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/ScopeExit.h"
10 #include "llvm/ADT/StringRef.h"
11
12 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h"
13 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h"
14 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h"
15 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h"
16 #include "lldb/Core/Debugger.h"
17 #include "lldb/Core/FileSpecList.h"
18 #include "lldb/Core/Module.h"
19 #include "lldb/Core/ModuleSpec.h"
20 #include "lldb/Core/PluginManager.h"
21 #include "lldb/Core/Progress.h"
22 #include "lldb/Core/Section.h"
23 #include "lldb/Core/StreamFile.h"
24 #include "lldb/Host/Host.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/LLDBLog.h"
40 #include "lldb/Utility/Log.h"
41 #include "lldb/Utility/RangeMap.h"
42 #include "lldb/Utility/RegisterValue.h"
43 #include "lldb/Utility/Status.h"
44 #include "lldb/Utility/StreamString.h"
45 #include "lldb/Utility/Timer.h"
46 #include "lldb/Utility/UUID.h"
47
48 #include "lldb/Host/SafeMachO.h"
49
50 #include "llvm/ADT/DenseSet.h"
51 #include "llvm/Support/FormatVariadic.h"
52 #include "llvm/Support/MemoryBuffer.h"
53
54 #include "ObjectFileMachO.h"
55
56 #if defined(__APPLE__)
57 #include <TargetConditionals.h>
58 // GetLLDBSharedCacheUUID() needs to call dlsym()
59 #include <dlfcn.h>
60 #include <mach/mach_init.h>
61 #include <mach/vm_map.h>
62 #include <lldb/Host/SafeMachO.h>
63 #endif
64
65 #ifndef __APPLE__
66 #include "Utility/UuidCompatibility.h"
67 #else
68 #include <uuid/uuid.h>
69 #endif
70
71 #include <bitset>
72 #include <memory>
73
74 // Unfortunately the signpost header pulls in the system MachO header, too.
75 #ifdef CPU_TYPE_ARM
76 #undef CPU_TYPE_ARM
77 #endif
78 #ifdef CPU_TYPE_ARM64
79 #undef CPU_TYPE_ARM64
80 #endif
81 #ifdef CPU_TYPE_ARM64_32
82 #undef CPU_TYPE_ARM64_32
83 #endif
84 #ifdef CPU_TYPE_I386
85 #undef CPU_TYPE_I386
86 #endif
87 #ifdef CPU_TYPE_X86_64
88 #undef CPU_TYPE_X86_64
89 #endif
90 #ifdef MH_DYLINKER
91 #undef MH_DYLINKER
92 #endif
93 #ifdef MH_OBJECT
94 #undef MH_OBJECT
95 #endif
96 #ifdef LC_VERSION_MIN_MACOSX
97 #undef LC_VERSION_MIN_MACOSX
98 #endif
99 #ifdef LC_VERSION_MIN_IPHONEOS
100 #undef LC_VERSION_MIN_IPHONEOS
101 #endif
102 #ifdef LC_VERSION_MIN_TVOS
103 #undef LC_VERSION_MIN_TVOS
104 #endif
105 #ifdef LC_VERSION_MIN_WATCHOS
106 #undef LC_VERSION_MIN_WATCHOS
107 #endif
108 #ifdef LC_BUILD_VERSION
109 #undef LC_BUILD_VERSION
110 #endif
111 #ifdef PLATFORM_MACOS
112 #undef PLATFORM_MACOS
113 #endif
114 #ifdef PLATFORM_MACCATALYST
115 #undef PLATFORM_MACCATALYST
116 #endif
117 #ifdef PLATFORM_IOS
118 #undef PLATFORM_IOS
119 #endif
120 #ifdef PLATFORM_IOSSIMULATOR
121 #undef PLATFORM_IOSSIMULATOR
122 #endif
123 #ifdef PLATFORM_TVOS
124 #undef PLATFORM_TVOS
125 #endif
126 #ifdef PLATFORM_TVOSSIMULATOR
127 #undef PLATFORM_TVOSSIMULATOR
128 #endif
129 #ifdef PLATFORM_WATCHOS
130 #undef PLATFORM_WATCHOS
131 #endif
132 #ifdef PLATFORM_WATCHOSSIMULATOR
133 #undef PLATFORM_WATCHOSSIMULATOR
134 #endif
135
136 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull
137 using namespace lldb;
138 using namespace lldb_private;
139 using namespace llvm::MachO;
140
141 LLDB_PLUGIN_DEFINE(ObjectFileMachO)
142
143 // Some structure definitions needed for parsing the dyld shared cache files
144 // found on iOS devices.
145
146 struct lldb_copy_dyld_cache_header_v1 {
147 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc.
148 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info
149 uint32_t mappingCount; // number of dyld_cache_mapping_info entries
150 uint32_t imagesOffset;
151 uint32_t imagesCount;
152 uint64_t dyldBaseAddress;
153 uint64_t codeSignatureOffset;
154 uint64_t codeSignatureSize;
155 uint64_t slideInfoOffset;
156 uint64_t slideInfoSize;
157 uint64_t localSymbolsOffset;
158 uint64_t localSymbolsSize;
159 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13
160 // and later
161 };
162
PrintRegisterValue(RegisterContext * reg_ctx,const char * name,const char * alt_name,size_t reg_byte_size,Stream & data)163 static void PrintRegisterValue(RegisterContext *reg_ctx, const char *name,
164 const char *alt_name, size_t reg_byte_size,
165 Stream &data) {
166 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name);
167 if (reg_info == nullptr)
168 reg_info = reg_ctx->GetRegisterInfoByName(alt_name);
169 if (reg_info) {
170 lldb_private::RegisterValue reg_value;
171 if (reg_ctx->ReadRegister(reg_info, reg_value)) {
172 if (reg_info->byte_size >= reg_byte_size)
173 data.Write(reg_value.GetBytes(), reg_byte_size);
174 else {
175 data.Write(reg_value.GetBytes(), reg_info->byte_size);
176 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; ++i)
177 data.PutChar(0);
178 }
179 return;
180 }
181 }
182 // Just write zeros if all else fails
183 for (size_t i = 0; i < reg_byte_size; ++i)
184 data.PutChar(0);
185 }
186
187 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 {
188 public:
RegisterContextDarwin_x86_64_Mach(lldb_private::Thread & thread,const DataExtractor & data)189 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread,
190 const DataExtractor &data)
191 : RegisterContextDarwin_x86_64(thread, 0) {
192 SetRegisterDataFrom_LC_THREAD(data);
193 }
194
InvalidateAllRegisters()195 void InvalidateAllRegisters() override {
196 // Do nothing... registers are always valid...
197 }
198
SetRegisterDataFrom_LC_THREAD(const DataExtractor & data)199 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) {
200 lldb::offset_t offset = 0;
201 SetError(GPRRegSet, Read, -1);
202 SetError(FPURegSet, Read, -1);
203 SetError(EXCRegSet, Read, -1);
204 bool done = false;
205
206 while (!done) {
207 int flavor = data.GetU32(&offset);
208 if (flavor == 0)
209 done = true;
210 else {
211 uint32_t i;
212 uint32_t count = data.GetU32(&offset);
213 switch (flavor) {
214 case GPRRegSet:
215 for (i = 0; i < count; ++i)
216 (&gpr.rax)[i] = data.GetU64(&offset);
217 SetError(GPRRegSet, Read, 0);
218 done = true;
219
220 break;
221 case FPURegSet:
222 // TODO: fill in FPU regs....
223 // SetError (FPURegSet, Read, -1);
224 done = true;
225
226 break;
227 case EXCRegSet:
228 exc.trapno = data.GetU32(&offset);
229 exc.err = data.GetU32(&offset);
230 exc.faultvaddr = data.GetU64(&offset);
231 SetError(EXCRegSet, Read, 0);
232 done = true;
233 break;
234 case 7:
235 case 8:
236 case 9:
237 // fancy flavors that encapsulate of the above flavors...
238 break;
239
240 default:
241 done = true;
242 break;
243 }
244 }
245 }
246 }
247
Create_LC_THREAD(Thread * thread,Stream & data)248 static bool Create_LC_THREAD(Thread *thread, Stream &data) {
249 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext());
250 if (reg_ctx_sp) {
251 RegisterContext *reg_ctx = reg_ctx_sp.get();
252
253 data.PutHex32(GPRRegSet); // Flavor
254 data.PutHex32(GPRWordCount);
255 PrintRegisterValue(reg_ctx, "rax", nullptr, 8, data);
256 PrintRegisterValue(reg_ctx, "rbx", nullptr, 8, data);
257 PrintRegisterValue(reg_ctx, "rcx", nullptr, 8, data);
258 PrintRegisterValue(reg_ctx, "rdx", nullptr, 8, data);
259 PrintRegisterValue(reg_ctx, "rdi", nullptr, 8, data);
260 PrintRegisterValue(reg_ctx, "rsi", nullptr, 8, data);
261 PrintRegisterValue(reg_ctx, "rbp", nullptr, 8, data);
262 PrintRegisterValue(reg_ctx, "rsp", nullptr, 8, data);
263 PrintRegisterValue(reg_ctx, "r8", nullptr, 8, data);
264 PrintRegisterValue(reg_ctx, "r9", nullptr, 8, data);
265 PrintRegisterValue(reg_ctx, "r10", nullptr, 8, data);
266 PrintRegisterValue(reg_ctx, "r11", nullptr, 8, data);
267 PrintRegisterValue(reg_ctx, "r12", nullptr, 8, data);
268 PrintRegisterValue(reg_ctx, "r13", nullptr, 8, data);
269 PrintRegisterValue(reg_ctx, "r14", nullptr, 8, data);
270 PrintRegisterValue(reg_ctx, "r15", nullptr, 8, data);
271 PrintRegisterValue(reg_ctx, "rip", nullptr, 8, data);
272 PrintRegisterValue(reg_ctx, "rflags", nullptr, 8, data);
273 PrintRegisterValue(reg_ctx, "cs", nullptr, 8, data);
274 PrintRegisterValue(reg_ctx, "fs", nullptr, 8, data);
275 PrintRegisterValue(reg_ctx, "gs", nullptr, 8, data);
276
277 // // Write out the FPU registers
278 // const size_t fpu_byte_size = sizeof(FPU);
279 // size_t bytes_written = 0;
280 // data.PutHex32 (FPURegSet);
281 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t));
282 // bytes_written += data.PutHex32(0); // uint32_t pad[0]
283 // bytes_written += data.PutHex32(0); // uint32_t pad[1]
284 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2,
285 // data); // uint16_t fcw; // "fctrl"
286 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2,
287 // data); // uint16_t fsw; // "fstat"
288 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1,
289 // data); // uint8_t ftw; // "ftag"
290 // bytes_written += data.PutHex8 (0); // uint8_t pad1;
291 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2,
292 // data); // uint16_t fop; // "fop"
293 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4,
294 // data); // uint32_t ip; // "fioff"
295 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2,
296 // data); // uint16_t cs; // "fiseg"
297 // bytes_written += data.PutHex16 (0); // uint16_t pad2;
298 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4,
299 // data); // uint32_t dp; // "fooff"
300 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2,
301 // data); // uint16_t ds; // "foseg"
302 // bytes_written += data.PutHex16 (0); // uint16_t pad3;
303 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4,
304 // data); // uint32_t mxcsr;
305 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL,
306 // 4, data);// uint32_t mxcsrmask;
307 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL,
308 // sizeof(MMSReg), data);
309 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL,
310 // sizeof(MMSReg), data);
311 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL,
312 // sizeof(MMSReg), data);
313 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL,
314 // sizeof(MMSReg), data);
315 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL,
316 // sizeof(MMSReg), data);
317 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL,
318 // sizeof(MMSReg), data);
319 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL,
320 // sizeof(MMSReg), data);
321 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL,
322 // sizeof(MMSReg), data);
323 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL,
324 // sizeof(XMMReg), data);
325 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL,
326 // sizeof(XMMReg), data);
327 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL,
328 // sizeof(XMMReg), data);
329 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL,
330 // sizeof(XMMReg), data);
331 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL,
332 // sizeof(XMMReg), data);
333 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL,
334 // sizeof(XMMReg), data);
335 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL,
336 // sizeof(XMMReg), data);
337 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL,
338 // sizeof(XMMReg), data);
339 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL,
340 // sizeof(XMMReg), data);
341 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL,
342 // sizeof(XMMReg), data);
343 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL,
344 // sizeof(XMMReg), data);
345 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL,
346 // sizeof(XMMReg), data);
347 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL,
348 // sizeof(XMMReg), data);
349 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL,
350 // sizeof(XMMReg), data);
351 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL,
352 // sizeof(XMMReg), data);
353 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL,
354 // sizeof(XMMReg), data);
355 //
356 // // Fill rest with zeros
357 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++
358 // i)
359 // data.PutChar(0);
360
361 // Write out the EXC registers
362 data.PutHex32(EXCRegSet);
363 data.PutHex32(EXCWordCount);
364 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data);
365 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data);
366 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 8, data);
367 return true;
368 }
369 return false;
370 }
371
372 protected:
DoReadGPR(lldb::tid_t tid,int flavor,GPR & gpr)373 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; }
374
DoReadFPU(lldb::tid_t tid,int flavor,FPU & fpu)375 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; }
376
DoReadEXC(lldb::tid_t tid,int flavor,EXC & exc)377 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; }
378
DoWriteGPR(lldb::tid_t tid,int flavor,const GPR & gpr)379 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override {
380 return 0;
381 }
382
DoWriteFPU(lldb::tid_t tid,int flavor,const FPU & fpu)383 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override {
384 return 0;
385 }
386
DoWriteEXC(lldb::tid_t tid,int flavor,const EXC & exc)387 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override {
388 return 0;
389 }
390 };
391
392 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 {
393 public:
RegisterContextDarwin_i386_Mach(lldb_private::Thread & thread,const DataExtractor & data)394 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread,
395 const DataExtractor &data)
396 : RegisterContextDarwin_i386(thread, 0) {
397 SetRegisterDataFrom_LC_THREAD(data);
398 }
399
InvalidateAllRegisters()400 void InvalidateAllRegisters() override {
401 // Do nothing... registers are always valid...
402 }
403
SetRegisterDataFrom_LC_THREAD(const DataExtractor & data)404 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) {
405 lldb::offset_t offset = 0;
406 SetError(GPRRegSet, Read, -1);
407 SetError(FPURegSet, Read, -1);
408 SetError(EXCRegSet, Read, -1);
409 bool done = false;
410
411 while (!done) {
412 int flavor = data.GetU32(&offset);
413 if (flavor == 0)
414 done = true;
415 else {
416 uint32_t i;
417 uint32_t count = data.GetU32(&offset);
418 switch (flavor) {
419 case GPRRegSet:
420 for (i = 0; i < count; ++i)
421 (&gpr.eax)[i] = data.GetU32(&offset);
422 SetError(GPRRegSet, Read, 0);
423 done = true;
424
425 break;
426 case FPURegSet:
427 // TODO: fill in FPU regs....
428 // SetError (FPURegSet, Read, -1);
429 done = true;
430
431 break;
432 case EXCRegSet:
433 exc.trapno = data.GetU32(&offset);
434 exc.err = data.GetU32(&offset);
435 exc.faultvaddr = data.GetU32(&offset);
436 SetError(EXCRegSet, Read, 0);
437 done = true;
438 break;
439 case 7:
440 case 8:
441 case 9:
442 // fancy flavors that encapsulate of the above flavors...
443 break;
444
445 default:
446 done = true;
447 break;
448 }
449 }
450 }
451 }
452
Create_LC_THREAD(Thread * thread,Stream & data)453 static bool Create_LC_THREAD(Thread *thread, Stream &data) {
454 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext());
455 if (reg_ctx_sp) {
456 RegisterContext *reg_ctx = reg_ctx_sp.get();
457
458 data.PutHex32(GPRRegSet); // Flavor
459 data.PutHex32(GPRWordCount);
460 PrintRegisterValue(reg_ctx, "eax", nullptr, 4, data);
461 PrintRegisterValue(reg_ctx, "ebx", nullptr, 4, data);
462 PrintRegisterValue(reg_ctx, "ecx", nullptr, 4, data);
463 PrintRegisterValue(reg_ctx, "edx", nullptr, 4, data);
464 PrintRegisterValue(reg_ctx, "edi", nullptr, 4, data);
465 PrintRegisterValue(reg_ctx, "esi", nullptr, 4, data);
466 PrintRegisterValue(reg_ctx, "ebp", nullptr, 4, data);
467 PrintRegisterValue(reg_ctx, "esp", nullptr, 4, data);
468 PrintRegisterValue(reg_ctx, "ss", nullptr, 4, data);
469 PrintRegisterValue(reg_ctx, "eflags", nullptr, 4, data);
470 PrintRegisterValue(reg_ctx, "eip", nullptr, 4, data);
471 PrintRegisterValue(reg_ctx, "cs", nullptr, 4, data);
472 PrintRegisterValue(reg_ctx, "ds", nullptr, 4, data);
473 PrintRegisterValue(reg_ctx, "es", nullptr, 4, data);
474 PrintRegisterValue(reg_ctx, "fs", nullptr, 4, data);
475 PrintRegisterValue(reg_ctx, "gs", nullptr, 4, data);
476
477 // Write out the EXC registers
478 data.PutHex32(EXCRegSet);
479 data.PutHex32(EXCWordCount);
480 PrintRegisterValue(reg_ctx, "trapno", nullptr, 4, data);
481 PrintRegisterValue(reg_ctx, "err", nullptr, 4, data);
482 PrintRegisterValue(reg_ctx, "faultvaddr", nullptr, 4, data);
483 return true;
484 }
485 return false;
486 }
487
488 protected:
DoReadGPR(lldb::tid_t tid,int flavor,GPR & gpr)489 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; }
490
DoReadFPU(lldb::tid_t tid,int flavor,FPU & fpu)491 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; }
492
DoReadEXC(lldb::tid_t tid,int flavor,EXC & exc)493 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; }
494
DoWriteGPR(lldb::tid_t tid,int flavor,const GPR & gpr)495 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override {
496 return 0;
497 }
498
DoWriteFPU(lldb::tid_t tid,int flavor,const FPU & fpu)499 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override {
500 return 0;
501 }
502
DoWriteEXC(lldb::tid_t tid,int flavor,const EXC & exc)503 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override {
504 return 0;
505 }
506 };
507
508 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm {
509 public:
RegisterContextDarwin_arm_Mach(lldb_private::Thread & thread,const DataExtractor & data)510 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread,
511 const DataExtractor &data)
512 : RegisterContextDarwin_arm(thread, 0) {
513 SetRegisterDataFrom_LC_THREAD(data);
514 }
515
InvalidateAllRegisters()516 void InvalidateAllRegisters() override {
517 // Do nothing... registers are always valid...
518 }
519
SetRegisterDataFrom_LC_THREAD(const DataExtractor & data)520 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) {
521 lldb::offset_t offset = 0;
522 SetError(GPRRegSet, Read, -1);
523 SetError(FPURegSet, Read, -1);
524 SetError(EXCRegSet, Read, -1);
525 bool done = false;
526
527 while (!done) {
528 int flavor = data.GetU32(&offset);
529 uint32_t count = data.GetU32(&offset);
530 lldb::offset_t next_thread_state = offset + (count * 4);
531 switch (flavor) {
532 case GPRAltRegSet:
533 case GPRRegSet:
534 // On ARM, the CPSR register is also included in the count but it is
535 // not included in gpr.r so loop until (count-1).
536 for (uint32_t i = 0; i < (count - 1); ++i) {
537 gpr.r[i] = data.GetU32(&offset);
538 }
539 // Save cpsr explicitly.
540 gpr.cpsr = data.GetU32(&offset);
541
542 SetError(GPRRegSet, Read, 0);
543 offset = next_thread_state;
544 break;
545
546 case FPURegSet: {
547 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0];
548 const int fpu_reg_buf_size = sizeof(fpu.floats);
549 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle,
550 fpu_reg_buf) == fpu_reg_buf_size) {
551 offset += fpu_reg_buf_size;
552 fpu.fpscr = data.GetU32(&offset);
553 SetError(FPURegSet, Read, 0);
554 } else {
555 done = true;
556 }
557 }
558 offset = next_thread_state;
559 break;
560
561 case EXCRegSet:
562 if (count == 3) {
563 exc.exception = data.GetU32(&offset);
564 exc.fsr = data.GetU32(&offset);
565 exc.far = data.GetU32(&offset);
566 SetError(EXCRegSet, Read, 0);
567 }
568 done = true;
569 offset = next_thread_state;
570 break;
571
572 // Unknown register set flavor, stop trying to parse.
573 default:
574 done = true;
575 }
576 }
577 }
578
Create_LC_THREAD(Thread * thread,Stream & data)579 static bool Create_LC_THREAD(Thread *thread, Stream &data) {
580 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext());
581 if (reg_ctx_sp) {
582 RegisterContext *reg_ctx = reg_ctx_sp.get();
583
584 data.PutHex32(GPRRegSet); // Flavor
585 data.PutHex32(GPRWordCount);
586 PrintRegisterValue(reg_ctx, "r0", nullptr, 4, data);
587 PrintRegisterValue(reg_ctx, "r1", nullptr, 4, data);
588 PrintRegisterValue(reg_ctx, "r2", nullptr, 4, data);
589 PrintRegisterValue(reg_ctx, "r3", nullptr, 4, data);
590 PrintRegisterValue(reg_ctx, "r4", nullptr, 4, data);
591 PrintRegisterValue(reg_ctx, "r5", nullptr, 4, data);
592 PrintRegisterValue(reg_ctx, "r6", nullptr, 4, data);
593 PrintRegisterValue(reg_ctx, "r7", nullptr, 4, data);
594 PrintRegisterValue(reg_ctx, "r8", nullptr, 4, data);
595 PrintRegisterValue(reg_ctx, "r9", nullptr, 4, data);
596 PrintRegisterValue(reg_ctx, "r10", nullptr, 4, data);
597 PrintRegisterValue(reg_ctx, "r11", nullptr, 4, data);
598 PrintRegisterValue(reg_ctx, "r12", nullptr, 4, data);
599 PrintRegisterValue(reg_ctx, "sp", nullptr, 4, data);
600 PrintRegisterValue(reg_ctx, "lr", nullptr, 4, data);
601 PrintRegisterValue(reg_ctx, "pc", nullptr, 4, data);
602 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data);
603
604 // Write out the EXC registers
605 // data.PutHex32 (EXCRegSet);
606 // data.PutHex32 (EXCWordCount);
607 // WriteRegister (reg_ctx, "exception", NULL, 4, data);
608 // WriteRegister (reg_ctx, "fsr", NULL, 4, data);
609 // WriteRegister (reg_ctx, "far", NULL, 4, data);
610 return true;
611 }
612 return false;
613 }
614
615 protected:
DoReadGPR(lldb::tid_t tid,int flavor,GPR & gpr)616 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; }
617
DoReadFPU(lldb::tid_t tid,int flavor,FPU & fpu)618 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; }
619
DoReadEXC(lldb::tid_t tid,int flavor,EXC & exc)620 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; }
621
DoReadDBG(lldb::tid_t tid,int flavor,DBG & dbg)622 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; }
623
DoWriteGPR(lldb::tid_t tid,int flavor,const GPR & gpr)624 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override {
625 return 0;
626 }
627
DoWriteFPU(lldb::tid_t tid,int flavor,const FPU & fpu)628 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override {
629 return 0;
630 }
631
DoWriteEXC(lldb::tid_t tid,int flavor,const EXC & exc)632 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override {
633 return 0;
634 }
635
DoWriteDBG(lldb::tid_t tid,int flavor,const DBG & dbg)636 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override {
637 return -1;
638 }
639 };
640
641 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 {
642 public:
RegisterContextDarwin_arm64_Mach(lldb_private::Thread & thread,const DataExtractor & data)643 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread,
644 const DataExtractor &data)
645 : RegisterContextDarwin_arm64(thread, 0) {
646 SetRegisterDataFrom_LC_THREAD(data);
647 }
648
InvalidateAllRegisters()649 void InvalidateAllRegisters() override {
650 // Do nothing... registers are always valid...
651 }
652
SetRegisterDataFrom_LC_THREAD(const DataExtractor & data)653 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) {
654 lldb::offset_t offset = 0;
655 SetError(GPRRegSet, Read, -1);
656 SetError(FPURegSet, Read, -1);
657 SetError(EXCRegSet, Read, -1);
658 bool done = false;
659 while (!done) {
660 int flavor = data.GetU32(&offset);
661 uint32_t count = data.GetU32(&offset);
662 lldb::offset_t next_thread_state = offset + (count * 4);
663 switch (flavor) {
664 case GPRRegSet:
665 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1
666 // 32-bit register)
667 if (count >= (33 * 2) + 1) {
668 for (uint32_t i = 0; i < 29; ++i)
669 gpr.x[i] = data.GetU64(&offset);
670 gpr.fp = data.GetU64(&offset);
671 gpr.lr = data.GetU64(&offset);
672 gpr.sp = data.GetU64(&offset);
673 gpr.pc = data.GetU64(&offset);
674 gpr.cpsr = data.GetU32(&offset);
675 SetError(GPRRegSet, Read, 0);
676 }
677 offset = next_thread_state;
678 break;
679 case FPURegSet: {
680 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0];
681 const int fpu_reg_buf_size = sizeof(fpu);
682 if (fpu_reg_buf_size == count * sizeof(uint32_t) &&
683 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle,
684 fpu_reg_buf) == fpu_reg_buf_size) {
685 SetError(FPURegSet, Read, 0);
686 } else {
687 done = true;
688 }
689 }
690 offset = next_thread_state;
691 break;
692 case EXCRegSet:
693 if (count == 4) {
694 exc.far = data.GetU64(&offset);
695 exc.esr = data.GetU32(&offset);
696 exc.exception = data.GetU32(&offset);
697 SetError(EXCRegSet, Read, 0);
698 }
699 offset = next_thread_state;
700 break;
701 default:
702 done = true;
703 break;
704 }
705 }
706 }
707
Create_LC_THREAD(Thread * thread,Stream & data)708 static bool Create_LC_THREAD(Thread *thread, Stream &data) {
709 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext());
710 if (reg_ctx_sp) {
711 RegisterContext *reg_ctx = reg_ctx_sp.get();
712
713 data.PutHex32(GPRRegSet); // Flavor
714 data.PutHex32(GPRWordCount);
715 PrintRegisterValue(reg_ctx, "x0", nullptr, 8, data);
716 PrintRegisterValue(reg_ctx, "x1", nullptr, 8, data);
717 PrintRegisterValue(reg_ctx, "x2", nullptr, 8, data);
718 PrintRegisterValue(reg_ctx, "x3", nullptr, 8, data);
719 PrintRegisterValue(reg_ctx, "x4", nullptr, 8, data);
720 PrintRegisterValue(reg_ctx, "x5", nullptr, 8, data);
721 PrintRegisterValue(reg_ctx, "x6", nullptr, 8, data);
722 PrintRegisterValue(reg_ctx, "x7", nullptr, 8, data);
723 PrintRegisterValue(reg_ctx, "x8", nullptr, 8, data);
724 PrintRegisterValue(reg_ctx, "x9", nullptr, 8, data);
725 PrintRegisterValue(reg_ctx, "x10", nullptr, 8, data);
726 PrintRegisterValue(reg_ctx, "x11", nullptr, 8, data);
727 PrintRegisterValue(reg_ctx, "x12", nullptr, 8, data);
728 PrintRegisterValue(reg_ctx, "x13", nullptr, 8, data);
729 PrintRegisterValue(reg_ctx, "x14", nullptr, 8, data);
730 PrintRegisterValue(reg_ctx, "x15", nullptr, 8, data);
731 PrintRegisterValue(reg_ctx, "x16", nullptr, 8, data);
732 PrintRegisterValue(reg_ctx, "x17", nullptr, 8, data);
733 PrintRegisterValue(reg_ctx, "x18", nullptr, 8, data);
734 PrintRegisterValue(reg_ctx, "x19", nullptr, 8, data);
735 PrintRegisterValue(reg_ctx, "x20", nullptr, 8, data);
736 PrintRegisterValue(reg_ctx, "x21", nullptr, 8, data);
737 PrintRegisterValue(reg_ctx, "x22", nullptr, 8, data);
738 PrintRegisterValue(reg_ctx, "x23", nullptr, 8, data);
739 PrintRegisterValue(reg_ctx, "x24", nullptr, 8, data);
740 PrintRegisterValue(reg_ctx, "x25", nullptr, 8, data);
741 PrintRegisterValue(reg_ctx, "x26", nullptr, 8, data);
742 PrintRegisterValue(reg_ctx, "x27", nullptr, 8, data);
743 PrintRegisterValue(reg_ctx, "x28", nullptr, 8, data);
744 PrintRegisterValue(reg_ctx, "fp", nullptr, 8, data);
745 PrintRegisterValue(reg_ctx, "lr", nullptr, 8, data);
746 PrintRegisterValue(reg_ctx, "sp", nullptr, 8, data);
747 PrintRegisterValue(reg_ctx, "pc", nullptr, 8, data);
748 PrintRegisterValue(reg_ctx, "cpsr", nullptr, 4, data);
749 data.PutHex32(0); // uint32_t pad at the end
750
751 // Write out the EXC registers
752 data.PutHex32(EXCRegSet);
753 data.PutHex32(EXCWordCount);
754 PrintRegisterValue(reg_ctx, "far", nullptr, 8, data);
755 PrintRegisterValue(reg_ctx, "esr", nullptr, 4, data);
756 PrintRegisterValue(reg_ctx, "exception", nullptr, 4, data);
757 return true;
758 }
759 return false;
760 }
761
762 protected:
DoReadGPR(lldb::tid_t tid,int flavor,GPR & gpr)763 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; }
764
DoReadFPU(lldb::tid_t tid,int flavor,FPU & fpu)765 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; }
766
DoReadEXC(lldb::tid_t tid,int flavor,EXC & exc)767 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; }
768
DoReadDBG(lldb::tid_t tid,int flavor,DBG & dbg)769 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; }
770
DoWriteGPR(lldb::tid_t tid,int flavor,const GPR & gpr)771 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override {
772 return 0;
773 }
774
DoWriteFPU(lldb::tid_t tid,int flavor,const FPU & fpu)775 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override {
776 return 0;
777 }
778
DoWriteEXC(lldb::tid_t tid,int flavor,const EXC & exc)779 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override {
780 return 0;
781 }
782
DoWriteDBG(lldb::tid_t tid,int flavor,const DBG & dbg)783 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override {
784 return -1;
785 }
786 };
787
MachHeaderSizeFromMagic(uint32_t magic)788 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) {
789 switch (magic) {
790 case MH_MAGIC:
791 case MH_CIGAM:
792 return sizeof(struct llvm::MachO::mach_header);
793
794 case MH_MAGIC_64:
795 case MH_CIGAM_64:
796 return sizeof(struct llvm::MachO::mach_header_64);
797 break;
798
799 default:
800 break;
801 }
802 return 0;
803 }
804
805 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008
806
807 char ObjectFileMachO::ID;
808
Initialize()809 void ObjectFileMachO::Initialize() {
810 PluginManager::RegisterPlugin(
811 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance,
812 CreateMemoryInstance, GetModuleSpecifications, SaveCore);
813 }
814
Terminate()815 void ObjectFileMachO::Terminate() {
816 PluginManager::UnregisterPlugin(CreateInstance);
817 }
818
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)819 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp,
820 DataBufferSP data_sp,
821 lldb::offset_t data_offset,
822 const FileSpec *file,
823 lldb::offset_t file_offset,
824 lldb::offset_t length) {
825 if (!data_sp) {
826 data_sp = MapFileData(*file, length, file_offset);
827 if (!data_sp)
828 return nullptr;
829 data_offset = 0;
830 }
831
832 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length))
833 return nullptr;
834
835 // Update the data to contain the entire file if it doesn't already
836 if (data_sp->GetByteSize() < length) {
837 data_sp = MapFileData(*file, length, file_offset);
838 if (!data_sp)
839 return nullptr;
840 data_offset = 0;
841 }
842 auto objfile_up = std::make_unique<ObjectFileMachO>(
843 module_sp, data_sp, data_offset, file, file_offset, length);
844 if (!objfile_up || !objfile_up->ParseHeader())
845 return nullptr;
846
847 return objfile_up.release();
848 }
849
CreateMemoryInstance(const lldb::ModuleSP & module_sp,WritableDataBufferSP data_sp,const ProcessSP & process_sp,lldb::addr_t header_addr)850 ObjectFile *ObjectFileMachO::CreateMemoryInstance(
851 const lldb::ModuleSP &module_sp, WritableDataBufferSP data_sp,
852 const ProcessSP &process_sp, lldb::addr_t header_addr) {
853 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) {
854 std::unique_ptr<ObjectFile> objfile_up(
855 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr));
856 if (objfile_up.get() && objfile_up->ParseHeader())
857 return objfile_up.release();
858 }
859 return nullptr;
860 }
861
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)862 size_t ObjectFileMachO::GetModuleSpecifications(
863 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp,
864 lldb::offset_t data_offset, lldb::offset_t file_offset,
865 lldb::offset_t length, lldb_private::ModuleSpecList &specs) {
866 const size_t initial_count = specs.GetSize();
867
868 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) {
869 DataExtractor data;
870 data.SetData(data_sp);
871 llvm::MachO::mach_header header;
872 if (ParseHeader(data, &data_offset, header)) {
873 size_t header_and_load_cmds =
874 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic);
875 if (header_and_load_cmds >= data_sp->GetByteSize()) {
876 data_sp = MapFileData(file, header_and_load_cmds, file_offset);
877 data.SetData(data_sp);
878 data_offset = MachHeaderSizeFromMagic(header.magic);
879 }
880 if (data_sp) {
881 ModuleSpec base_spec;
882 base_spec.GetFileSpec() = file;
883 base_spec.SetObjectOffset(file_offset);
884 base_spec.SetObjectSize(length);
885 GetAllArchSpecs(header, data, data_offset, base_spec, specs);
886 }
887 }
888 }
889 return specs.GetSize() - initial_count;
890 }
891
GetSegmentNameTEXT()892 ConstString ObjectFileMachO::GetSegmentNameTEXT() {
893 static ConstString g_segment_name_TEXT("__TEXT");
894 return g_segment_name_TEXT;
895 }
896
GetSegmentNameDATA()897 ConstString ObjectFileMachO::GetSegmentNameDATA() {
898 static ConstString g_segment_name_DATA("__DATA");
899 return g_segment_name_DATA;
900 }
901
GetSegmentNameDATA_DIRTY()902 ConstString ObjectFileMachO::GetSegmentNameDATA_DIRTY() {
903 static ConstString g_segment_name("__DATA_DIRTY");
904 return g_segment_name;
905 }
906
GetSegmentNameDATA_CONST()907 ConstString ObjectFileMachO::GetSegmentNameDATA_CONST() {
908 static ConstString g_segment_name("__DATA_CONST");
909 return g_segment_name;
910 }
911
GetSegmentNameOBJC()912 ConstString ObjectFileMachO::GetSegmentNameOBJC() {
913 static ConstString g_segment_name_OBJC("__OBJC");
914 return g_segment_name_OBJC;
915 }
916
GetSegmentNameLINKEDIT()917 ConstString ObjectFileMachO::GetSegmentNameLINKEDIT() {
918 static ConstString g_section_name_LINKEDIT("__LINKEDIT");
919 return g_section_name_LINKEDIT;
920 }
921
GetSegmentNameDWARF()922 ConstString ObjectFileMachO::GetSegmentNameDWARF() {
923 static ConstString g_section_name("__DWARF");
924 return g_section_name;
925 }
926
GetSectionNameEHFrame()927 ConstString ObjectFileMachO::GetSectionNameEHFrame() {
928 static ConstString g_section_name_eh_frame("__eh_frame");
929 return g_section_name_eh_frame;
930 }
931
MagicBytesMatch(DataBufferSP data_sp,lldb::addr_t data_offset,lldb::addr_t data_length)932 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP data_sp,
933 lldb::addr_t data_offset,
934 lldb::addr_t data_length) {
935 DataExtractor data;
936 data.SetData(data_sp, data_offset, data_length);
937 lldb::offset_t offset = 0;
938 uint32_t magic = data.GetU32(&offset);
939 return MachHeaderSizeFromMagic(magic) != 0;
940 }
941
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)942 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp,
943 DataBufferSP data_sp,
944 lldb::offset_t data_offset,
945 const FileSpec *file,
946 lldb::offset_t file_offset,
947 lldb::offset_t length)
948 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset),
949 m_mach_segments(), m_mach_sections(), m_entry_point_address(),
950 m_thread_context_offsets(), m_thread_context_offsets_valid(false),
951 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) {
952 ::memset(&m_header, 0, sizeof(m_header));
953 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab));
954 }
955
ObjectFileMachO(const lldb::ModuleSP & module_sp,lldb::WritableDataBufferSP header_data_sp,const lldb::ProcessSP & process_sp,lldb::addr_t header_addr)956 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp,
957 lldb::WritableDataBufferSP header_data_sp,
958 const lldb::ProcessSP &process_sp,
959 lldb::addr_t header_addr)
960 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp),
961 m_mach_segments(), m_mach_sections(), m_entry_point_address(),
962 m_thread_context_offsets(), m_thread_context_offsets_valid(false),
963 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) {
964 ::memset(&m_header, 0, sizeof(m_header));
965 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab));
966 }
967
ParseHeader(DataExtractor & data,lldb::offset_t * data_offset_ptr,llvm::MachO::mach_header & header)968 bool ObjectFileMachO::ParseHeader(DataExtractor &data,
969 lldb::offset_t *data_offset_ptr,
970 llvm::MachO::mach_header &header) {
971 data.SetByteOrder(endian::InlHostByteOrder());
972 // Leave magic in the original byte order
973 header.magic = data.GetU32(data_offset_ptr);
974 bool can_parse = false;
975 bool is_64_bit = false;
976 switch (header.magic) {
977 case MH_MAGIC:
978 data.SetByteOrder(endian::InlHostByteOrder());
979 data.SetAddressByteSize(4);
980 can_parse = true;
981 break;
982
983 case MH_MAGIC_64:
984 data.SetByteOrder(endian::InlHostByteOrder());
985 data.SetAddressByteSize(8);
986 can_parse = true;
987 is_64_bit = true;
988 break;
989
990 case MH_CIGAM:
991 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig
992 ? eByteOrderLittle
993 : eByteOrderBig);
994 data.SetAddressByteSize(4);
995 can_parse = true;
996 break;
997
998 case MH_CIGAM_64:
999 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig
1000 ? eByteOrderLittle
1001 : eByteOrderBig);
1002 data.SetAddressByteSize(8);
1003 is_64_bit = true;
1004 can_parse = true;
1005 break;
1006
1007 default:
1008 break;
1009 }
1010
1011 if (can_parse) {
1012 data.GetU32(data_offset_ptr, &header.cputype, 6);
1013 if (is_64_bit)
1014 *data_offset_ptr += 4;
1015 return true;
1016 } else {
1017 memset(&header, 0, sizeof(header));
1018 }
1019 return false;
1020 }
1021
ParseHeader()1022 bool ObjectFileMachO::ParseHeader() {
1023 ModuleSP module_sp(GetModule());
1024 if (!module_sp)
1025 return false;
1026
1027 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
1028 bool can_parse = false;
1029 lldb::offset_t offset = 0;
1030 m_data.SetByteOrder(endian::InlHostByteOrder());
1031 // Leave magic in the original byte order
1032 m_header.magic = m_data.GetU32(&offset);
1033 switch (m_header.magic) {
1034 case MH_MAGIC:
1035 m_data.SetByteOrder(endian::InlHostByteOrder());
1036 m_data.SetAddressByteSize(4);
1037 can_parse = true;
1038 break;
1039
1040 case MH_MAGIC_64:
1041 m_data.SetByteOrder(endian::InlHostByteOrder());
1042 m_data.SetAddressByteSize(8);
1043 can_parse = true;
1044 break;
1045
1046 case MH_CIGAM:
1047 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig
1048 ? eByteOrderLittle
1049 : eByteOrderBig);
1050 m_data.SetAddressByteSize(4);
1051 can_parse = true;
1052 break;
1053
1054 case MH_CIGAM_64:
1055 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig
1056 ? eByteOrderLittle
1057 : eByteOrderBig);
1058 m_data.SetAddressByteSize(8);
1059 can_parse = true;
1060 break;
1061
1062 default:
1063 break;
1064 }
1065
1066 if (can_parse) {
1067 m_data.GetU32(&offset, &m_header.cputype, 6);
1068
1069 ModuleSpecList all_specs;
1070 ModuleSpec base_spec;
1071 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic),
1072 base_spec, all_specs);
1073
1074 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) {
1075 ArchSpec mach_arch =
1076 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture();
1077
1078 // Check if the module has a required architecture
1079 const ArchSpec &module_arch = module_sp->GetArchitecture();
1080 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch))
1081 continue;
1082
1083 if (SetModulesArchitecture(mach_arch)) {
1084 const size_t header_and_lc_size =
1085 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic);
1086 if (m_data.GetByteSize() < header_and_lc_size) {
1087 DataBufferSP data_sp;
1088 ProcessSP process_sp(m_process_wp.lock());
1089 if (process_sp) {
1090 data_sp = ReadMemory(process_sp, m_memory_addr, header_and_lc_size);
1091 } else {
1092 // Read in all only the load command data from the file on disk
1093 data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset);
1094 if (data_sp->GetByteSize() != header_and_lc_size)
1095 continue;
1096 }
1097 if (data_sp)
1098 m_data.SetData(data_sp);
1099 }
1100 }
1101 return true;
1102 }
1103 // None found.
