeObjectFile;
5750 
5751   case MH_EXECUTE:
5752     return eTypeExecutable; // 0x2u
5753   case MH_FVMLIB:
5754     return eTypeSharedLibrary; // 0x3u
5755   case MH_CORE:
5756     return eTypeCoreFile; // 0x4u
5757   case MH_PRELOAD:
5758     return eTypeSharedLibrary; // 0x5u
5759   case MH_DYLIB:
5760     return eTypeSharedLibrary; // 0x6u
5761   case MH_DYLINKER:
5762     return eTypeDynamicLinker; // 0x7u
5763   case MH_BUNDLE:
5764     return eTypeSharedLibrary; // 0x8u
5765   case MH_DYLIB_STUB:
5766     return eTypeStubLibrary; // 0x9u
5767   case MH_DSYM:
5768     return eTypeDebugInfo; // 0xAu
5769   case MH_KEXT_BUNDLE:
5770     return eTypeSharedLibrary; // 0xBu
5771   default:
5772     break;
5773   }
5774   return eTypeUnknown;
5775 }
5776 
CalculateStrata()5777 ObjectFile::Strata ObjectFileMachO::CalculateStrata() {
5778   switch (m_header.filetype) {
5779   case MH_OBJECT: // 0x1u
5780   {
5781     // 32 bit kexts are just object files, but they do have a valid
5782     // UUID load command.
5783     if (GetUUID()) {
5784       // this checking for the UUID load command is not enough we could
5785       // eventually look for the symbol named "OSKextGetCurrentIdentifier" as
5786       // this is required of kexts
5787       if (m_type == eTypeInvalid)
5788         m_type = eTypeSharedLibrary;
5789 
5790       return eStrataKernel;
5791     }
5792   }
5793     return eStrataUnknown;
5794 
5795   case MH_EXECUTE: // 0x2u
5796     // Check for the MH_DYLDLINK bit in the flags
5797     if (m_header.flags & MH_DYLDLINK) {
5798       return eStrataUser;
5799     } else {
5800       SectionList *section_list = GetSectionList();
5801       if (section_list) {
5802         static ConstString g_kld_section_name("__KLD");
5803         if (section_list->FindSectionByName(g_kld_section_name))
5804           return eStrataKernel;
5805       }
5806     }
5807     return eStrataRawImage;
5808 
5809   case MH_FVMLIB:
5810     return eStrataUser; // 0x3u
5811   case MH_CORE:
5812     return eStrataUnknown; // 0x4u
5813   case MH_PRELOAD:
5814     return eStrataRawImage; // 0x5u
5815   case MH_DYLIB:
5816     return eStrataUser; // 0x6u
5817   case MH_DYLINKER:
5818     return eStrataUser; // 0x7u
5819   case MH_BUNDLE:
5820     return eStrataUser; // 0x8u
5821   case MH_DYLIB_STUB:
5822     return eStrataUser; // 0x9u
5823   case MH_DSYM:
5824     return eStrataUnknown; // 0xAu
5825   case MH_KEXT_BUNDLE:
5826     return eStrataKernel; // 0xBu
5827   default:
5828     break;
5829   }
5830   return eStrataUnknown;
5831 }
5832 
GetVersion()5833 llvm::VersionTuple ObjectFileMachO::GetVersion() {
5834   ModuleSP module_sp(GetModule());
5835   if (module_sp) {
5836     std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5837     llvm::MachO::dylib_command load_cmd;
5838     lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
5839     uint32_t version_cmd = 0;
5840     uint64_t version = 0;
5841     uint32_t i;
5842     for (i = 0; i < m_header.ncmds; ++i) {
5843       const lldb::offset_t cmd_offset = offset;
5844       if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr)
5845         break;
5846 
5847       if (load_cmd.cmd == LC_ID_DYLIB) {
5848         if (version_cmd == 0) {
5849           version_cmd = load_cmd.cmd;
5850           if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == nullptr)
5851             break;
5852           version = load_cmd.dylib.current_version;
5853         }
5854         break; // Break for now unless there is another more complete version
5855                // number load command in the future.
5856       }
5857       offset = cmd_offset + load_cmd.cmdsize;
5858     }
5859 
5860     if (version_cmd == LC_ID_DYLIB) {
5861       unsigned major = (version & 0xFFFF0000ull) >> 16;
5862       unsigned minor = (version & 0x0000FF00ull) >> 8;
5863       unsigned subminor = (version & 0x000000FFull);
5864       return llvm::VersionTuple(major, minor, subminor);
5865     }
5866   }
5867   return llvm::VersionTuple();
5868 }
5869 
GetArchitecture()5870 ArchSpec ObjectFileMachO::GetArchitecture() {
5871   ModuleSP module_sp(GetModule());
5872   ArchSpec arch;
5873   if (module_sp) {
5874     std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5875 
5876     return GetArchitecture(module_sp, m_header, m_data,
5877                            MachHeaderSizeFromMagic(m_header.magic));
5878   }
5879   return arch;
5880 }
5881 
GetProcessSharedCacheUUID(Process * process,addr_t & base_addr,UUID & uuid)5882 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process,
5883                                                 addr_t &base_addr, UUID &uuid) {
5884   uuid.Clear();
5885   base_addr = LLDB_INVALID_ADDRESS;
5886   if (process && process->GetDynamicLoader()) {
5887     DynamicLoader *dl = process->GetDynamicLoader();
5888     LazyBool using_shared_cache;
5889     LazyBool private_shared_cache;
5890     dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache,
5891                                   private_shared_cache);
5892   }
5893   Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS |
5894                                                   LIBLLDB_LOG_PROCESS));
5895   LLDB_LOGF(
5896       log,
5897       "inferior process shared cache has a UUID of %s, base address 0x%" PRIx64,
5898       uuid.GetAsString().c_str(), base_addr);
5899 }
5900 
5901 // From dyld SPI header dyld_process_info.h
5902 typedef void *dyld_process_info;
5903 struct lldb_copy__dyld_process_cache_info {
5904   uuid_t cacheUUID;          // UUID of cache used by process
5905   uint64_t cacheBaseAddress; // load address of dyld shared cache
5906   bool noCache;              // process is running without a dyld cache
5907   bool privateCache; // process is using a private copy of its dyld cache
5908 };
5909 
5910 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with
5911 // llvm enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile
5912 // errors. So we need to use the actual underlying types of task_t and
5913 // kern_return_t below.