1104 return false;
1105 } else {
1106 memset(&m_header, 0, sizeof(struct llvm::MachO::mach_header));
1107 }
1108 return false;
1109 }
1110
GetByteOrder() const1111 ByteOrder ObjectFileMachO::GetByteOrder() const {
1112 return m_data.GetByteOrder();
1113 }
1114
IsExecutable() const1115 bool ObjectFileMachO::IsExecutable() const {
1116 return m_header.filetype == MH_EXECUTE;
1117 }
1118
IsDynamicLoader() const1119 bool ObjectFileMachO::IsDynamicLoader() const {
1120 return m_header.filetype == MH_DYLINKER;
1121 }
1122
IsSharedCacheBinary() const1123 bool ObjectFileMachO::IsSharedCacheBinary() const {
1124 return m_header.flags & MH_DYLIB_IN_CACHE;
1125 }
1126
GetAddressByteSize() const1127 uint32_t ObjectFileMachO::GetAddressByteSize() const {
1128 return m_data.GetAddressByteSize();
1129 }
1130
GetAddressClass(lldb::addr_t file_addr)1131 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) {
1132 Symtab *symtab = GetSymtab();
1133 if (!symtab)
1134 return AddressClass::eUnknown;
1135
1136 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr);
1137 if (symbol) {
1138 if (symbol->ValueIsAddress()) {
1139 SectionSP section_sp(symbol->GetAddressRef().GetSection());
1140 if (section_sp) {
1141 const lldb::SectionType section_type = section_sp->GetType();
1142 switch (section_type) {
1143 case eSectionTypeInvalid:
1144 return AddressClass::eUnknown;
1145
1146 case eSectionTypeCode:
1147 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) {
1148 // For ARM we have a bit in the n_desc field of the symbol that
1149 // tells us ARM/Thumb which is bit 0x0008.
1150 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB)
1151 return AddressClass::eCodeAlternateISA;
1152 }
1153 return AddressClass::eCode;
1154
1155 case eSectionTypeContainer:
1156 return AddressClass::eUnknown;
1157
1158 case eSectionTypeData:
1159 case eSectionTypeDataCString:
1160 case eSectionTypeDataCStringPointers:
1161 case eSectionTypeDataSymbolAddress:
1162 case eSectionTypeData4:
1163 case eSectionTypeData8:
1164 case eSectionTypeData16:
1165 case eSectionTypeDataPointers:
1166 case eSectionTypeZeroFill:
1167 case eSectionTypeDataObjCMessageRefs:
1168 case eSectionTypeDataObjCCFStrings:
1169 case eSectionTypeGoSymtab:
1170 return AddressClass::eData;
1171
1172 case eSectionTypeDebug:
1173 case eSectionTypeDWARFDebugAbbrev:
1174 case eSectionTypeDWARFDebugAbbrevDwo:
1175 case eSectionTypeDWARFDebugAddr:
1176 case eSectionTypeDWARFDebugAranges:
1177 case eSectionTypeDWARFDebugCuIndex:
1178 case eSectionTypeDWARFDebugFrame:
1179 case eSectionTypeDWARFDebugInfo:
1180 case eSectionTypeDWARFDebugInfoDwo:
1181 case eSectionTypeDWARFDebugLine:
1182 case eSectionTypeDWARFDebugLineStr:
1183 case eSectionTypeDWARFDebugLoc:
1184 case eSectionTypeDWARFDebugLocDwo:
1185 case eSectionTypeDWARFDebugLocLists:
1186 case eSectionTypeDWARFDebugLocListsDwo:
1187 case eSectionTypeDWARFDebugMacInfo:
1188 case eSectionTypeDWARFDebugMacro:
1189 case eSectionTypeDWARFDebugNames:
1190 case eSectionTypeDWARFDebugPubNames:
1191 case eSectionTypeDWARFDebugPubTypes:
1192 case eSectionTypeDWARFDebugRanges:
1193 case eSectionTypeDWARFDebugRngLists:
1194 case eSectionTypeDWARFDebugRngListsDwo:
1195 case eSectionTypeDWARFDebugStr:
1196 case eSectionTypeDWARFDebugStrDwo:
1197 case eSectionTypeDWARFDebugStrOffsets:
1198 case eSectionTypeDWARFDebugStrOffsetsDwo:
1199 case eSectionTypeDWARFDebugTuIndex:
1200 case eSectionTypeDWARFDebugTypes:
1201 case eSectionTypeDWARFDebugTypesDwo:
1202 case eSectionTypeDWARFAppleNames:
1203 case eSectionTypeDWARFAppleTypes:
1204 case eSectionTypeDWARFAppleNamespaces:
1205 case eSectionTypeDWARFAppleObjC:
1206 case eSectionTypeDWARFGNUDebugAltLink:
1207 return AddressClass::eDebug;
1208
1209 case eSectionTypeEHFrame:
1210 case eSectionTypeARMexidx:
1211 case eSectionTypeARMextab:
1212 case eSectionTypeCompactUnwind:
1213 return AddressClass::eRuntime;
1214
1215 case eSectionTypeAbsoluteAddress:
1216 case eSectionTypeELFSymbolTable:
1217 case eSectionTypeELFDynamicSymbols:
1218 case eSectionTypeELFRelocationEntries:
1219 case eSectionTypeELFDynamicLinkInfo:
1220 case eSectionTypeOther:
1221 return AddressClass::eUnknown;
1222 }
1223 }
1224 }
1225
1226 const SymbolType symbol_type = symbol->GetType();
1227 switch (symbol_type) {
1228 case eSymbolTypeAny:
1229 return AddressClass::eUnknown;
1230 case eSymbolTypeAbsolute:
1231 return AddressClass::eUnknown;
1232
1233 case eSymbolTypeCode:
1234 case eSymbolTypeTrampoline:
1235 case eSymbolTypeResolver:
1236 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) {
1237 // For ARM we have a bit in the n_desc field of the symbol that tells
1238 // us ARM/Thumb which is bit 0x0008.
1239 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB)
1240 return AddressClass::eCodeAlternateISA;
1241 }
1242 return AddressClass::eCode;
1243
1244 case eSymbolTypeData:
1245 return AddressClass::eData;
1246 case eSymbolTypeRuntime:
1247 return AddressClass::eRuntime;
1248 case eSymbolTypeException:
1249 return AddressClass::eRuntime;
1250 case eSymbolTypeSourceFile:
1251 return AddressClass::eDebug;
1252 case eSymbolTypeHeaderFile:
1253 return AddressClass::eDebug;
1254 case eSymbolTypeObjectFile:
1255 return AddressClass::eDebug;
1256 case eSymbolTypeCommonBlock:
1257 return AddressClass::eDebug;
1258 case eSymbolTypeBlock:
1259 return AddressClass::eDebug;
1260 case eSymbolTypeLocal:
1261 return AddressClass::eData;
1262 case eSymbolTypeParam:
1263 return AddressClass::eData;
1264 case eSymbolTypeVariable:
1265 return AddressClass::eData;
1266 case eSymbolTypeVariableType:
1267 return AddressClass::eDebug;
1268 case eSymbolTypeLineEntry:
1269 return AddressClass::eDebug;
1270 case eSymbolTypeLineHeader:
1271 return AddressClass::eDebug;
1272 case eSymbolTypeScopeBegin:
1273 return AddressClass::eDebug;
1274 case eSymbolTypeScopeEnd:
1275 return AddressClass::eDebug;
1276 case eSymbolTypeAdditional:
1277 return AddressClass::eUnknown;
1278 case eSymbolTypeCompiler:
1279 return AddressClass::eDebug;
1280 case eSymbolTypeInstrumentation:
1281 return AddressClass::eDebug;
1282 case eSymbolTypeUndefined:
1283 return AddressClass::eUnknown;
1284 case eSymbolTypeObjCClass:
1285 return AddressClass::eRuntime;
1286 case eSymbolTypeObjCMetaClass:
1287 return AddressClass::eRuntime;
1288 case eSymbolTypeObjCIVar:
1289 return AddressClass::eRuntime;
1290 case eSymbolTypeReExported:
1291 return AddressClass::eRuntime;
1292 }
1293 }
1294 return AddressClass::eUnknown;
1295 }
1296
IsStripped()1297 bool ObjectFileMachO::IsStripped() {
1298 if (m_dysymtab.cmd == 0) {
1299 ModuleSP module_sp(GetModule());
1300 if (module_sp) {
1301 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
1302 for (uint32_t i = 0; i < m_header.ncmds; ++i) {
1303 const lldb::offset_t load_cmd_offset = offset;
1304
1305 llvm::MachO::load_command lc = {};
1306 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr)
1307 break;
1308 if (lc.cmd == LC_DYSYMTAB) {
1309 m_dysymtab.cmd = lc.cmd;
1310 m_dysymtab.cmdsize = lc.cmdsize;
1311 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym,
1312 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) ==
1313 nullptr) {
1314 // Clear m_dysymtab if we were unable to read all items from the
1315 // load command
1316 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab));
1317 }
1318 }
1319 offset = load_cmd_offset + lc.cmdsize;
1320 }
1321 }
1322 }
1323 if (m_dysymtab.cmd)
1324 return m_dysymtab.nlocalsym <= 1;
1325 return false;
1326 }
1327
GetEncryptedFileRanges()1328 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() {
1329 EncryptedFileRanges result;
1330 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
1331
1332 llvm::MachO::encryption_info_command encryption_cmd;
1333 for (uint32_t i = 0; i < m_header.ncmds; ++i) {
1334 const lldb::offset_t load_cmd_offset = offset;
1335 if (m_data.GetU32(&offset, &encryption_cmd, 2) == nullptr)
1336 break;
1337
1338 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the
1339 // 3 fields we care about, so treat them the same.
1340 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO ||
1341 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) {
1342 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) {
1343 if (encryption_cmd.cryptid != 0) {
1344 EncryptedFileRanges::Entry entry;
1345 entry.SetRangeBase(encryption_cmd.cryptoff);
1346 entry.SetByteSize(encryption_cmd.cryptsize);
1347 result.Append(entry);
1348 }
1349 }
1350 }
1351 offset = load_cmd_offset + encryption_cmd.cmdsize;
1352 }
1353
1354 return result;
1355 }
1356
SanitizeSegmentCommand(llvm::MachO::segment_command_64 & seg_cmd,uint32_t cmd_idx)1357 void ObjectFileMachO::SanitizeSegmentCommand(
1358 llvm::MachO::segment_command_64 &seg_cmd, uint32_t cmd_idx) {
1359 if (m_length == 0 || seg_cmd.filesize == 0)
1360 return;
1361
1362 if (IsSharedCacheBinary() && !IsInMemory()) {
1363 // In shared cache images, the load commands are relative to the
1364 // shared cache file, and not the specific image we are
1365 // examining. Let's fix this up so that it looks like a normal
1366 // image.
1367 if (strncmp(seg_cmd.segname, "__TEXT", sizeof(seg_cmd.segname)) == 0)
1368 m_text_address = seg_cmd.vmaddr;
1369 if (strncmp(seg_cmd.segname, "__LINKEDIT", sizeof(seg_cmd.segname)) == 0)
1370 m_linkedit_original_offset = seg_cmd.fileoff;
1371
1372 seg_cmd.fileoff = seg_cmd.vmaddr - m_text_address;
1373 }
1374
1375 if (seg_cmd.fileoff > m_length) {
1376 // We have a load command that says it extends past the end of the file.
1377 // This is likely a corrupt file. We don't have any way to return an error
1378 // condition here (this method was likely invoked from something like
1379 // ObjectFile::GetSectionList()), so we just null out the section contents,
1380 // and dump a message to stdout. The most common case here is core file
1381 // debugging with a truncated file.
1382 const char *lc_segment_name =
1383 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT";
1384 GetModule()->ReportWarning(
1385 "load command %u %s has a fileoff (0x%" PRIx64
1386 ") that extends beyond the end of the file (0x%" PRIx64
1387 "), ignoring this section",
1388 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length);
1389
1390 seg_cmd.fileoff = 0;
1391 seg_cmd.filesize = 0;
1392 }
1393
1394 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) {
1395 // We have a load command that says it extends past the end of the file.
1396 // This is likely a corrupt file. We don't have any way to return an error
1397 // condition here (this method was likely invoked from something like
1398 // ObjectFile::GetSectionList()), so we just null out the section contents,
1399 // and dump a message to stdout. The most common case here is core file
1400 // debugging with a truncated file.
1401 const char *lc_segment_name =
1402 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT";
1403 GetModule()->ReportWarning(
1404 "load command %u %s has a fileoff + filesize (0x%" PRIx64
1405 ") that extends beyond the end of the file (0x%" PRIx64
1406 "), the segment will be truncated to match",
1407 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length);
1408
1409 // Truncate the length
1410 seg_cmd.filesize = m_length - seg_cmd.fileoff;
1411 }
1412 }
1413
1414 static uint32_t
GetSegmentPermissions(const llvm::MachO::segment_command_64 & seg_cmd)1415 GetSegmentPermissions(const llvm::MachO::segment_command_64 &seg_cmd) {
1416 uint32_t result = 0;
1417 if (seg_cmd.initprot & VM_PROT_READ)
1418 result |= ePermissionsReadable;
1419 if (seg_cmd.initprot & VM_PROT_WRITE)
1420 result |= ePermissionsWritable;
1421 if (seg_cmd.initprot & VM_PROT_EXECUTE)
1422 result |= ePermissionsExecutable;
1423 return result;
1424 }
1425
GetSectionType(uint32_t flags,ConstString section_name)1426 static lldb::SectionType GetSectionType(uint32_t flags,
1427 ConstString section_name) {
1428
1429 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS))
1430 return eSectionTypeCode;
1431
1432 uint32_t mach_sect_type = flags & SECTION_TYPE;
1433 static ConstString g_sect_name_objc_data("__objc_data");
1434 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs");
1435 static ConstString g_sect_name_objc_selrefs("__objc_selrefs");
1436 static ConstString g_sect_name_objc_classrefs("__objc_classrefs");
1437 static ConstString g_sect_name_objc_superrefs("__objc_superrefs");
1438 static ConstString g_sect_name_objc_const("__objc_const");
1439 static ConstString g_sect_name_objc_classlist("__objc_classlist");
1440 static ConstString g_sect_name_cfstring("__cfstring");
1441
1442 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev");
1443 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges");
1444 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame");
1445 static ConstString g_sect_name_dwarf_debug_info("__debug_info");
1446 static ConstString g_sect_name_dwarf_debug_line("__debug_line");
1447 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc");
1448 static ConstString g_sect_name_dwarf_debug_loclists("__debug_loclists");
1449 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo");
1450 static ConstString g_sect_name_dwarf_debug_names("__debug_names");
1451 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames");
1452 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes");
1453 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges");
1454 static ConstString g_sect_name_dwarf_debug_str("__debug_str");
1455 static ConstString g_sect_name_dwarf_debug_types("__debug_types");
1456 static ConstString g_sect_name_dwarf_apple_names("__apple_names");
1457 static ConstString g_sect_name_dwarf_apple_types("__apple_types");
1458 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac");
1459 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc");
1460 static ConstString g_sect_name_eh_frame("__eh_frame");
1461 static ConstString g_sect_name_compact_unwind("__unwind_info");
1462 static ConstString g_sect_name_text("__text");
1463 static ConstString g_sect_name_data("__data");
1464 static ConstString g_sect_name_go_symtab("__gosymtab");
1465
1466 if (section_name == g_sect_name_dwarf_debug_abbrev)
1467 return eSectionTypeDWARFDebugAbbrev;
1468 if (section_name == g_sect_name_dwarf_debug_aranges)
1469 return eSectionTypeDWARFDebugAranges;
1470 if (section_name == g_sect_name_dwarf_debug_frame)
1471 return eSectionTypeDWARFDebugFrame;
1472 if (section_name == g_sect_name_dwarf_debug_info)
1473 return eSectionTypeDWARFDebugInfo;
1474 if (section_name == g_sect_name_dwarf_debug_line)
1475 return eSectionTypeDWARFDebugLine;
1476 if (section_name == g_sect_name_dwarf_debug_loc)
1477 return eSectionTypeDWARFDebugLoc;
1478 if (section_name == g_sect_name_dwarf_debug_loclists)
1479 return eSectionTypeDWARFDebugLocLists;
1480 if (section_name == g_sect_name_dwarf_debug_macinfo)
1481 return eSectionTypeDWARFDebugMacInfo;
1482 if (section_name == g_sect_name_dwarf_debug_names)
1483 return eSectionTypeDWARFDebugNames;
1484 if (section_name == g_sect_name_dwarf_debug_pubnames)
1485 return eSectionTypeDWARFDebugPubNames;
1486 if (section_name == g_sect_name_dwarf_debug_pubtypes)
1487 return eSectionTypeDWARFDebugPubTypes;
1488 if (section_name == g_sect_name_dwarf_debug_ranges)
1489 return eSectionTypeDWARFDebugRanges;
1490 if (section_name == g_sect_name_dwarf_debug_str)
1491 return eSectionTypeDWARFDebugStr;
1492 if (section_name == g_sect_name_dwarf_debug_types)
1493 return eSectionTypeDWARFDebugTypes;
1494 if (section_name == g_sect_name_dwarf_apple_names)
1495 return eSectionTypeDWARFAppleNames;
1496 if (section_name == g_sect_name_dwarf_apple_types)
1497 return eSectionTypeDWARFAppleTypes;
1498 if (section_name == g_sect_name_dwarf_apple_namespaces)
1499 return eSectionTypeDWARFAppleNamespaces;
1500 if (section_name == g_sect_name_dwarf_apple_objc)
1501 return eSectionTypeDWARFAppleObjC;
1502 if (section_name == g_sect_name_objc_selrefs)
1503 return eSectionTypeDataCStringPointers;
1504 if (section_name == g_sect_name_objc_msgrefs)
1505 return eSectionTypeDataObjCMessageRefs;
1506 if (section_name == g_sect_name_eh_frame)
1507 return eSectionTypeEHFrame;
1508 if (section_name == g_sect_name_compact_unwind)
1509 return eSectionTypeCompactUnwind;
1510 if (section_name == g_sect_name_cfstring)
1511 return eSectionTypeDataObjCCFStrings;
1512 if (section_name == g_sect_name_go_symtab)
1513 return eSectionTypeGoSymtab;
1514 if (section_name == g_sect_name_objc_data ||
1515 section_name == g_sect_name_objc_classrefs ||
1516 section_name == g_sect_name_objc_superrefs ||
1517 section_name == g_sect_name_objc_const ||
1518 section_name == g_sect_name_objc_classlist) {
1519 return eSectionTypeDataPointers;
1520 }
1521
1522 switch (mach_sect_type) {
1523 // TODO: categorize sections by other flags for regular sections
1524 case S_REGULAR:
1525 if (section_name == g_sect_name_text)
1526 return eSectionTypeCode;
1527 if (section_name == g_sect_name_data)
1528 return eSectionTypeData;
1529 return eSectionTypeOther;
1530 case S_ZEROFILL:
1531 return eSectionTypeZeroFill;
1532 case S_CSTRING_LITERALS: // section with only literal C strings
1533 return eSectionTypeDataCString;
1534 case S_4BYTE_LITERALS: // section with only 4 byte literals
1535 return eSectionTypeData4;
1536 case S_8BYTE_LITERALS: // section with only 8 byte literals
1537 return eSectionTypeData8;
1538 case S_LITERAL_POINTERS: // section with only pointers to literals
1539 return eSectionTypeDataPointers;
1540 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers
1541 return eSectionTypeDataPointers;
1542 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers
1543 return eSectionTypeDataPointers;
1544 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in
1545 // the reserved2 field
1546 return eSectionTypeCode;
1547 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for
1548 // initialization
1549 return eSectionTypeDataPointers;
1550 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for
1551 // termination
1552 return eSectionTypeDataPointers;
1553 case S_COALESCED:
1554 return eSectionTypeOther;
1555 case S_GB_ZEROFILL:
1556 return eSectionTypeZeroFill;
1557 case S_INTERPOSING: // section with only pairs of function pointers for
1558 // interposing
1559 return eSectionTypeCode;
1560 case S_16BYTE_LITERALS: // section with only 16 byte literals
1561 return eSectionTypeData16;
1562 case S_DTRACE_DOF:
1563 return eSectionTypeDebug;
1564 case S_LAZY_DYLIB_SYMBOL_POINTERS:
1565 return eSectionTypeDataPointers;
1566 default:
1567 return eSectionTypeOther;
1568 }
1569 }
1570
1571 struct ObjectFileMachO::SegmentParsingContext {
1572 const EncryptedFileRanges EncryptedRanges;
1573 lldb_private::SectionList &UnifiedList;
1574 uint32_t NextSegmentIdx = 0;
1575 uint32_t NextSectionIdx = 0;
1576 bool FileAddressesChanged = false;
1577
SegmentParsingContextObjectFileMachO::SegmentParsingContext1578 SegmentParsingContext(EncryptedFileRanges EncryptedRanges,
1579 lldb_private::SectionList &UnifiedList)
1580 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {}
1581 };
1582
ProcessSegmentCommand(const llvm::MachO::load_command & load_cmd_,lldb::offset_t offset,uint32_t cmd_idx,SegmentParsingContext & context)1583 void ObjectFileMachO::ProcessSegmentCommand(
1584 const llvm::MachO::load_command &load_cmd_, lldb::offset_t offset,
1585 uint32_t cmd_idx, SegmentParsingContext &context) {
1586 llvm::MachO::segment_command_64 load_cmd;
1587 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_));
1588
1589 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16))
1590 return;
1591
1592 ModuleSP module_sp = GetModule();
1593 const bool is_core = GetType() == eTypeCoreFile;
1594 const bool is_dsym = (m_header.filetype == MH_DSYM);
1595 bool add_section = true;
1596 bool add_to_unified = true;
1597 ConstString const_segname(
1598 load_cmd.segname, strnlen(load_cmd.segname, sizeof(load_cmd.segname)));
1599
1600 SectionSP unified_section_sp(
1601 context.UnifiedList.FindSectionByName(const_segname));
1602 if (is_dsym && unified_section_sp) {
1603 if (const_segname == GetSegmentNameLINKEDIT()) {
1604 // We need to keep the __LINKEDIT segment private to this object file
1605 // only
1606 add_to_unified = false;
1607 } else {
1608 // This is the dSYM file and this section has already been created by the
1609 // object file, no need to create it.
1610 add_section = false;
1611 }
1612 }
1613 load_cmd.vmaddr = m_data.GetAddress(&offset);
1614 load_cmd.vmsize = m_data.GetAddress(&offset);
1615 load_cmd.fileoff = m_data.GetAddress(&offset);
1616 load_cmd.filesize = m_data.GetAddress(&offset);
1617 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4))
1618 return;
1619
1620 SanitizeSegmentCommand(load_cmd, cmd_idx);
1621
1622 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd);
1623 const bool segment_is_encrypted =
1624 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0;
1625
1626 // Keep a list of mach segments around in case we need to get at data that
1627 // isn't stored in the abstracted Sections.
1628 m_mach_segments.push_back(load_cmd);
1629
1630 // Use a segment ID of the segment index shifted left by 8 so they never
1631 // conflict with any of the sections.
1632 SectionSP segment_sp;
1633 if (add_section && (const_segname || is_core)) {
1634 segment_sp = std::make_shared<Section>(
1635 module_sp, // Module to which this section belongs
1636 this, // Object file to which this sections belongs
1637 ++context.NextSegmentIdx
1638 << 8, // Section ID is the 1 based segment index
1639 // shifted right by 8 bits as not to collide with any of the 256
1640 // section IDs that are possible
1641 const_segname, // Name of this section
1642 eSectionTypeContainer, // This section is a container of other
1643 // sections.
1644 load_cmd.vmaddr, // File VM address == addresses as they are
1645 // found in the object file
1646 load_cmd.vmsize, // VM size in bytes of this section
1647 load_cmd.fileoff, // Offset to the data for this section in
1648 // the file
1649 load_cmd.filesize, // Size in bytes of this section as found
1650 // in the file
1651 0, // Segments have no alignment information
1652 load_cmd.flags); // Flags for this section
1653
1654 segment_sp->SetIsEncrypted(segment_is_encrypted);
1655 m_sections_up->AddSection(segment_sp);
1656 segment_sp->SetPermissions(segment_permissions);
1657 if (add_to_unified)
1658 context.UnifiedList.AddSection(segment_sp);
1659 } else if (unified_section_sp) {
1660 // If this is a dSYM and the file addresses in the dSYM differ from the
1661 // file addresses in the ObjectFile, we must use the file base address for
1662 // the Section from the dSYM for the DWARF to resolve correctly.
1663 // This only happens with binaries in the shared cache in practice;
1664 // normally a mismatch like this would give a binary & dSYM that do not
1665 // match UUIDs. When a binary is included in the shared cache, its
1666 // segments are rearranged to optimize the shared cache, so its file
1667 // addresses will differ from what the ObjectFile had originally,
1668 // and what the dSYM has.
1669 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) {
1670 Log *log = GetLog(LLDBLog::Symbols);
1671 if (log) {
1672 log->Printf(
1673 "Installing dSYM's %s segment file address over ObjectFile's "
1674 "so symbol table/debug info resolves correctly for %s",
1675 const_segname.AsCString(),
1676 module_sp->GetFileSpec().GetFilename().AsCString());
1677 }
1678
1679 // Make sure we've parsed the symbol table from the ObjectFile before
1680 // we go around changing its Sections.
1681 module_sp->GetObjectFile()->GetSymtab();
1682 // eh_frame would present the same problems but we parse that on a per-
1683 // function basis as-needed so it's more difficult to remove its use of
1684 // the Sections. Realistically, the environments where this code path
1685 // will be taken will not have eh_frame sections.
1686
1687 unified_section_sp->SetFileAddress(load_cmd.vmaddr);
1688
1689 // Notify the module that the section addresses have been changed once
1690 // we're done so any file-address caches can be updated.
1691 context.FileAddressesChanged = true;
1692 }
1693 m_sections_up->AddSection(unified_section_sp);
1694 }
1695
1696 llvm::MachO::section_64 sect64;
1697 ::memset(§64, 0, sizeof(sect64));
1698 // Push a section into our mach sections for the section at index zero
1699 // (NO_SECT) if we don't have any mach sections yet...
1700 if (m_mach_sections.empty())
1701 m_mach_sections.push_back(sect64);
1702 uint32_t segment_sect_idx;
1703 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1;
1704
1705 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8;
1706 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects;
1707 ++segment_sect_idx) {
1708 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname,
1709 sizeof(sect64.sectname)) == nullptr)
1710 break;
1711 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname,
1712 sizeof(sect64.segname)) == nullptr)
1713 break;
1714 sect64.addr = m_data.GetAddress(&offset);
1715 sect64.size = m_data.GetAddress(&offset);
1716
1717 if (m_data.GetU32(&offset, §64.offset, num_u32s) == nullptr)
1718 break;
1719
1720 if (IsSharedCacheBinary() && !IsInMemory()) {
1721 sect64.offset = sect64.addr - m_text_address;
1722 }
1723
1724 // Keep a list of mach sections around in case we need to get at data that
1725 // isn't stored in the abstracted Sections.
1726 m_mach_sections.push_back(sect64);
1727
1728 if (add_section) {
1729 ConstString section_name(
1730 sect64.sectname, strnlen(sect64.sectname, sizeof(sect64.sectname)));
1731 if (!const_segname) {
1732 // We have a segment with no name so we need to conjure up segments
1733 // that correspond to the section's segname if there isn't already such
1734 // a section. If there is such a section, we resize the section so that
1735 // it spans all sections. We also mark these sections as fake so
1736 // address matches don't hit if they land in the gaps between the child
1737 // sections.
1738 const_segname.SetTrimmedCStringWithLength(sect64.segname,
1739 sizeof(sect64.segname));
1740 segment_sp = context.UnifiedList.FindSectionByName(const_segname);
1741 if (segment_sp.get()) {
1742 Section *segment = segment_sp.get();
1743 // Grow the section size as needed.
1744 const lldb::addr_t sect64_min_addr = sect64.addr;
1745 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size;
1746 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize();
1747 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress();
1748 const lldb::addr_t curr_seg_max_addr =
1749 curr_seg_min_addr + curr_seg_byte_size;
1750 if (sect64_min_addr >= curr_seg_min_addr) {
1751 const lldb::addr_t new_seg_byte_size =
1752 sect64_max_addr - curr_seg_min_addr;
1753 // Only grow the section size if needed
1754 if (new_seg_byte_size > curr_seg_byte_size)
1755 segment->SetByteSize(new_seg_byte_size);
1756 } else {
1757 // We need to change the base address of the segment and adjust the
1758 // child section offsets for all existing children.