5914 extern "C" unsigned int /*task_t*/ mach_task_self();
5915 
GetLLDBSharedCacheUUID(addr_t & base_addr,UUID & uuid)5916 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) {
5917   uuid.Clear();
5918   base_addr = LLDB_INVALID_ADDRESS;
5919 
5920 #if defined(__APPLE__)
5921   uint8_t *(*dyld_get_all_image_infos)(void);
5922   dyld_get_all_image_infos =
5923       (uint8_t * (*)()) dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos");
5924   if (dyld_get_all_image_infos) {
5925     uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos();
5926     if (dyld_all_image_infos_address) {
5927       uint32_t *version = (uint32_t *)
5928           dyld_all_image_infos_address; // version <mach-o/dyld_images.h>
5929       if (*version >= 13) {
5930         uuid_t *sharedCacheUUID_address = 0;
5931         int wordsize = sizeof(uint8_t *);
5932         if (wordsize == 8) {
5933           sharedCacheUUID_address =
5934               (uuid_t *)((uint8_t *)dyld_all_image_infos_address +
5935                          160); // sharedCacheUUID <mach-o/dyld_images.h>
5936           if (*version >= 15)
5937             base_addr =
5938                 *(uint64_t
5939                       *)((uint8_t *)dyld_all_image_infos_address +
5940                          176); // sharedCacheBaseAddress <mach-o/dyld_images.h>
5941         } else {
5942           sharedCacheUUID_address =
5943               (uuid_t *)((uint8_t *)dyld_all_image_infos_address +
5944                          84); // sharedCacheUUID <mach-o/dyld_images.h>
5945           if (*version >= 15) {
5946             base_addr = 0;
5947             base_addr =
5948                 *(uint32_t
5949                       *)((uint8_t *)dyld_all_image_infos_address +
5950                          100); // sharedCacheBaseAddress <mach-o/dyld_images.h>
5951           }
5952         }
5953         uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t));
5954       }
5955     }
5956   } else {
5957     // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI
5958     dyld_process_info (*dyld_process_info_create)(
5959         unsigned int /* task_t */ task, uint64_t timestamp,
5960         unsigned int /*kern_return_t*/ *kernelError);
5961     void (*dyld_process_info_get_cache)(void *info, void *cacheInfo);
5962     void (*dyld_process_info_release)(dyld_process_info info);
5963 
5964     dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t,
5965                                           unsigned int /*kern_return_t*/ *))
5966         dlsym(RTLD_DEFAULT, "_dyld_process_info_create");
5967     dyld_process_info_get_cache = (void (*)(void *, void *))dlsym(
5968         RTLD_DEFAULT, "_dyld_process_info_get_cache");
5969     dyld_process_info_release =
5970         (void (*)(void *))dlsym(RTLD_DEFAULT, "_dyld_process_info_release");
5971 
5972     if (dyld_process_info_create && dyld_process_info_get_cache) {
5973       unsigned int /*kern_return_t */ kern_ret;
5974       dyld_process_info process_info =
5975           dyld_process_info_create(::mach_task_self(), 0, &kern_ret);
5976       if (process_info) {
5977         struct lldb_copy__dyld_process_cache_info sc_info;
5978         memset(&sc_info, 0, sizeof(struct lldb_copy__dyld_process_cache_info));
5979         dyld_process_info_get_cache(process_info, &sc_info);
5980         if (sc_info.cacheBaseAddress != 0) {
5981           base_addr = sc_info.cacheBaseAddress;
5982           uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t));
5983         }
5984         dyld_process_info_release(process_info);
5985       }
5986     }
5987   }
5988   Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS |
5989                                                   LIBLLDB_LOG_PROCESS));
5990   if (log && uuid.IsValid())
5991     LLDB_LOGF(log,
5992               "lldb's in-memory shared cache has a UUID of %s base address of "
5993               "0x%" PRIx64,
5994               uuid.GetAsString().c_str(), base_addr);
5995 #endif
5996 }
5997 
GetMinimumOSVersion()5998 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() {
5999   if (!m_min_os_version) {
6000     lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
6001     for (uint32_t i = 0; i < m_header.ncmds; ++i) {
6002       const lldb::offset_t load_cmd_offset = offset;
6003 
6004       llvm::MachO::version_min_command lc;
6005       if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr)
6006         break;
6007       if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX ||
6008           lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS ||
6009           lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS ||
6010           lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) {
6011         if (m_data.GetU32(&offset, &lc.version,
6012                           (sizeof(lc) / sizeof(uint32_t)) - 2)) {
6013           const uint32_t xxxx = lc.version >> 16;
6014           const uint32_t yy = (lc.version >> 8) & 0xffu;
6015           const uint32_t zz = lc.version & 0xffu;
6016           if (xxxx) {
6017             m_min_os_version = llvm::VersionTuple(xxxx, yy, zz);
6018             break;
6019           }
6020         }
6021       } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) {
6022         // struct build_version_command {
6023         //     uint32_t    cmd;            /* LC_BUILD_VERSION */
6024         //     uint32_t    cmdsize;        /* sizeof(struct
6025         //     build_version_command) plus */
6026         //                                 /* ntools * sizeof(struct
6027         //                                 build_tool_version) */
6028         //     uint32_t    platform;       /* platform */
6029         //     uint32_t    minos;          /* X.Y.Z is encoded in nibbles
6030         //     xxxx.yy.zz */ uint32_t    sdk;            /* X.Y.Z is encoded in
6031         //     nibbles xxxx.yy.zz */ uint32_t    ntools;         /* number of
6032         //     tool entries following this */
6033         // };
6034 
6035         offset += 4; // skip platform
6036         uint32_t minos = m_data.GetU32(&offset);
6037 
6038         const uint32_t xxxx = minos >> 16;
6039         const uint32_t yy = (minos >> 8) & 0xffu;
6040         const uint32_t zz = minos & 0xffu;
6041         if (xxxx) {
6042           m_min_os_version = llvm::VersionTuple(xxxx, yy, zz);
6043           break;
6044         }
6045       }
6046 
6047       offset = load_cmd_offset + lc.cmdsize;
6048     }
6049 
6050     if (!m_min_os_version) {
6051       // Set version to an empty value so we don't keep trying to
6052       m_min_os_version = llvm::VersionTuple();
6053     }
6054   }
6055 
6056   return *m_min_os_version;
6057 }
6058 
GetSDKVersion()6059 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() {
6060   if (!m_sdk_versions.hasValue()) {
6061     lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
6062     for (uint32_t i = 0; i < m_header.ncmds; ++i) {
6063       const lldb::offset_t load_cmd_offset = offset;
6064 
6065       llvm::MachO::version_min_command lc;
6066       if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr)
6067         break;
6068       if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX ||
6069           lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS ||
6070           lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS ||
6071           lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) {
6072         if (m_data.GetU32(&offset, &lc.version,
6073                           (sizeof(lc) / sizeof(uint32_t)) - 2)) {
6074           const uint32_t xxxx = lc.sdk >> 16;
6075           const uint32_t yy = (lc.sdk >> 8) & 0xffu;
6076           const uint32_t zz = lc.sdk & 0xffu;
6077           if (xxxx) {
6078             m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz);