1759 const lldb::addr_t slide_amount =
1760 sect64_min_addr - curr_seg_min_addr;
1761 segment->Slide(slide_amount, false);
1762 segment->GetChildren().Slide(-slide_amount, false);
1763 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr);
1764 }
1765
1766 // Grow the section size as needed.
1767 if (sect64.offset) {
1768 const lldb::addr_t segment_min_file_offset =
1769 segment->GetFileOffset();
1770 const lldb::addr_t segment_max_file_offset =
1771 segment_min_file_offset + segment->GetFileSize();
1772
1773 const lldb::addr_t section_min_file_offset = sect64.offset;
1774 const lldb::addr_t section_max_file_offset =
1775 section_min_file_offset + sect64.size;
1776 const lldb::addr_t new_file_offset =
1777 std::min(section_min_file_offset, segment_min_file_offset);
1778 const lldb::addr_t new_file_size =
1779 std::max(section_max_file_offset, segment_max_file_offset) -
1780 new_file_offset;
1781 segment->SetFileOffset(new_file_offset);
1782 segment->SetFileSize(new_file_size);
1783 }
1784 } else {
1785 // Create a fake section for the section's named segment
1786 segment_sp = std::make_shared<Section>(
1787 segment_sp, // Parent section
1788 module_sp, // Module to which this section belongs
1789 this, // Object file to which this section belongs
1790 ++context.NextSegmentIdx
1791 << 8, // Section ID is the 1 based segment index
1792 // shifted right by 8 bits as not to
1793 // collide with any of the 256 section IDs
1794 // that are possible
1795 const_segname, // Name of this section
1796 eSectionTypeContainer, // This section is a container of
1797 // other sections.
1798 sect64.addr, // File VM address == addresses as they are
1799 // found in the object file
1800 sect64.size, // VM size in bytes of this section
1801 sect64.offset, // Offset to the data for this section in
1802 // the file
1803 sect64.offset ? sect64.size : 0, // Size in bytes of
1804 // this section as
1805 // found in the file
1806 sect64.align,
1807 load_cmd.flags); // Flags for this section
1808 segment_sp->SetIsFake(true);
1809 segment_sp->SetPermissions(segment_permissions);
1810 m_sections_up->AddSection(segment_sp);
1811 if (add_to_unified)
1812 context.UnifiedList.AddSection(segment_sp);
1813 segment_sp->SetIsEncrypted(segment_is_encrypted);
1814 }
1815 }
1816 assert(segment_sp.get());
1817
1818 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name);
1819
1820 SectionSP section_sp(new Section(
1821 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name,
1822 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size,
1823 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align,
1824 sect64.flags));
1825 // Set the section to be encrypted to match the segment
1826
1827 bool section_is_encrypted = false;
1828 if (!segment_is_encrypted && load_cmd.filesize != 0)
1829 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains(
1830 sect64.offset) != nullptr;
1831
1832 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted);
1833 section_sp->SetPermissions(segment_permissions);
1834 segment_sp->GetChildren().AddSection(section_sp);
1835
1836 if (segment_sp->IsFake()) {
1837 segment_sp.reset();
1838 const_segname.Clear();
1839 }
1840 }
1841 }
1842 if (segment_sp && is_dsym) {
1843 if (first_segment_sectID <= context.NextSectionIdx) {
1844 lldb::user_id_t sect_uid;
1845 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx;
1846 ++sect_uid) {
1847 SectionSP curr_section_sp(
1848 segment_sp->GetChildren().FindSectionByID(sect_uid));
1849 SectionSP next_section_sp;
1850 if (sect_uid + 1 <= context.NextSectionIdx)
1851 next_section_sp =
1852 segment_sp->GetChildren().FindSectionByID(sect_uid + 1);
1853
1854 if (curr_section_sp.get()) {
1855 if (curr_section_sp->GetByteSize() == 0) {
1856 if (next_section_sp.get() != nullptr)
1857 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() -
1858 curr_section_sp->GetFileAddress());
1859 else
1860 curr_section_sp->SetByteSize(load_cmd.vmsize);
1861 }
1862 }
1863 }
1864 }
1865 }
1866 }
1867
ProcessDysymtabCommand(const llvm::MachO::load_command & load_cmd,lldb::offset_t offset)1868 void ObjectFileMachO::ProcessDysymtabCommand(
1869 const llvm::MachO::load_command &load_cmd, lldb::offset_t offset) {
1870 m_dysymtab.cmd = load_cmd.cmd;
1871 m_dysymtab.cmdsize = load_cmd.cmdsize;
1872 m_data.GetU32(&offset, &m_dysymtab.ilocalsym,
1873 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2);
1874 }
1875
CreateSections(SectionList & unified_section_list)1876 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) {
1877 if (m_sections_up)
1878 return;
1879
1880 m_sections_up = std::make_unique<SectionList>();
1881
1882 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
1883 // bool dump_sections = false;
1884 ModuleSP module_sp(GetModule());
1885
1886 offset = MachHeaderSizeFromMagic(m_header.magic);
1887
1888 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list);
1889 llvm::MachO::load_command load_cmd;
1890 for (uint32_t i = 0; i < m_header.ncmds; ++i) {
1891 const lldb::offset_t load_cmd_offset = offset;
1892 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr)
1893 break;
1894
1895 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64)
1896 ProcessSegmentCommand(load_cmd, offset, i, context);
1897 else if (load_cmd.cmd == LC_DYSYMTAB)
1898 ProcessDysymtabCommand(load_cmd, offset);
1899
1900 offset = load_cmd_offset + load_cmd.cmdsize;
1901 }
1902
1903 if (context.FileAddressesChanged && module_sp)
1904 module_sp->SectionFileAddressesChanged();
1905 }
1906
1907 class MachSymtabSectionInfo {
1908 public:
MachSymtabSectionInfo(SectionList * section_list)1909 MachSymtabSectionInfo(SectionList *section_list)
1910 : m_section_list(section_list), m_section_infos() {
1911 // Get the number of sections down to a depth of 1 to include all segments
1912 // and their sections, but no other sections that may be added for debug
1913 // map or
1914 m_section_infos.resize(section_list->GetNumSections(1));
1915 }
1916
GetSection(uint8_t n_sect,addr_t file_addr)1917 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) {
1918 if (n_sect == 0)
1919 return SectionSP();
1920 if (n_sect < m_section_infos.size()) {
1921 if (!m_section_infos[n_sect].section_sp) {
1922 SectionSP section_sp(m_section_list->FindSectionByID(n_sect));
1923 m_section_infos[n_sect].section_sp = section_sp;
1924 if (section_sp) {
1925 m_section_infos[n_sect].vm_range.SetBaseAddress(
1926 section_sp->GetFileAddress());
1927 m_section_infos[n_sect].vm_range.SetByteSize(
1928 section_sp->GetByteSize());
1929 } else {
1930 std::string filename = "<unknown>";
1931 SectionSP first_section_sp(m_section_list->GetSectionAtIndex(0));
1932 if (first_section_sp)
1933 filename = first_section_sp->GetObjectFile()->GetFileSpec().GetPath();
1934
1935 Debugger::ReportError(
1936 llvm::formatv("unable to find section {0} for a symbol in "
1937 "{1}, corrupt file?",
1938 n_sect, filename));
1939 }
1940 }
1941 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) {
1942 // Symbol is in section.
1943 return m_section_infos[n_sect].section_sp;
1944 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 &&
1945 m_section_infos[n_sect].vm_range.GetBaseAddress() ==
1946 file_addr) {
1947 // Symbol is in section with zero size, but has the same start address
1948 // as the section. This can happen with linker symbols (symbols that
1949 // start with the letter 'l' or 'L'.
1950 return m_section_infos[n_sect].section_sp;
1951 }
1952 }
1953 return m_section_list->FindSectionContainingFileAddress(file_addr);
1954 }
1955
1956 protected:
1957 struct SectionInfo {
SectionInfoMachSymtabSectionInfo::SectionInfo1958 SectionInfo() : vm_range(), section_sp() {}
1959
1960 VMRange vm_range;
1961 SectionSP section_sp;
1962 };
1963 SectionList *m_section_list;
1964 std::vector<SectionInfo> m_section_infos;
1965 };
1966
1967 #define TRIE_SYMBOL_IS_THUMB (1ULL << 63)
1968 struct TrieEntry {
DumpTrieEntry1969 void Dump() const {
1970 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"",
1971 static_cast<unsigned long long>(address),
1972 static_cast<unsigned long long>(flags),
1973 static_cast<unsigned long long>(other), name.GetCString());
1974 if (import_name)
1975 printf(" -> \"%s\"\n", import_name.GetCString());
1976 else
1977 printf("\n");
1978 }
1979 ConstString name;
1980 uint64_t address = LLDB_INVALID_ADDRESS;
1981 uint64_t flags =
1982 0; // EXPORT_SYMBOL_FLAGS_REEXPORT, EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER,
1983 // TRIE_SYMBOL_IS_THUMB
1984 uint64_t other = 0;
1985 ConstString import_name;
1986 };
1987
1988 struct TrieEntryWithOffset {
1989 lldb::offset_t nodeOffset;
1990 TrieEntry entry;
1991
TrieEntryWithOffsetTrieEntryWithOffset1992 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {}
1993
DumpTrieEntryWithOffset1994 void Dump(uint32_t idx) const {
1995 printf("[%3u] 0x%16.16llx: ", idx,
1996 static_cast<unsigned long long>(nodeOffset));
1997 entry.Dump();
1998 }
1999
operator <TrieEntryWithOffset2000 bool operator<(const TrieEntryWithOffset &other) const {
2001 return (nodeOffset < other.nodeOffset);
2002 }
2003 };
2004
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)2005 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset,
2006 const bool is_arm, addr_t text_seg_base_addr,
2007 std::vector<llvm::StringRef> &nameSlices,
2008 std::set<lldb::addr_t> &resolver_addresses,
2009 std::vector<TrieEntryWithOffset> &reexports,
2010 std::vector<TrieEntryWithOffset> &ext_symbols) {
2011 if (!data.ValidOffset(offset))
2012 return true;
2013
2014 // Terminal node -- end of a branch, possibly add this to
2015 // the symbol table or resolver table.
2016 const uint64_t terminalSize = data.GetULEB128(&offset);
2017 lldb::offset_t children_offset = offset + terminalSize;
2018 if (terminalSize != 0) {
2019 TrieEntryWithOffset e(offset);
2020 e.entry.flags = data.GetULEB128(&offset);
2021 const char *import_name = nullptr;
2022 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) {
2023 e.entry.address = 0;
2024 e.entry.other = data.GetULEB128(&offset); // dylib ordinal
2025 import_name = data.GetCStr(&offset);
2026 } else {
2027 e.entry.address = data.GetULEB128(&offset);
2028 if (text_seg_base_addr != LLDB_INVALID_ADDRESS)
2029 e.entry.address += text_seg_base_addr;
2030 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) {
2031 e.entry.other = data.GetULEB128(&offset);
2032 uint64_t resolver_addr = e.entry.other;
2033 if (text_seg_base_addr != LLDB_INVALID_ADDRESS)
2034 resolver_addr += text_seg_base_addr;
2035 if (is_arm)
2036 resolver_addr &= THUMB_ADDRESS_BIT_MASK;
2037 resolver_addresses.insert(resolver_addr);
2038 } else
2039 e.entry.other = 0;
2040 }
2041 bool add_this_entry = false;
2042 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT) &&
2043 import_name && import_name[0]) {
2044 // add symbols that are reexport symbols with a valid import name.
2045 add_this_entry = true;
2046 } else if (e.entry.flags == 0 &&
2047 (import_name == nullptr || import_name[0] == '\0')) {
2048 // add externally visible symbols, in case the nlist record has
2049 // been stripped/omitted.
2050 add_this_entry = true;
2051 }
2052 if (add_this_entry) {
2053 std::string name;
2054 if (!nameSlices.empty()) {
2055 for (auto name_slice : nameSlices)
2056 name.append(name_slice.data(), name_slice.size());
2057 }
2058 if (name.size() > 1) {
2059 // Skip the leading '_'
2060 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1);
2061 }
2062 if (import_name) {
2063 // Skip the leading '_'
2064 e.entry.import_name.SetCString(import_name + 1);
2065 }
2066 if (Flags(e.entry.flags).Test(EXPORT_SYMBOL_FLAGS_REEXPORT)) {
2067 reexports.push_back(e);
2068 } else {
2069 if (is_arm && (e.entry.address & 1)) {
2070 e.entry.flags |= TRIE_SYMBOL_IS_THUMB;
2071 e.entry.address &= THUMB_ADDRESS_BIT_MASK;
2072 }
2073 ext_symbols.push_back(e);
2074 }
2075 }
2076 }
2077
2078 const uint8_t childrenCount = data.GetU8(&children_offset);
2079 for (uint8_t i = 0; i < childrenCount; ++i) {
2080 const char *cstr = data.GetCStr(&children_offset);
2081 if (cstr)
2082 nameSlices.push_back(llvm::StringRef(cstr));
2083 else
2084 return false; // Corrupt data
2085 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset);
2086 if (childNodeOffset) {
2087 if (!ParseTrieEntries(data, childNodeOffset, is_arm, text_seg_base_addr,
2088 nameSlices, resolver_addresses, reexports,
2089 ext_symbols)) {
2090 return false;
2091 }
2092 }
2093 nameSlices.pop_back();
2094 }
2095 return true;
2096 }
2097
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)2098 static SymbolType GetSymbolType(const char *&symbol_name,
2099 bool &demangled_is_synthesized,
2100 const SectionSP &text_section_sp,
2101 const SectionSP &data_section_sp,
2102 const SectionSP &data_dirty_section_sp,
2103 const SectionSP &data_const_section_sp,
2104 const SectionSP &symbol_section) {
2105 SymbolType type = eSymbolTypeInvalid;
2106
2107 const char *symbol_sect_name = symbol_section->GetName().AsCString();
2108 if (symbol_section->IsDescendant(text_section_sp.get())) {
2109 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS |
2110 S_ATTR_SELF_MODIFYING_CODE |
2111 S_ATTR_SOME_INSTRUCTIONS))
2112 type = eSymbolTypeData;
2113 else
2114 type = eSymbolTypeCode;
2115 } else if (symbol_section->IsDescendant(data_section_sp.get()) ||
2116 symbol_section->IsDescendant(data_dirty_section_sp.get()) ||
2117 symbol_section->IsDescendant(data_const_section_sp.get())) {
2118 if (symbol_sect_name &&
2119 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) {
2120 type = eSymbolTypeRuntime;
2121
2122 if (symbol_name) {
2123 llvm::StringRef symbol_name_ref(symbol_name);
2124 if (symbol_name_ref.startswith("OBJC_")) {
2125 static const llvm::StringRef g_objc_v2_prefix_class("OBJC_CLASS_$_");
2126 static const llvm::StringRef g_objc_v2_prefix_metaclass(
2127 "OBJC_METACLASS_$_");
2128 static const llvm::StringRef g_objc_v2_prefix_ivar("OBJC_IVAR_$_");
2129 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) {
2130 symbol_name = symbol_name + g_objc_v2_prefix_class.size();
2131 type = eSymbolTypeObjCClass;
2132 demangled_is_synthesized = true;
2133 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) {
2134 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size();
2135 type = eSymbolTypeObjCMetaClass;
2136 demangled_is_synthesized = true;
2137 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) {
2138 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size();
2139 type = eSymbolTypeObjCIVar;
2140 demangled_is_synthesized = true;
2141 }
2142 }
2143 }
2144 } else if (symbol_sect_name &&
2145 ::strstr(symbol_sect_name, "__gcc_except_tab") ==
2146 symbol_sect_name) {
2147 type = eSymbolTypeException;
2148 } else {
2149 type = eSymbolTypeData;
2150 }
2151 } else if (symbol_sect_name &&
2152 ::strstr(symbol_sect_name, "__IMPORT") == symbol_sect_name) {
2153 type = eSymbolTypeTrampoline;
2154 }
2155 return type;
2156 }
2157
2158 // 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)2159 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache,
2160 const ByteOrder byte_order,
2161 const uint32_t addr_byte_size) {
2162 UUID dsc_uuid;
2163 DataBufferSP DscData = MapFileData(
2164 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0);
2165 if (!DscData)
2166 return dsc_uuid;
2167 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size);
2168
2169 char version_str[7];
2170 lldb::offset_t offset = 0;
2171 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6);
2172 version_str[6] = '\0';
2173 if (strcmp(version_str, "dyld_v") == 0) {
2174 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid);
2175 dsc_uuid = UUID::fromOptionalData(
2176 dsc_header_data.GetData(&offset, sizeof(uuid_t)), sizeof(uuid_t));
2177 }
2178 Log *log = GetLog(LLDBLog::Symbols);
2179 if (log && dsc_uuid.IsValid()) {
2180 LLDB_LOGF(log, "Shared cache %s has UUID %s",
2181 dyld_shared_cache.GetPath().c_str(),
2182 dsc_uuid.GetAsString().c_str());
2183 }
2184 return dsc_uuid;
2185 }
2186
2187 static llvm::Optional<struct nlist_64>
ParseNList(DataExtractor & nlist_data,lldb::offset_t & nlist_data_offset,size_t nlist_byte_size)2188 ParseNList(DataExtractor &nlist_data, lldb::offset_t &nlist_data_offset,
2189 size_t nlist_byte_size) {
2190 struct nlist_64 nlist;
2191 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size))
2192 return {};
2193 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset);
2194 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset);
2195 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset);
2196 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset);
2197 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset);
2198 return nlist;
2199 }
2200
2201 enum { DebugSymbols = true, NonDebugSymbols = false };
2202
ParseSymtab(Symtab & symtab)2203 void ObjectFileMachO::ParseSymtab(Symtab &symtab) {
2204 ModuleSP module_sp(GetModule());
2205 if (!module_sp)
2206 return;
2207
2208 const FileSpec &file = m_file ? m_file : module_sp->GetFileSpec();
2209 const char *file_name = file.GetFilename().AsCString("<Unknown>");
2210 LLDB_SCOPED_TIMERF("ObjectFileMachO::ParseSymtab () module = %s", file_name);
2211 Progress progress(llvm::formatv("Parsing symbol table for {0}", file_name));
2212
2213 llvm::MachO::symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0};
2214 llvm::MachO::linkedit_data_command function_starts_load_command = {0, 0, 0, 0};
2215 llvm::MachO::linkedit_data_command exports_trie_load_command = {0, 0, 0, 0};
2216 llvm::MachO::dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
2217 llvm::MachO::dysymtab_command dysymtab = m_dysymtab;
2218 // The data element of type bool indicates that this entry is thumb
2219 // code.
2220 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts;
2221
2222 // Record the address of every function/data that we add to the symtab.
2223 // We add symbols to the table in the order of most information (nlist
2224 // records) to least (function starts), and avoid duplicating symbols
2225 // via this set.
2226 llvm::DenseSet<addr_t> symbols_added;
2227
2228 // We are using a llvm::DenseSet for "symbols_added" so we must be sure we
2229 // do not add the tombstone or empty keys to the set.
2230 auto add_symbol_addr = [&symbols_added](lldb::addr_t file_addr) {
2231 // Don't add the tombstone or empty keys.
2232 if (file_addr == UINT64_MAX || file_addr == UINT64_MAX - 1)
2233 return;
2234 symbols_added.insert(file_addr);
2235 };
2236 FunctionStarts function_starts;
2237 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
2238 uint32_t i;
2239 FileSpecList dylib_files;
2240 Log *log = GetLog(LLDBLog::Symbols);
2241 llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_");
2242 llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_");
2243 llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_");
2244 UUID image_uuid;
2245
2246 for (i = 0; i < m_header.ncmds; ++i) {
2247 const lldb::offset_t cmd_offset = offset;
2248 // Read in the load command and load command size
2249 llvm::MachO::load_command lc;
2250 if (m_data.GetU32(&offset, &lc, 2) == nullptr)
2251 break;
2252 // Watch for the symbol table load command
2253 switch (lc.cmd) {
2254 case LC_SYMTAB:
2255 symtab_load_command.cmd = lc.cmd;
2256 symtab_load_command.cmdsize = lc.cmdsize;
2257 // Read in the rest of the symtab load command
2258 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) ==
2259 nullptr) // fill in symoff, nsyms, stroff, strsize fields
2260 return;
2261 break;
2262
2263 case LC_DYLD_INFO:
2264 case LC_DYLD_INFO_ONLY:
2265 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) {
2266 dyld_info.cmd = lc.cmd;
2267 dyld_info.cmdsize = lc.cmdsize;
2268 } else {
2269 memset(&dyld_info, 0, sizeof(dyld_info));
2270 }
2271 break;
2272
2273 case LC_LOAD_DYLIB:
2274 case LC_LOAD_WEAK_DYLIB:
2275 case LC_REEXPORT_DYLIB:
2276 case LC_LOADFVMLIB:
2277 case LC_LOAD_UPWARD_DYLIB: {
2278 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset);
2279 const char *path = m_data.PeekCStr(name_offset);
2280 if (path) {
2281 FileSpec file_spec(path);
2282 // Strip the path if there is @rpath, @executable, etc so we just use
2283 // the basename
2284 if (path[0] == '@')
2285 file_spec.GetDirectory().Clear();
2286
2287 if (lc.cmd == LC_REEXPORT_DYLIB) {
2288 m_reexported_dylibs.AppendIfUnique(file_spec);
2289 }
2290
2291 dylib_files.Append(file_spec);
2292 }
2293 } break;
2294
2295 case LC_DYLD_EXPORTS_TRIE:
2296 exports_trie_load_command.cmd = lc.cmd;
2297 exports_trie_load_command.cmdsize = lc.cmdsize;
2298 if (m_data.GetU32(&offset, &exports_trie_load_command.dataoff, 2) ==
2299 nullptr) // fill in offset and size fields
2300 memset(&exports_trie_load_command, 0,
2301 sizeof(exports_trie_load_command));
2302 break;
2303 case LC_FUNCTION_STARTS:
2304 function_starts_load_command.cmd = lc.cmd;
2305 function_starts_load_command.cmdsize = lc.cmdsize;
2306 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) ==
2307 nullptr) // fill in data offset and size fields
2308 memset(&function_starts_load_command, 0,
2309 sizeof(function_starts_load_command));
2310 break;
2311
2312 case LC_UUID: {
2313 const uint8_t *uuid_bytes = m_data.PeekData(offset, 16);
2314
2315 if (uuid_bytes)
2316 image_uuid = UUID::fromOptionalData(uuid_bytes, 16);
2317 break;
2318 }
2319
2320 default:
2321 break;
2322 }
2323 offset = cmd_offset + lc.cmdsize;
2324 }
2325
2326 if (!symtab_load_command.cmd)
2327 return;
2328
2329 SectionList *section_list = GetSectionList();
2330 if (section_list == nullptr)
2331 return;
2332
2333 const uint32_t addr_byte_size = m_data.GetAddressByteSize();
2334 const ByteOrder byte_order = m_data.GetByteOrder();
2335 bool bit_width_32 = addr_byte_size == 4;
2336 const size_t nlist_byte_size =
2337 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64);
2338
2339 DataExtractor nlist_data(nullptr, 0, byte_order, addr_byte_size);
2340 DataExtractor strtab_data(nullptr, 0, byte_order, addr_byte_size);
2341 DataExtractor function_starts_data(nullptr, 0, byte_order, addr_byte_size);
2342 DataExtractor indirect_symbol_index_data(nullptr, 0, byte_order,
2343 addr_byte_size);
2344 DataExtractor dyld_trie_data(nullptr, 0, byte_order, addr_byte_size);
2345
2346 const addr_t nlist_data_byte_size =
2347 symtab_load_command.nsyms * nlist_byte_size;
2348 const addr_t strtab_data_byte_size = symtab_load_command.strsize;
2349 addr_t strtab_addr = LLDB_INVALID_ADDRESS;
2350
2351 ProcessSP process_sp(m_process_wp.lock());
2352 Process *process = process_sp.get();
2353
2354 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete;
2355 bool is_shared_cache_image = IsSharedCacheBinary();
2356 bool is_local_shared_cache_image = is_shared_cache_image && !IsInMemory();
2357 SectionSP linkedit_section_sp(
2358 section_list->FindSectionByName(GetSegmentNameLINKEDIT()));
2359
2360 if (process && m_header.filetype != llvm::MachO::MH_OBJECT &&
2361 !is_local_shared_cache_image) {
2362 Target &target = process->GetTarget();
2363
2364 memory_module_load_level = target.GetMemoryModuleLoadLevel();
2365
2366 // Reading mach file from memory in a process or core file...
2367
2368 if (linkedit_section_sp) {
2369 addr_t linkedit_load_addr =
2370 linkedit_section_sp->GetLoadBaseAddress(&target);
2371 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) {
2372 // We might be trying to access the symbol table before the
2373 // __LINKEDIT's load address has been set in the target. We can't
2374 // fail to read the symbol table, so calculate the right address
2375 // manually
2376 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage(
2377 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get());
2378 }
2379
2380 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset();
2381 const addr_t symoff_addr = linkedit_load_addr +
2382 symtab_load_command.symoff -
2383 linkedit_file_offset;
2384 strtab_addr = linkedit_load_addr + symtab_load_command.stroff -
2385 linkedit_file_offset;
2386
2387 // Always load dyld - the dynamic linker - from memory if we didn't
2388 // find a binary anywhere else. lldb will not register
2389 // dylib/framework/bundle loads/unloads if we don't have the dyld
2390 // symbols, we force dyld to load from memory despite the user's
2391 // target.memory-module-load-level setting.
2392 if (memory_module_load_level == eMemoryModuleLoadLevelComplete ||
2393 m_header.filetype == llvm::MachO::MH_DYLINKER) {
2394 DataBufferSP nlist_data_sp(
2395 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size));
2396 if (nlist_data_sp)
2397 nlist_data.SetData(nlist_data_sp, 0, nlist_data_sp->GetByteSize());
2398 if (dysymtab.nindirectsyms != 0) {
2399 const addr_t indirect_syms_addr = linkedit_load_addr +
2400 dysymtab.indirectsymoff -
2401 linkedit_file_offset;
2402 DataBufferSP indirect_syms_data_sp(ReadMemory(
2403 process_sp, indirect_syms_addr, dysymtab.nindirectsyms * 4));
2404 if (indirect_syms_data_sp)
2405 indirect_symbol_index_data.SetData(
2406 indirect_syms_data_sp, 0,
2407 indirect_syms_data_sp->GetByteSize());
2408 // If this binary is outside the shared cache,
2409 // cache the string table.
2410 // Binaries in the shared cache all share a giant string table,
2411 // and we can't share the string tables across multiple
2412 // ObjectFileMachO's, so we'd end up re-reading this mega-strtab
2413 // for every binary in the shared cache - it would be a big perf
2414 // problem. For binaries outside the shared cache, it's faster to
2415 // read the entire strtab at once instead of piece-by-piece as we
2416 // process the nlist records.
2417 if (!is_shared_cache_image) {
2418 DataBufferSP strtab_data_sp(
2419 ReadMemory(process_sp, strtab_addr, strtab_data_byte_size));
2420 if (strtab_data_sp) {
2421 strtab_data.SetData(strtab_data_sp, 0,
2422 strtab_data_sp->GetByteSize());
2423 }
2424 }
2425 }
2426 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) {
2427 if (function_starts_load_command.cmd) {
2428 const addr_t func_start_addr =
2429 linkedit_load_addr + function_starts_load_command.dataoff -
2430 linkedit_file_offset;
2431 DataBufferSP func_start_data_sp(
2432 ReadMemory(process_sp, func_start_addr,
2433 function_starts_load_command.datasize));
2434 if (func_start_data_sp)
2435 function_starts_data.SetData(func_start_data_sp, 0,
2436 func_start_data_sp->GetByteSize());
2437 }
2438 }
2439 }
2440 }
2441 } else {
2442 if (is_local_shared_cache_image) {
2443 // The load commands in shared cache images are relative to the
2444 // beginning of the shared cache, not the library image. The
2445 // data we get handed when creating the ObjectFileMachO starts
2446 // at the beginning of a specific library and spans to the end
2447 // of the cache to be able to reach the shared LINKEDIT
2448 // segments. We need to convert the load command offsets to be
2449 // relative to the beginning of our specific image.
2450 lldb::addr_t linkedit_offset = linkedit_section_sp->GetFileOffset();
2451 lldb::offset_t linkedit_slide =
2452 linkedit_offset - m_linkedit_original_offset;
2453 symtab_load_command.symoff += linkedit_slide;
2454 symtab_load_command.stroff += linkedit_slide;
2455 dyld_info.export_off += linkedit_slide;
2456 dysymtab.indirectsymoff += linkedit_slide;
2457 function_starts_load_command.dataoff += linkedit_slide;
2458 exports_trie_load_command.dataoff += linkedit_slide;
2459 }
2460
2461 nlist_data.SetData(m_data, symtab_load_command.symoff,
2462 nlist_data_byte_size);
2463 strtab_data.SetData(m_data, symtab_load_command.stroff,
2464 strtab_data_byte_size);
2465
2466 // We shouldn't have exports data from both the LC_DYLD_INFO command
2467 // AND the LC_DYLD_EXPORTS_TRIE command in the same binary:
2468 lldbassert(!((dyld_info.export_size > 0)
2469 && (exports_trie_load_command.datasize > 0)));
2470 if (dyld_info.export_size > 0) {
2471 dyld_trie_data.SetData(m_data, dyld_info.export_off,
2472 dyld_info.export_size);
2473 } else if (exports_trie_load_command.datasize > 0) {
2474 dyld_trie_data.SetData(m_data, exports_trie_load_command.dataoff,
2475 exports_trie_load_command.datasize);
2476 }
2477
2478 if (dysymtab.nindirectsyms != 0) {
2479 indirect_symbol_index_data.SetData(m_data, dysymtab.indirectsymoff,
2480 dysymtab.nindirectsyms * 4);
2481 }
2482 if (function_starts_load_command.cmd) {
2483 function_starts_data.SetData(m_data, function_starts_load_command.dataoff,
2484 function_starts_load_command.datasize);
2485 }
2486 }
2487
2488 const bool have_strtab_data = strtab_data.GetByteSize() > 0;
2489
2490 ConstString g_segment_name_TEXT = GetSegmentNameTEXT();
2491 ConstString g_segment_name_DATA = GetSegmentNameDATA();
2492 ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY();
2493 ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST();
2494 ConstString g_segment_name_OBJC = GetSegmentNameOBJC();
2495 ConstString g_section_name_eh_frame = GetSectionNameEHFrame();
2496 SectionSP text_section_sp(
2497 section_list->FindSectionByName(g_segment_name_TEXT));
2498 SectionSP data_section_sp(
2499 section_list->FindSectionByName(g_segment_name_DATA));
2500 SectionSP data_dirty_section_sp(
2501 section_list->FindSectionByName(g_segment_name_DATA_DIRTY));
2502 SectionSP data_const_section_sp(
2503 section_list->FindSectionByName(g_segment_name_DATA_CONST));
2504 SectionSP objc_section_sp(
2505 section_list->FindSectionByName(g_segment_name_OBJC));
2506 SectionSP eh_frame_section_sp;
2507 if (text_section_sp.get())
2508 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName(
2509 g_section_name_eh_frame);
2510 else
2511 eh_frame_section_sp =
2512 section_list->FindSectionByName(g_section_name_eh_frame);
2513
2514 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM);
2515 const bool always_thumb = GetArchitecture().IsAlwaysThumbInstructions();
2516
2517 // lldb works best if it knows the start address of all functions in a
2518 // module. Linker symbols or debug info are normally the best source of
2519 // information for start addr / size but they may be stripped in a released
2520 // binary. Two additional sources of information exist in Mach-O binaries:
2521 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each
2522 // function's start address in the
2523 // binary, relative to the text section.
2524 // eh_frame - the eh_frame FDEs have the start addr & size of
2525 // each function
2526 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on
2527 // all modern binaries.
2528 // Binaries built to run on older releases may need to use eh_frame
2529 // information.
2530
2531 if (text_section_sp && function_starts_data.GetByteSize()) {
2532 FunctionStarts::Entry function_start_entry;
2533 function_start_entry.data = false;
2534 lldb::offset_t function_start_offset = 0;
2535 function_start_entry.addr = text_section_sp->GetFileAddress();
2536 uint64_t delta;
2537 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) >
2538 0) {
2539 // Now append the current entry
2540 function_start_entry.addr += delta;
2541 if (is_arm) {
2542 if (function_start_entry.addr & 1) {
2543 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK;
2544 function_start_entry.data = true;
2545 } else if (always_thumb) {
2546 function_start_entry.data = true;
2547 }
2548 }
2549 function_starts.Append(function_start_entry);
2550 }
2551 } else {
2552 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the
2553 // load command claiming an eh_frame but it doesn't actually have the
2554 // eh_frame content. And if we have a dSYM, we don't need to do any of
2555 // this fill-in-the-missing-symbols works anyway - the debug info should
2556 // give us all the functions in the module.
2557 if (text_section_sp.get() && eh_frame_section_sp.get() &&
2558 m_type != eTypeDebugInfo) {
2559 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp,
2560 DWARFCallFrameInfo::EH);
2561 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions;
2562 eh_frame.GetFunctionAddressAndSizeVector(functions);
2563 addr_t text_base_addr = text_section_sp->GetFileAddress();
2564 size_t count = functions.GetSize();
2565 for (size_t i = 0; i < count; ++i) {
2566 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func =
2567 functions.GetEntryAtIndex(i);
2568 if (func) {
2569 FunctionStarts::Entry function_start_entry;
2570 function_start_entry.addr = func->base - text_base_addr;
2571 if (is_arm) {
2572 if (function_start_entry.addr & 1) {
2573 function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK;
2574 function_start_entry.data = true;
2575 } else if (always_thumb) {
2576 function_start_entry.data = true;
2577 }
2578 }
2579 function_starts.Append(function_start_entry);
2580 }
2581 }
2582 }
2583 }
2584
2585 const size_t function_starts_count = function_starts.GetSize();
2586
2587 // For user process binaries (executables, dylibs, frameworks, bundles), if
2588 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're
2589 // going to assume the binary has been stripped. Don't allow assembly
2590 // language instruction emulation because we don't know proper function
2591 // start boundaries.
2592 //
2593 // For all other types of binaries (kernels, stand-alone bare board
2594 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame
2595 // sections - we should not make any assumptions about them based on that.