6079             break;
6080           } else {
6081             GetModule()->ReportWarning("minimum OS version load command with "
6082                                        "invalid (0) version found.");
6083           }
6084         }
6085       }
6086       offset = load_cmd_offset + lc.cmdsize;
6087     }
6088 
6089     if (!m_sdk_versions.hasValue()) {
6090       offset = MachHeaderSizeFromMagic(m_header.magic);
6091       for (uint32_t i = 0; i < m_header.ncmds; ++i) {
6092         const lldb::offset_t load_cmd_offset = offset;
6093 
6094         llvm::MachO::version_min_command lc;
6095         if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr)
6096           break;
6097         if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) {
6098           // struct build_version_command {
6099           //     uint32_t    cmd;            /* LC_BUILD_VERSION */
6100           //     uint32_t    cmdsize;        /* sizeof(struct
6101           //     build_version_command) plus */
6102           //                                 /* ntools * sizeof(struct
6103           //                                 build_tool_version) */
6104           //     uint32_t    platform;       /* platform */
6105           //     uint32_t    minos;          /* X.Y.Z is encoded in nibbles
6106           //     xxxx.yy.zz */ uint32_t    sdk;            /* X.Y.Z is encoded
6107           //     in nibbles xxxx.yy.zz */ uint32_t    ntools;         /* number
6108           //     of tool entries following this */
6109           // };
6110 
6111           offset += 4; // skip platform
6112           uint32_t minos = m_data.GetU32(&offset);
6113 
6114           const uint32_t xxxx = minos >> 16;
6115           const uint32_t yy = (minos >> 8) & 0xffu;
6116           const uint32_t zz = minos & 0xffu;
6117           if (xxxx) {
6118             m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz);
6119             break;
6120           }
6121         }
6122         offset = load_cmd_offset + lc.cmdsize;
6123       }
6124     }
6125 
6126     if (!m_sdk_versions.hasValue())
6127       m_sdk_versions = llvm::VersionTuple();
6128   }
6129 
6130   return m_sdk_versions.getValue();
6131 }
6132 
GetIsDynamicLinkEditor()6133 bool ObjectFileMachO::GetIsDynamicLinkEditor() {
6134   return m_header.filetype == llvm::MachO::MH_DYLINKER;
6135 }
6136 
AllowAssemblyEmulationUnwindPlans()6137 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() {
6138   return m_allow_assembly_emulation_unwind_plans;
6139 }
6140 
6141 // PluginInterface protocol
GetPluginName()6142 lldb_private::ConstString ObjectFileMachO::GetPluginName() {
6143   return GetPluginNameStatic();
6144 }
6145 
GetPluginVersion()6146 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; }
6147 
GetMachHeaderSection()6148 Section *ObjectFileMachO::GetMachHeaderSection() {
6149   // Find the first address of the mach header which is the first non-zero file
6150   // sized section whose file offset is zero. This is the base file address of
6151   // the mach-o file which can be subtracted from the vmaddr of the other
6152   // segments found in memory and added to the load address
6153   ModuleSP module_sp = GetModule();
6154   if (!module_sp)
6155     return nullptr;
6156   SectionList *section_list = GetSectionList();
6157   if (!section_list)
6158     return nullptr;
6159   const size_t num_sections = section_list->GetSize();
6160   for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) {
6161     Section *section = section_list->GetSectionAtIndex(sect_idx).get();
6162     if (section->GetFileOffset() == 0 && SectionIsLoadable(section))
6163       return section;
6164   }
6165   return nullptr;
6166 }
6167 
SectionIsLoadable(const Section * section)6168 bool ObjectFileMachO::SectionIsLoadable(const Section *section) {
6169   if (!section)
6170     return false;
6171   const bool is_dsym = (m_header.filetype == MH_DSYM);
6172   if (section->GetFileSize() == 0 && !is_dsym)
6173     return false;
6174   if (section->IsThreadSpecific())
6175     return false;
6176   if (GetModule().get() != section->GetModule().get())
6177     return false;
6178   // Be careful with __LINKEDIT and __DWARF segments
6179   if (section->GetName() == GetSegmentNameLINKEDIT() ||
6180       section->GetName() == GetSegmentNameDWARF()) {
6181     // Only map __LINKEDIT and __DWARF if we have an in memory image and
6182     // this isn't a kernel binary like a kext or mach_kernel.
6183     const bool is_memory_image = (bool)m_process_wp.lock();
6184     const Strata strata = GetStrata();
6185     if (is_memory_image == false || strata == eStrataKernel)
6186       return false;
6187   }
6188   return true;
6189 }
6190 
CalculateSectionLoadAddressForMemoryImage(lldb::addr_t header_load_address,const Section * header_section,const Section * section)6191 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage(
6192     lldb::addr_t header_load_address, const Section *header_section,
6193     const Section *section) {
6194   ModuleSP module_sp = GetModule();
6195   if (module_sp && header_section && section &&
6196       header_load_address != LLDB_INVALID_ADDRESS) {
6197     lldb::addr_t file_addr = header_section->GetFileAddress();
6198     if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section))
6199       return section->GetFileAddress() - file_addr + header_load_address;
6200   }
6201   return LLDB_INVALID_ADDRESS;
6202 }
6203 
SetLoadAddress(Target & target,lldb::addr_t value,bool value_is_offset)6204 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value,
6205                                      bool value_is_offset) {
6206   ModuleSP module_sp = GetModule();
6207   if (!module_sp)
6208     return false;
6209 
6210   SectionList *section_list = GetSectionList();
6211   if (!section_list)
6212     return false;
6213 
6214   size_t num_loaded_sections = 0;
6215   const size_t num_sections = section_list->GetSize();
6216 
6217   if (value_is_offset) {
6218     // "value" is an offset to apply to each top level segment
6219     for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) {
6220       // Iterate through the object file sections to find all of the
6221       // sections that size on disk (to avoid __PAGEZERO) and load them
6222       SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx));
6223       if (SectionIsLoadable(section_sp.get()))
6224         if (target.GetSectionLoadList().SetSectionLoadAddress(
6225                 section_sp, section_sp->GetFileAddress() + value))
6226           ++num_loaded_sections;
6227     }
6228   } else {
6229     // "value" is the new base address of the mach_header, adjust each
6230     // section accordingly
6231 
6232     Section *mach_header_section = GetMachHeaderSection();
6233     if (mach_header_section) {
6234       for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) {
6235         SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx));
6236 
6237         lldb::addr_t section_load_addr =
6238             CalculateSectionLoadAddressForMemoryImage(
6239                 value, mach_header_section, section_sp.get());
6240         if (section_load_addr != LLDB_INVALID_ADDRESS) {
6241           if (target.GetSectionLoadList().SetSectionLoadAddress(
6242                   section_sp, section_load_addr))
6243             ++num_loaded_sections;
6244         }
6245       }
6246     }
6247   }
6248   return num_loaded_sections > 0;
6249 }
6250 
6251 struct all_image_infos_header {
6252   uint32_t version;         // currently 1
6253   uint32_t imgcount;        // number of binary images
6254   uint64_t entries_fileoff; // file offset in the corefile of where the array of
6255                             // struct entry's begin.
6256   uint32_t entries_size;    // size of 'struct entry'.
6257   uint32_t unused;
6258 };
6259 
6260 struct image_entry {
6261   uint64_t filepath_offset;  // offset in corefile to c-string of the file path,
6262                              // UINT64_MAX if unavailable.
6263   uuid_t uuid;               // uint8_t[16].  should be set to all zeroes if
6264                              // uuid is unknown.
6265   uint64_t load_address;     // UINT64_MAX if unknown.
6266   uint64_t seg_addrs_offset; // offset to the array of struct segment_vmaddr's.
6267   uint32_t segment_count;    // The number of segments for this binary.
6268   uint32_t unused;
6269 
image_entryimage_entry6270   image_entry() {
6271     filepath_offset = UINT64_MAX;
6272     memset(&uuid, 0, sizeof(uuid_t));
6273     segment_count = 0;
6274     load_address = UINT64_MAX;
6275     seg_addrs_offset = UINT64_MAX;
6276     unused = 0;
6277   }
image_entryimage_entry6278   image_entry(const image_entry &rhs) {
6279     filepath_offset = rhs.filepath_offset;
6280     memcpy(&uuid, &rhs.uuid, sizeof(uuid_t));
6281     segment_count = rhs.segment_count;
6282     seg_addrs_offset = rhs.seg_addrs_offset;
6283     load_address = rhs.load_address;
6284     unused = rhs.unused;
6285   }
6286 };
6287 
6288 struct segment_vmaddr {
6289   char segname[16];
6290   uint64_t vmaddr;
6291   uint64_t unused;
6292 
segment_vmaddrsegment_vmaddr6293   segment_vmaddr() {
6294     memset(&segname, 0, 16);
6295     vmaddr = UINT64_MAX;
6296     unused = 0;
6297   }
segment_vmaddrsegment_vmaddr6298   segment_vmaddr(const segment_vmaddr &rhs) {
6299     memcpy(&segname, &rhs.segname, 16);
6300     vmaddr = rhs.vmaddr;
6301     unused = rhs.unused;
6302   }
6303 };
6304 
6305 // Write the payload for the "all image infos" LC_NOTE into
6306 // the supplied all_image_infos_payload, assuming that this
6307 // will be written into the corefile starting at
6308 // initial_file_offset.
6309 //
6310 // The placement of this payload is a little tricky.  We're
6311 // laying this out as
6312 //
6313 // 1. header (struct all_image_info_header)
6314 // 2. Array of fixed-size (struct image_entry)'s, one
6315 //    per binary image present in the process.
6316 // 3. Arrays of (struct segment_vmaddr)'s, a varying number
6317 //    for each binary image.
6318 // 4. Variable length c-strings of binary image filepaths,
6319 //    one per binary.
6320 //
6321 // To compute where everything will be laid out in the
6322 // payload, we need to iterate over the images and calculate
6323 // how many segment_vmaddr structures each image will need,
6324 // and how long each image's filepath c-string is. There
6325 // are some multiple passes over the image list while calculating
6326 // everything.
6327 
6328 static offset_t
CreateAllImageInfosPayload(const lldb::ProcessSP & process_sp,offset_t initial_file_offset,StreamString & all_image_infos_payload)6329 CreateAllImageInfosPayload(const lldb::ProcessSP &process_sp,
6330                            offset_t initial_file_offset,
6331                            StreamString &all_image_infos_payload) {
6332   Target &target = process_sp->GetTarget();
6333   const ModuleList &modules = target.GetImages();
6334   size_t modules_count = modules.GetSize();
6335 
6336   std::set<std::string> executing_uuids;
6337   ThreadList &thread_list(process_sp->GetThreadList());
6338   for (uint32_t i = 0; i < thread_list.GetSize(); i++) {
6339     ThreadSP thread_sp = thread_list.GetThreadAtIndex(i);
6340     uint32_t stack_frame_count = thread_sp->GetStackFrameCount();
6341     for (uint32_t j = 0; j < stack_frame_count; j++) {
6342       StackFrameSP stack_frame_sp = thread_sp->GetStackFrameAtIndex(j);
6343       Address pc = stack_frame_sp->GetFrameCodeAddress();
6344       ModuleSP module_sp = pc.GetModule();
6345       if (module_sp) {
6346         UUID uuid = module_sp->GetUUID();
6347         if (uuid.IsValid()) {
6348           executing_uuids.insert(uuid.GetAsString());
6349         }
6350       }
6351     }
6352   }
6353 
6354   struct all_image_infos_header infos;
6355   infos.version = 1;
6356   infos.imgcount = modules_count;
6357   infos.entries_size = sizeof(image_entry);
6358   infos.entries_fileoff = initial_file_offset + sizeof(all_image_infos_header);
6359   infos.unused = 0;
6360 
6361   all_image_infos_payload.PutHex32(infos.version);
6362   all_image_infos_payload.PutHex32(infos.imgcount);
6363   all_image_infos_payload.PutHex64(infos.entries_fileoff);
6364   all_image_infos_payload.PutHex32(infos.entries_size);
6365   all_image_infos_payload.PutHex32(infos.unused);
6366 
6367   // First create the structures for all of the segment name+vmaddr vectors
6368   // for each module, so we will know the size of them as we add the
6369   // module entries.
6370   std::vector<std::vector<segment_vmaddr>> modules_segment_vmaddrs;
6371   for (size_t i = 0; i < modules_count; i++) {
6372     ModuleSP module = modules.GetModuleAtIndex(i);
6373 
6374     SectionList *sections = module->GetSectionList();
6375     size_t sections_count = sections->GetSize();
6376     std::vector<segment_vmaddr> segment_vmaddrs;
6377     for (size_t j = 0; j < sections_count; j++) {
6378       SectionSP section = sections->GetSectionAtIndex(j);
6379       if (!section->GetParent().get()) {
6380         addr_t vmaddr = section->GetLoadBaseAddress(&target);
6381         if (vmaddr == LLDB_INVALID_ADDRESS)
6382           continue;
6383         ConstString name = section->GetName();
6384         segment_vmaddr seg_vmaddr;
6385         strncpy(seg_vmaddr.segname, name.AsCString(),
6386                 sizeof(seg_vmaddr.segname));
6387         seg_vmaddr.vmaddr = vmaddr;
6388         seg_vmaddr.unused = 0;
6389         segment_vmaddrs.push_back(seg_vmaddr);
6390       }
6391     }
6392     modules_segment_vmaddrs.push_back(segment_vmaddrs);
6393   }
6394 
6395   offset_t size_of_vmaddr_structs = 0;
6396   for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) {
6397     size_of_vmaddr_structs +=
6398         modules_segment_vmaddrs[i].size() * sizeof(segment_vmaddr);
6399   }
6400 
6401   offset_t size_of_filepath_cstrings = 0;
6402   for (size_t i = 0; i < modules_count; i++) {
6403     ModuleSP module_sp = modules.GetModuleAtIndex(i);
6404     size_of_filepath_cstrings += module_sp->GetFileSpec().GetPath().size() + 1;
6405   }
6406 
6407   // Calculate the file offsets of our "all image infos" payload in the
6408   // corefile. initial_file_offset the original value passed in to this method.