2596 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) {
2597 m_allow_assembly_emulation_unwind_plans = false;
2598 Log *unwind_or_symbol_log(GetLog(LLDBLog::Symbols | LLDBLog::Unwind));
2599
2600 if (unwind_or_symbol_log)
2601 module_sp->LogMessage(
2602 unwind_or_symbol_log,
2603 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds");
2604 }
2605
2606 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get()
2607 ? eh_frame_section_sp->GetID()
2608 : static_cast<user_id_t>(NO_SECT);
2609
2610 uint32_t N_SO_index = UINT32_MAX;
2611
2612 MachSymtabSectionInfo section_info(section_list);
2613 std::vector<uint32_t> N_FUN_indexes;
2614 std::vector<uint32_t> N_NSYM_indexes;
2615 std::vector<uint32_t> N_INCL_indexes;
2616 std::vector<uint32_t> N_BRAC_indexes;
2617 std::vector<uint32_t> N_COMM_indexes;
2618 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap;
2619 typedef llvm::DenseMap<uint32_t, uint32_t> NListIndexToSymbolIndexMap;
2620 typedef llvm::DenseMap<const char *, uint32_t> ConstNameToSymbolIndexMap;
2621 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx;
2622 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx;
2623 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx;
2624 // Any symbols that get merged into another will get an entry in this map
2625 // so we know
2626 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx;
2627 uint32_t nlist_idx = 0;
2628 Symbol *symbol_ptr = nullptr;
2629
2630 uint32_t sym_idx = 0;
2631 Symbol *sym = nullptr;
2632 size_t num_syms = 0;
2633 std::string memory_symbol_name;
2634 uint32_t unmapped_local_symbols_found = 0;
2635
2636 std::vector<TrieEntryWithOffset> reexport_trie_entries;
2637 std::vector<TrieEntryWithOffset> external_sym_trie_entries;
2638 std::set<lldb::addr_t> resolver_addresses;
2639
2640 if (dyld_trie_data.GetByteSize() > 0) {
2641 ConstString text_segment_name("__TEXT");
2642 SectionSP text_segment_sp =
2643 GetSectionList()->FindSectionByName(text_segment_name);
2644 lldb::addr_t text_segment_file_addr = LLDB_INVALID_ADDRESS;
2645 if (text_segment_sp)
2646 text_segment_file_addr = text_segment_sp->GetFileAddress();
2647 std::vector<llvm::StringRef> nameSlices;
2648 ParseTrieEntries(dyld_trie_data, 0, is_arm, text_segment_file_addr,
2649 nameSlices, resolver_addresses, reexport_trie_entries,
2650 external_sym_trie_entries);
2651 }
2652
2653 typedef std::set<ConstString> IndirectSymbols;
2654 IndirectSymbols indirect_symbol_names;
2655
2656 #if TARGET_OS_IPHONE
2657
2658 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been
2659 // optimized by moving LOCAL symbols out of the memory mapped portion of
2660 // the DSC. The symbol information has all been retained, but it isn't
2661 // available in the normal nlist data. However, there *are* duplicate
2662 // entries of *some*
2663 // LOCAL symbols in the normal nlist data. To handle this situation
2664 // correctly, we must first attempt
2665 // to parse any DSC unmapped symbol information. If we find any, we set a
2666 // flag that tells the normal nlist parser to ignore all LOCAL symbols.
2667
2668 if (IsSharedCacheBinary()) {
2669 // Before we can start mapping the DSC, we need to make certain the
2670 // target process is actually using the cache we can find.
2671
2672 // Next we need to determine the correct path for the dyld shared cache.
2673
2674 ArchSpec header_arch = GetArchitecture();
2675
2676 UUID dsc_uuid;
2677 UUID process_shared_cache_uuid;
2678 addr_t process_shared_cache_base_addr;
2679
2680 if (process) {
2681 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr,
2682 process_shared_cache_uuid);
2683 }
2684
2685 __block bool found_image = false;
2686 __block void *nlist_buffer = nullptr;
2687 __block unsigned nlist_count = 0;
2688 __block char *string_table = nullptr;
2689 __block vm_offset_t vm_nlist_memory = 0;
2690 __block mach_msg_type_number_t vm_nlist_bytes_read = 0;
2691 __block vm_offset_t vm_string_memory = 0;
2692 __block mach_msg_type_number_t vm_string_bytes_read = 0;
2693
2694 auto _ = llvm::make_scope_exit(^{
2695 if (vm_nlist_memory)
2696 vm_deallocate(mach_task_self(), vm_nlist_memory, vm_nlist_bytes_read);
2697 if (vm_string_memory)
2698 vm_deallocate(mach_task_self(), vm_string_memory, vm_string_bytes_read);
2699 });
2700
2701 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap;
2702 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName;
2703 UndefinedNameToDescMap undefined_name_to_desc;
2704 SymbolIndexToName reexport_shlib_needs_fixup;
2705
2706 dyld_for_each_installed_shared_cache(^(dyld_shared_cache_t shared_cache) {
2707 uuid_t cache_uuid;
2708 dyld_shared_cache_copy_uuid(shared_cache, &cache_uuid);
2709 if (found_image)
2710 return;
2711
2712 if (process_shared_cache_uuid.IsValid() &&
2713 process_shared_cache_uuid != UUID::fromOptionalData(&cache_uuid, 16))
2714 return;
2715
2716 dyld_shared_cache_for_each_image(shared_cache, ^(dyld_image_t image) {
2717 uuid_t dsc_image_uuid;
2718 if (found_image)
2719 return;
2720
2721 dyld_image_copy_uuid(image, &dsc_image_uuid);
2722 if (image_uuid != UUID::fromOptionalData(dsc_image_uuid, 16))
2723 return;
2724
2725 found_image = true;
2726
2727 // Compute the size of the string table. We need to ask dyld for a
2728 // new SPI to avoid this step.
2729 dyld_image_local_nlist_content_4Symbolication(
2730 image, ^(const void *nlistStart, uint64_t nlistCount,
2731 const char *stringTable) {
2732 if (!nlistStart || !nlistCount)
2733 return;
2734
2735 // The buffers passed here are valid only inside the block.
2736 // Use vm_read to make a cheap copy of them available for our
2737 // processing later.
2738 kern_return_t ret =
2739 vm_read(mach_task_self(), (vm_address_t)nlistStart,
2740 nlist_byte_size * nlistCount, &vm_nlist_memory,
2741 &vm_nlist_bytes_read);
2742 if (ret != KERN_SUCCESS)
2743 return;
2744 assert(vm_nlist_bytes_read == nlist_byte_size * nlistCount);
2745
2746 // We don't know the size of the string table. It's cheaper
2747 // to map the whol VM region than to determine the size by
2748 // parsing all teh nlist entries.
2749 vm_address_t string_address = (vm_address_t)stringTable;
2750 vm_size_t region_size;
2751 mach_msg_type_number_t info_count = VM_REGION_BASIC_INFO_COUNT_64;
2752 vm_region_basic_info_data_t info;
2753 memory_object_name_t object;
2754 ret = vm_region_64(mach_task_self(), &string_address,
2755 ®ion_size, VM_REGION_BASIC_INFO_64,
2756 (vm_region_info_t)&info, &info_count, &object);
2757 if (ret != KERN_SUCCESS)
2758 return;
2759
2760 ret = vm_read(mach_task_self(), (vm_address_t)stringTable,
2761 region_size -
2762 ((vm_address_t)stringTable - string_address),
2763 &vm_string_memory, &vm_string_bytes_read);
2764 if (ret != KERN_SUCCESS)
2765 return;
2766
2767 nlist_buffer = (void *)vm_nlist_memory;
2768 string_table = (char *)vm_string_memory;
2769 nlist_count = nlistCount;
2770 });
2771 });
2772 });
2773 if (nlist_buffer) {
2774 DataExtractor dsc_local_symbols_data(nlist_buffer,
2775 nlist_count * nlist_byte_size,
2776 byte_order, addr_byte_size);
2777 unmapped_local_symbols_found = nlist_count;
2778
2779 // The normal nlist code cannot correctly size the Symbols
2780 // array, we need to allocate it here.
2781 sym = symtab.Resize(
2782 symtab_load_command.nsyms + m_dysymtab.nindirectsyms +
2783 unmapped_local_symbols_found - m_dysymtab.nlocalsym);
2784 num_syms = symtab.GetNumSymbols();
2785
2786 lldb::offset_t nlist_data_offset = 0;
2787
2788 for (uint32_t nlist_index = 0;
2789 nlist_index < nlist_count;
2790 nlist_index++) {
2791 /////////////////////////////
2792 {
2793 llvm::Optional<struct nlist_64> nlist_maybe =
2794 ParseNList(dsc_local_symbols_data, nlist_data_offset,
2795 nlist_byte_size);
2796 if (!nlist_maybe)
2797 break;
2798 struct nlist_64 nlist = *nlist_maybe;
2799
2800 SymbolType type = eSymbolTypeInvalid;
2801 const char *symbol_name = string_table + nlist.n_strx;
2802
2803 if (symbol_name == NULL) {
2804 // No symbol should be NULL, even the symbols with no
2805 // string values should have an offset zero which
2806 // points to an empty C-string
2807 Debugger::ReportError(llvm::formatv(
2808 "DSC unmapped local symbol[{0}] has invalid "
2809 "string table offset {1:x} in {2}, ignoring symbol",
2810 nlist_index, nlist.n_strx,
2811 module_sp->GetFileSpec().GetPath());
2812 continue;
2813 }
2814 if (symbol_name[0] == '\0')
2815 symbol_name = NULL;
2816
2817 const char *symbol_name_non_abi_mangled = NULL;
2818
2819 SectionSP symbol_section;
2820 uint32_t symbol_byte_size = 0;
2821 bool add_nlist = true;
2822 bool is_debug = ((nlist.n_type & N_STAB) != 0);
2823 bool demangled_is_synthesized = false;
2824 bool is_gsym = false;
2825 bool set_value = true;
2826
2827 assert(sym_idx < num_syms);
2828
2829 sym[sym_idx].SetDebug(is_debug);
2830
2831 if (is_debug) {
2832 switch (nlist.n_type) {
2833 case N_GSYM:
2834 // global symbol: name,,NO_SECT,type,0
2835 // Sometimes the N_GSYM value contains the address.
2836
2837 // FIXME: In the .o files, we have a GSYM and a debug
2838 // symbol for all the ObjC data. They
2839 // have the same address, but we want to ensure that
2840 // we always find only the real symbol, 'cause we
2841 // don't currently correctly attribute the
2842 // GSYM one to the ObjCClass/Ivar/MetaClass
2843 // symbol type. This is a temporary hack to make
2844 // sure the ObjectiveC symbols get treated correctly.
2845 // To do this right, we should coalesce all the GSYM
2846 // & global symbols that have the same address.
2847
2848 is_gsym = true;
2849 sym[sym_idx].SetExternal(true);
2850
2851 if (symbol_name && symbol_name[0] == '_' &&
2852 symbol_name[1] == 'O') {
2853 llvm::StringRef symbol_name_ref(symbol_name);
2854 if (symbol_name_ref.startswith(
2855 g_objc_v2_prefix_class)) {
2856 symbol_name_non_abi_mangled = symbol_name + 1;
2857 symbol_name =
2858 symbol_name + g_objc_v2_prefix_class.size();
2859 type = eSymbolTypeObjCClass;
2860 demangled_is_synthesized = true;
2861
2862 } else if (symbol_name_ref.startswith(
2863 g_objc_v2_prefix_metaclass)) {
2864 symbol_name_non_abi_mangled = symbol_name + 1;
2865 symbol_name =
2866 symbol_name + g_objc_v2_prefix_metaclass.size();
2867 type = eSymbolTypeObjCMetaClass;
2868 demangled_is_synthesized = true;
2869 } else if (symbol_name_ref.startswith(
2870 g_objc_v2_prefix_ivar)) {
2871 symbol_name_non_abi_mangled = symbol_name + 1;
2872 symbol_name =
2873 symbol_name + g_objc_v2_prefix_ivar.size();
2874 type = eSymbolTypeObjCIVar;
2875 demangled_is_synthesized = true;
2876 }
2877 } else {
2878 if (nlist.n_value != 0)
2879 symbol_section = section_info.GetSection(
2880 nlist.n_sect, nlist.n_value);
2881 type = eSymbolTypeData;
2882 }
2883 break;
2884
2885 case N_FNAME:
2886 // procedure name (f77 kludge): name,,NO_SECT,0,0
2887 type = eSymbolTypeCompiler;
2888 break;
2889
2890 case N_FUN:
2891 // procedure: name,,n_sect,linenumber,address
2892 if (symbol_name) {
2893 type = eSymbolTypeCode;
2894 symbol_section = section_info.GetSection(
2895 nlist.n_sect, nlist.n_value);
2896
2897 N_FUN_addr_to_sym_idx.insert(
2898 std::make_pair(nlist.n_value, sym_idx));
2899 // We use the current number of symbols in the
2900 // symbol table in lieu of using nlist_idx in case
2901 // we ever start trimming entries out
2902 N_FUN_indexes.push_back(sym_idx);
2903 } else {
2904 type = eSymbolTypeCompiler;
2905
2906 if (!N_FUN_indexes.empty()) {
2907 // Copy the size of the function into the
2908 // original
2909 // STAB entry so we don't have
2910 // to hunt for it later
2911 symtab.SymbolAtIndex(N_FUN_indexes.back())
2912 ->SetByteSize(nlist.n_value);
2913 N_FUN_indexes.pop_back();
2914 // We don't really need the end function STAB as
2915 // it contains the size which we already placed
2916 // with the original symbol, so don't add it if
2917 // we want a minimal symbol table
2918 add_nlist = false;
2919 }
2920 }
2921 break;
2922
2923 case N_STSYM:
2924 // static symbol: name,,n_sect,type,address
2925 N_STSYM_addr_to_sym_idx.insert(
2926 std::make_pair(nlist.n_value, sym_idx));
2927 symbol_section = section_info.GetSection(nlist.n_sect,
2928 nlist.n_value);
2929 if (symbol_name && symbol_name[0]) {
2930 type = ObjectFile::GetSymbolTypeFromName(
2931 symbol_name + 1, eSymbolTypeData);
2932 }
2933 break;
2934
2935 case N_LCSYM:
2936 // .lcomm symbol: name,,n_sect,type,address
2937 symbol_section = section_info.GetSection(nlist.n_sect,
2938 nlist.n_value);
2939 type = eSymbolTypeCommonBlock;
2940 break;
2941
2942 case N_BNSYM:
2943 // We use the current number of symbols in the symbol
2944 // table in lieu of using nlist_idx in case we ever
2945 // start trimming entries out Skip these if we want
2946 // minimal symbol tables
2947 add_nlist = false;
2948 break;
2949
2950 case N_ENSYM:
2951 // Set the size of the N_BNSYM to the terminating
2952 // index of this N_ENSYM so that we can always skip
2953 // the entire symbol if we need to navigate more
2954 // quickly at the source level when parsing STABS
2955 // Skip these if we want minimal symbol tables
2956 add_nlist = false;
2957 break;
2958
2959 case N_OPT:
2960 // emitted with gcc2_compiled and in gcc source
2961 type = eSymbolTypeCompiler;
2962 break;
2963
2964 case N_RSYM:
2965 // register sym: name,,NO_SECT,type,register
2966 type = eSymbolTypeVariable;
2967 break;
2968
2969 case N_SLINE:
2970 // src line: 0,,n_sect,linenumber,address
2971 symbol_section = section_info.GetSection(nlist.n_sect,
2972 nlist.n_value);
2973 type = eSymbolTypeLineEntry;
2974 break;
2975
2976 case N_SSYM:
2977 // structure elt: name,,NO_SECT,type,struct_offset
2978 type = eSymbolTypeVariableType;
2979 break;
2980
2981 case N_SO:
2982 // source file name
2983 type = eSymbolTypeSourceFile;
2984 if (symbol_name == NULL) {
2985 add_nlist = false;
2986 if (N_SO_index != UINT32_MAX) {
2987 // Set the size of the N_SO to the terminating
2988 // index of this N_SO so that we can always skip
2989 // the entire N_SO if we need to navigate more
2990 // quickly at the source level when parsing STABS
2991 symbol_ptr = symtab.SymbolAtIndex(N_SO_index);
2992 symbol_ptr->SetByteSize(sym_idx);
2993 symbol_ptr->SetSizeIsSibling(true);
2994 }
2995 N_NSYM_indexes.clear();
2996 N_INCL_indexes.clear();
2997 N_BRAC_indexes.clear();
2998 N_COMM_indexes.clear();
2999 N_FUN_indexes.clear();
3000 N_SO_index = UINT32_MAX;
3001 } else {
3002 // We use the current number of symbols in the
3003 // symbol table in lieu of using nlist_idx in case
3004 // we ever start trimming entries out
3005 const bool N_SO_has_full_path = symbol_name[0] == '/';
3006 if (N_SO_has_full_path) {
3007 if ((N_SO_index == sym_idx - 1) &&
3008 ((sym_idx - 1) < num_syms)) {
3009 // We have two consecutive N_SO entries where
3010 // the first contains a directory and the
3011 // second contains a full path.
3012 sym[sym_idx - 1].GetMangled().SetValue(
3013 ConstString(symbol_name), false);
3014 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1;
3015 add_nlist = false;
3016 } else {
3017 // This is the first entry in a N_SO that
3018 // contains a directory or
3019 // a full path to the source file
3020 N_SO_index = sym_idx;
3021 }
3022 } else if ((N_SO_index == sym_idx - 1) &&
3023 ((sym_idx - 1) < num_syms)) {
3024 // This is usually the second N_SO entry that
3025 // contains just the filename, so here we combine
3026 // it with the first one if we are minimizing the
3027 // symbol table
3028 const char *so_path = sym[sym_idx - 1]
3029 .GetMangled()
3030 .GetDemangledName()
3031 .AsCString();
3032 if (so_path && so_path[0]) {
3033 std::string full_so_path(so_path);
3034 const size_t double_slash_pos =
3035 full_so_path.find("//");
3036 if (double_slash_pos != std::string::npos) {
3037 // The linker has been generating bad N_SO
3038 // entries with doubled up paths
3039 // in the format "%s%s" where the first
3040 // string in the DW_AT_comp_dir, and the
3041 // second is the directory for the source
3042 // file so you end up with a path that looks
3043 // like "/tmp/src//tmp/src/"
3044 FileSpec so_dir(so_path);
3045 if (!FileSystem::Instance().Exists(so_dir)) {
3046 so_dir.SetFile(
3047 &full_so_path[double_slash_pos + 1],
3048 FileSpec::Style::native);
3049 if (FileSystem::Instance().Exists(so_dir)) {
3050 // Trim off the incorrect path
3051 full_so_path.erase(0, double_slash_pos + 1);
3052 }
3053 }
3054 }
3055 if (*full_so_path.rbegin() != '/')
3056 full_so_path += '/';
3057 full_so_path += symbol_name;
3058 sym[sym_idx - 1].GetMangled().SetValue(
3059 ConstString(full_so_path.c_str()), false);
3060 add_nlist = false;
3061 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1;
3062 }
3063 } else {
3064 // This could be a relative path to a N_SO
3065 N_SO_index = sym_idx;
3066 }
3067 }
3068 break;
3069
3070 case N_OSO:
3071 // object file name: name,,0,0,st_mtime
3072 type = eSymbolTypeObjectFile;
3073 break;
3074
3075 case N_LSYM:
3076 // local sym: name,,NO_SECT,type,offset
3077 type = eSymbolTypeLocal;
3078 break;
3079
3080 // INCL scopes
3081 case N_BINCL:
3082 // include file beginning: name,,NO_SECT,0,sum We use
3083 // the current number of symbols in the symbol table
3084 // in lieu of using nlist_idx in case we ever start
3085 // trimming entries out
3086 N_INCL_indexes.push_back(sym_idx);
3087 type = eSymbolTypeScopeBegin;
3088 break;
3089
3090 case N_EINCL:
3091 // include file end: name,,NO_SECT,0,0
3092 // Set the size of the N_BINCL to the terminating
3093 // index of this N_EINCL so that we can always skip
3094 // the entire symbol if we need to navigate more
3095 // quickly at the source level when parsing STABS
3096 if (!N_INCL_indexes.empty()) {
3097 symbol_ptr =
3098 symtab.SymbolAtIndex(N_INCL_indexes.back());
3099 symbol_ptr->SetByteSize(sym_idx + 1);
3100 symbol_ptr->SetSizeIsSibling(true);
3101 N_INCL_indexes.pop_back();
3102 }
3103 type = eSymbolTypeScopeEnd;
3104 break;
3105
3106 case N_SOL:
3107 // #included file name: name,,n_sect,0,address
3108 type = eSymbolTypeHeaderFile;
3109
3110 // We currently don't use the header files on darwin
3111 add_nlist = false;
3112 break;
3113
3114 case N_PARAMS:
3115 // compiler parameters: name,,NO_SECT,0,0
3116 type = eSymbolTypeCompiler;
3117 break;
3118
3119 case N_VERSION:
3120 // compiler version: name,,NO_SECT,0,0
3121 type = eSymbolTypeCompiler;
3122 break;
3123
3124 case N_OLEVEL:
3125 // compiler -O level: name,,NO_SECT,0,0
3126 type = eSymbolTypeCompiler;
3127 break;
3128
3129 case N_PSYM:
3130 // parameter: name,,NO_SECT,type,offset
3131 type = eSymbolTypeVariable;
3132 break;
3133
3134 case N_ENTRY:
3135 // alternate entry: name,,n_sect,linenumber,address
3136 symbol_section = section_info.GetSection(nlist.n_sect,
3137 nlist.n_value);
3138 type = eSymbolTypeLineEntry;
3139 break;
3140
3141 // Left and Right Braces
3142 case N_LBRAC:
3143 // left bracket: 0,,NO_SECT,nesting level,address We
3144 // use the current number of symbols in the symbol
3145 // table in lieu of using nlist_idx in case we ever
3146 // start trimming entries out
3147 symbol_section = section_info.GetSection(nlist.n_sect,
3148 nlist.n_value);
3149 N_BRAC_indexes.push_back(sym_idx);
3150 type = eSymbolTypeScopeBegin;
3151 break;
3152
3153 case N_RBRAC:
3154 // right bracket: 0,,NO_SECT,nesting level,address
3155 // Set the size of the N_LBRAC to the terminating
3156 // index of this N_RBRAC so that we can always skip
3157 // the entire symbol if we need to navigate more
3158 // quickly at the source level when parsing STABS
3159 symbol_section = section_info.GetSection(nlist.n_sect,
3160 nlist.n_value);
3161 if (!N_BRAC_indexes.empty()) {
3162 symbol_ptr =
3163 symtab.SymbolAtIndex(N_BRAC_indexes.back());
3164 symbol_ptr->SetByteSize(sym_idx + 1);
3165 symbol_ptr->SetSizeIsSibling(true);
3166 N_BRAC_indexes.pop_back();
3167 }
3168 type = eSymbolTypeScopeEnd;
3169 break;
3170
3171 case N_EXCL:
3172 // deleted include file: name,,NO_SECT,0,sum
3173 type = eSymbolTypeHeaderFile;
3174 break;
3175
3176 // COMM scopes
3177 case N_BCOMM:
3178 // begin common: name,,NO_SECT,0,0
3179 // We use the current number of symbols in the symbol
3180 // table in lieu of using nlist_idx in case we ever
3181 // start trimming entries out
3182 type = eSymbolTypeScopeBegin;
3183 N_COMM_indexes.push_back(sym_idx);
3184 break;
3185
3186 case N_ECOML:
3187 // end common (local name): 0,,n_sect,0,address
3188 symbol_section = section_info.GetSection(nlist.n_sect,
3189 nlist.n_value);
3190 // Fall through
3191
3192 case N_ECOMM:
3193 // end common: name,,n_sect,0,0
3194 // Set the size of the N_BCOMM to the terminating
3195 // index of this N_ECOMM/N_ECOML so that we can
3196 // always skip the entire symbol if we need to
3197 // navigate more quickly at the source level when
3198 // parsing STABS
3199 if (!N_COMM_indexes.empty()) {
3200 symbol_ptr =
3201 symtab.SymbolAtIndex(N_COMM_indexes.back());
3202 symbol_ptr->SetByteSize(sym_idx + 1);
3203 symbol_ptr->SetSizeIsSibling(true);
3204 N_COMM_indexes.pop_back();
3205 }
3206 type = eSymbolTypeScopeEnd;
3207 break;
3208
3209 case N_LENG:
3210 // second stab entry with length information
3211 type = eSymbolTypeAdditional;
3212 break;
3213
3214 default:
3215 break;
3216 }
3217 } else {
3218 // uint8_t n_pext = N_PEXT & nlist.n_type;
3219 uint8_t n_type = N_TYPE & nlist.n_type;
3220 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0);
3221
3222 switch (n_type) {
3223 case N_INDR: {
3224 const char *reexport_name_cstr =
3225 strtab_data.PeekCStr(nlist.n_value);
3226 if (reexport_name_cstr && reexport_name_cstr[0]) {
3227 type = eSymbolTypeReExported;
3228 ConstString reexport_name(
3229 reexport_name_cstr +
3230 ((reexport_name_cstr[0] == '_') ? 1 : 0));
3231 sym[sym_idx].SetReExportedSymbolName(reexport_name);
3232 set_value = false;
3233 reexport_shlib_needs_fixup[sym_idx] = reexport_name;
3234 indirect_symbol_names.insert(ConstString(
3235 symbol_name + ((symbol_name[0] == '_') ? 1 : 0)));
3236 } else
3237 type = eSymbolTypeUndefined;
3238 } break;
3239
3240 case N_UNDF:
3241 if (symbol_name && symbol_name[0]) {
3242 ConstString undefined_name(
3243 symbol_name + ((symbol_name[0] == '_') ? 1 : 0));
3244 undefined_name_to_desc[undefined_name] = nlist.n_desc;
3245 }
3246 // Fall through
3247 case N_PBUD:
3248 type = eSymbolTypeUndefined;
3249 break;
3250
3251 case N_ABS:
3252 type = eSymbolTypeAbsolute;
3253 break;
3254
3255 case N_SECT: {
3256 symbol_section = section_info.GetSection(nlist.n_sect,
3257 nlist.n_value);
3258
3259 if (symbol_section == NULL) {
3260 // TODO: warn about this?
3261 add_nlist = false;
3262 break;
3263 }
3264
3265 if (TEXT_eh_frame_sectID == nlist.n_sect) {
3266 type = eSymbolTypeException;
3267 } else {
3268 uint32_t section_type =
3269 symbol_section->Get() & SECTION_TYPE;
3270
3271 switch (section_type) {
3272 case S_CSTRING_LITERALS:
3273 type = eSymbolTypeData;
3274 break; // section with only literal C strings
3275 case S_4BYTE_LITERALS:
3276 type = eSymbolTypeData;
3277 break; // section with only 4 byte literals
3278 case S_8BYTE_LITERALS:
3279 type = eSymbolTypeData;
3280 break; // section with only 8 byte literals
3281 case S_LITERAL_POINTERS:
3282 type = eSymbolTypeTrampoline;
3283 break; // section with only pointers to literals
3284 case S_NON_LAZY_SYMBOL_POINTERS:
3285 type = eSymbolTypeTrampoline;
3286 break; // section with only non-lazy symbol
3287 // pointers
3288 case S_LAZY_SYMBOL_POINTERS:
3289 type = eSymbolTypeTrampoline;
3290 break; // section with only lazy symbol pointers
3291 case S_SYMBOL_STUBS:
3292 type = eSymbolTypeTrampoline;
3293 break; // section with only symbol stubs, byte
3294 // size of stub in the reserved2 field
3295 case S_MOD_INIT_FUNC_POINTERS:
3296 type = eSymbolTypeCode;
3297 break; // section with only function pointers for
3298 // initialization
3299 case S_MOD_TERM_FUNC_POINTERS:
3300 type = eSymbolTypeCode;
3301 break; // section with only function pointers for
3302 // termination
3303 case S_INTERPOSING:
3304 type = eSymbolTypeTrampoline;
3305 break; // section with only pairs of function
3306 // pointers for interposing
3307 case S_16BYTE_LITERALS:
3308 type = eSymbolTypeData;
3309 break; // section with only 16 byte literals
3310 case S_DTRACE_DOF:
3311 type = eSymbolTypeInstrumentation;
3312 break;
3313 case S_LAZY_DYLIB_SYMBOL_POINTERS:
3314 type = eSymbolTypeTrampoline;
3315 break;
3316 default:
3317 switch (symbol_section->GetType()) {
3318 case lldb::eSectionTypeCode:
3319 type = eSymbolTypeCode;
3320 break;
3321 case eSectionTypeData:
3322 case eSectionTypeDataCString: // Inlined C string
3323 // data
3324 case eSectionTypeDataCStringPointers: // Pointers
3325 // to C
3326 // string
3327 // data
3328 case eSectionTypeDataSymbolAddress: // Address of
3329 // a symbol in
3330 // the symbol
3331 // table
3332 case eSectionTypeData4:
3333 case eSectionTypeData8:
3334 case eSectionTypeData16:
3335 type = eSymbolTypeData;
3336 break;
3337 default:
3338 break;
3339 }
3340 break;
3341 }
3342
3343 if (type == eSymbolTypeInvalid) {
3344 const char *symbol_sect_name =
3345 symbol_section->GetName().AsCString();
3346 if (symbol_section->IsDescendant(
3347 text_section_sp.get())) {
3348 if (symbol_section->IsClear(
3349 S_ATTR_PURE_INSTRUCTIONS |
3350 S_ATTR_SELF_MODIFYING_CODE |
3351 S_ATTR_SOME_INSTRUCTIONS))
3352 type = eSymbolTypeData;
3353 else
3354 type = eSymbolTypeCode;
3355 } else if (symbol_section->IsDescendant(
3356 data_section_sp.get()) ||
3357 symbol_section->IsDescendant(
3358 data_dirty_section_sp.get()) ||
3359 symbol_section->IsDescendant(
3360 data_const_section_sp.get())) {
3361 if (symbol_sect_name &&
3362 ::strstr(symbol_sect_name, "__objc") ==
3363 symbol_sect_name) {
3364 type = eSymbolTypeRuntime;
3365
3366 if (symbol_name) {
3367 llvm::StringRef symbol_name_ref(symbol_name);
3368 if (symbol_name_ref.startswith("_OBJC_")) {
3369 llvm::StringRef
3370 g_objc_v2_prefix_class(
3371 "_OBJC_CLASS_$_");
3372 llvm::StringRef
3373 g_objc_v2_prefix_metaclass(
3374 "_OBJC_METACLASS_$_");
3375 llvm::StringRef
3376 g_objc_v2_prefix_ivar("_OBJC_IVAR_$_");
3377 if (symbol_name_ref.startswith(
3378 g_objc_v2_prefix_class)) {
3379 symbol_name_non_abi_mangled =
3380 symbol_name + 1;
3381 symbol_name =
3382 symbol_name +
3383 g_objc_v2_prefix_class.size();
3384 type = eSymbolTypeObjCClass;
3385 demangled_is_synthesized = true;
3386 } else if (
3387 symbol_name_ref.startswith(
3388 g_objc_v2_prefix_metaclass)) {
3389 symbol_name_non_abi_mangled =
3390 symbol_name + 1;
3391 symbol_name =
3392 symbol_name +
3393 g_objc_v2_prefix_metaclass.size();
3394 type = eSymbolTypeObjCMetaClass;
3395 demangled_is_synthesized = true;
3396 } else if (symbol_name_ref.startswith(
3397 g_objc_v2_prefix_ivar)) {
3398 symbol_name_non_abi_mangled =
3399 symbol_name + 1;
3400 symbol_name =
3401 symbol_name +
3402 g_objc_v2_prefix_ivar.size();
3403 type = eSymbolTypeObjCIVar;
3404 demangled_is_synthesized = true;
3405 }
3406 }
3407 }
3408 } else if (symbol_sect_name &&
3409 ::strstr(symbol_sect_name,
3410 "__gcc_except_tab") ==
3411 symbol_sect_name) {
3412 type = eSymbolTypeException;
3413 } else {
3414 type = eSymbolTypeData;
3415 }
3416 } else if (symbol_sect_name &&
3417 ::strstr(symbol_sect_name, "__IMPORT") ==
3418 symbol_sect_name) {
3419 type = eSymbolTypeTrampoline;
3420 } else if (symbol_section->IsDescendant(
3421 objc_section_sp.get())) {
3422 type = eSymbolTypeRuntime;
3423 if (symbol_name && symbol_name[0] == '.') {
3424 llvm::StringRef symbol_name_ref(symbol_name);
3425 llvm::StringRef
3426 g_objc_v1_prefix_class(".objc_class_name_");
3427 if (symbol_name_ref.startswith(
3428 g_objc_v1_prefix_class)) {
3429 symbol_name_non_abi_mangled = symbol_name;
3430 symbol_name = symbol_name +
3431 g_objc_v1_prefix_class.size();
3432 type = eSymbolTypeObjCClass;
3433 demangled_is_synthesized = true;
3434 }
3435 }
3436 }
3437 }
3438 }
3439 } break;
3440 }
3441 }
3442
3443 if (add_nlist) {
3444 uint64_t symbol_value = nlist.n_value;
3445 if (symbol_name_non_abi_mangled) {
3446 sym[sym_idx].GetMangled().SetMangledName(
3447 ConstString(symbol_name_non_abi_mangled));
3448 sym[sym_idx].GetMangled().SetDemangledName(
3449 ConstString(symbol_name));
3450 } else {
3451 bool symbol_name_is_mangled = false;
3452
3453 if (symbol_name && symbol_name[0] == '_') {
3454 symbol_name_is_mangled = symbol_name[1] == '_';
3455 symbol_name++; // Skip the leading underscore
3456 }
3457
3458 if (symbol_name) {
3459 ConstString const_symbol_name(symbol_name);
3460 sym[sym_idx].GetMangled().SetValue(
3461 const_symbol_name, symbol_name_is_mangled);
3462 if (is_gsym && is_debug) {
3463 const char *gsym_name =
3464 sym[sym_idx]
3465 .GetMangled()
3466 .GetName(Mangled::ePreferMangled)
3467 .GetCString();
3468 if (gsym_name)
3469 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx;
3470 }
3471 }
3472 }
3473 if (symbol_section) {
3474 const addr_t section_file_addr =
3475 symbol_section->GetFileAddress();
3476 if (symbol_byte_size == 0 &&
3477 function_starts_count > 0) {
3478 addr_t symbol_lookup_file_addr = nlist.n_value;
3479 // Do an exact address match for non-ARM addresses,
3480 // else get the closest since the symbol might be a
3481 // thumb symbol which has an address with bit zero
3482 // set
3483 FunctionStarts::Entry *func_start_entry =
3484 function_starts.FindEntry(symbol_lookup_file_addr,
3485 !is_arm);
3486 if (is_arm && func_start_entry) {
3487 // Verify that the function start address is the
3488 // symbol address (ARM) or the symbol address + 1
3489 // (thumb)
3490 if (func_start_entry->addr !=
3491 symbol_lookup_file_addr &&
3492 func_start_entry->addr !=
3493 (symbol_lookup_file_addr + 1)) {
3494 // Not the right entry, NULL it out...