6409 
6410   offset_t start_of_entries =
6411       initial_file_offset + sizeof(all_image_infos_header);
6412   offset_t start_of_seg_vmaddrs =
6413       start_of_entries + sizeof(image_entry) * modules_count;
6414   offset_t start_of_filenames = start_of_seg_vmaddrs + size_of_vmaddr_structs;
6415 
6416   offset_t final_file_offset = start_of_filenames + size_of_filepath_cstrings;
6417 
6418   // Now write the one-per-module 'struct image_entry' into the
6419   // StringStream; keep track of where the struct segment_vmaddr
6420   // entries for each module will end up in the corefile.
6421 
6422   offset_t current_string_offset = start_of_filenames;
6423   offset_t current_segaddrs_offset = start_of_seg_vmaddrs;
6424   std::vector<struct image_entry> image_entries;
6425   for (size_t i = 0; i < modules_count; i++) {
6426     ModuleSP module_sp = modules.GetModuleAtIndex(i);
6427 
6428     struct image_entry ent;
6429     memcpy(&ent.uuid, module_sp->GetUUID().GetBytes().data(), sizeof(ent.uuid));
6430     if (modules_segment_vmaddrs[i].size() > 0) {
6431       ent.segment_count = modules_segment_vmaddrs[i].size();
6432       ent.seg_addrs_offset = current_segaddrs_offset;
6433     }
6434     ent.filepath_offset = current_string_offset;
6435     ObjectFile *objfile = module_sp->GetObjectFile();
6436     if (objfile) {
6437       Address base_addr(objfile->GetBaseAddress());
6438       if (base_addr.IsValid()) {
6439         ent.load_address = base_addr.GetLoadAddress(&target);
6440       }
6441     }
6442 
6443     all_image_infos_payload.PutHex64(ent.filepath_offset);
6444     all_image_infos_payload.PutRawBytes(ent.uuid, sizeof(ent.uuid));
6445     all_image_infos_payload.PutHex64(ent.load_address);
6446     all_image_infos_payload.PutHex64(ent.seg_addrs_offset);
6447     all_image_infos_payload.PutHex32(ent.segment_count);
6448 
6449     if (executing_uuids.find(module_sp->GetUUID().GetAsString()) !=
6450         executing_uuids.end())
6451       all_image_infos_payload.PutHex32(1);
6452     else
6453       all_image_infos_payload.PutHex32(0);
6454 
6455     current_segaddrs_offset += ent.segment_count * sizeof(segment_vmaddr);
6456     current_string_offset += module_sp->GetFileSpec().GetPath().size() + 1;
6457   }
6458 
6459   // Now write the struct segment_vmaddr entries into the StringStream.
6460 
6461   for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) {
6462     if (modules_segment_vmaddrs[i].size() == 0)
6463       continue;
6464     for (struct segment_vmaddr segvm : modules_segment_vmaddrs[i]) {
6465       all_image_infos_payload.PutRawBytes(segvm.segname, sizeof(segvm.segname));
6466       all_image_infos_payload.PutHex64(segvm.vmaddr);
6467       all_image_infos_payload.PutHex64(segvm.unused);
6468     }
6469   }
6470 
6471   for (size_t i = 0; i < modules_count; i++) {
6472     ModuleSP module_sp = modules.GetModuleAtIndex(i);
6473     std::string filepath = module_sp->GetFileSpec().GetPath();
6474     all_image_infos_payload.PutRawBytes(filepath.data(), filepath.size() + 1);
6475   }
6476 
6477   return final_file_offset;
6478 }
6479 
6480 // Temp struct used to combine contiguous memory regions with
6481 // identical permissions.
6482 struct page_object {
6483   addr_t addr;
6484   addr_t size;
6485   uint32_t prot;
6486 };
6487 
SaveCore(const lldb::ProcessSP & process_sp,const FileSpec & outfile,lldb::SaveCoreStyle & core_style,Status & error)6488 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp,
6489                                const FileSpec &outfile,
6490                                lldb::SaveCoreStyle &core_style, Status &error) {
6491   if (!process_sp)
6492     return false;
6493 
6494   // For Mach-O, we can only create full corefiles or dirty-page-only
6495   // corefiles.  The default is dirty-page-only.
6496   if (core_style != SaveCoreStyle::eSaveCoreFull) {
6497     core_style = SaveCoreStyle::eSaveCoreDirtyOnly;
6498   } else {
6499     core_style = SaveCoreStyle::eSaveCoreFull;
6500   }
6501 
6502   Target &target = process_sp->GetTarget();
6503   const ArchSpec target_arch = target.GetArchitecture();
6504   const llvm::Triple &target_triple = target_arch.GetTriple();
6505   if (target_triple.getVendor() == llvm::Triple::Apple &&
6506       (target_triple.getOS() == llvm::Triple::MacOSX ||
6507        target_triple.getOS() == llvm::Triple::IOS ||
6508        target_triple.getOS() == llvm::Triple::WatchOS ||
6509        target_triple.getOS() == llvm::Triple::TvOS)) {
6510     // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS))
6511     // {
6512     bool make_core = false;
6513     switch (target_arch.GetMachine()) {
6514     case llvm::Triple::aarch64:
6515     case llvm::Triple::aarch64_32:
6516     case llvm::Triple::arm:
6517     case llvm::Triple::thumb:
6518     case llvm::Triple::x86:
6519     case llvm::Triple::x86_64:
6520       make_core = true;
6521       break;
6522     default:
6523       error.SetErrorStringWithFormat("unsupported core architecture: %s",
6524                                      target_triple.str().c_str());
6525       break;
6526     }
6527 
6528     if (make_core) {
6529       std::vector<llvm::MachO::segment_command_64> segment_load_commands;
6530       //                uint32_t range_info_idx = 0;
6531       MemoryRegionInfo range_info;
6532       Status range_error = process_sp->GetMemoryRegionInfo(0, range_info);
6533       const uint32_t addr_byte_size = target_arch.GetAddressByteSize();
6534       const ByteOrder byte_order = target_arch.GetByteOrder();
6535       std::vector<page_object> pages_to_copy;
6536 
6537       if (range_error.Success()) {
6538         while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) {
6539           // Calculate correct protections
6540           uint32_t prot = 0;
6541           if (range_info.GetReadable() == MemoryRegionInfo::eYes)
6542             prot |= VM_PROT_READ;
6543           if (range_info.GetWritable() == MemoryRegionInfo::eYes)
6544             prot |= VM_PROT_WRITE;
6545           if (range_info.GetExecutable() == MemoryRegionInfo::eYes)
6546             prot |= VM_PROT_EXECUTE;
6547 
6548           const addr_t addr = range_info.GetRange().GetRangeBase();
6549           const addr_t size = range_info.GetRange().GetByteSize();
6550 
6551           if (size == 0)
6552             break;
6553 
6554           if (prot != 0) {
6555             addr_t pagesize = range_info.GetPageSize();
6556             const llvm::Optional<std::vector<addr_t>> &dirty_page_list =
6557                 range_info.GetDirtyPageList();
6558             if (core_style == SaveCoreStyle::eSaveCoreDirtyOnly &&
6559                 dirty_page_list.hasValue()) {
6560               core_style = SaveCoreStyle::eSaveCoreDirtyOnly;
6561               for (addr_t dirtypage : dirty_page_list.getValue()) {
6562                 page_object obj;
6563                 obj.addr = dirtypage;
6564                 obj.size = pagesize;
6565                 obj.prot = prot;
6566                 pages_to_copy.push_back(obj);
6567               }
6568             } else {
6569               page_object obj;
6570               obj.addr = addr;
6571               obj.size = size;
6572               obj.prot = prot;
6573               pages_to_copy.push_back(obj);
6574             }
6575           }
6576 
6577           range_error = process_sp->GetMemoryRegionInfo(
6578               range_info.GetRange().GetRangeEnd(), range_info);
6579           if (range_error.Fail())
6580             break;
6581         }
6582 
6583         // Combine contiguous entries that have the same
6584         // protections so we don't have an excess of
6585         // load commands.
6586         std::vector<page_object> combined_page_objects;
6587         page_object last_obj;
6588         last_obj.addr = LLDB_INVALID_ADDRESS;
6589         for (page_object obj : pages_to_copy) {
6590           if (last_obj.addr == LLDB_INVALID_ADDRESS) {
6591             last_obj = obj;
6592             continue;
6593           }
6594           if (last_obj.addr + last_obj.size == obj.addr &&
6595               last_obj.prot == obj.prot) {
6596             last_obj.size += obj.size;
6597             continue;
6598           }
6599           combined_page_objects.push_back(last_obj);
6600           last_obj = obj;
6601         }
6602 
6603         for (page_object obj : combined_page_objects) {
6604           uint32_t cmd_type = LC_SEGMENT_64;
6605           uint32_t segment_size = sizeof(llvm::MachO::segment_command_64);
6606           if (addr_byte_size == 4) {
6607             cmd_type = LC_SEGMENT;
6608             segment_size = sizeof(llvm::MachO::segment_command);
6609           }
6610           llvm::MachO::segment_command_64 segment = {
6611               cmd_type,     // uint32_t cmd;
6612               segment_size, // uint32_t cmdsize;
6613               {0},          // char segname[16];
6614               obj.addr,     // uint64_t vmaddr;    // uint32_t for 32-bit
6615                             // Mach-O
6616               obj.size,     // uint64_t vmsize;    // uint32_t for 32-bit
6617                             // Mach-O
6618               0,            // uint64_t fileoff;   // uint32_t for 32-bit Mach-O
6619               obj.size,     // uint64_t filesize;  // uint32_t for 32-bit
6620                             // Mach-O
6621               obj.prot,     // uint32_t maxprot;
6622               obj.prot,     // uint32_t initprot;
6623               0,            // uint32_t nsects;
6624               0};           // uint32_t flags;
6625           segment_load_commands.push_back(segment);
6626         }
6627 
6628         StreamString buffer(Stream::eBinary, addr_byte_size, byte_order);
6629 
6630         llvm::MachO::mach_header_64 mach_header;
6631         if (addr_byte_size == 8) {
6632           mach_header.magic = MH_MAGIC_64;
6633         } else {
6634           mach_header.magic = MH_MAGIC;
6635         }
6636         mach_header.cputype = target_arch.GetMachOCPUType();
6637         mach_header.cpusubtype = target_arch.GetMachOCPUSubType();
6638         mach_header.filetype = MH_CORE;
6639         mach_header.ncmds = segment_load_commands.size();
6640         mach_header.flags = 0;
6641         mach_header.reserved = 0;
6642         ThreadList &thread_list = process_sp->GetThreadList();
6643         const uint32_t num_threads = thread_list.GetSize();
6644 
6645         // Make an array of LC_THREAD data items. Each one contains the
6646         // contents of the LC_THREAD load command. The data doesn't contain
6647         // the load command + load command size, we will add the load command
6648         // and load command size as we emit the data.
6649         std::vector<StreamString> LC_THREAD_datas(num_threads);
6650         for (auto &LC_THREAD_data : LC_THREAD_datas) {
6651           LC_THREAD_data.GetFlags().Set(Stream::eBinary);
6652           LC_THREAD_data.SetAddressByteSize(addr_byte_size);
6653           LC_THREAD_data.SetByteOrder(byte_order);
6654         }
6655         for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) {
6656           ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx));
6657           if (thread_sp) {
6658             switch (mach_header.cputype) {
6659             case llvm::MachO::CPU_TYPE_ARM64:
6660             case llvm::MachO::CPU_TYPE_ARM64_32:
6661               RegisterContextDarwin_arm64_Mach::Create_LC_THREAD(
6662                   thread_sp.get(), LC_THREAD_datas[thread_idx]);
6663               break;
6664 
6665             case llvm::MachO::CPU_TYPE_ARM:
6666               RegisterContextDarwin_arm_Mach::Create_LC_THREAD(
6667                   thread_sp.get(), LC_THREAD_datas[thread_idx]);
6668               break;
6669 
6670             case llvm::MachO::CPU_TYPE_I386:
6671               RegisterContextDarwin_i386_Mach::Create_LC_THREAD(
6672                   thread_sp.get(), LC_THREAD_datas[thread_idx]);
6673               break;
6674 
6675             case llvm::MachO::CPU_TYPE_X86_64:
6676               RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD(
6677                   thread_sp.get(), LC_THREAD_datas[thread_idx]);
6678               break;
6679             }
6680           }
6681         }
6682 
6683         // The size of the load command is the size of the segments...
6684         if (addr_byte_size == 8) {
6685           mach_header.sizeofcmds = segment_load_commands.size() *
6686                                    sizeof(llvm::MachO::segment_command_64);
6687         } else {
6688           mach_header.sizeofcmds = segment_load_commands.size() *
6689                                    sizeof(llvm::MachO::segment_command);
6690         }
6691 
6692         // and the size of all LC_THREAD load command
6693         for (const auto &LC_THREAD_data : LC_THREAD_datas) {
6694           ++mach_header.ncmds;
6695           mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize();
6696         }
6697 
6698         // Bits will be set to indicate which bits are NOT used in
6699         // addressing in this process or 0 for unknown.