3495 func_start_entry = NULL;
3496 }
3497 }
3498 if (func_start_entry) {
3499 func_start_entry->data = true;
3500
3501 addr_t symbol_file_addr = func_start_entry->addr;
3502 uint32_t symbol_flags = 0;
3503 if (is_arm) {
3504 if (symbol_file_addr & 1)
3505 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB;
3506 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK;
3507 }
3508
3509 const FunctionStarts::Entry *next_func_start_entry =
3510 function_starts.FindNextEntry(func_start_entry);
3511 const addr_t section_end_file_addr =
3512 section_file_addr +
3513 symbol_section->GetByteSize();
3514 if (next_func_start_entry) {
3515 addr_t next_symbol_file_addr =
3516 next_func_start_entry->addr;
3517 // Be sure the clear the Thumb address bit when
3518 // we calculate the size from the current and
3519 // next address
3520 if (is_arm)
3521 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK;
3522 symbol_byte_size = std::min<lldb::addr_t>(
3523 next_symbol_file_addr - symbol_file_addr,
3524 section_end_file_addr - symbol_file_addr);
3525 } else {
3526 symbol_byte_size =
3527 section_end_file_addr - symbol_file_addr;
3528 }
3529 }
3530 }
3531 symbol_value -= section_file_addr;
3532 }
3533
3534 if (is_debug == false) {
3535 if (type == eSymbolTypeCode) {
3536 // See if we can find a N_FUN entry for any code
3537 // symbols. If we do find a match, and the name
3538 // matches, then we can merge the two into just the
3539 // function symbol to avoid duplicate entries in
3540 // the symbol table
3541 auto range =
3542 N_FUN_addr_to_sym_idx.equal_range(nlist.n_value);
3543 if (range.first != range.second) {
3544 bool found_it = false;
3545 for (auto pos = range.first; pos != range.second;
3546 ++pos) {
3547 if (sym[sym_idx].GetMangled().GetName(
3548 Mangled::ePreferMangled) ==
3549 sym[pos->second].GetMangled().GetName(
3550 Mangled::ePreferMangled)) {
3551 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second;
3552 // We just need the flags from the linker
3553 // symbol, so put these flags
3554 // into the N_FUN flags to avoid duplicate
3555 // symbols in the symbol table
3556 sym[pos->second].SetExternal(
3557 sym[sym_idx].IsExternal());
3558 sym[pos->second].SetFlags(nlist.n_type << 16 |
3559 nlist.n_desc);
3560 if (resolver_addresses.find(nlist.n_value) !=
3561 resolver_addresses.end())
3562 sym[pos->second].SetType(eSymbolTypeResolver);
3563 sym[sym_idx].Clear();
3564 found_it = true;
3565 break;
3566 }
3567 }
3568 if (found_it)
3569 continue;
3570 } else {
3571 if (resolver_addresses.find(nlist.n_value) !=
3572 resolver_addresses.end())
3573 type = eSymbolTypeResolver;
3574 }
3575 } else if (type == eSymbolTypeData ||
3576 type == eSymbolTypeObjCClass ||
3577 type == eSymbolTypeObjCMetaClass ||
3578 type == eSymbolTypeObjCIVar) {
3579 // See if we can find a N_STSYM entry for any data
3580 // symbols. If we do find a match, and the name
3581 // matches, then we can merge the two into just the
3582 // Static symbol to avoid duplicate entries in the
3583 // symbol table
3584 auto range = N_STSYM_addr_to_sym_idx.equal_range(
3585 nlist.n_value);
3586 if (range.first != range.second) {
3587 bool found_it = false;
3588 for (auto pos = range.first; pos != range.second;
3589 ++pos) {
3590 if (sym[sym_idx].GetMangled().GetName(
3591 Mangled::ePreferMangled) ==
3592 sym[pos->second].GetMangled().GetName(
3593 Mangled::ePreferMangled)) {
3594 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second;
3595 // We just need the flags from the linker
3596 // symbol, so put these flags
3597 // into the N_STSYM flags to avoid duplicate
3598 // symbols in the symbol table
3599 sym[pos->second].SetExternal(
3600 sym[sym_idx].IsExternal());
3601 sym[pos->second].SetFlags(nlist.n_type << 16 |
3602 nlist.n_desc);
3603 sym[sym_idx].Clear();
3604 found_it = true;
3605 break;
3606 }
3607 }
3608 if (found_it)
3609 continue;
3610 } else {
3611 const char *gsym_name =
3612 sym[sym_idx]
3613 .GetMangled()
3614 .GetName(Mangled::ePreferMangled)
3615 .GetCString();
3616 if (gsym_name) {
3617 // Combine N_GSYM stab entries with the non
3618 // stab symbol
3619 ConstNameToSymbolIndexMap::const_iterator pos =
3620 N_GSYM_name_to_sym_idx.find(gsym_name);
3621 if (pos != N_GSYM_name_to_sym_idx.end()) {
3622 const uint32_t GSYM_sym_idx = pos->second;
3623 m_nlist_idx_to_sym_idx[nlist_idx] =
3624 GSYM_sym_idx;
3625 // Copy the address, because often the N_GSYM
3626 // address has an invalid address of zero
3627 // when the global is a common symbol
3628 sym[GSYM_sym_idx].GetAddressRef().SetSection(
3629 symbol_section);
3630 sym[GSYM_sym_idx].GetAddressRef().SetOffset(
3631 symbol_value);
3632 add_symbol_addr(sym[GSYM_sym_idx]
3633 .GetAddress()
3634 .GetFileAddress());
3635 // We just need the flags from the linker
3636 // symbol, so put these flags
3637 // into the N_GSYM flags to avoid duplicate
3638 // symbols in the symbol table
3639 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 |
3640 nlist.n_desc);
3641 sym[sym_idx].Clear();
3642 continue;
3643 }
3644 }
3645 }
3646 }
3647 }
3648
3649 sym[sym_idx].SetID(nlist_idx);
3650 sym[sym_idx].SetType(type);
3651 if (set_value) {
3652 sym[sym_idx].GetAddressRef().SetSection(symbol_section);
3653 sym[sym_idx].GetAddressRef().SetOffset(symbol_value);
3654 add_symbol_addr(
3655 sym[sym_idx].GetAddress().GetFileAddress());
3656 }
3657 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc);
3658
3659 if (symbol_byte_size > 0)
3660 sym[sym_idx].SetByteSize(symbol_byte_size);
3661
3662 if (demangled_is_synthesized)
3663 sym[sym_idx].SetDemangledNameIsSynthesized(true);
3664 ++sym_idx;
3665 } else {
3666 sym[sym_idx].Clear();
3667 }
3668 }
3669 /////////////////////////////
3670 }
3671 }
3672
3673 for (const auto &pos : reexport_shlib_needs_fixup) {
3674 const auto undef_pos = undefined_name_to_desc.find(pos.second);
3675 if (undef_pos != undefined_name_to_desc.end()) {
3676 const uint8_t dylib_ordinal =
3677 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second);
3678 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize())
3679 sym[pos.first].SetReExportedSymbolSharedLibrary(
3680 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1));
3681 }
3682 }
3683 }
3684
3685 #endif
3686 lldb::offset_t nlist_data_offset = 0;
3687
3688 if (nlist_data.GetByteSize() > 0) {
3689
3690 // If the sym array was not created while parsing the DSC unmapped
3691 // symbols, create it now.
3692 if (sym == nullptr) {
3693 sym =
3694 symtab.Resize(symtab_load_command.nsyms + m_dysymtab.nindirectsyms);
3695 num_syms = symtab.GetNumSymbols();
3696 }
3697
3698 if (unmapped_local_symbols_found) {
3699 assert(m_dysymtab.ilocalsym == 0);
3700 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size);
3701 nlist_idx = m_dysymtab.nlocalsym;
3702 } else {
3703 nlist_idx = 0;
3704 }
3705
3706 typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap;
3707 typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName;
3708 UndefinedNameToDescMap undefined_name_to_desc;
3709 SymbolIndexToName reexport_shlib_needs_fixup;
3710
3711 // Symtab parsing is a huge mess. Everything is entangled and the code
3712 // requires access to a ridiculous amount of variables. LLDB depends
3713 // heavily on the proper merging of symbols and to get that right we need
3714 // to make sure we have parsed all the debug symbols first. Therefore we
3715 // invoke the lambda twice, once to parse only the debug symbols and then
3716 // once more to parse the remaining symbols.
3717 auto ParseSymbolLambda = [&](struct nlist_64 &nlist, uint32_t nlist_idx,
3718 bool debug_only) {
3719 const bool is_debug = ((nlist.n_type & N_STAB) != 0);
3720 if (is_debug != debug_only)
3721 return true;
3722
3723 const char *symbol_name_non_abi_mangled = nullptr;
3724 const char *symbol_name = nullptr;
3725
3726 if (have_strtab_data) {
3727 symbol_name = strtab_data.PeekCStr(nlist.n_strx);
3728
3729 if (symbol_name == nullptr) {
3730 // No symbol should be NULL, even the symbols with no string values
3731 // should have an offset zero which points to an empty C-string
3732 Debugger::ReportError(llvm::formatv(
3733 "symbol[{0}] has invalid string table offset {1:x} in {2}, "
3734 "ignoring symbol",
3735 nlist_idx, nlist.n_strx, module_sp->GetFileSpec().GetPath()));
3736 return true;
3737 }
3738 if (symbol_name[0] == '\0')
3739 symbol_name = nullptr;
3740 } else {
3741 const addr_t str_addr = strtab_addr + nlist.n_strx;
3742 Status str_error;
3743 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name,
3744 str_error))
3745 symbol_name = memory_symbol_name.c_str();
3746 }
3747
3748 SymbolType type = eSymbolTypeInvalid;
3749 SectionSP symbol_section;
3750 lldb::addr_t symbol_byte_size = 0;
3751 bool add_nlist = true;
3752 bool is_gsym = false;
3753 bool demangled_is_synthesized = false;
3754 bool set_value = true;
3755
3756 assert(sym_idx < num_syms);
3757 sym[sym_idx].SetDebug(is_debug);
3758
3759 if (is_debug) {
3760 switch (nlist.n_type) {
3761 case N_GSYM:
3762 // global symbol: name,,NO_SECT,type,0
3763 // Sometimes the N_GSYM value contains the address.
3764
3765 // FIXME: In the .o files, we have a GSYM and a debug symbol for all
3766 // the ObjC data. They
3767 // have the same address, but we want to ensure that we always find
3768 // only the real symbol, 'cause we don't currently correctly
3769 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol
3770 // type. This is a temporary hack to make sure the ObjectiveC
3771 // symbols get treated correctly. To do this right, we should
3772 // coalesce all the GSYM & global symbols that have the same
3773 // address.
3774 is_gsym = true;
3775 sym[sym_idx].SetExternal(true);
3776
3777 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') {
3778 llvm::StringRef symbol_name_ref(symbol_name);
3779 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) {
3780 symbol_name_non_abi_mangled = symbol_name + 1;
3781 symbol_name = symbol_name + g_objc_v2_prefix_class.size();
3782 type = eSymbolTypeObjCClass;
3783 demangled_is_synthesized = true;
3784
3785 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) {
3786 symbol_name_non_abi_mangled = symbol_name + 1;
3787 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size();
3788 type = eSymbolTypeObjCMetaClass;
3789 demangled_is_synthesized = true;
3790 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) {
3791 symbol_name_non_abi_mangled = symbol_name + 1;
3792 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size();
3793 type = eSymbolTypeObjCIVar;
3794 demangled_is_synthesized = true;
3795 }
3796 } else {
3797 if (nlist.n_value != 0)
3798 symbol_section =
3799 section_info.GetSection(nlist.n_sect, nlist.n_value);
3800 type = eSymbolTypeData;
3801 }
3802 break;
3803
3804 case N_FNAME:
3805 // procedure name (f77 kludge): name,,NO_SECT,0,0
3806 type = eSymbolTypeCompiler;
3807 break;
3808
3809 case N_FUN:
3810 // procedure: name,,n_sect,linenumber,address
3811 if (symbol_name) {
3812 type = eSymbolTypeCode;
3813 symbol_section =
3814 section_info.GetSection(nlist.n_sect, nlist.n_value);
3815
3816 N_FUN_addr_to_sym_idx.insert(
3817 std::make_pair(nlist.n_value, sym_idx));
3818 // We use the current number of symbols in the symbol table in
3819 // lieu of using nlist_idx in case we ever start trimming entries
3820 // out
3821 N_FUN_indexes.push_back(sym_idx);
3822 } else {
3823 type = eSymbolTypeCompiler;
3824
3825 if (!N_FUN_indexes.empty()) {
3826 // Copy the size of the function into the original STAB entry
3827 // so we don't have to hunt for it later
3828 symtab.SymbolAtIndex(N_FUN_indexes.back())
3829 ->SetByteSize(nlist.n_value);
3830 N_FUN_indexes.pop_back();
3831 // We don't really need the end function STAB as it contains
3832 // the size which we already placed with the original symbol,
3833 // so don't add it if we want a minimal symbol table
3834 add_nlist = false;
3835 }
3836 }
3837 break;
3838
3839 case N_STSYM:
3840 // static symbol: name,,n_sect,type,address
3841 N_STSYM_addr_to_sym_idx.insert(
3842 std::make_pair(nlist.n_value, sym_idx));
3843 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value);
3844 if (symbol_name && symbol_name[0]) {
3845 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1,
3846 eSymbolTypeData);
3847 }
3848 break;
3849
3850 case N_LCSYM:
3851 // .lcomm symbol: name,,n_sect,type,address
3852 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value);
3853 type = eSymbolTypeCommonBlock;
3854 break;
3855
3856 case N_BNSYM:
3857 // We use the current number of symbols in the symbol table in lieu
3858 // of using nlist_idx in case we ever start trimming entries out
3859 // Skip these if we want minimal symbol tables
3860 add_nlist = false;
3861 break;
3862
3863 case N_ENSYM:
3864 // Set the size of the N_BNSYM to the terminating index of this
3865 // N_ENSYM so that we can always skip the entire symbol if we need
3866 // to navigate more quickly at the source level when parsing STABS
3867 // Skip these if we want minimal symbol tables
3868 add_nlist = false;
3869 break;
3870
3871 case N_OPT:
3872 // emitted with gcc2_compiled and in gcc source
3873 type = eSymbolTypeCompiler;
3874 break;
3875
3876 case N_RSYM:
3877 // register sym: name,,NO_SECT,type,register
3878 type = eSymbolTypeVariable;
3879 break;
3880
3881 case N_SLINE:
3882 // src line: 0,,n_sect,linenumber,address
3883 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value);
3884 type = eSymbolTypeLineEntry;
3885 break;
3886
3887 case N_SSYM:
3888 // structure elt: name,,NO_SECT,type,struct_offset
3889 type = eSymbolTypeVariableType;
3890 break;
3891
3892 case N_SO:
3893 // source file name
3894 type = eSymbolTypeSourceFile;
3895 if (symbol_name == nullptr) {
3896 add_nlist = false;
3897 if (N_SO_index != UINT32_MAX) {
3898 // Set the size of the N_SO to the terminating index of this
3899 // N_SO so that we can always skip the entire N_SO if we need
3900 // to navigate more quickly at the source level when parsing
3901 // STABS
3902 symbol_ptr = symtab.SymbolAtIndex(N_SO_index);
3903 symbol_ptr->SetByteSize(sym_idx);
3904 symbol_ptr->SetSizeIsSibling(true);
3905 }
3906 N_NSYM_indexes.clear();
3907 N_INCL_indexes.clear();
3908 N_BRAC_indexes.clear();
3909 N_COMM_indexes.clear();
3910 N_FUN_indexes.clear();
3911 N_SO_index = UINT32_MAX;
3912 } else {
3913 // We use the current number of symbols in the symbol table in
3914 // lieu of using nlist_idx in case we ever start trimming entries
3915 // out
3916 const bool N_SO_has_full_path = symbol_name[0] == '/';
3917 if (N_SO_has_full_path) {
3918 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) {
3919 // We have two consecutive N_SO entries where the first
3920 // contains a directory and the second contains a full path.
3921 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name),
3922 false);
3923 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1;
3924 add_nlist = false;
3925 } else {
3926 // This is the first entry in a N_SO that contains a
3927 // directory or a full path to the source file
3928 N_SO_index = sym_idx;
3929 }
3930 } else if ((N_SO_index == sym_idx - 1) &&
3931 ((sym_idx - 1) < num_syms)) {
3932 // This is usually the second N_SO entry that contains just the
3933 // filename, so here we combine it with the first one if we are
3934 // minimizing the symbol table
3935 const char *so_path =
3936 sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString();
3937 if (so_path && so_path[0]) {
3938 std::string full_so_path(so_path);
3939 const size_t double_slash_pos = full_so_path.find("//");
3940 if (double_slash_pos != std::string::npos) {
3941 // The linker has been generating bad N_SO entries with
3942 // doubled up paths in the format "%s%s" where the first
3943 // string in the DW_AT_comp_dir, and the second is the
3944 // directory for the source file so you end up with a path
3945 // that looks like "/tmp/src//tmp/src/"
3946 FileSpec so_dir(so_path);
3947 if (!FileSystem::Instance().Exists(so_dir)) {
3948 so_dir.SetFile(&full_so_path[double_slash_pos + 1],
3949 FileSpec::Style::native);
3950 if (FileSystem::Instance().Exists(so_dir)) {
3951 // Trim off the incorrect path
3952 full_so_path.erase(0, double_slash_pos + 1);
3953 }
3954 }
3955 }
3956 if (*full_so_path.rbegin() != '/')
3957 full_so_path += '/';
3958 full_so_path += symbol_name;
3959 sym[sym_idx - 1].GetMangled().SetValue(
3960 ConstString(full_so_path.c_str()), false);
3961 add_nlist = false;
3962 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1;
3963 }
3964 } else {
3965 // This could be a relative path to a N_SO
3966 N_SO_index = sym_idx;
3967 }
3968 }
3969 break;
3970
3971 case N_OSO:
3972 // object file name: name,,0,0,st_mtime
3973 type = eSymbolTypeObjectFile;
3974 break;
3975
3976 case N_LSYM:
3977 // local sym: name,,NO_SECT,type,offset
3978 type = eSymbolTypeLocal;
3979 break;
3980
3981 // INCL scopes
3982 case N_BINCL:
3983 // include file beginning: name,,NO_SECT,0,sum We use the current
3984 // number of symbols in the symbol table in lieu of using nlist_idx
3985 // in case we ever start trimming entries out
3986 N_INCL_indexes.push_back(sym_idx);
3987 type = eSymbolTypeScopeBegin;
3988 break;
3989
3990 case N_EINCL:
3991 // include file end: name,,NO_SECT,0,0
3992 // Set the size of the N_BINCL to the terminating index of this
3993 // N_EINCL so that we can always skip the entire symbol if we need
3994 // to navigate more quickly at the source level when parsing STABS
3995 if (!N_INCL_indexes.empty()) {
3996 symbol_ptr = symtab.SymbolAtIndex(N_INCL_indexes.back());
3997 symbol_ptr->SetByteSize(sym_idx + 1);
3998 symbol_ptr->SetSizeIsSibling(true);
3999 N_INCL_indexes.pop_back();
4000 }
4001 type = eSymbolTypeScopeEnd;
4002 break;
4003
4004 case N_SOL:
4005 // #included file name: name,,n_sect,0,address
4006 type = eSymbolTypeHeaderFile;
4007
4008 // We currently don't use the header files on darwin
4009 add_nlist = false;
4010 break;
4011
4012 case N_PARAMS:
4013 // compiler parameters: name,,NO_SECT,0,0
4014 type = eSymbolTypeCompiler;
4015 break;
4016
4017 case N_VERSION:
4018 // compiler version: name,,NO_SECT,0,0
4019 type = eSymbolTypeCompiler;
4020 break;
4021
4022 case N_OLEVEL:
4023 // compiler -O level: name,,NO_SECT,0,0
4024 type = eSymbolTypeCompiler;
4025 break;
4026
4027 case N_PSYM:
4028 // parameter: name,,NO_SECT,type,offset
4029 type = eSymbolTypeVariable;
4030 break;
4031
4032 case N_ENTRY:
4033 // alternate entry: name,,n_sect,linenumber,address
4034 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value);
4035 type = eSymbolTypeLineEntry;
4036 break;
4037
4038 // Left and Right Braces
4039 case N_LBRAC:
4040 // left bracket: 0,,NO_SECT,nesting level,address We use the
4041 // current number of symbols in the symbol table in lieu of using
4042 // nlist_idx in case we ever start trimming entries out
4043 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value);
4044 N_BRAC_indexes.push_back(sym_idx);
4045 type = eSymbolTypeScopeBegin;
4046 break;
4047
4048 case N_RBRAC:
4049 // right bracket: 0,,NO_SECT,nesting level,address Set the size of
4050 // the N_LBRAC to the terminating index of this N_RBRAC so that we
4051 // can always skip the entire symbol if we need to navigate more
4052 // quickly at the source level when parsing STABS
4053 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value);
4054 if (!N_BRAC_indexes.empty()) {
4055 symbol_ptr = symtab.SymbolAtIndex(N_BRAC_indexes.back());
4056 symbol_ptr->SetByteSize(sym_idx + 1);
4057 symbol_ptr->SetSizeIsSibling(true);
4058 N_BRAC_indexes.pop_back();
4059 }
4060 type = eSymbolTypeScopeEnd;
4061 break;
4062
4063 case N_EXCL:
4064 // deleted include file: name,,NO_SECT,0,sum
4065 type = eSymbolTypeHeaderFile;
4066 break;
4067
4068 // COMM scopes
4069 case N_BCOMM:
4070 // begin common: name,,NO_SECT,0,0
4071 // We use the current number of symbols in the symbol table in lieu
4072 // of using nlist_idx in case we ever start trimming entries out
4073 type = eSymbolTypeScopeBegin;
4074 N_COMM_indexes.push_back(sym_idx);
4075 break;
4076
4077 case N_ECOML:
4078 // end common (local name): 0,,n_sect,0,address
4079 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value);
4080 LLVM_FALLTHROUGH;
4081
4082 case N_ECOMM:
4083 // end common: name,,n_sect,0,0
4084 // Set the size of the N_BCOMM to the terminating index of this
4085 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if
4086 // we need to navigate more quickly at the source level when
4087 // parsing STABS
4088 if (!N_COMM_indexes.empty()) {
4089 symbol_ptr = symtab.SymbolAtIndex(N_COMM_indexes.back());
4090 symbol_ptr->SetByteSize(sym_idx + 1);
4091 symbol_ptr->SetSizeIsSibling(true);
4092 N_COMM_indexes.pop_back();
4093 }
4094 type = eSymbolTypeScopeEnd;
4095 break;
4096
4097 case N_LENG:
4098 // second stab entry with length information
4099 type = eSymbolTypeAdditional;
4100 break;
4101
4102 default:
4103 break;
4104 }
4105 } else {
4106 uint8_t n_type = N_TYPE & nlist.n_type;
4107 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0);
4108
4109 switch (n_type) {
4110 case N_INDR: {
4111 const char *reexport_name_cstr = strtab_data.PeekCStr(nlist.n_value);
4112 if (reexport_name_cstr && reexport_name_cstr[0]) {
4113 type = eSymbolTypeReExported;
4114 ConstString reexport_name(reexport_name_cstr +
4115 ((reexport_name_cstr[0] == '_') ? 1 : 0));
4116 sym[sym_idx].SetReExportedSymbolName(reexport_name);
4117 set_value = false;
4118 reexport_shlib_needs_fixup[sym_idx] = reexport_name;
4119 indirect_symbol_names.insert(
4120 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0)));
4121 } else
4122 type = eSymbolTypeUndefined;
4123 } break;
4124
4125 case N_UNDF:
4126 if (symbol_name && symbol_name[0]) {
4127 ConstString undefined_name(symbol_name +
4128 ((symbol_name[0] == '_') ? 1 : 0));
4129 undefined_name_to_desc[undefined_name] = nlist.n_desc;
4130 }
4131 LLVM_FALLTHROUGH;
4132
4133 case N_PBUD:
4134 type = eSymbolTypeUndefined;
4135 break;
4136
4137 case N_ABS:
4138 type = eSymbolTypeAbsolute;
4139 break;
4140
4141 case N_SECT: {
4142 symbol_section = section_info.GetSection(nlist.n_sect, nlist.n_value);
4143
4144 if (!symbol_section) {
4145 // TODO: warn about this?
4146 add_nlist = false;
4147 break;
4148 }
4149
4150 if (TEXT_eh_frame_sectID == nlist.n_sect) {
4151 type = eSymbolTypeException;
4152 } else {
4153 uint32_t section_type = symbol_section->Get() & SECTION_TYPE;
4154
4155 switch (section_type) {
4156 case S_CSTRING_LITERALS:
4157 type = eSymbolTypeData;
4158 break; // section with only literal C strings
4159 case S_4BYTE_LITERALS:
4160 type = eSymbolTypeData;
4161 break; // section with only 4 byte literals
4162 case S_8BYTE_LITERALS:
4163 type = eSymbolTypeData;
4164 break; // section with only 8 byte literals
4165 case S_LITERAL_POINTERS:
4166 type = eSymbolTypeTrampoline;
4167 break; // section with only pointers to literals
4168 case S_NON_LAZY_SYMBOL_POINTERS:
4169 type = eSymbolTypeTrampoline;
4170 break; // section with only non-lazy symbol pointers
4171 case S_LAZY_SYMBOL_POINTERS:
4172 type = eSymbolTypeTrampoline;
4173 break; // section with only lazy symbol pointers
4174 case S_SYMBOL_STUBS:
4175 type = eSymbolTypeTrampoline;
4176 break; // section with only symbol stubs, byte size of stub in
4177 // the reserved2 field
4178 case S_MOD_INIT_FUNC_POINTERS:
4179 type = eSymbolTypeCode;
4180 break; // section with only function pointers for initialization
4181 case S_MOD_TERM_FUNC_POINTERS:
4182 type = eSymbolTypeCode;
4183 break; // section with only function pointers for termination
4184 case S_INTERPOSING:
4185 type = eSymbolTypeTrampoline;
4186 break; // section with only pairs of function pointers for
4187 // interposing
4188 case S_16BYTE_LITERALS:
4189 type = eSymbolTypeData;
4190 break; // section with only 16 byte literals
4191 case S_DTRACE_DOF:
4192 type = eSymbolTypeInstrumentation;
4193 break;
4194 case S_LAZY_DYLIB_SYMBOL_POINTERS:
4195 type = eSymbolTypeTrampoline;
4196 break;
4197 default:
4198 switch (symbol_section->GetType()) {
4199 case lldb::eSectionTypeCode:
4200 type = eSymbolTypeCode;
4201 break;
4202 case eSectionTypeData:
4203 case eSectionTypeDataCString: // Inlined C string data
4204 case eSectionTypeDataCStringPointers: // Pointers to C string
4205 // data
4206 case eSectionTypeDataSymbolAddress: // Address of a symbol in
4207 // the symbol table
4208 case eSectionTypeData4:
4209 case eSectionTypeData8:
4210 case eSectionTypeData16:
4211 type = eSymbolTypeData;
4212 break;
4213 default:
4214 break;
4215 }
4216 break;
4217 }
4218
4219 if (type == eSymbolTypeInvalid) {
4220 const char *symbol_sect_name =
4221 symbol_section->GetName().AsCString();
4222 if (symbol_section->IsDescendant(text_section_sp.get())) {
4223 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS |
4224 S_ATTR_SELF_MODIFYING_CODE |
4225 S_ATTR_SOME_INSTRUCTIONS))
4226 type = eSymbolTypeData;
4227 else
4228 type = eSymbolTypeCode;
4229 } else if (symbol_section->IsDescendant(data_section_sp.get()) ||
4230 symbol_section->IsDescendant(
4231 data_dirty_section_sp.get()) ||
4232 symbol_section->IsDescendant(
4233 data_const_section_sp.get())) {
4234 if (symbol_sect_name &&
4235 ::strstr(symbol_sect_name, "__objc") == symbol_sect_name) {
4236 type = eSymbolTypeRuntime;
4237
4238 if (symbol_name) {
4239 llvm::StringRef symbol_name_ref(symbol_name);
4240 if (symbol_name_ref.startswith("_OBJC_")) {
4241 llvm::StringRef g_objc_v2_prefix_class(
4242 "_OBJC_CLASS_$_");
4243 llvm::StringRef g_objc_v2_prefix_metaclass(
4244 "_OBJC_METACLASS_$_");
4245 llvm::StringRef g_objc_v2_prefix_ivar(
4246 "_OBJC_IVAR_$_");
4247 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) {
4248 symbol_name_non_abi_mangled = symbol_name + 1;
4249 symbol_name =
4250 symbol_name + g_objc_v2_prefix_class.size();
4251 type = eSymbolTypeObjCClass;
4252 demangled_is_synthesized = true;
4253 } else if (symbol_name_ref.startswith(
4254 g_objc_v2_prefix_metaclass)) {
4255 symbol_name_non_abi_mangled = symbol_name + 1;
4256 symbol_name =
4257 symbol_name + g_objc_v2_prefix_metaclass.size();
4258 type = eSymbolTypeObjCMetaClass;
4259 demangled_is_synthesized = true;
4260 } else if (symbol_name_ref.startswith(
4261 g_objc_v2_prefix_ivar)) {
4262 symbol_name_non_abi_mangled = symbol_name + 1;
4263 symbol_name =
4264 symbol_name + g_objc_v2_prefix_ivar.size();
4265 type = eSymbolTypeObjCIVar;
4266 demangled_is_synthesized = true;
4267 }
4268 }
4269 }
4270 } else if (symbol_sect_name &&
4271 ::strstr(symbol_sect_name, "__gcc_except_tab") ==
4272 symbol_sect_name) {
4273 type = eSymbolTypeException;
4274 } else {
4275 type = eSymbolTypeData;
4276 }
4277 } else if (symbol_sect_name &&
4278 ::strstr(symbol_sect_name, "__IMPORT") ==
4279 symbol_sect_name) {
4280 type = eSymbolTypeTrampoline;
4281 } else if (symbol_section->IsDescendant(objc_section_sp.get())) {
4282 type = eSymbolTypeRuntime;
4283 if (symbol_name && symbol_name[0] == '.') {
4284 llvm::StringRef symbol_name_ref(symbol_name);
4285 llvm::StringRef g_objc_v1_prefix_class(
4286 ".objc_class_name_");
4287 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) {
4288 symbol_name_non_abi_mangled = symbol_name;
4289 symbol_name = symbol_name + g_objc_v1_prefix_class.size();
4290 type = eSymbolTypeObjCClass;
4291 demangled_is_synthesized = true;
4292 }
4293 }
4294 }
4295 }
4296 }
4297 } break;
4298 }
4299 }
4300
4301 if (!add_nlist) {
4302 sym[sym_idx].Clear();
4303 return true;
4304 }
4305
4306 uint64_t symbol_value = nlist.n_value;
4307
4308 if (symbol_name_non_abi_mangled) {
4309 sym[sym_idx].GetMangled().SetMangledName(
4310 ConstString(symbol_name_non_abi_mangled));
4311 sym[sym_idx].GetMangled().SetDemangledName(ConstString(symbol_name));
4312 } else {
4313 bool symbol_name_is_mangled = false;
4314
4315 if (symbol_name && symbol_name[0] == '_') {
4316 symbol_name_is_mangled = symbol_name[1] == '_';
4317 symbol_name++; // Skip the leading underscore
4318 }
4319
4320 if (symbol_name) {
4321 ConstString const_symbol_name(symbol_name);
4322 sym[sym_idx].GetMangled().SetValue(const_symbol_name,
4323 symbol_name_is_mangled);
4324 }
4325 }
4326
4327 if (is_gsym) {
4328 const char *gsym_name = sym[sym_idx]
4329 .GetMangled()
4330 .GetName(Mangled::ePreferMangled)
4331 .GetCString();
4332 if (gsym_name)
4333 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx;
4334 }
4335
4336 if (symbol_section) {
4337 const addr_t section_file_addr = symbol_section->GetFileAddress();
4338 if (symbol_byte_size == 0 && function_starts_count > 0) {
4339 addr_t symbol_lookup_file_addr = nlist.n_value;
4340 // Do an exact address match for non-ARM addresses, else get the
4341 // closest since the symbol might be a thumb symbol which has an
4342 // address with bit zero set.
4343 FunctionStarts::Entry *func_start_entry =
4344 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm);
4345 if (is_arm && func_start_entry) {
4346 // Verify that the function start address is the symbol address
4347 // (ARM) or the symbol address + 1 (thumb).
4348 if (func_start_entry->addr != symbol_lookup_file_addr &&
4349 func_start_entry->addr != (symbol_lookup_file_addr + 1)) {
4350 // Not the right entry, NULL it out...
4351 func_start_entry = nullptr;
4352 }
4353 }
4354 if (func_start_entry) {
4355 func_start_entry->data = true;
4356
4357 addr_t symbol_file_addr = func_start_entry->addr;
4358 if (is_arm)
4359 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK;
4360
4361 const FunctionStarts::Entry *next_func_start_entry =
4362 function_starts.FindNextEntry(func_start_entry);
4363 const addr_t section_end_file_addr =
4364 section_file_addr + symbol_section->GetByteSize();
4365 if (next_func_start_entry) {
4366 addr_t next_symbol_file_addr = next_func_start_entry->addr;
4367 // Be sure the clear the Thumb address bit when we calculate the
4368 // size from the current and next address
4369 if (is_arm)
4370 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK;
4371 symbol_byte_size = std::min<lldb::addr_t>(
4372 next_symbol_file_addr - symbol_file_addr,
4373 section_end_file_addr - symbol_file_addr);
4374 } else {
4375 symbol_byte_size = section_end_file_addr - symbol_file_addr;
4376 }
4377 }
4378 }
4379 symbol_value -= section_file_addr;
4380 }
4381
4382 if (!is_debug) {
4383 if (type == eSymbolTypeCode) {
4384 // See if we can find a N_FUN entry for any code symbols. If we do
4385 // find a match, and the name matches, then we can merge the two into
4386 // just the function symbol to avoid duplicate entries in the symbol
4387 // table.
4388 std::pair<ValueToSymbolIndexMap::const_iterator,
4389 ValueToSymbolIndexMap::const_iterator>
4390 range;
4391 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value);
4392 if (range.first != range.second) {
4393 for (ValueToSymbolIndexMap::const_iterator pos = range.first;
4394 pos != range.second; ++pos) {
4395 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) ==
4396 sym[pos->second].GetMangled().GetName(
4397 Mangled::ePreferMangled)) {
4398 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second;
4399 // We just need the flags from the linker symbol, so put these
4400 // flags into the N_FUN flags to avoid duplicate symbols in the
4401 // symbol table.
4402 sym[pos->second].SetExternal(sym[sym_idx].IsExternal());
4403 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc);
4404 if (resolver_addresses.find(nlist.n_value) !=
4405 resolver_addresses.end())
4406 sym[pos->second].SetType(eSymbolTypeResolver);
4407 sym[sym_idx].Clear();
4408 return true;
4409 }
4410 }
4411 } else {
4412 if (resolver_addresses.find(nlist.n_value) !=
4413 resolver_addresses.end())
4414 type = eSymbolTypeResolver;
4415 }
4416 } else if (type == eSymbolTypeData || type == eSymbolTypeObjCClass ||
4417 type == eSymbolTypeObjCMetaClass ||
4418 type == eSymbolTypeObjCIVar) {
4419 // See if we can find a N_STSYM entry for any data symbols. If we do
4420 // find a match, and the name matches, then we can merge the two into
4421 // just the Static symbol to avoid duplicate entries in the symbol
4422 // table.