6700         uint64_t address_mask = process_sp->GetCodeAddressMask();
6701         if (address_mask != 0) {
6702           // LC_NOTE "addrable bits"
6703           mach_header.ncmds++;
6704           mach_header.sizeofcmds += sizeof(llvm::MachO::note_command);
6705         }
6706 
6707         // LC_NOTE "all image infos"
6708         mach_header.ncmds++;
6709         mach_header.sizeofcmds += sizeof(llvm::MachO::note_command);
6710 
6711         // Write the mach header
6712         buffer.PutHex32(mach_header.magic);
6713         buffer.PutHex32(mach_header.cputype);
6714         buffer.PutHex32(mach_header.cpusubtype);
6715         buffer.PutHex32(mach_header.filetype);
6716         buffer.PutHex32(mach_header.ncmds);
6717         buffer.PutHex32(mach_header.sizeofcmds);
6718         buffer.PutHex32(mach_header.flags);
6719         if (addr_byte_size == 8) {
6720           buffer.PutHex32(mach_header.reserved);
6721         }
6722 
6723         // Skip the mach header and all load commands and align to the next
6724         // 0x1000 byte boundary
6725         addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds;
6726 
6727         file_offset = llvm::alignTo(file_offset, 16);
6728         std::vector<std::unique_ptr<LCNoteEntry>> lc_notes;
6729 
6730         // Add "addrable bits" LC_NOTE when an address mask is available
6731         if (address_mask != 0) {
6732           std::unique_ptr<LCNoteEntry> addrable_bits_lcnote_up(
6733               new LCNoteEntry(addr_byte_size, byte_order));
6734           addrable_bits_lcnote_up->name = "addrable bits";
6735           addrable_bits_lcnote_up->payload_file_offset = file_offset;
6736           int bits = std::bitset<64>(~address_mask).count();
6737           addrable_bits_lcnote_up->payload.PutHex32(3); // version
6738           addrable_bits_lcnote_up->payload.PutHex32(
6739               bits); // # of bits used for addressing
6740           addrable_bits_lcnote_up->payload.PutHex64(0); // unused
6741 
6742           file_offset += addrable_bits_lcnote_up->payload.GetSize();
6743 
6744           lc_notes.push_back(std::move(addrable_bits_lcnote_up));
6745         }
6746 
6747         // Add "all image infos" LC_NOTE
6748         std::unique_ptr<LCNoteEntry> all_image_infos_lcnote_up(
6749             new LCNoteEntry(addr_byte_size, byte_order));
6750         all_image_infos_lcnote_up->name = "all image infos";
6751         all_image_infos_lcnote_up->payload_file_offset = file_offset;
6752         file_offset = CreateAllImageInfosPayload(
6753             process_sp, file_offset, all_image_infos_lcnote_up->payload);
6754         lc_notes.push_back(std::move(all_image_infos_lcnote_up));
6755 
6756         // Add LC_NOTE load commands
6757         for (auto &lcnote : lc_notes) {
6758           // Add the LC_NOTE load command to the file.
6759           buffer.PutHex32(LC_NOTE);
6760           buffer.PutHex32(sizeof(llvm::MachO::note_command));
6761           char namebuf[16];
6762           memset(namebuf, 0, sizeof(namebuf));
6763           // this is the uncommon case where strncpy is exactly
6764           // the right one, doesn't need to be nul terminated.
6765           strncpy(namebuf, lcnote->name.c_str(), sizeof(namebuf));
6766           buffer.PutRawBytes(namebuf, sizeof(namebuf));
6767           buffer.PutHex64(lcnote->payload_file_offset);
6768           buffer.PutHex64(lcnote->payload.GetSize());
6769         }
6770 
6771         // Align to 4096-byte page boundary for the LC_SEGMENTs.
6772         file_offset = llvm::alignTo(file_offset, 4096);
6773 
6774         for (auto &segment : segment_load_commands) {
6775           segment.fileoff = file_offset;
6776           file_offset += segment.filesize;
6777         }
6778 
6779         // Write out all of the LC_THREAD load commands
6780         for (const auto &LC_THREAD_data : LC_THREAD_datas) {
6781           const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize();
6782           buffer.PutHex32(LC_THREAD);
6783           buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data
6784           buffer.Write(LC_THREAD_data.GetString().data(), LC_THREAD_data_size);
6785         }
6786 
6787         // Write out all of the segment load commands
6788         for (const auto &segment : segment_load_commands) {
6789           buffer.PutHex32(segment.cmd);
6790           buffer.PutHex32(segment.cmdsize);
6791           buffer.PutRawBytes(segment.segname, sizeof(segment.segname));
6792           if (addr_byte_size == 8) {
6793             buffer.PutHex64(segment.vmaddr);
6794             buffer.PutHex64(segment.vmsize);
6795             buffer.PutHex64(segment.fileoff);
6796             buffer.PutHex64(segment.filesize);
6797           } else {
6798             buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr));
6799             buffer.PutHex32(static_cast<uint32_t>(segment.vmsize));
6800             buffer.PutHex32(static_cast<uint32_t>(segment.fileoff));
6801             buffer.PutHex32(static_cast<uint32_t>(segment.filesize));
6802           }
6803           buffer.PutHex32(segment.maxprot);
6804           buffer.PutHex32(segment.initprot);
6805           buffer.PutHex32(segment.nsects);
6806           buffer.PutHex32(segment.flags);
6807         }
6808 
6809         std::string core_file_path(outfile.GetPath());
6810         auto core_file = FileSystem::Instance().Open(
6811             outfile, File::eOpenOptionWrite | File::eOpenOptionTruncate |
6812                          File::eOpenOptionCanCreate);
6813         if (!core_file) {
6814           error = core_file.takeError();
6815         } else {
6816           // Read 1 page at a time
6817           uint8_t bytes[0x1000];
6818           // Write the mach header and load commands out to the core file
6819           size_t bytes_written = buffer.GetString().size();
6820           error =
6821               core_file.get()->Write(buffer.GetString().data(), bytes_written);
6822           if (error.Success()) {
6823 
6824             for (auto &lcnote : lc_notes) {
6825               if (core_file.get()->SeekFromStart(lcnote->payload_file_offset) ==
6826                   -1) {
6827                 error.SetErrorStringWithFormat("Unable to seek to corefile pos "
6828                                                "to write '%s' LC_NOTE payload",
6829                                                lcnote->name.c_str());
6830                 return false;
6831               }
6832               bytes_written = lcnote->payload.GetSize();
6833               error = core_file.get()->Write(lcnote->payload.GetData(),
6834                                              bytes_written);
6835               if (!error.Success())
6836                 return false;
6837             }
6838 
6839             // Now write the file data for all memory segments in the process
6840             for (const auto &segment : segment_load_commands) {
6841               if (core_file.get()->SeekFromStart(segment.fileoff) == -1) {
6842                 error.SetErrorStringWithFormat(
6843                     "unable to seek to offset 0x%" PRIx64 " in '%s'",
6844                     segment.fileoff, core_file_path.c_str());
6845                 break;
6846               }
6847 
6848               target.GetDebugger().GetAsyncOutputStream()->Printf(
6849                   "Saving %" PRId64
6850                   " bytes of data for memory region at 0x%" PRIx64 "\n",
6851                   segment.vmsize, segment.vmaddr);
6852               addr_t bytes_left = segment.vmsize;
6853               addr_t addr = segment.vmaddr;
6854               Status memory_read_error;
6855               while (bytes_left > 0 && error.Success()) {
6856                 const size_t bytes_to_read =
6857                     bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left;
6858 
6859                 // In a savecore setting, we don't really care about caching,
6860                 // as the data is dumped and very likely never read again,
6861                 // so we call ReadMemoryFromInferior to bypass it.