4423 std::pair<ValueToSymbolIndexMap::const_iterator,
4424 ValueToSymbolIndexMap::const_iterator>
4425 range;
4426 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value);
4427 if (range.first != range.second) {
4428 for (ValueToSymbolIndexMap::const_iterator pos = range.first;
4429 pos != range.second; ++pos) {
4430 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) ==
4431 sym[pos->second].GetMangled().GetName(
4432 Mangled::ePreferMangled)) {
4433 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second;
4434 // We just need the flags from the linker symbol, so put these
4435 // flags into the N_STSYM flags to avoid duplicate symbols in
4436 // the symbol table.
4437 sym[pos->second].SetExternal(sym[sym_idx].IsExternal());
4438 sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc);
4439 sym[sym_idx].Clear();
4440 return true;
4441 }
4442 }
4443 } else {
4444 // Combine N_GSYM stab entries with the non stab symbol.
4445 const char *gsym_name = sym[sym_idx]
4446 .GetMangled()
4447 .GetName(Mangled::ePreferMangled)
4448 .GetCString();
4449 if (gsym_name) {
4450 ConstNameToSymbolIndexMap::const_iterator pos =
4451 N_GSYM_name_to_sym_idx.find(gsym_name);
4452 if (pos != N_GSYM_name_to_sym_idx.end()) {
4453 const uint32_t GSYM_sym_idx = pos->second;
4454 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx;
4455 // Copy the address, because often the N_GSYM address has an
4456 // invalid address of zero when the global is a common symbol.
4457 sym[GSYM_sym_idx].GetAddressRef().SetSection(symbol_section);
4458 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value);
4459 add_symbol_addr(
4460 sym[GSYM_sym_idx].GetAddress().GetFileAddress());
4461 // We just need the flags from the linker symbol, so put these
4462 // flags into the N_GSYM flags to avoid duplicate symbols in
4463 // the symbol table.
4464 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc);
4465 sym[sym_idx].Clear();
4466 return true;
4467 }
4468 }
4469 }
4470 }
4471 }
4472
4473 sym[sym_idx].SetID(nlist_idx);
4474 sym[sym_idx].SetType(type);
4475 if (set_value) {
4476 sym[sym_idx].GetAddressRef().SetSection(symbol_section);
4477 sym[sym_idx].GetAddressRef().SetOffset(symbol_value);
4478 if (symbol_section)
4479 add_symbol_addr(sym[sym_idx].GetAddress().GetFileAddress());
4480 }
4481 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc);
4482 if (nlist.n_desc & N_WEAK_REF)
4483 sym[sym_idx].SetIsWeak(true);
4484
4485 if (symbol_byte_size > 0)
4486 sym[sym_idx].SetByteSize(symbol_byte_size);
4487
4488 if (demangled_is_synthesized)
4489 sym[sym_idx].SetDemangledNameIsSynthesized(true);
4490
4491 ++sym_idx;
4492 return true;
4493 };
4494
4495 // First parse all the nlists but don't process them yet. See the next
4496 // comment for an explanation why.
4497 std::vector<struct nlist_64> nlists;
4498 nlists.reserve(symtab_load_command.nsyms);
4499 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) {
4500 if (auto nlist =
4501 ParseNList(nlist_data, nlist_data_offset, nlist_byte_size))
4502 nlists.push_back(*nlist);
4503 else
4504 break;
4505 }
4506
4507 // Now parse all the debug symbols. This is needed to merge non-debug
4508 // symbols in the next step. Non-debug symbols are always coalesced into
4509 // the debug symbol. Doing this in one step would mean that some symbols
4510 // won't be merged.
4511 nlist_idx = 0;
4512 for (auto &nlist : nlists) {
4513 if (!ParseSymbolLambda(nlist, nlist_idx++, DebugSymbols))
4514 break;
4515 }
4516
4517 // Finally parse all the non debug symbols.
4518 nlist_idx = 0;
4519 for (auto &nlist : nlists) {
4520 if (!ParseSymbolLambda(nlist, nlist_idx++, NonDebugSymbols))
4521 break;
4522 }
4523
4524 for (const auto &pos : reexport_shlib_needs_fixup) {
4525 const auto undef_pos = undefined_name_to_desc.find(pos.second);
4526 if (undef_pos != undefined_name_to_desc.end()) {
4527 const uint8_t dylib_ordinal =
4528 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second);
4529 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize())
4530 sym[pos.first].SetReExportedSymbolSharedLibrary(
4531 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1));
4532 }
4533 }
4534 }
4535
4536 // Count how many trie symbols we'll add to the symbol table
4537 int trie_symbol_table_augment_count = 0;
4538 for (auto &e : external_sym_trie_entries) {
4539 if (symbols_added.find(e.entry.address) == symbols_added.end())
4540 trie_symbol_table_augment_count++;
4541 }
4542
4543 if (num_syms < sym_idx + trie_symbol_table_augment_count) {
4544 num_syms = sym_idx + trie_symbol_table_augment_count;
4545 sym = symtab.Resize(num_syms);
4546 }
4547 uint32_t synthetic_sym_id = symtab_load_command.nsyms;
4548
4549 // Add symbols from the trie to the symbol table.
4550 for (auto &e : external_sym_trie_entries) {
4551 if (symbols_added.contains(e.entry.address))
4552 continue;
4553
4554 // Find the section that this trie address is in, use that to annotate
4555 // symbol type as we add the trie address and name to the symbol table.
4556 Address symbol_addr;
4557 if (module_sp->ResolveFileAddress(e.entry.address, symbol_addr)) {
4558 SectionSP symbol_section(symbol_addr.GetSection());
4559 const char *symbol_name = e.entry.name.GetCString();
4560 bool demangled_is_synthesized = false;
4561 SymbolType type =
4562 GetSymbolType(symbol_name, demangled_is_synthesized, text_section_sp,
4563 data_section_sp, data_dirty_section_sp,
4564 data_const_section_sp, symbol_section);
4565
4566 sym[sym_idx].SetType(type);
4567 if (symbol_section) {
4568 sym[sym_idx].SetID(synthetic_sym_id++);
4569 sym[sym_idx].GetMangled().SetMangledName(ConstString(symbol_name));
4570 if (demangled_is_synthesized)
4571 sym[sym_idx].SetDemangledNameIsSynthesized(true);
4572 sym[sym_idx].SetIsSynthetic(true);
4573 sym[sym_idx].SetExternal(true);
4574 sym[sym_idx].GetAddressRef() = symbol_addr;
4575 add_symbol_addr(symbol_addr.GetFileAddress());
4576 if (e.entry.flags & TRIE_SYMBOL_IS_THUMB)
4577 sym[sym_idx].SetFlags(MACHO_NLIST_ARM_SYMBOL_IS_THUMB);
4578 ++sym_idx;
4579 }
4580 }
4581 }
4582
4583 if (function_starts_count > 0) {
4584 uint32_t num_synthetic_function_symbols = 0;
4585 for (i = 0; i < function_starts_count; ++i) {
4586 if (symbols_added.find(function_starts.GetEntryRef(i).addr) ==
4587 symbols_added.end())
4588 ++num_synthetic_function_symbols;
4589 }
4590
4591 if (num_synthetic_function_symbols > 0) {
4592 if (num_syms < sym_idx + num_synthetic_function_symbols) {
4593 num_syms = sym_idx + num_synthetic_function_symbols;
4594 sym = symtab.Resize(num_syms);
4595 }
4596 for (i = 0; i < function_starts_count; ++i) {
4597 const FunctionStarts::Entry *func_start_entry =
4598 function_starts.GetEntryAtIndex(i);
4599 if (symbols_added.find(func_start_entry->addr) == symbols_added.end()) {
4600 addr_t symbol_file_addr = func_start_entry->addr;
4601 uint32_t symbol_flags = 0;
4602 if (func_start_entry->data)
4603 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB;
4604 Address symbol_addr;
4605 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) {
4606 SectionSP symbol_section(symbol_addr.GetSection());
4607 uint32_t symbol_byte_size = 0;
4608 if (symbol_section) {
4609 const addr_t section_file_addr = symbol_section->GetFileAddress();
4610 const FunctionStarts::Entry *next_func_start_entry =
4611 function_starts.FindNextEntry(func_start_entry);
4612 const addr_t section_end_file_addr =
4613 section_file_addr + symbol_section->GetByteSize();
4614 if (next_func_start_entry) {
4615 addr_t next_symbol_file_addr = next_func_start_entry->addr;
4616 if (is_arm)
4617 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK;
4618 symbol_byte_size = std::min<lldb::addr_t>(
4619 next_symbol_file_addr - symbol_file_addr,
4620 section_end_file_addr - symbol_file_addr);
4621 } else {
4622 symbol_byte_size = section_end_file_addr - symbol_file_addr;
4623 }
4624 sym[sym_idx].SetID(synthetic_sym_id++);
4625 // Don't set the name for any synthetic symbols, the Symbol
4626 // object will generate one if needed when the name is accessed
4627 // via accessors.
4628 sym[sym_idx].GetMangled().SetDemangledName(ConstString());
4629 sym[sym_idx].SetType(eSymbolTypeCode);
4630 sym[sym_idx].SetIsSynthetic(true);
4631 sym[sym_idx].GetAddressRef() = symbol_addr;
4632 add_symbol_addr(symbol_addr.GetFileAddress());
4633 if (symbol_flags)
4634 sym[sym_idx].SetFlags(symbol_flags);
4635 if (symbol_byte_size)
4636 sym[sym_idx].SetByteSize(symbol_byte_size);
4637 ++sym_idx;
4638 }
4639 }
4640 }
4641 }
4642 }
4643 }
4644
4645 // Trim our symbols down to just what we ended up with after removing any
4646 // symbols.
4647 if (sym_idx < num_syms) {
4648 num_syms = sym_idx;
4649 sym = symtab.Resize(num_syms);
4650 }
4651
4652 // Now synthesize indirect symbols
4653 if (m_dysymtab.nindirectsyms != 0) {
4654 if (indirect_symbol_index_data.GetByteSize()) {
4655 NListIndexToSymbolIndexMap::const_iterator end_index_pos =
4656 m_nlist_idx_to_sym_idx.end();
4657
4658 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size();
4659 ++sect_idx) {
4660 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) ==
4661 S_SYMBOL_STUBS) {
4662 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2;
4663 if (symbol_stub_byte_size == 0)
4664 continue;
4665
4666 const uint32_t num_symbol_stubs =
4667 m_mach_sections[sect_idx].size / symbol_stub_byte_size;
4668
4669 if (num_symbol_stubs == 0)
4670 continue;
4671
4672 const uint32_t symbol_stub_index_offset =
4673 m_mach_sections[sect_idx].reserved1;
4674 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) {
4675 const uint32_t symbol_stub_index =
4676 symbol_stub_index_offset + stub_idx;
4677 const lldb::addr_t symbol_stub_addr =
4678 m_mach_sections[sect_idx].addr +
4679 (stub_idx * symbol_stub_byte_size);
4680 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4;
4681 if (indirect_symbol_index_data.ValidOffsetForDataOfSize(
4682 symbol_stub_offset, 4)) {
4683 const uint32_t stub_sym_id =
4684 indirect_symbol_index_data.GetU32(&symbol_stub_offset);
4685 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL))
4686 continue;
4687
4688 NListIndexToSymbolIndexMap::const_iterator index_pos =
4689 m_nlist_idx_to_sym_idx.find(stub_sym_id);
4690 Symbol *stub_symbol = nullptr;
4691 if (index_pos != end_index_pos) {
4692 // We have a remapping from the original nlist index to a
4693 // current symbol index, so just look this up by index
4694 stub_symbol = symtab.SymbolAtIndex(index_pos->second);
4695 } else {
4696 // We need to lookup a symbol using the original nlist symbol
4697 // index since this index is coming from the S_SYMBOL_STUBS
4698 stub_symbol = symtab.FindSymbolByID(stub_sym_id);
4699 }
4700
4701 if (stub_symbol) {
4702 Address so_addr(symbol_stub_addr, section_list);
4703
4704 if (stub_symbol->GetType() == eSymbolTypeUndefined) {
4705 // Change the external symbol into a trampoline that makes
4706 // sense These symbols were N_UNDF N_EXT, and are useless
4707 // to us, so we can re-use them so we don't have to make up
4708 // a synthetic symbol for no good reason.
4709 if (resolver_addresses.find(symbol_stub_addr) ==
4710 resolver_addresses.end())
4711 stub_symbol->SetType(eSymbolTypeTrampoline);
4712 else
4713 stub_symbol->SetType(eSymbolTypeResolver);
4714 stub_symbol->SetExternal(false);
4715 stub_symbol->GetAddressRef() = so_addr;
4716 stub_symbol->SetByteSize(symbol_stub_byte_size);
4717 } else {
4718 // Make a synthetic symbol to describe the trampoline stub
4719 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled());
4720 if (sym_idx >= num_syms) {
4721 sym = symtab.Resize(++num_syms);
4722 stub_symbol = nullptr; // this pointer no longer valid
4723 }
4724 sym[sym_idx].SetID(synthetic_sym_id++);
4725 sym[sym_idx].GetMangled() = stub_symbol_mangled_name;
4726 if (resolver_addresses.find(symbol_stub_addr) ==
4727 resolver_addresses.end())
4728 sym[sym_idx].SetType(eSymbolTypeTrampoline);
4729 else
4730 sym[sym_idx].SetType(eSymbolTypeResolver);
4731 sym[sym_idx].SetIsSynthetic(true);
4732 sym[sym_idx].GetAddressRef() = so_addr;
4733 add_symbol_addr(so_addr.GetFileAddress());
4734 sym[sym_idx].SetByteSize(symbol_stub_byte_size);
4735 ++sym_idx;
4736 }
4737 } else {
4738 if (log)
4739 log->Warning("symbol stub referencing symbol table symbol "
4740 "%u that isn't in our minimal symbol table, "
4741 "fix this!!!",
4742 stub_sym_id);
4743 }
4744 }
4745 }
4746 }
4747 }
4748 }
4749 }
4750
4751 if (!reexport_trie_entries.empty()) {
4752 for (const auto &e : reexport_trie_entries) {
4753 if (e.entry.import_name) {
4754 // Only add indirect symbols from the Trie entries if we didn't have
4755 // a N_INDR nlist entry for this already
4756 if (indirect_symbol_names.find(e.entry.name) ==
4757 indirect_symbol_names.end()) {
4758 // Make a synthetic symbol to describe re-exported symbol.
4759 if (sym_idx >= num_syms)
4760 sym = symtab.Resize(++num_syms);
4761 sym[sym_idx].SetID(synthetic_sym_id++);
4762 sym[sym_idx].GetMangled() = Mangled(e.entry.name);
4763 sym[sym_idx].SetType(eSymbolTypeReExported);
4764 sym[sym_idx].SetIsSynthetic(true);
4765 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name);
4766 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) {
4767 sym[sym_idx].SetReExportedSymbolSharedLibrary(
4768 dylib_files.GetFileSpecAtIndex(e.entry.other - 1));
4769 }
4770 ++sym_idx;
4771 }
4772 }
4773 }
4774 }
4775 }
4776
4777 void ObjectFileMachO::Dump(Stream *s) {
4778 ModuleSP module_sp(GetModule());
4779 if (module_sp) {
4780 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
4781 s->Printf("%p: ", static_cast<void *>(this));
4782 s->Indent();
4783 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64)
4784 s->PutCString("ObjectFileMachO64");
4785 else
4786 s->PutCString("ObjectFileMachO32");
4787
4788 *s << ", file = '" << m_file;
4789 ModuleSpecList all_specs;
4790 ModuleSpec base_spec;
4791 GetAllArchSpecs(m_header, m_data, MachHeaderSizeFromMagic(m_header.magic),
4792 base_spec, all_specs);
4793 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) {
4794 *s << "', triple";
4795 if (e)
4796 s->Printf("[%d]", i);
4797 *s << " = ";
4798 *s << all_specs.GetModuleSpecRefAtIndex(i)
4799 .GetArchitecture()
4800 .GetTriple()
4801 .getTriple();
4802 }
4803 *s << "\n";
4804 SectionList *sections = GetSectionList();
4805 if (sections)
4806 sections->Dump(s->AsRawOstream(), s->GetIndentLevel(), nullptr, true,
4807 UINT32_MAX);
4808
4809 if (m_symtab_up)
4810 m_symtab_up->Dump(s, nullptr, eSortOrderNone);
4811 }
4812 }
4813
4814 UUID ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header,
4815 const lldb_private::DataExtractor &data,
4816 lldb::offset_t lc_offset) {
4817 uint32_t i;
4818 llvm::MachO::uuid_command load_cmd;
4819
4820 lldb::offset_t offset = lc_offset;
4821 for (i = 0; i < header.ncmds; ++i) {
4822 const lldb::offset_t cmd_offset = offset;
4823 if (data.GetU32(&offset, &load_cmd, 2) == nullptr)
4824 break;
4825
4826 if (load_cmd.cmd == LC_UUID) {
4827 const uint8_t *uuid_bytes = data.PeekData(offset, 16);
4828
4829 if (uuid_bytes) {
4830 // OpenCL on Mac OS X uses the same UUID for each of its object files.
4831 // We pretend these object files have no UUID to prevent crashing.
4832
4833 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8,
4834 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63,
4835 0xbb, 0x14, 0xf0, 0x0d};
4836
4837 if (!memcmp(uuid_bytes, opencl_uuid, 16))
4838 return UUID();
4839
4840 return UUID::fromOptionalData(uuid_bytes, 16);
4841 }
4842 return UUID();
4843 }
4844 offset = cmd_offset + load_cmd.cmdsize;
4845 }
4846 return UUID();
4847 }
4848
4849 static llvm::StringRef GetOSName(uint32_t cmd) {
4850 switch (cmd) {
4851 case llvm::MachO::LC_VERSION_MIN_IPHONEOS:
4852 return llvm::Triple::getOSTypeName(llvm::Triple::IOS);
4853 case llvm::MachO::LC_VERSION_MIN_MACOSX:
4854 return llvm::Triple::getOSTypeName(llvm::Triple::MacOSX);
4855 case llvm::MachO::LC_VERSION_MIN_TVOS:
4856 return llvm::Triple::getOSTypeName(llvm::Triple::TvOS);
4857 case llvm::MachO::LC_VERSION_MIN_WATCHOS:
4858 return llvm::Triple::getOSTypeName(llvm::Triple::WatchOS);
4859 default:
4860 llvm_unreachable("unexpected LC_VERSION load command");
4861 }
4862 }
4863
4864 namespace {
4865 struct OSEnv {
4866 llvm::StringRef os_type;
4867 llvm::StringRef environment;
4868 OSEnv(uint32_t cmd) {
4869 switch (cmd) {
4870 case llvm::MachO::PLATFORM_MACOS:
4871 os_type = llvm::Triple::getOSTypeName(llvm::Triple::MacOSX);
4872 return;
4873 case llvm::MachO::PLATFORM_IOS:
4874 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS);
4875 return;
4876 case llvm::MachO::PLATFORM_TVOS:
4877 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS);
4878 return;
4879 case llvm::MachO::PLATFORM_WATCHOS:
4880 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS);
4881 return;
4882 // TODO: add BridgeOS & DriverKit once in llvm/lib/Support/Triple.cpp
4883 // NEED_BRIDGEOS_TRIPLE
4884 // case llvm::MachO::PLATFORM_BRIDGEOS:
4885 // os_type = llvm::Triple::getOSTypeName(llvm::Triple::BridgeOS);
4886 // return;
4887 // case llvm::MachO::PLATFORM_DRIVERKIT:
4888 // os_type = llvm::Triple::getOSTypeName(llvm::Triple::DriverKit);
4889 // return;
4890 case llvm::MachO::PLATFORM_MACCATALYST:
4891 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS);
4892 environment = llvm::Triple::getEnvironmentTypeName(llvm::Triple::MacABI);
4893 return;
4894 case llvm::MachO::PLATFORM_IOSSIMULATOR:
4895 os_type = llvm::Triple::getOSTypeName(llvm::Triple::IOS);
4896 environment =
4897 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator);
4898 return;
4899 case llvm::MachO::PLATFORM_TVOSSIMULATOR:
4900 os_type = llvm::Triple::getOSTypeName(llvm::Triple::TvOS);
4901 environment =
4902 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator);
4903 return;
4904 case llvm::MachO::PLATFORM_WATCHOSSIMULATOR:
4905 os_type = llvm::Triple::getOSTypeName(llvm::Triple::WatchOS);
4906 environment =
4907 llvm::Triple::getEnvironmentTypeName(llvm::Triple::Simulator);
4908 return;
4909 default: {
4910 Log *log(GetLog(LLDBLog::Symbols | LLDBLog::Process));
4911 LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION");
4912 }
4913 }
4914 }
4915 };
4916
4917 struct MinOS {
4918 uint32_t major_version, minor_version, patch_version;
4919 MinOS(uint32_t version)
4920 : major_version(version >> 16), minor_version((version >> 8) & 0xffu),
4921 patch_version(version & 0xffu) {}
4922 };
4923 } // namespace
4924
4925 void ObjectFileMachO::GetAllArchSpecs(const llvm::MachO::mach_header &header,
4926 const lldb_private::DataExtractor &data,
4927 lldb::offset_t lc_offset,
4928 ModuleSpec &base_spec,
4929 lldb_private::ModuleSpecList &all_specs) {
4930 auto &base_arch = base_spec.GetArchitecture();
4931 base_arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype);
4932 if (!base_arch.IsValid())
4933 return;
4934
4935 bool found_any = false;
4936 auto add_triple = [&](const llvm::Triple &triple) {
4937 auto spec = base_spec;
4938 spec.GetArchitecture().GetTriple() = triple;
4939 if (spec.GetArchitecture().IsValid()) {
4940 spec.GetUUID() = ObjectFileMachO::GetUUID(header, data, lc_offset);
4941 all_specs.Append(spec);
4942 found_any = true;
4943 }
4944 };
4945
4946 // Set OS to an unspecified unknown or a "*" so it can match any OS
4947 llvm::Triple base_triple = base_arch.GetTriple();
4948 base_triple.setOS(llvm::Triple::UnknownOS);
4949 base_triple.setOSName(llvm::StringRef());
4950
4951 if (header.filetype == MH_PRELOAD) {
4952 if (header.cputype == CPU_TYPE_ARM) {
4953 // If this is a 32-bit arm binary, and it's a standalone binary, force
4954 // the Vendor to Apple so we don't accidentally pick up the generic
4955 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the
4956 // frame pointer register; most other armv7 ABIs use a combination of
4957 // r7 and r11.
4958 base_triple.setVendor(llvm::Triple::Apple);
4959 } else {
4960 // Set vendor to an unspecified unknown or a "*" so it can match any
4961 // vendor This is required for correct behavior of EFI debugging on
4962 // x86_64
4963 base_triple.setVendor(llvm::Triple::UnknownVendor);
4964 base_triple.setVendorName(llvm::StringRef());
4965 }
4966 return add_triple(base_triple);
4967 }
4968
4969 llvm::MachO::load_command load_cmd;
4970
4971 // See if there is an LC_VERSION_MIN_* load command that can give
4972 // us the OS type.
4973 lldb::offset_t offset = lc_offset;
4974 for (uint32_t i = 0; i < header.ncmds; ++i) {
4975 const lldb::offset_t cmd_offset = offset;
4976 if (data.GetU32(&offset, &load_cmd, 2) == nullptr)
4977 break;
4978
4979 llvm::MachO::version_min_command version_min;
4980 switch (load_cmd.cmd) {
4981 case llvm::MachO::LC_VERSION_MIN_MACOSX:
4982 case llvm::MachO::LC_VERSION_MIN_IPHONEOS:
4983 case llvm::MachO::LC_VERSION_MIN_TVOS:
4984 case llvm::MachO::LC_VERSION_MIN_WATCHOS: {
4985 if (load_cmd.cmdsize != sizeof(version_min))
4986 break;
4987 if (data.ExtractBytes(cmd_offset, sizeof(version_min),
4988 data.GetByteOrder(), &version_min) == 0)
4989 break;
4990 MinOS min_os(version_min.version);
4991 llvm::SmallString<32> os_name;
4992 llvm::raw_svector_ostream os(os_name);
4993 os << GetOSName(load_cmd.cmd) << min_os.major_version << '.'
4994 << min_os.minor_version << '.' << min_os.patch_version;
4995
4996 auto triple = base_triple;
4997 triple.setOSName(os.str());
4998
4999 // Disambiguate legacy simulator platforms.
5000 if (load_cmd.cmd != llvm::MachO::LC_VERSION_MIN_MACOSX &&
5001 (base_triple.getArch() == llvm::Triple::x86_64 ||
5002 base_triple.getArch() == llvm::Triple::x86)) {
5003 // The combination of legacy LC_VERSION_MIN load command and
5004 // x86 architecture always indicates a simulator environment.
5005 // The combination of LC_VERSION_MIN and arm architecture only
5006 // appears for native binaries. Back-deploying simulator
5007 // binaries on Apple Silicon Macs use the modern unambigous
5008 // LC_BUILD_VERSION load commands; no special handling required.
5009 triple.setEnvironment(llvm::Triple::Simulator);
5010 }
5011 add_triple(triple);
5012 break;
5013 }
5014 default:
5015 break;
5016 }
5017
5018 offset = cmd_offset + load_cmd.cmdsize;
5019 }
5020
5021 // See if there are LC_BUILD_VERSION load commands that can give
5022 // us the OS type.
5023 offset = lc_offset;
5024 for (uint32_t i = 0; i < header.ncmds; ++i) {
5025 const lldb::offset_t cmd_offset = offset;
5026 if (data.GetU32(&offset, &load_cmd, 2) == nullptr)
5027 break;
5028
5029 do {
5030 if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) {
5031 llvm::MachO::build_version_command build_version;
5032 if (load_cmd.cmdsize < sizeof(build_version)) {
5033 // Malformed load command.
5034 break;
5035 }
5036 if (data.ExtractBytes(cmd_offset, sizeof(build_version),
5037 data.GetByteOrder(), &build_version) == 0)
5038 break;
5039 MinOS min_os(build_version.minos);
5040 OSEnv os_env(build_version.platform);
5041 llvm::SmallString<16> os_name;
5042 llvm::raw_svector_ostream os(os_name);
5043 os << os_env.os_type << min_os.major_version << '.'
5044 << min_os.minor_version << '.' << min_os.patch_version;
5045 auto triple = base_triple;
5046 triple.setOSName(os.str());
5047 os_name.clear();
5048 if (!os_env.environment.empty())
5049 triple.setEnvironmentName(os_env.environment);
5050 add_triple(triple);
5051 }
5052 } while (false);
5053 offset = cmd_offset + load_cmd.cmdsize;
5054 }
5055
5056 if (!found_any) {
5057 add_triple(base_triple);
5058 }
5059 }
5060
5061 ArchSpec ObjectFileMachO::GetArchitecture(
5062 ModuleSP module_sp, const llvm::MachO::mach_header &header,
5063 const lldb_private::DataExtractor &data, lldb::offset_t lc_offset) {
5064 ModuleSpecList all_specs;
5065 ModuleSpec base_spec;
5066 GetAllArchSpecs(header, data, MachHeaderSizeFromMagic(header.magic),
5067 base_spec, all_specs);
5068
5069 // If the object file offers multiple alternative load commands,
5070 // pick the one that matches the module.
5071 if (module_sp) {
5072 const ArchSpec &module_arch = module_sp->GetArchitecture();
5073 for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) {
5074 ArchSpec mach_arch =
5075 all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture();
5076 if (module_arch.IsCompatibleMatch(mach_arch))
5077 return mach_arch;
5078 }
5079 }
5080
5081 // Return the first arch we found.
5082 if (all_specs.GetSize() == 0)
5083 return {};
5084 return all_specs.GetModuleSpecRefAtIndex(0).GetArchitecture();
5085 }
5086
5087 UUID ObjectFileMachO::GetUUID() {
5088 ModuleSP module_sp(GetModule());
5089 if (module_sp) {
5090 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5091 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
5092 return GetUUID(m_header, m_data, offset);
5093 }
5094 return UUID();
5095 }
5096
5097 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) {
5098 uint32_t count = 0;
5099 ModuleSP module_sp(GetModule());
5100 if (module_sp) {
5101 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5102 llvm::MachO::load_command load_cmd;
5103 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
5104 std::vector<std::string> rpath_paths;
5105 std::vector<std::string> rpath_relative_paths;
5106 std::vector<std::string> at_exec_relative_paths;
5107 uint32_t i;
5108 for (i = 0; i < m_header.ncmds; ++i) {
5109 const uint32_t cmd_offset = offset;
5110 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr)
5111 break;
5112
5113 switch (load_cmd.cmd) {
5114 case LC_RPATH:
5115 case LC_LOAD_DYLIB:
5116 case LC_LOAD_WEAK_DYLIB:
5117 case LC_REEXPORT_DYLIB:
5118 case LC_LOAD_DYLINKER:
5119 case LC_LOADFVMLIB:
5120 case LC_LOAD_UPWARD_DYLIB: {
5121 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset);
5122 const char *path = m_data.PeekCStr(name_offset);
5123 if (path) {
5124 if (load_cmd.cmd == LC_RPATH)
5125 rpath_paths.push_back(path);
5126 else {
5127 if (path[0] == '@') {
5128 if (strncmp(path, "@rpath", strlen("@rpath")) == 0)
5129 rpath_relative_paths.push_back(path + strlen("@rpath"));
5130 else if (strncmp(path, "@executable_path",
5131 strlen("@executable_path")) == 0)
5132 at_exec_relative_paths.push_back(path +
5133 strlen("@executable_path"));
5134 } else {
5135 FileSpec file_spec(path);
5136 if (files.AppendIfUnique(file_spec))
5137 count++;
5138 }
5139 }
5140 }
5141 } break;
5142
5143 default:
5144 break;
5145 }
5146 offset = cmd_offset + load_cmd.cmdsize;
5147 }
5148
5149 FileSpec this_file_spec(m_file);
5150 FileSystem::Instance().Resolve(this_file_spec);
5151
5152 if (!rpath_paths.empty()) {
5153 // Fixup all LC_RPATH values to be absolute paths
5154 std::string loader_path("@loader_path");
5155 std::string executable_path("@executable_path");
5156 for (auto &rpath : rpath_paths) {
5157 if (llvm::StringRef(rpath).startswith(loader_path)) {
5158 rpath.erase(0, loader_path.size());
5159 rpath.insert(0, this_file_spec.GetDirectory().GetCString());
5160 } else if (llvm::StringRef(rpath).startswith(executable_path)) {
5161 rpath.erase(0, executable_path.size());
5162 rpath.insert(0, this_file_spec.GetDirectory().GetCString());
5163 }
5164 }
5165
5166 for (const auto &rpath_relative_path : rpath_relative_paths) {
5167 for (const auto &rpath : rpath_paths) {
5168 std::string path = rpath;
5169 path += rpath_relative_path;
5170 // It is OK to resolve this path because we must find a file on disk
5171 // for us to accept it anyway if it is rpath relative.
5172 FileSpec file_spec(path);
5173 FileSystem::Instance().Resolve(file_spec);
5174 if (FileSystem::Instance().Exists(file_spec) &&
5175 files.AppendIfUnique(file_spec)) {
5176 count++;
5177 break;
5178 }
5179 }
5180 }
5181 }
5182
5183 // We may have @executable_paths but no RPATHS. Figure those out here.
5184 // Only do this if this object file is the executable. We have no way to
5185 // get back to the actual executable otherwise, so we won't get the right
5186 // path.
5187 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) {
5188 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent();
5189 for (const auto &at_exec_relative_path : at_exec_relative_paths) {
5190 FileSpec file_spec =
5191 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path);
5192 if (FileSystem::Instance().Exists(file_spec) &&
5193 files.AppendIfUnique(file_spec))
5194 count++;
5195 }
5196 }
5197 }
5198 return count;
5199 }
5200
5201 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() {
5202 // If the object file is not an executable it can't hold the entry point.
5203 // m_entry_point_address is initialized to an invalid address, so we can just
5204 // return that. If m_entry_point_address is valid it means we've found it
5205 // already, so return the cached value.
5206
5207 if ((!IsExecutable() && !IsDynamicLoader()) ||
5208 m_entry_point_address.IsValid()) {
5209 return m_entry_point_address;
5210 }
5211
5212 // Otherwise, look for the UnixThread or Thread command. The data for the
5213 // Thread command is given in /usr/include/mach-o.h, but it is basically:
5214 //
5215 // uint32_t flavor - this is the flavor argument you would pass to
5216 // thread_get_state
5217 // uint32_t count - this is the count of longs in the thread state data
5218 // struct XXX_thread_state state - this is the structure from
5219 // <machine/thread_status.h> corresponding to the flavor.
5220 // <repeat this trio>
5221 //
5222 // So we just keep reading the various register flavors till we find the GPR
5223 // one, then read the PC out of there.
5224 // FIXME: We will need to have a "RegisterContext data provider" class at some
5225 // point that can get all the registers
5226 // out of data in this form & attach them to a given thread. That should
5227 // underlie the MacOS X User process plugin, and we'll also need it for the
5228 // MacOS X Core File process plugin. When we have that we can also use it
5229 // here.
5230 //
5231 // For now we hard-code the offsets and flavors we need:
5232 //
5233 //
5234
5235 ModuleSP module_sp(GetModule());
5236 if (module_sp) {
5237 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5238 llvm::MachO::load_command load_cmd;
5239 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
5240 uint32_t i;
5241 lldb::addr_t start_address = LLDB_INVALID_ADDRESS;
5242 bool done = false;
5243
5244 for (i = 0; i < m_header.ncmds; ++i) {
5245 const lldb::offset_t cmd_offset = offset;
5246 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr)
5247 break;
5248
5249 switch (load_cmd.cmd) {
5250 case LC_UNIXTHREAD:
5251 case LC_THREAD: {
5252 while (offset < cmd_offset + load_cmd.cmdsize) {
5253 uint32_t flavor = m_data.GetU32(&offset);
5254 uint32_t count = m_data.GetU32(&offset);
5255 if (count == 0) {
5256 // We've gotten off somehow, log and exit;
5257 return m_entry_point_address;
5258 }
5259
5260 switch (m_header.cputype) {
5261 case llvm::MachO::CPU_TYPE_ARM:
5262 if (flavor == 1 ||
5263 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32
5264 // from mach/arm/thread_status.h
5265 {
5266 offset += 60; // This is the offset of pc in the GPR thread state
5267 // data structure.
5268 start_address = m_data.GetU32(&offset);
5269 done = true;
5270 }
5271 break;
5272 case llvm::MachO::CPU_TYPE_ARM64:
5273 case llvm::MachO::CPU_TYPE_ARM64_32:
5274 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h
5275 {
5276 offset += 256; // This is the offset of pc in the GPR thread state
5277 // data structure.
5278 start_address = m_data.GetU64(&offset);
5279 done = true;
5280 }
5281 break;
5282 case llvm::MachO::CPU_TYPE_I386:
5283 if (flavor ==
5284 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h
5285 {
5286 offset += 40; // This is the offset of eip in the GPR thread state
5287 // data structure.
5288 start_address = m_data.GetU32(&offset);
5289 done = true;
5290 }
5291 break;
5292 case llvm::MachO::CPU_TYPE_X86_64:
5293 if (flavor ==
5294 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h
5295 {
5296 offset += 16 * 8; // This is the offset of rip in the GPR thread
5297 // state data structure.
5298 start_address = m_data.GetU64(&offset);
5299 done = true;
5300 }
5301 break;
5302 default:
5303 return m_entry_point_address;
5304 }
5305 // Haven't found the GPR flavor yet, skip over the data for this
5306 // flavor:
5307 if (done)
5308 break;
5309 offset += count * 4;
5310 }
5311 } break;
5312 case LC_MAIN: {
5313 ConstString text_segment_name("__TEXT");
5314 uint64_t entryoffset = m_data.GetU64(&offset);
5315 SectionSP text_segment_sp =
5316 GetSectionList()->FindSectionByName(text_segment_name);
5317 if (text_segment_sp) {
5318 done = true;
5319 start_address = text_segment_sp->GetFileAddress() + entryoffset;
5320 }
5321 } break;
5322
5323 default:
5324 break;
5325 }
5326 if (done)
5327 break;
5328
5329 // Go to the next load command:
5330 offset = cmd_offset + load_cmd.cmdsize;
5331 }
5332
5333 if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) {
5334 if (GetSymtab()) {
5335 Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType(
5336 ConstString("_dyld_start"), SymbolType::eSymbolTypeCode,
5337 Symtab::eDebugAny, Symtab::eVisibilityAny);
5338 if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) {
5339 start_address = dyld_start_sym->GetAddress().GetFileAddress();
5340 }
5341 }
5342 }
5343
5344 if (start_address != LLDB_INVALID_ADDRESS) {
5345 // We got the start address from the load commands, so now resolve that
5346 // address in the sections of this ObjectFile:
5347 if (!m_entry_point_address.ResolveAddressUsingFileSections(
5348 start_address, GetSectionList())) {
5349 m_entry_point_address.Clear();
5350 }
5351 } else {
5352 // We couldn't read the UnixThread load command - maybe it wasn't there.
5353 // As a fallback look for the "start" symbol in the main executable.
5354
5355 ModuleSP module_sp(GetModule());
5356
5357 if (module_sp) {
5358 SymbolContextList contexts;
5359 SymbolContext context;
5360 module_sp->FindSymbolsWithNameAndType(ConstString("start"),
5361 eSymbolTypeCode, contexts);
5362 if (contexts.GetSize()) {
5363 if (contexts.GetContextAtIndex(0, context))
5364 m_entry_point_address = context.symbol->GetAddress();
5365 }
5366 }
5367 }
5368 }
5369
5370 return m_entry_point_address;
5371 }
5372
5373 lldb_private::Address ObjectFileMachO::GetBaseAddress() {
5374 lldb_private::Address header_addr;
5375 SectionList *section_list = GetSectionList();
5376 if (section_list) {
5377 SectionSP text_segment_sp(
5378 section_list->FindSectionByName(GetSegmentNameTEXT()));
5379 if (text_segment_sp) {
5380 header_addr.SetSection(text_segment_sp);
5381 header_addr.SetOffset(0);
5382 }
5383 }
5384 return header_addr;
5385 }
5386
5387 uint32_t ObjectFileMachO::GetNumThreadContexts() {
5388 ModuleSP module_sp(GetModule());
5389 if (module_sp) {
5390 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5391 if (!m_thread_context_offsets_valid) {
5392 m_thread_context_offsets_valid = true;
5393 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
5394 FileRangeArray::Entry file_range;
5395 llvm::MachO::thread_command thread_cmd;
5396 for (uint32_t i = 0; i < m_header.ncmds; ++i) {
5397 const uint32_t cmd_offset = offset;
5398 if (m_data.GetU32(&offset, &thread_cmd, 2) == nullptr)
5399 break;
5400
5401 if (thread_cmd.cmd == LC_THREAD) {
5402 file_range.SetRangeBase(offset);
5403 file_range.SetByteSize(thread_cmd.cmdsize - 8);
5404 m_thread_context_offsets.Append(file_range);
5405 }
5406 offset = cmd_offset + thread_cmd.cmdsize;
5407 }
5408 }
5409 }
5410 return m_thread_context_offsets.GetSize();
5411 }
5412
5413 std::string ObjectFileMachO::GetIdentifierString() {
5414 std::string result;
5415 ModuleSP module_sp(GetModule());
5416 if (module_sp) {
5417 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5418
5419 // First, look over the load commands for an LC_NOTE load command with
5420 // data_owner string "kern ver str" & use that if found.
5421 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
5422 for (uint32_t i = 0; i < m_header.ncmds; ++i) {
5423 const uint32_t cmd_offset = offset;
5424 llvm::MachO::load_command lc = {};
5425 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr)
5426 break;
5427 if (lc.cmd == LC_NOTE) {
5428 char data_owner[17];
5429 m_data.CopyData(offset, 16, data_owner);
5430 data_owner[16] = '\0';
5431 offset += 16;
5432 uint64_t fileoff = m_data.GetU64_unchecked(&offset);
5433 uint64_t size = m_data.GetU64_unchecked(&offset);
5434
5435 // "kern ver str" has a uint32_t version and then a nul terminated
5436 // c-string.
5437 if (strcmp("kern ver str", data_owner) == 0) {
5438 offset = fileoff;
5439 uint32_t version;
5440 if (m_data.GetU32(&offset, &version, 1) != nullptr) {
5441 if (version == 1) {
5442 uint32_t strsize = size - sizeof(uint32_t);
5443 char *buf = (char *)malloc(strsize);
5444 if (buf) {
5445 m_data.CopyData(offset, strsize, buf);
5446 buf[strsize - 1] = '\0';
5447 result = buf;
5448 if (buf)
5449 free(buf);
5450 return result;
5451 }
5452 }
5453 }
5454 }
5455 }
5456 offset = cmd_offset + lc.cmdsize;
5457 }
5458
5459 // Second, make a pass over the load commands looking for an obsolete
5460 // LC_IDENT load command.
5461 offset = MachHeaderSizeFromMagic(m_header.magic);
5462 for (uint32_t i = 0; i < m_header.ncmds; ++i) {
5463 const uint32_t cmd_offset = offset;
5464 llvm::MachO::ident_command ident_command;
5465 if (m_data.GetU32(&offset, &ident_command, 2) == nullptr)
5466 break;
5467 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) {
5468 char *buf = (char *)malloc(ident_command.cmdsize);
5469 if (buf != nullptr && m_data.CopyData(offset, ident_command.cmdsize,
5470 buf) == ident_command.cmdsize) {
5471 buf[ident_command.cmdsize - 1] = '\0';
5472 result = buf;
5473 }
5474 if (buf)
5475 free(buf);
5476 }
5477 offset = cmd_offset + ident_command.cmdsize;
5478 }
5479 }
5480 return result;
5481 }
5482
5483 addr_t ObjectFileMachO::GetAddressMask() {
5484 addr_t mask = 0;
5485 ModuleSP module_sp(GetModule());
5486 if (module_sp) {
5487 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5488 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
5489 for (uint32_t i = 0; i < m_header.ncmds; ++i) {
5490 const uint32_t cmd_offset = offset;
5491 llvm::MachO::load_command lc = {};
5492 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr)
5493 break;
5494 if (lc.cmd == LC_NOTE) {
5495 char data_owner[17];
5496 m_data.CopyData(offset, 16, data_owner);
5497 data_owner[16] = '\0';
5498 offset += 16;
5499 uint64_t fileoff = m_data.GetU64_unchecked(&offset);
5500
5501 // "addrable bits" has a uint32_t version and a uint32_t
5502 // number of bits used in addressing.
5503 if (strcmp("addrable bits", data_owner) == 0) {
5504 offset = fileoff;
5505 uint32_t version;
5506 if (m_data.GetU32(&offset, &version, 1) != nullptr) {
5507 if (version == 3) {
5508 uint32_t num_addr_bits = m_data.GetU32_unchecked(&offset);
5509 if (num_addr_bits != 0) {
5510 mask = ~((1ULL << num_addr_bits) - 1);
5511 }
5512 break;
5513 }
5514 }
5515 }
5516 }
5517 offset = cmd_offset + lc.cmdsize;
5518 }
5519 }
5520 return mask;
5521 }
5522
5523 bool ObjectFileMachO::GetCorefileMainBinaryInfo(addr_t &value,
5524 bool &value_is_offset,
5525 UUID &uuid,
5526 ObjectFile::BinaryType &type) {
5527 value = LLDB_INVALID_ADDRESS;
5528 value_is_offset = false;
5529 uuid.Clear();
5530 uint32_t log2_pagesize = 0; // not currently passed up to caller
5531 uint32_t platform = 0; // not currently passed up to caller
5532 ModuleSP module_sp(GetModule());
5533 if (module_sp) {
5534 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5535 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
5536 for (uint32_t i = 0; i < m_header.ncmds; ++i) {
5537 const uint32_t cmd_offset = offset;
5538 llvm::MachO::load_command lc = {};
5539 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr)
5540 break;
5541 if (lc.cmd == LC_NOTE) {
5542 char data_owner[17];
5543 memset(data_owner, 0, sizeof(data_owner));
5544 m_data.CopyData(offset, 16, data_owner);
5545 offset += 16;
5546 uint64_t fileoff = m_data.GetU64_unchecked(&offset);
5547 uint64_t size = m_data.GetU64_unchecked(&offset);
5548
5549 // struct main_bin_spec
5550 // {
5551 // uint32_t version; // currently 2
5552 // uint32_t type; // 0 == unspecified, 1 == kernel,
5553 // // 2 == user process,
5554 // // 3 == standalone binary
5555 // uint64_t address; // UINT64_MAX if address not specified
5556 // uint64_t slide; // slide, UINT64_MAX if unspecified
5557 // // 0 if no slide needs to be applied to
5558 // // file address
5559 // uuid_t uuid; // all zero's if uuid not specified
5560 // uint32_t log2_pagesize; // process page size in log base 2,
5561 // // e.g. 4k pages are 12.
5562 // // 0 for unspecified
5563 // uint32_t platform; // The Mach-O platform for this corefile.
5564 // // 0 for unspecified.
5565 // // The values are defined in
5566 // // <mach-o/loader.h>, PLATFORM_*.
5567 // } __attribute((packed));
5568
5569 // "main bin spec" (main binary specification) data payload is
5570 // formatted:
5571 // uint32_t version [currently 1]
5572 // uint32_t type [0 == unspecified, 1 == kernel,
5573 // 2 == user process, 3 == firmware ]
5574 // uint64_t address [ UINT64_MAX if address not specified ]
5575 // uuid_t uuid [ all zero's if uuid not specified ]
5576 // uint32_t log2_pagesize [ process page size in log base
5577 // 2, e.g. 4k pages are 12.
5578 // 0 for unspecified ]
5579 // uint32_t unused [ for alignment ]
5580
5581 if (strcmp("main bin spec", data_owner) == 0 && size >= 32) {
5582 offset = fileoff;
5583 uint32_t version;
5584 if (m_data.GetU32(&offset, &version, 1) != nullptr && version <= 2) {
5585 uint32_t binspec_type = 0;
5586 uuid_t raw_uuid;
5587 memset(raw_uuid, 0, sizeof(uuid_t));
5588
5589 if (!m_data.GetU32(&offset, &binspec_type, 1))
5590 return false;
5591 if (!m_data.GetU64(&offset, &value, 1))
5592 return false;
5593 uint64_t slide = LLDB_INVALID_ADDRESS;
5594 if (version > 1 && !m_data.GetU64(&offset, &slide, 1))
5595 return false;
5596 if (value == LLDB_INVALID_ADDRESS &&
5597 slide != LLDB_INVALID_ADDRESS) {
5598 value = slide;
5599 value_is_offset = true;
5600 }
5601
5602 if (m_data.CopyData(offset, sizeof(uuid_t), raw_uuid) != 0) {
5603 uuid = UUID::fromOptionalData(raw_uuid, sizeof(uuid_t));
5604 // convert the "main bin spec" type into our
5605 // ObjectFile::BinaryType enum
5606 switch (binspec_type) {
5607 case 0:
5608 type = eBinaryTypeUnknown;
5609 break;
5610 case 1:
5611 type = eBinaryTypeKernel;
5612 break;
5613 case 2:
5614 type = eBinaryTypeUser;
5615 break;
5616 case 3:
5617 type = eBinaryTypeStandalone;
5618 break;
5619 }
5620 if (!m_data.GetU32(&offset, &log2_pagesize, 1))
5621 return false;
5622 if (version > 1 && !m_data.GetU32(&offset, &platform, 1))
5623 return false;
5624 return true;
5625 }
5626 }
5627 }
5628 }
5629 offset = cmd_offset + lc.cmdsize;
5630 }
5631 }
5632 return false;
5633 }
5634
5635 lldb::RegisterContextSP
5636 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx,
5637 lldb_private::Thread &thread) {
5638 lldb::RegisterContextSP reg_ctx_sp;
5639
5640 ModuleSP module_sp(GetModule());
5641 if (module_sp) {
5642 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5643 if (!m_thread_context_offsets_valid)
5644 GetNumThreadContexts();
5645
5646 const FileRangeArray::Entry *thread_context_file_range =
5647 m_thread_context_offsets.GetEntryAtIndex(idx);
5648 if (thread_context_file_range) {
5649
5650 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(),
5651 thread_context_file_range->GetByteSize());
5652
5653 switch (m_header.cputype) {
5654 case llvm::MachO::CPU_TYPE_ARM64:
5655 case llvm::MachO::CPU_TYPE_ARM64_32:
5656 reg_ctx_sp =
5657 std::make_shared<RegisterContextDarwin_arm64_Mach>(thread, data);
5658 break;
5659
5660 case llvm::MachO::CPU_TYPE_ARM:
5661 reg_ctx_sp =
5662 std::make_shared<RegisterContextDarwin_arm_Mach>(thread, data);
5663 break;
5664
5665 case llvm::MachO::CPU_TYPE_I386:
5666 reg_ctx_sp =
5667 std::make_shared<RegisterContextDarwin_i386_Mach>(thread, data);
5668 break;
5669
5670 case llvm::MachO::CPU_TYPE_X86_64:
5671 reg_ctx_sp =
5672 std::make_shared<RegisterContextDarwin_x86_64_Mach>(thread, data);
5673 break;
5674 }
5675 }
5676 }
5677 return reg_ctx_sp;
5678 }
5679
5680 ObjectFile::Type ObjectFileMachO::CalculateType() {
5681 switch (m_header.filetype) {
5682 case MH_OBJECT: // 0x1u
5683 if (GetAddressByteSize() == 4) {
5684 // 32 bit kexts are just object files, but they do have a valid
5685 // UUID load command.
5686 if (GetUUID()) {
5687 // this checking for the UUID load command is not enough we could
5688 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as
5689 // this is required of kexts
5690 if (m_strata == eStrataInvalid)
5691 m_strata = eStrataKernel;
5692 return eTypeSharedLibrary;
5693 }
5694 }
5695 return eTypeObjectFile;
5696
5697 case MH_EXECUTE:
5698 return eTypeExecutable; // 0x2u
5699 case MH_FVMLIB:
5700 return eTypeSharedLibrary; // 0x3u
5701 case MH_CORE:
5702 return eTypeCoreFile; // 0x4u
5703 case MH_PRELOAD:
5704 return eTypeSharedLibrary; // 0x5u
5705 case MH_DYLIB:
5706 return eTypeSharedLibrary; // 0x6u
5707 case MH_DYLINKER:
5708 return eTypeDynamicLinker; // 0x7u
5709 case MH_BUNDLE:
5710 return eTypeSharedLibrary; // 0x8u
5711 case MH_DYLIB_STUB:
5712 return eTypeStubLibrary; // 0x9u
5713 case MH_DSYM:
5714 return eTypeDebugInfo; // 0xAu
5715 case MH_KEXT_BUNDLE:
5716 return eTypeSharedLibrary; // 0xBu
5717 default:
5718 break;
5719 }
5720 return eTypeUnknown;
5721 }
5722
5723 ObjectFile::Strata ObjectFileMachO::CalculateStrata() {
5724 switch (m_header.filetype) {
5725 case MH_OBJECT: // 0x1u
5726 {
5727 // 32 bit kexts are just object files, but they do have a valid
5728 // UUID load command.
5729 if (GetUUID()) {
5730 // this checking for the UUID load command is not enough we could
5731 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as
5732 // this is required of kexts
5733 if (m_type == eTypeInvalid)
5734 m_type = eTypeSharedLibrary;
5735
5736 return eStrataKernel;
5737 }
5738 }
5739 return eStrataUnknown;
5740
5741 case MH_EXECUTE: // 0x2u
5742 // Check for the MH_DYLDLINK bit in the flags
5743 if (m_header.flags & MH_DYLDLINK) {
5744 return eStrataUser;
5745 } else {
5746 SectionList *section_list = GetSectionList();
5747 if (section_list) {
5748 static ConstString g_kld_section_name("__KLD");
5749 if (section_list->FindSectionByName(g_kld_section_name))
5750 return eStrataKernel;
5751 }
5752 }
5753 return eStrataRawImage;
5754
5755 case MH_FVMLIB:
5756 return eStrataUser; // 0x3u
5757 case MH_CORE:
5758 return eStrataUnknown; // 0x4u
5759 case MH_PRELOAD:
5760 return eStrataRawImage; // 0x5u
5761 case MH_DYLIB:
5762 return eStrataUser; // 0x6u
5763 case MH_DYLINKER:
5764 return eStrataUser; // 0x7u
5765 case MH_BUNDLE:
5766 return eStrataUser; // 0x8u
5767 case MH_DYLIB_STUB:
5768 return eStrataUser; // 0x9u
5769 case MH_DSYM:
5770 return eStrataUnknown; // 0xAu
5771 case MH_KEXT_BUNDLE:
5772 return eStrataKernel; // 0xBu
5773 default:
5774 break;
5775 }
5776 return eStrataUnknown;
5777 }
5778
5779 llvm::VersionTuple ObjectFileMachO::GetVersion() {
5780 ModuleSP module_sp(GetModule());
5781 if (module_sp) {
5782 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5783 llvm::MachO::dylib_command load_cmd;
5784 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
5785 uint32_t version_cmd = 0;
5786 uint64_t version = 0;
5787 uint32_t i;
5788 for (i = 0; i < m_header.ncmds; ++i) {
5789 const lldb::offset_t cmd_offset = offset;
5790 if (m_data.GetU32(&offset, &load_cmd, 2) == nullptr)
5791 break;
5792
5793 if (load_cmd.cmd == LC_ID_DYLIB) {
5794 if (version_cmd == 0) {
5795 version_cmd = load_cmd.cmd;
5796 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == nullptr)
5797 break;
5798 version = load_cmd.dylib.current_version;
5799 }
5800 break; // Break for now unless there is another more complete version
5801 // number load command in the future.
5802 }
5803 offset = cmd_offset + load_cmd.cmdsize;
5804 }
5805
5806 if (version_cmd == LC_ID_DYLIB) {
5807 unsigned major = (version & 0xFFFF0000ull) >> 16;
5808 unsigned minor = (version & 0x0000FF00ull) >> 8;
5809 unsigned subminor = (version & 0x000000FFull);
5810 return llvm::VersionTuple(major, minor, subminor);
5811 }
5812 }
5813 return llvm::VersionTuple();
5814 }
5815
5816 ArchSpec ObjectFileMachO::GetArchitecture() {
5817 ModuleSP module_sp(GetModule());
5818 ArchSpec arch;
5819 if (module_sp) {
5820 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex());
5821
5822 return GetArchitecture(module_sp, m_header, m_data,
5823 MachHeaderSizeFromMagic(m_header.magic));
5824 }
5825 return arch;
5826 }
5827
5828 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process,
5829 addr_t &base_addr, UUID &uuid) {
5830 uuid.Clear();
5831 base_addr = LLDB_INVALID_ADDRESS;
5832 if (process && process->GetDynamicLoader()) {
5833 DynamicLoader *dl = process->GetDynamicLoader();
5834 LazyBool using_shared_cache;
5835 LazyBool private_shared_cache;
5836 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache,
5837 private_shared_cache);
5838 }
5839 Log *log(GetLog(LLDBLog::Symbols | LLDBLog::Process));
5840 LLDB_LOGF(
5841 log,
5842 "inferior process shared cache has a UUID of %s, base address 0x%" PRIx64,
5843 uuid.GetAsString().c_str(), base_addr);
5844 }
5845
5846 // From dyld SPI header dyld_process_info.h
5847 typedef void *dyld_process_info;
5848 struct lldb_copy__dyld_process_cache_info {
5849 uuid_t cacheUUID; // UUID of cache used by process
5850 uint64_t cacheBaseAddress; // load address of dyld shared cache
5851 bool noCache; // process is running without a dyld cache
5852 bool privateCache; // process is using a private copy of its dyld cache
5853 };
5854
5855 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with
5856 // llvm enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile
5857 // errors. So we need to use the actual underlying types of task_t and
5858 // kern_return_t below.
5859 extern "C" unsigned int /*task_t*/ mach_task_self();
5860
5861 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) {
5862 uuid.Clear();
5863 base_addr = LLDB_INVALID_ADDRESS;
5864
5865 #if defined(__APPLE__)
5866 uint8_t *(*dyld_get_all_image_infos)(void);
5867 dyld_get_all_image_infos =
5868 (uint8_t * (*)()) dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos");
5869 if (dyld_get_all_image_infos) {
5870 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos();
5871 if (dyld_all_image_infos_address) {
5872 uint32_t *version = (uint32_t *)
5873 dyld_all_image_infos_address; // version <mach-o/dyld_images.h>
5874 if (*version >= 13) {
5875 uuid_t *sharedCacheUUID_address = 0;
5876 int wordsize = sizeof(uint8_t *);
5877 if (wordsize == 8) {
5878 sharedCacheUUID_address =
5879 (uuid_t *)((uint8_t *)dyld_all_image_infos_address +
5880 160); // sharedCacheUUID <mach-o/dyld_images.h>
5881 if (*version >= 15)
5882 base_addr =
5883 *(uint64_t
5884 *)((uint8_t *)dyld_all_image_infos_address +
5885 176); // sharedCacheBaseAddress <mach-o/dyld_images.h>
5886 } else {
5887 sharedCacheUUID_address =
5888 (uuid_t *)((uint8_t *)dyld_all_image_infos_address +
5889 84); // sharedCacheUUID <mach-o/dyld_images.h>
5890 if (*version >= 15) {
5891 base_addr = 0;
5892 base_addr =
5893 *(uint32_t
5894 *)((uint8_t *)dyld_all_image_infos_address +
5895 100); // sharedCacheBaseAddress <mach-o/dyld_images.h>
5896 }
5897 }
5898 uuid = UUID::fromOptionalData(sharedCacheUUID_address, sizeof(uuid_t));
5899 }
5900 }
5901 } else {
5902 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI
5903 dyld_process_info (*dyld_process_info_create)(
5904 unsigned int /* task_t */ task, uint64_t timestamp,
5905 unsigned int /*kern_return_t*/ *kernelError);
5906 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo);
5907 void (*dyld_process_info_release)(dyld_process_info info);
5908
5909 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t,
5910 unsigned int /*kern_return_t*/ *))
5911 dlsym(RTLD_DEFAULT, "_dyld_process_info_create");
5912 dyld_process_info_get_cache = (void (*)(void *, void *))dlsym(
5913 RTLD_DEFAULT, "_dyld_process_info_get_cache");
5914 dyld_process_info_release =
5915 (void (*)(void *))dlsym(RTLD_DEFAULT, "_dyld_process_info_release");
5916
5917 if (dyld_process_info_create && dyld_process_info_get_cache) {
5918 unsigned int /*kern_return_t */ kern_ret;
5919 dyld_process_info process_info =
5920 dyld_process_info_create(::mach_task_self(), 0, &kern_ret);
5921 if (process_info) {
5922 struct lldb_copy__dyld_process_cache_info sc_info;
5923 memset(&sc_info, 0, sizeof(struct lldb_copy__dyld_process_cache_info));
5924 dyld_process_info_get_cache(process_info, &sc_info);
5925 if (sc_info.cacheBaseAddress != 0) {
5926 base_addr = sc_info.cacheBaseAddress;
5927 uuid = UUID::fromOptionalData(sc_info.cacheUUID, sizeof(uuid_t));
5928 }
5929 dyld_process_info_release(process_info);
5930 }
5931 }
5932 }
5933 Log *log(GetLog(LLDBLog::Symbols | LLDBLog::Process));
5934 if (log && uuid.IsValid())
5935 LLDB_LOGF(log,
5936 "lldb's in-memory shared cache has a UUID of %s base address of "
5937 "0x%" PRIx64,
5938 uuid.GetAsString().c_str(), base_addr);
5939 #endif
5940 }
5941
5942 llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() {
5943 if (!m_min_os_version) {
5944 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
5945 for (uint32_t i = 0; i < m_header.ncmds; ++i) {
5946 const lldb::offset_t load_cmd_offset = offset;
5947
5948 llvm::MachO::version_min_command lc = {};
5949 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr)
5950 break;
5951 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX ||
5952 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS ||
5953 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS ||
5954 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) {
5955 if (m_data.GetU32(&offset, &lc.version,
5956 (sizeof(lc) / sizeof(uint32_t)) - 2)) {
5957 const uint32_t xxxx = lc.version >> 16;
5958 const uint32_t yy = (lc.version >> 8) & 0xffu;
5959 const uint32_t zz = lc.version & 0xffu;
5960 if (xxxx) {
5961 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz);
5962 break;
5963 }
5964 }
5965 } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) {
5966 // struct build_version_command {
5967 // uint32_t cmd; /* LC_BUILD_VERSION */
5968 // uint32_t cmdsize; /* sizeof(struct
5969 // build_version_command) plus */
5970 // /* ntools * sizeof(struct
5971 // build_tool_version) */
5972 // uint32_t platform; /* platform */
5973 // uint32_t minos; /* X.Y.Z is encoded in nibbles
5974 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded in
5975 // nibbles xxxx.yy.zz */ uint32_t ntools; /* number of
5976 // tool entries following this */
5977 // };
5978
5979 offset += 4; // skip platform
5980 uint32_t minos = m_data.GetU32(&offset);
5981
5982 const uint32_t xxxx = minos >> 16;
5983 const uint32_t yy = (minos >> 8) & 0xffu;
5984 const uint32_t zz = minos & 0xffu;
5985 if (xxxx) {
5986 m_min_os_version = llvm::VersionTuple(xxxx, yy, zz);
5987 break;
5988 }
5989 }
5990
5991 offset = load_cmd_offset + lc.cmdsize;
5992 }
5993
5994 if (!m_min_os_version) {
5995 // Set version to an empty value so we don't keep trying to
5996 m_min_os_version = llvm::VersionTuple();
5997 }
5998 }
5999
6000 return *m_min_os_version;
6001 }
6002
6003 llvm::VersionTuple ObjectFileMachO::GetSDKVersion() {
6004 if (!m_sdk_versions) {
6005 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
6006 for (uint32_t i = 0; i < m_header.ncmds; ++i) {
6007 const lldb::offset_t load_cmd_offset = offset;
6008
6009 llvm::MachO::version_min_command lc = {};
6010 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr)
6011 break;
6012 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX ||
6013 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS ||
6014 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS ||
6015 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) {
6016 if (m_data.GetU32(&offset, &lc.version,
6017 (sizeof(lc) / sizeof(uint32_t)) - 2)) {
6018 const uint32_t xxxx = lc.sdk >> 16;
6019 const uint32_t yy = (lc.sdk >> 8) & 0xffu;
6020 const uint32_t zz = lc.sdk & 0xffu;
6021 if (xxxx) {
6022 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz);
6023 break;
6024 } else {
6025 GetModule()->ReportWarning("minimum OS version load command with "
6026 "invalid (0) version found.");
6027 }
6028 }
6029 }
6030 offset = load_cmd_offset + lc.cmdsize;
6031 }
6032
6033 if (!m_sdk_versions) {
6034 offset = MachHeaderSizeFromMagic(m_header.magic);
6035 for (uint32_t i = 0; i < m_header.ncmds; ++i) {
6036 const lldb::offset_t load_cmd_offset = offset;
6037
6038 llvm::MachO::version_min_command lc = {};
6039 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr)
6040 break;
6041 if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) {
6042 // struct build_version_command {
6043 // uint32_t cmd; /* LC_BUILD_VERSION */
6044 // uint32_t cmdsize; /* sizeof(struct
6045 // build_version_command) plus */
6046 // /* ntools * sizeof(struct
6047 // build_tool_version) */
6048 // uint32_t platform; /* platform */
6049 // uint32_t minos; /* X.Y.Z is encoded in nibbles
6050 // xxxx.yy.zz */ uint32_t sdk; /* X.Y.Z is encoded
6051 // in nibbles xxxx.yy.zz */ uint32_t ntools; /* number
6052 // of tool entries following this */
6053 // };
6054
6055 offset += 4; // skip platform
6056 uint32_t minos = m_data.GetU32(&offset);
6057
6058 const uint32_t xxxx = minos >> 16;
6059 const uint32_t yy = (minos >> 8) & 0xffu;
6060 const uint32_t zz = minos & 0xffu;
6061 if (xxxx) {
6062 m_sdk_versions = llvm::VersionTuple(xxxx, yy, zz);
6063 break;
6064 }
6065 }
6066 offset = load_cmd_offset + lc.cmdsize;
6067 }
6068 }
6069
6070 if (!m_sdk_versions)
6071 m_sdk_versions = llvm::VersionTuple();
6072 }
6073
6074 return *m_sdk_versions;
6075 }
6076
6077 bool ObjectFileMachO::GetIsDynamicLinkEditor() {
6078 return m_header.filetype == llvm::MachO::MH_DYLINKER;
6079 }
6080
6081 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() {
6082 return m_allow_assembly_emulation_unwind_plans;
6083 }
6084
6085 Section *ObjectFileMachO::GetMachHeaderSection() {
6086 // Find the first address of the mach header which is the first non-zero file
6087 // sized section whose file offset is zero. This is the base file address of
6088 // the mach-o file which can be subtracted from the vmaddr of the other
6089 // segments found in memory and added to the load address
6090 ModuleSP module_sp = GetModule();
6091 if (!module_sp)
6092 return nullptr;
6093 SectionList *section_list = GetSectionList();
6094 if (!section_list)
6095 return nullptr;
6096 const size_t num_sections = section_list->GetSize();
6097 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) {
6098 Section *section = section_list->GetSectionAtIndex(sect_idx).get();
6099 if (section->GetFileOffset() == 0 && SectionIsLoadable(section))
6100 return section;
6101 }
6102
6103 // We may have a binary in the shared cache that has a non-zero
6104 // file address for its first segment, traditionally the __TEXT segment.
6105 // Search for it by name and return it as our next best guess.
6106 SectionSP text_segment_sp =
6107 GetSectionList()->FindSectionByName(GetSegmentNameTEXT());
6108 if (text_segment_sp.get() && SectionIsLoadable(text_segment_sp.get()))
6109 return text_segment_sp.get();
6110
6111 return nullptr;
6112 }
6113
6114 bool ObjectFileMachO::SectionIsLoadable(const Section *section) {
6115 if (!section)
6116 return false;
6117 const bool is_dsym = (m_header.filetype == MH_DSYM);
6118 if (section->GetFileSize() == 0 && !is_dsym)
6119 return false;
6120 if (section->IsThreadSpecific())
6121 return false;
6122 if (GetModule().get() != section->GetModule().get())
6123 return false;
6124 // Be careful with __LINKEDIT and __DWARF segments
6125 if (section->GetName() == GetSegmentNameLINKEDIT() ||
6126 section->GetName() == GetSegmentNameDWARF()) {
6127 // Only map __LINKEDIT and __DWARF if we have an in memory image and
6128 // this isn't a kernel binary like a kext or mach_kernel.
6129 const bool is_memory_image = (bool)m_process_wp.lock();
6130 const Strata strata = GetStrata();
6131 if (is_memory_image == false || strata == eStrataKernel)
6132 return false;
6133 }
6134 return true;
6135 }
6136
6137 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage(
6138 lldb::addr_t header_load_address, const Section *header_section,
6139 const Section *section) {
6140 ModuleSP module_sp = GetModule();
6141 if (module_sp && header_section && section &&
6142 header_load_address != LLDB_INVALID_ADDRESS) {
6143 lldb::addr_t file_addr = header_section->GetFileAddress();
6144 if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section))
6145 return section->GetFileAddress() - file_addr + header_load_address;
6146 }
6147 return LLDB_INVALID_ADDRESS;
6148 }
6149
6150 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value,
6151 bool value_is_offset) {
6152 ModuleSP module_sp = GetModule();
6153 if (!module_sp)
6154 return false;
6155
6156 SectionList *section_list = GetSectionList();
6157 if (!section_list)
6158 return false;
6159
6160 size_t num_loaded_sections = 0;
6161 const size_t num_sections = section_list->GetSize();
6162
6163 if (value_is_offset) {
6164 // "value" is an offset to apply to each top level segment
6165 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) {
6166 // Iterate through the object file sections to find all of the
6167 // sections that size on disk (to avoid __PAGEZERO) and load them
6168 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx));
6169 if (SectionIsLoadable(section_sp.get()))
6170 if (target.GetSectionLoadList().SetSectionLoadAddress(
6171 section_sp, section_sp->GetFileAddress() + value))
6172 ++num_loaded_sections;
6173 }
6174 } else {
6175 // "value" is the new base address of the mach_header, adjust each
6176 // section accordingly
6177
6178 Section *mach_header_section = GetMachHeaderSection();
6179 if (mach_header_section) {
6180 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) {
6181 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx));
6182
6183 lldb::addr_t section_load_addr =
6184 CalculateSectionLoadAddressForMemoryImage(
6185 value, mach_header_section, section_sp.get());
6186 if (section_load_addr != LLDB_INVALID_ADDRESS) {
6187 if (target.GetSectionLoadList().SetSectionLoadAddress(
6188 section_sp, section_load_addr))
6189 ++num_loaded_sections;
6190 }
6191 }
6192 }
6193 }
6194 return num_loaded_sections > 0;
6195 }
6196
6197 struct all_image_infos_header {
6198 uint32_t version; // currently 1
6199 uint32_t imgcount; // number of binary images
6200 uint64_t entries_fileoff; // file offset in the corefile of where the array of
6201 // struct entry's begin.
6202 uint32_t entries_size; // size of 'struct entry'.
6203 uint32_t unused;
6204 };
6205
6206 struct image_entry {
6207 uint64_t filepath_offset; // offset in corefile to c-string of the file path,
6208 // UINT64_MAX if unavailable.