6862                 const size_t bytes_read = process_sp->ReadMemoryFromInferior(
6863                     addr, bytes, bytes_to_read, memory_read_error);
6864 
6865                 if (bytes_read == bytes_to_read) {
6866                   size_t bytes_written = bytes_read;
6867                   error = core_file.get()->Write(bytes, bytes_written);
6868                   bytes_left -= bytes_read;
6869                   addr += bytes_read;
6870                 } else {
6871                   // Some pages within regions are not readable, those should
6872                   // be zero filled
6873                   memset(bytes, 0, bytes_to_read);
6874                   size_t bytes_written = bytes_to_read;
6875                   error = core_file.get()->Write(bytes, bytes_written);
6876                   bytes_left -= bytes_to_read;
6877                   addr += bytes_to_read;
6878                 }
6879               }
6880             }
6881           }
6882         }
6883       } else {
6884         error.SetErrorString(
6885             "process doesn't support getting memory region info");
6886       }
6887     }
6888     return true; // This is the right plug to handle saving core files for
6889                  // this process
6890   }
6891   return false;
6892 }
6893 
6894 ObjectFileMachO::MachOCorefileAllImageInfos
GetCorefileAllImageInfos()6895 ObjectFileMachO::GetCorefileAllImageInfos() {
6896   MachOCorefileAllImageInfos image_infos;
6897 
6898   // Look for an "all image infos" LC_NOTE.
6899   lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
6900   for (uint32_t i = 0; i < m_header.ncmds; ++i) {
6901     const uint32_t cmd_offset = offset;
6902     llvm::MachO::load_command lc;
6903     if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr)
6904       break;
6905     if (lc.cmd == LC_NOTE) {
6906       char data_owner[17];
6907       m_data.CopyData(offset, 16, data_owner);
6908       data_owner[16] = '\0';
6909       offset += 16;
6910       uint64_t fileoff = m_data.GetU64_unchecked(&offset);
6911       offset += 4; /* size unused */
6912 
6913       if (strcmp("all image infos", data_owner) == 0) {
6914         offset = fileoff;
6915         // Read the struct all_image_infos_header.
6916         uint32_t version = m_data.GetU32(&offset);
6917         if (version != 1) {
6918           return image_infos;
6919         }
6920         uint32_t imgcount = m_data.GetU32(&offset);
6921         uint64_t entries_fileoff = m_data.GetU64(&offset);
6922         offset += 4; // uint32_t entries_size;
6923         offset += 4; // uint32_t unused;
6924 
6925         offset = entries_fileoff;
6926         for (uint32_t i = 0; i < imgcount; i++) {
6927           // Read the struct image_entry.
6928           offset_t filepath_offset = m_data.GetU64(&offset);
6929           uuid_t uuid;
6930           memcpy(&uuid, m_data.GetData(&offset, sizeof(uuid_t)),
6931                  sizeof(uuid_t));
6932           uint64_t load_address = m_data.GetU64(&offset);
6933           offset_t seg_addrs_offset = m_data.GetU64(&offset);
6934           uint32_t segment_count = m_data.GetU32(&offset);
6935           uint32_t currently_executing = m_data.GetU32(&offset);
6936 
6937           MachOCorefileImageEntry image_entry;
6938           image_entry.filename = (const char *)m_data.GetCStr(&filepath_offset);
6939           image_entry.uuid = UUID::fromData(uuid, sizeof(uuid_t));
6940           image_entry.load_address = load_address;
6941           image_entry.currently_executing = currently_executing;
6942 
6943           offset_t seg_vmaddrs_offset = seg_addrs_offset;
6944           for (uint32_t j = 0; j < segment_count; j++) {
6945             char segname[17];
6946             m_data.CopyData(seg_vmaddrs_offset, 16, segname);
6947             segname[16] = '\0';
6948             seg_vmaddrs_offset += 16;
6949             uint64_t vmaddr = m_data.GetU64(&seg_vmaddrs_offset);
6950             seg_vmaddrs_offset += 8; /* unused */
6951 
6952             std::tuple<ConstString, addr_t> new_seg{ConstString(segname),
6953                                                     vmaddr};
6954             image_entry.segment_load_addresses.push_back(new_seg);
6955           }
6956           image_infos.all_image_infos.push_back(image_entry);
6957         }
6958       }
6959     }
6960     offset = cmd_offset + lc.cmdsize;
6961   }
6962 
6963   return image_infos;
6964 }
6965 
LoadCoreFileImages(lldb_private::Process & process)6966 bool ObjectFileMachO::LoadCoreFileImages(lldb_private::Process &process) {
6967   MachOCorefileAllImageInfos image_infos = GetCorefileAllImageInfos();
6968   bool added_images = false;
6969   if (image_infos.IsValid()) {
6970     for (const MachOCorefileImageEntry &image : image_infos.all_image_infos) {
6971       ModuleSpec module_spec;
6972       module_spec.GetUUID() = image.uuid;
6973       module_spec.GetFileSpec() = FileSpec(image.filename.c_str());
6974       if (image.currently_executing) {
6975         Symbols::DownloadObjectAndSymbolFile(module_spec, true);
6976         if (FileSystem::Instance().Exists(module_spec.GetFileSpec())) {
6977           process.GetTarget().GetOrCreateModule(module_spec, false);
6978         }
6979       }
6980       Status error;
6981       ModuleSP module_sp =
6982           process.GetTarget().GetOrCreateModule(module_spec, false, &error);
6983       if (!module_sp.get() || !module_sp->GetObjectFile()) {
6984         if (image.load_address != LLDB_INVALID_ADDRESS) {
6985           module_sp = process.ReadModuleFromMemory(module_spec.GetFileSpec(),
6986                                                    image.load_address);
6987         }
6988       }
6989       if (module_sp.get() && module_sp->GetObjectFile()) {
6990         added_images = true;
6991         if (module_sp->GetObjectFile()->GetType() ==
6992             ObjectFile::eTypeExecutable) {
6993           process.GetTarget().SetExecutableModule(module_sp, eLoadDependentsNo);
6994         }
6995         for (auto name_vmaddr_tuple : image.segment_load_addresses) {
6996           SectionList *sectlist = module_sp->GetObjectFile()->GetSectionList();
6997           if (sectlist) {
6998             SectionSP sect_sp =
6999                 sectlist->FindSectionByName(std::get<0>(name_vmaddr_tuple));
7000             if (sect_sp) {
7001               process.GetTarget().SetSectionLoadAddress(
7002                   sect_sp, std::get<1>(name_vmaddr_tuple));
7003             }
7004           }
7005         }
7006       }
7007     }
7008   }
7009   return added_images;
7010 }
7011