6209 uuid_t uuid; // uint8_t[16]. should be set to all zeroes if
6210 // uuid is unknown.
6211 uint64_t load_address; // UINT64_MAX if unknown.
6212 uint64_t seg_addrs_offset; // offset to the array of struct segment_vmaddr's.
6213 uint32_t segment_count; // The number of segments for this binary.
6214 uint32_t unused;
6215
6216 image_entry() {
6217 filepath_offset = UINT64_MAX;
6218 memset(&uuid, 0, sizeof(uuid_t));
6219 segment_count = 0;
6220 load_address = UINT64_MAX;
6221 seg_addrs_offset = UINT64_MAX;
6222 unused = 0;
6223 }
6224 image_entry(const image_entry &rhs) {
6225 filepath_offset = rhs.filepath_offset;
6226 memcpy(&uuid, &rhs.uuid, sizeof(uuid_t));
6227 segment_count = rhs.segment_count;
6228 seg_addrs_offset = rhs.seg_addrs_offset;
6229 load_address = rhs.load_address;
6230 unused = rhs.unused;
6231 }
6232 };
6233
6234 struct segment_vmaddr {
6235 char segname[16];
6236 uint64_t vmaddr;
6237 uint64_t unused;
6238
6239 segment_vmaddr() {
6240 memset(&segname, 0, 16);
6241 vmaddr = UINT64_MAX;
6242 unused = 0;
6243 }
6244 segment_vmaddr(const segment_vmaddr &rhs) {
6245 memcpy(&segname, &rhs.segname, 16);
6246 vmaddr = rhs.vmaddr;
6247 unused = rhs.unused;
6248 }
6249 };
6250
6251 // Write the payload for the "all image infos" LC_NOTE into
6252 // the supplied all_image_infos_payload, assuming that this
6253 // will be written into the corefile starting at
6254 // initial_file_offset.
6255 //
6256 // The placement of this payload is a little tricky. We're
6257 // laying this out as
6258 //
6259 // 1. header (struct all_image_info_header)
6260 // 2. Array of fixed-size (struct image_entry)'s, one
6261 // per binary image present in the process.
6262 // 3. Arrays of (struct segment_vmaddr)'s, a varying number
6263 // for each binary image.
6264 // 4. Variable length c-strings of binary image filepaths,
6265 // one per binary.
6266 //
6267 // To compute where everything will be laid out in the
6268 // payload, we need to iterate over the images and calculate
6269 // how many segment_vmaddr structures each image will need,
6270 // and how long each image's filepath c-string is. There
6271 // are some multiple passes over the image list while calculating
6272 // everything.
6273
6274 static offset_t CreateAllImageInfosPayload(
6275 const lldb::ProcessSP &process_sp, offset_t initial_file_offset,
6276 StreamString &all_image_infos_payload, SaveCoreStyle core_style) {
6277 Target &target = process_sp->GetTarget();
6278 ModuleList modules = target.GetImages();
6279
6280 // stack-only corefiles have no reason to include binaries that
6281 // are not executing; we're trying to make the smallest corefile
6282 // we can, so leave the rest out.
6283 if (core_style == SaveCoreStyle::eSaveCoreStackOnly)
6284 modules.Clear();
6285
6286 std::set<std::string> executing_uuids;
6287 ThreadList &thread_list(process_sp->GetThreadList());
6288 for (uint32_t i = 0; i < thread_list.GetSize(); i++) {
6289 ThreadSP thread_sp = thread_list.GetThreadAtIndex(i);
6290 uint32_t stack_frame_count = thread_sp->GetStackFrameCount();
6291 for (uint32_t j = 0; j < stack_frame_count; j++) {
6292 StackFrameSP stack_frame_sp = thread_sp->GetStackFrameAtIndex(j);
6293 Address pc = stack_frame_sp->GetFrameCodeAddress();
6294 ModuleSP module_sp = pc.GetModule();
6295 if (module_sp) {
6296 UUID uuid = module_sp->GetUUID();
6297 if (uuid.IsValid()) {
6298 executing_uuids.insert(uuid.GetAsString());
6299 modules.AppendIfNeeded(module_sp);
6300 }
6301 }
6302 }
6303 }
6304 size_t modules_count = modules.GetSize();
6305
6306 struct all_image_infos_header infos;
6307 infos.version = 1;
6308 infos.imgcount = modules_count;
6309 infos.entries_size = sizeof(image_entry);
6310 infos.entries_fileoff = initial_file_offset + sizeof(all_image_infos_header);
6311 infos.unused = 0;
6312
6313 all_image_infos_payload.PutHex32(infos.version);
6314 all_image_infos_payload.PutHex32(infos.imgcount);
6315 all_image_infos_payload.PutHex64(infos.entries_fileoff);
6316 all_image_infos_payload.PutHex32(infos.entries_size);
6317 all_image_infos_payload.PutHex32(infos.unused);
6318
6319 // First create the structures for all of the segment name+vmaddr vectors
6320 // for each module, so we will know the size of them as we add the
6321 // module entries.
6322 std::vector<std::vector<segment_vmaddr>> modules_segment_vmaddrs;
6323 for (size_t i = 0; i < modules_count; i++) {
6324 ModuleSP module = modules.GetModuleAtIndex(i);
6325
6326 SectionList *sections = module->GetSectionList();
6327 size_t sections_count = sections->GetSize();
6328 std::vector<segment_vmaddr> segment_vmaddrs;
6329 for (size_t j = 0; j < sections_count; j++) {
6330 SectionSP section = sections->GetSectionAtIndex(j);
6331 if (!section->GetParent().get()) {
6332 addr_t vmaddr = section->GetLoadBaseAddress(&target);
6333 if (vmaddr == LLDB_INVALID_ADDRESS)
6334 continue;
6335 ConstString name = section->GetName();
6336 segment_vmaddr seg_vmaddr;
6337 strncpy(seg_vmaddr.segname, name.AsCString(),
6338 sizeof(seg_vmaddr.segname));
6339 seg_vmaddr.vmaddr = vmaddr;
6340 seg_vmaddr.unused = 0;
6341 segment_vmaddrs.push_back(seg_vmaddr);
6342 }
6343 }
6344 modules_segment_vmaddrs.push_back(segment_vmaddrs);
6345 }
6346
6347 offset_t size_of_vmaddr_structs = 0;
6348 for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) {
6349 size_of_vmaddr_structs +=
6350 modules_segment_vmaddrs[i].size() * sizeof(segment_vmaddr);
6351 }
6352
6353 offset_t size_of_filepath_cstrings = 0;
6354 for (size_t i = 0; i < modules_count; i++) {
6355 ModuleSP module_sp = modules.GetModuleAtIndex(i);
6356 size_of_filepath_cstrings += module_sp->GetFileSpec().GetPath().size() + 1;
6357 }
6358
6359 // Calculate the file offsets of our "all image infos" payload in the
6360 // corefile. initial_file_offset the original value passed in to this method.
6361
6362 offset_t start_of_entries =
6363 initial_file_offset + sizeof(all_image_infos_header);
6364 offset_t start_of_seg_vmaddrs =
6365 start_of_entries + sizeof(image_entry) * modules_count;
6366 offset_t start_of_filenames = start_of_seg_vmaddrs + size_of_vmaddr_structs;
6367
6368 offset_t final_file_offset = start_of_filenames + size_of_filepath_cstrings;
6369
6370 // Now write the one-per-module 'struct image_entry' into the
6371 // StringStream; keep track of where the struct segment_vmaddr
6372 // entries for each module will end up in the corefile.
6373
6374 offset_t current_string_offset = start_of_filenames;
6375 offset_t current_segaddrs_offset = start_of_seg_vmaddrs;
6376 std::vector<struct image_entry> image_entries;
6377 for (size_t i = 0; i < modules_count; i++) {
6378 ModuleSP module_sp = modules.GetModuleAtIndex(i);
6379
6380 struct image_entry ent;
6381 memcpy(&ent.uuid, module_sp->GetUUID().GetBytes().data(), sizeof(ent.uuid));
6382 if (modules_segment_vmaddrs[i].size() > 0) {
6383 ent.segment_count = modules_segment_vmaddrs[i].size();
6384 ent.seg_addrs_offset = current_segaddrs_offset;
6385 }
6386 ent.filepath_offset = current_string_offset;
6387 ObjectFile *objfile = module_sp->GetObjectFile();
6388 if (objfile) {
6389 Address base_addr(objfile->GetBaseAddress());
6390 if (base_addr.IsValid()) {
6391 ent.load_address = base_addr.GetLoadAddress(&target);
6392 }
6393 }
6394
6395 all_image_infos_payload.PutHex64(ent.filepath_offset);
6396 all_image_infos_payload.PutRawBytes(ent.uuid, sizeof(ent.uuid));
6397 all_image_infos_payload.PutHex64(ent.load_address);
6398 all_image_infos_payload.PutHex64(ent.seg_addrs_offset);
6399 all_image_infos_payload.PutHex32(ent.segment_count);
6400
6401 if (executing_uuids.find(module_sp->GetUUID().GetAsString()) !=
6402 executing_uuids.end())
6403 all_image_infos_payload.PutHex32(1);
6404 else
6405 all_image_infos_payload.PutHex32(0);
6406
6407 current_segaddrs_offset += ent.segment_count * sizeof(segment_vmaddr);
6408 current_string_offset += module_sp->GetFileSpec().GetPath().size() + 1;
6409 }
6410
6411 // Now write the struct segment_vmaddr entries into the StringStream.
6412
6413 for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) {
6414 if (modules_segment_vmaddrs[i].size() == 0)
6415 continue;
6416 for (struct segment_vmaddr segvm : modules_segment_vmaddrs[i]) {
6417 all_image_infos_payload.PutRawBytes(segvm.segname, sizeof(segvm.segname));
6418 all_image_infos_payload.PutHex64(segvm.vmaddr);
6419 all_image_infos_payload.PutHex64(segvm.unused);
6420 }
6421 }
6422
6423 for (size_t i = 0; i < modules_count; i++) {
6424 ModuleSP module_sp = modules.GetModuleAtIndex(i);
6425 std::string filepath = module_sp->GetFileSpec().GetPath();
6426 all_image_infos_payload.PutRawBytes(filepath.data(), filepath.size() + 1);
6427 }
6428
6429 return final_file_offset;
6430 }
6431
6432 // Temp struct used to combine contiguous memory regions with
6433 // identical permissions.
6434 struct page_object {
6435 addr_t addr;
6436 addr_t size;
6437 uint32_t prot;
6438 };
6439
6440 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp,
6441 const FileSpec &outfile,
6442 lldb::SaveCoreStyle &core_style, Status &error) {
6443 if (!process_sp)
6444 return false;
6445
6446 // Default on macOS is to create a dirty-memory-only corefile.
6447 if (core_style == SaveCoreStyle::eSaveCoreUnspecified) {
6448 core_style = SaveCoreStyle::eSaveCoreDirtyOnly;
6449 }
6450
6451 Target &target = process_sp->GetTarget();
6452 const ArchSpec target_arch = target.GetArchitecture();
6453 const llvm::Triple &target_triple = target_arch.GetTriple();
6454 if (target_triple.getVendor() == llvm::Triple::Apple &&
6455 (target_triple.getOS() == llvm::Triple::MacOSX ||
6456 target_triple.getOS() == llvm::Triple::IOS ||
6457 target_triple.getOS() == llvm::Triple::WatchOS ||
6458 target_triple.getOS() == llvm::Triple::TvOS)) {
6459 // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS))
6460 // {
6461 bool make_core = false;
6462 switch (target_arch.GetMachine()) {
6463 case llvm::Triple::aarch64:
6464 case llvm::Triple::aarch64_32:
6465 case llvm::Triple::arm:
6466 case llvm::Triple::thumb:
6467 case llvm::Triple::x86:
6468 case llvm::Triple::x86_64:
6469 make_core = true;
6470 break;
6471 default:
6472 error.SetErrorStringWithFormat("unsupported core architecture: %s",
6473 target_triple.str().c_str());
6474 break;
6475 }
6476
6477 if (make_core) {
6478 std::vector<llvm::MachO::segment_command_64> segment_load_commands;
6479 // uint32_t range_info_idx = 0;
6480 MemoryRegionInfo range_info;
6481 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info);
6482 const uint32_t addr_byte_size = target_arch.GetAddressByteSize();
6483 const ByteOrder byte_order = target_arch.GetByteOrder();
6484 std::vector<page_object> pages_to_copy;
6485
6486 if (range_error.Success()) {
6487 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) {
6488 // Calculate correct protections
6489 uint32_t prot = 0;
6490 if (range_info.GetReadable() == MemoryRegionInfo::eYes)
6491 prot |= VM_PROT_READ;
6492 if (range_info.GetWritable() == MemoryRegionInfo::eYes)
6493 prot |= VM_PROT_WRITE;
6494 if (range_info.GetExecutable() == MemoryRegionInfo::eYes)
6495 prot |= VM_PROT_EXECUTE;
6496
6497 const addr_t addr = range_info.GetRange().GetRangeBase();
6498 const addr_t size = range_info.GetRange().GetByteSize();
6499
6500 if (size == 0)
6501 break;
6502
6503 bool include_this_region = true;
6504 bool dirty_pages_only = false;
6505 if (core_style == SaveCoreStyle::eSaveCoreStackOnly) {
6506 dirty_pages_only = true;
6507 if (range_info.IsStackMemory() != MemoryRegionInfo::eYes) {
6508 include_this_region = false;
6509 }
6510 }
6511 if (core_style == SaveCoreStyle::eSaveCoreDirtyOnly) {
6512 dirty_pages_only = true;
6513 }
6514
6515 if (prot != 0 && include_this_region) {
6516 addr_t pagesize = range_info.GetPageSize();
6517 const llvm::Optional<std::vector<addr_t>> &dirty_page_list =
6518 range_info.GetDirtyPageList();
6519 if (dirty_pages_only && dirty_page_list) {
6520 for (addr_t dirtypage : dirty_page_list.value()) {
6521 page_object obj;
6522 obj.addr = dirtypage;
6523 obj.size = pagesize;
6524 obj.prot = prot;
6525 pages_to_copy.push_back(obj);
6526 }
6527 } else {
6528 page_object obj;
6529 obj.addr = addr;
6530 obj.size = size;
6531 obj.prot = prot;
6532 pages_to_copy.push_back(obj);
6533 }
6534 }
6535
6536 range_error = process_sp->GetMemoryRegionInfo(
6537 range_info.GetRange().GetRangeEnd(), range_info);
6538 if (range_error.Fail())
6539 break;
6540 }
6541
6542 // Combine contiguous entries that have the same
6543 // protections so we don't have an excess of
6544 // load commands.
6545 std::vector<page_object> combined_page_objects;
6546 page_object last_obj;
6547 last_obj.addr = LLDB_INVALID_ADDRESS;
6548 last_obj.size = 0;
6549 for (page_object obj : pages_to_copy) {
6550 if (last_obj.addr == LLDB_INVALID_ADDRESS) {
6551 last_obj = obj;
6552 continue;
6553 }
6554 if (last_obj.addr + last_obj.size == obj.addr &&
6555 last_obj.prot == obj.prot) {
6556 last_obj.size += obj.size;
6557 continue;
6558 }
6559 combined_page_objects.push_back(last_obj);
6560 last_obj = obj;
6561 }
6562 // Add the last entry we were looking to combine
6563 // on to the array.
6564 if (last_obj.addr != LLDB_INVALID_ADDRESS && last_obj.size != 0)
6565 combined_page_objects.push_back(last_obj);
6566
6567 for (page_object obj : combined_page_objects) {
6568 uint32_t cmd_type = LC_SEGMENT_64;
6569 uint32_t segment_size = sizeof(llvm::MachO::segment_command_64);
6570 if (addr_byte_size == 4) {
6571 cmd_type = LC_SEGMENT;
6572 segment_size = sizeof(llvm::MachO::segment_command);
6573 }
6574 llvm::MachO::segment_command_64 segment = {
6575 cmd_type, // uint32_t cmd;
6576 segment_size, // uint32_t cmdsize;
6577 {0}, // char segname[16];
6578 obj.addr, // uint64_t vmaddr; // uint32_t for 32-bit
6579 // Mach-O
6580 obj.size, // uint64_t vmsize; // uint32_t for 32-bit
6581 // Mach-O
6582 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O
6583 obj.size, // uint64_t filesize; // uint32_t for 32-bit
6584 // Mach-O
6585 obj.prot, // uint32_t maxprot;
6586 obj.prot, // uint32_t initprot;
6587 0, // uint32_t nsects;
6588 0}; // uint32_t flags;
6589 segment_load_commands.push_back(segment);
6590 }
6591
6592 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order);
6593
6594 llvm::MachO::mach_header_64 mach_header;
6595 if (addr_byte_size == 8) {
6596 mach_header.magic = MH_MAGIC_64;
6597 } else {
6598 mach_header.magic = MH_MAGIC;
6599 }
6600 mach_header.cputype = target_arch.GetMachOCPUType();
6601 mach_header.cpusubtype = target_arch.GetMachOCPUSubType();
6602 mach_header.filetype = MH_CORE;
6603 mach_header.ncmds = segment_load_commands.size();
6604 mach_header.flags = 0;
6605 mach_header.reserved = 0;
6606 ThreadList &thread_list = process_sp->GetThreadList();
6607 const uint32_t num_threads = thread_list.GetSize();
6608
6609 // Make an array of LC_THREAD data items. Each one contains the
6610 // contents of the LC_THREAD load command. The data doesn't contain
6611 // the load command + load command size, we will add the load command
6612 // and load command size as we emit the data.
6613 std::vector<StreamString> LC_THREAD_datas(num_threads);
6614 for (auto &LC_THREAD_data : LC_THREAD_datas) {
6615 LC_THREAD_data.GetFlags().Set(Stream::eBinary);
6616 LC_THREAD_data.SetAddressByteSize(addr_byte_size);
6617 LC_THREAD_data.SetByteOrder(byte_order);
6618 }
6619 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) {
6620 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx));
6621 if (thread_sp) {
6622 switch (mach_header.cputype) {
6623 case llvm::MachO::CPU_TYPE_ARM64:
6624 case llvm::MachO::CPU_TYPE_ARM64_32:
6625 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD(
6626 thread_sp.get(), LC_THREAD_datas[thread_idx]);
6627 break;
6628
6629 case llvm::MachO::CPU_TYPE_ARM:
6630 RegisterContextDarwin_arm_Mach::Create_LC_THREAD(
6631 thread_sp.get(), LC_THREAD_datas[thread_idx]);
6632 break;
6633
6634 case llvm::MachO::CPU_TYPE_I386:
6635 RegisterContextDarwin_i386_Mach::Create_LC_THREAD(
6636 thread_sp.get(), LC_THREAD_datas[thread_idx]);
6637 break;
6638
6639 case llvm::MachO::CPU_TYPE_X86_64:
6640 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD(
6641 thread_sp.get(), LC_THREAD_datas[thread_idx]);
6642 break;
6643 }
6644 }
6645 }
6646
6647 // The size of the load command is the size of the segments...
6648 if (addr_byte_size == 8) {
6649 mach_header.sizeofcmds = segment_load_commands.size() *
6650 sizeof(llvm::MachO::segment_command_64);
6651 } else {
6652 mach_header.sizeofcmds = segment_load_commands.size() *
6653 sizeof(llvm::MachO::segment_command);
6654 }
6655
6656 // and the size of all LC_THREAD load command
6657 for (const auto &LC_THREAD_data : LC_THREAD_datas) {
6658 ++mach_header.ncmds;
6659 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize();
6660 }
6661
6662 // Bits will be set to indicate which bits are NOT used in
6663 // addressing in this process or 0 for unknown.
6664 uint64_t address_mask = process_sp->GetCodeAddressMask();
6665 if (address_mask != 0) {
6666 // LC_NOTE "addrable bits"
6667 mach_header.ncmds++;
6668 mach_header.sizeofcmds += sizeof(llvm::MachO::note_command);
6669 }
6670
6671 // LC_NOTE "all image infos"
6672 mach_header.ncmds++;
6673 mach_header.sizeofcmds += sizeof(llvm::MachO::note_command);
6674
6675 // Write the mach header
6676 buffer.PutHex32(mach_header.magic);
6677 buffer.PutHex32(mach_header.cputype);
6678 buffer.PutHex32(mach_header.cpusubtype);
6679 buffer.PutHex32(mach_header.filetype);
6680 buffer.PutHex32(mach_header.ncmds);
6681 buffer.PutHex32(mach_header.sizeofcmds);
6682 buffer.PutHex32(mach_header.flags);
6683 if (addr_byte_size == 8) {
6684 buffer.PutHex32(mach_header.reserved);
6685 }
6686
6687 // Skip the mach header and all load commands and align to the next
6688 // 0x1000 byte boundary
6689 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds;
6690
6691 file_offset = llvm::alignTo(file_offset, 16);
6692 std::vector<std::unique_ptr<LCNoteEntry>> lc_notes;
6693
6694 // Add "addrable bits" LC_NOTE when an address mask is available
6695 if (address_mask != 0) {
6696 std::unique_ptr<LCNoteEntry> addrable_bits_lcnote_up(
6697 new LCNoteEntry(addr_byte_size, byte_order));
6698 addrable_bits_lcnote_up->name = "addrable bits";
6699 addrable_bits_lcnote_up->payload_file_offset = file_offset;
6700 int bits = std::bitset<64>(~address_mask).count();
6701 addrable_bits_lcnote_up->payload.PutHex32(3); // version
6702 addrable_bits_lcnote_up->payload.PutHex32(
6703 bits); // # of bits used for addressing
6704 addrable_bits_lcnote_up->payload.PutHex64(0); // unused
6705
6706 file_offset += addrable_bits_lcnote_up->payload.GetSize();
6707
6708 lc_notes.push_back(std::move(addrable_bits_lcnote_up));
6709 }
6710
6711 // Add "all image infos" LC_NOTE
6712 std::unique_ptr<LCNoteEntry> all_image_infos_lcnote_up(
6713 new LCNoteEntry(addr_byte_size, byte_order));
6714 all_image_infos_lcnote_up->name = "all image infos";
6715 all_image_infos_lcnote_up->payload_file_offset = file_offset;
6716 file_offset = CreateAllImageInfosPayload(
6717 process_sp, file_offset, all_image_infos_lcnote_up->payload,
6718 core_style);
6719 lc_notes.push_back(std::move(all_image_infos_lcnote_up));
6720
6721 // Add LC_NOTE load commands
6722 for (auto &lcnote : lc_notes) {
6723 // Add the LC_NOTE load command to the file.
6724 buffer.PutHex32(LC_NOTE);
6725 buffer.PutHex32(sizeof(llvm::MachO::note_command));
6726 char namebuf[16];
6727 memset(namebuf, 0, sizeof(namebuf));
6728 // this is the uncommon case where strncpy is exactly
6729 // the right one, doesn't need to be nul terminated.
6730 strncpy(namebuf, lcnote->name.c_str(), sizeof(namebuf));
6731 buffer.PutRawBytes(namebuf, sizeof(namebuf));
6732 buffer.PutHex64(lcnote->payload_file_offset);
6733 buffer.PutHex64(lcnote->payload.GetSize());
6734 }
6735
6736 // Align to 4096-byte page boundary for the LC_SEGMENTs.
6737 file_offset = llvm::alignTo(file_offset, 4096);
6738
6739 for (auto &segment : segment_load_commands) {
6740 segment.fileoff = file_offset;
6741 file_offset += segment.filesize;
6742 }
6743
6744 // Write out all of the LC_THREAD load commands
6745 for (const auto &LC_THREAD_data : LC_THREAD_datas) {
6746 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize();
6747 buffer.PutHex32(LC_THREAD);
6748 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data
6749 buffer.Write(LC_THREAD_data.GetString().data(), LC_THREAD_data_size);
6750 }
6751
6752 // Write out all of the segment load commands
6753 for (const auto &segment : segment_load_commands) {
6754 buffer.PutHex32(segment.cmd);
6755 buffer.PutHex32(segment.cmdsize);
6756 buffer.PutRawBytes(segment.segname, sizeof(segment.segname));
6757 if (addr_byte_size == 8) {
6758 buffer.PutHex64(segment.vmaddr);
6759 buffer.PutHex64(segment.vmsize);
6760 buffer.PutHex64(segment.fileoff);
6761 buffer.PutHex64(segment.filesize);
6762 } else {
6763 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr));
6764 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize));
6765 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff));
6766 buffer.PutHex32(static_cast<uint32_t>(segment.filesize));
6767 }
6768 buffer.PutHex32(segment.maxprot);
6769 buffer.PutHex32(segment.initprot);
6770 buffer.PutHex32(segment.nsects);
6771 buffer.PutHex32(segment.flags);
6772 }
6773
6774 std::string core_file_path(outfile.GetPath());
6775 auto core_file = FileSystem::Instance().Open(
6776 outfile, File::eOpenOptionWriteOnly | File::eOpenOptionTruncate |
6777 File::eOpenOptionCanCreate);
6778 if (!core_file) {
6779 error = core_file.takeError();
6780 } else {
6781 // Read 1 page at a time
6782 uint8_t bytes[0x1000];
6783 // Write the mach header and load commands out to the core file
6784 size_t bytes_written = buffer.GetString().size();
6785 error =
6786 core_file.get()->Write(buffer.GetString().data(), bytes_written);
6787 if (error.Success()) {
6788
6789 for (auto &lcnote : lc_notes) {
6790 if (core_file.get()->SeekFromStart(lcnote->payload_file_offset) ==
6791 -1) {
6792 error.SetErrorStringWithFormat("Unable to seek to corefile pos "
6793 "to write '%s' LC_NOTE payload",
6794 lcnote->name.c_str());
6795 return false;
6796 }
6797 bytes_written = lcnote->payload.GetSize();
6798 error = core_file.get()->Write(lcnote->payload.GetData(),
6799 bytes_written);
6800 if (!error.Success())
6801 return false;
6802 }
6803
6804 // Now write the file data for all memory segments in the process
6805 for (const auto &segment : segment_load_commands) {
6806 if (core_file.get()->SeekFromStart(segment.fileoff) == -1) {
6807 error.SetErrorStringWithFormat(
6808 "unable to seek to offset 0x%" PRIx64 " in '%s'",
6809 segment.fileoff, core_file_path.c_str());
6810 break;
6811 }
6812
6813 target.GetDebugger().GetAsyncOutputStream()->Printf(
6814 "Saving %" PRId64
6815 " bytes of data for memory region at 0x%" PRIx64 "\n",
6816 segment.vmsize, segment.vmaddr);
6817 addr_t bytes_left = segment.vmsize;
6818 addr_t addr = segment.vmaddr;
6819 Status memory_read_error;
6820 while (bytes_left > 0 && error.Success()) {
6821 const size_t bytes_to_read =
6822 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left;
6823
6824 // In a savecore setting, we don't really care about caching,
6825 // as the data is dumped and very likely never read again,
6826 // so we call ReadMemoryFromInferior to bypass it.
6827 const size_t bytes_read = process_sp->ReadMemoryFromInferior(
6828 addr, bytes, bytes_to_read, memory_read_error);
6829
6830 if (bytes_read == bytes_to_read) {
6831 size_t bytes_written = bytes_read;
6832 error = core_file.get()->Write(bytes, bytes_written);
6833 bytes_left -= bytes_read;
6834 addr += bytes_read;
6835 } else {
6836 // Some pages within regions are not readable, those should
6837 // be zero filled
6838 memset(bytes, 0, bytes_to_read);
6839 size_t bytes_written = bytes_to_read;
6840 error = core_file.get()->Write(bytes, bytes_written);
6841 bytes_left -= bytes_to_read;
6842 addr += bytes_to_read;
6843 }
6844 }
6845 }
6846 }
6847 }
6848 } else {
6849 error.SetErrorString(
6850 "process doesn't support getting memory region info");
6851 }
6852 }
6853 return true; // This is the right plug to handle saving core files for
6854 // this process
6855 }
6856 return false;
6857 }
6858
6859 ObjectFileMachO::MachOCorefileAllImageInfos
6860 ObjectFileMachO::GetCorefileAllImageInfos() {
6861 MachOCorefileAllImageInfos image_infos;
6862
6863 // Look for an "all image infos" LC_NOTE.
6864 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic);
6865 for (uint32_t i = 0; i < m_header.ncmds; ++i) {
6866 const uint32_t cmd_offset = offset;
6867 llvm::MachO::load_command lc = {};
6868 if (m_data.GetU32(&offset, &lc.cmd, 2) == nullptr)
6869 break;
6870 if (lc.cmd == LC_NOTE) {
6871 char data_owner[17];
6872 m_data.CopyData(offset, 16, data_owner);
6873 data_owner[16] = '\0';
6874 offset += 16;
6875 uint64_t fileoff = m_data.GetU64_unchecked(&offset);
6876 offset += 4; /* size unused */
6877
6878 if (strcmp("all image infos", data_owner) == 0) {
6879 offset = fileoff;
6880 // Read the struct all_image_infos_header.
6881 uint32_t version = m_data.GetU32(&offset);
6882 if (version != 1) {
6883 return image_infos;
6884 }
6885 uint32_t imgcount = m_data.GetU32(&offset);
6886 uint64_t entries_fileoff = m_data.GetU64(&offset);
6887 offset += 4; // uint32_t entries_size;
6888 offset += 4; // uint32_t unused;
6889
6890 offset = entries_fileoff;
6891 for (uint32_t i = 0; i < imgcount; i++) {
6892 // Read the struct image_entry.
6893 offset_t filepath_offset = m_data.GetU64(&offset);
6894 uuid_t uuid;
6895 memcpy(&uuid, m_data.GetData(&offset, sizeof(uuid_t)),
6896 sizeof(uuid_t));
6897 uint64_t load_address = m_data.GetU64(&offset);
6898 offset_t seg_addrs_offset = m_data.GetU64(&offset);
6899 uint32_t segment_count = m_data.GetU32(&offset);
6900 uint32_t currently_executing = m_data.GetU32(&offset);
6901
6902 MachOCorefileImageEntry image_entry;
6903 image_entry.filename = (const char *)m_data.GetCStr(&filepath_offset);
6904 image_entry.uuid = UUID::fromData(uuid, sizeof(uuid_t));
6905 image_entry.load_address = load_address;
6906 image_entry.currently_executing = currently_executing;
6907
6908 offset_t seg_vmaddrs_offset = seg_addrs_offset;
6909 for (uint32_t j = 0; j < segment_count; j++) {
6910 char segname[17];
6911 m_data.CopyData(seg_vmaddrs_offset, 16, segname);
6912 segname[16] = '\0';
6913 seg_vmaddrs_offset += 16;
6914 uint64_t vmaddr = m_data.GetU64(&seg_vmaddrs_offset);
6915 seg_vmaddrs_offset += 8; /* unused */
6916
6917 std::tuple<ConstString, addr_t> new_seg{ConstString(segname),
6918 vmaddr};
6919 image_entry.segment_load_addresses.push_back(new_seg);
6920 }
6921 image_infos.all_image_infos.push_back(image_entry);
6922 }
6923 } else if (strcmp("load binary", data_owner) == 0) {
6924 uint32_t version = m_data.GetU32(&fileoff);
6925 if (version == 1) {
6926 uuid_t uuid;
6927 memcpy(&uuid, m_data.GetData(&fileoff, sizeof(uuid_t)),
6928 sizeof(uuid_t));
6929 uint64_t load_address = m_data.GetU64(&fileoff);
6930 uint64_t slide = m_data.GetU64(&fileoff);
6931 std::string filename = m_data.GetCStr(&fileoff);
6932
6933 MachOCorefileImageEntry image_entry;
6934 image_entry.filename = filename;
6935 image_entry.uuid = UUID::fromData(uuid, sizeof(uuid_t));
6936 image_entry.load_address = load_address;
6937 image_entry.slide = slide;
6938 image_infos.all_image_infos.push_back(image_entry);
6939 }
6940 }
6941 }
6942 offset = cmd_offset + lc.cmdsize;
6943 }
6944
6945 return image_infos;
6946 }
6947
6948 bool ObjectFileMachO::LoadCoreFileImages(lldb_private::Process &process) {
6949 MachOCorefileAllImageInfos image_infos = GetCorefileAllImageInfos();
6950 Log *log = GetLog(LLDBLog::DynamicLoader);
6951
6952 ModuleList added_modules;
6953 for (const MachOCorefileImageEntry &image : image_infos.all_image_infos) {
6954 ModuleSpec module_spec;
6955 module_spec.GetUUID() = image.uuid;
6956 if (image.filename.empty()) {
6957 char namebuf[80];
6958 if (image.load_address != LLDB_INVALID_ADDRESS)
6959 snprintf(namebuf, sizeof(namebuf), "mem-image-0x%" PRIx64,
6960 image.load_address);
6961 else
6962 snprintf(namebuf, sizeof(namebuf), "mem-image+0x%" PRIx64, image.slide);
6963 module_spec.GetFileSpec() = FileSpec(namebuf);
6964 } else {
6965 module_spec.GetFileSpec() = FileSpec(image.filename.c_str());
6966 }
6967 if (image.currently_executing) {
6968 Status error;
6969 Symbols::DownloadObjectAndSymbolFile(module_spec, error, true);
6970 if (FileSystem::Instance().Exists(module_spec.GetFileSpec())) {
6971 process.GetTarget().GetOrCreateModule(module_spec, false);
6972 }
6973 }
6974 Status error;
6975 ModuleSP module_sp =
6976 process.GetTarget().GetOrCreateModule(module_spec, false, &error);
6977 if (!module_sp.get() || !module_sp->GetObjectFile()) {
6978 if (image.load_address != LLDB_INVALID_ADDRESS) {
6979 module_sp = process.ReadModuleFromMemory(module_spec.GetFileSpec(),
6980 image.load_address);
6981 }
6982 }
6983 if (module_sp.get()) {
6984 // Will call ModulesDidLoad with all modules once they've all
6985 // been added to the Target with load addresses. Don't notify
6986 // here, before the load address is set.
6987 const bool notify = false;
6988 process.GetTarget().GetImages().AppendIfNeeded(module_sp, notify);
6989 added_modules.Append(module_sp, notify);
6990 if (image.segment_load_addresses.size() > 0) {
6991 if (log) {
6992 std::string uuidstr = image.uuid.GetAsString();
6993 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' "
6994 "UUID %s with section load addresses",
6995 image.filename.c_str(), uuidstr.c_str());
6996 }
6997 for (auto name_vmaddr_tuple : image.segment_load_addresses) {
6998 SectionList *sectlist = module_sp->GetObjectFile()->GetSectionList();
6999 if (sectlist) {
7000 SectionSP sect_sp =
7001 sectlist->FindSectionByName(std::get<0>(name_vmaddr_tuple));
7002 if (sect_sp) {
7003 process.GetTarget().SetSectionLoadAddress(
7004 sect_sp, std::get<1>(name_vmaddr_tuple));
7005 }
7006 }
7007 }
7008 } else if (image.load_address != LLDB_INVALID_ADDRESS) {
7009 if (log) {
7010 std::string uuidstr = image.uuid.GetAsString();
7011 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' "
7012 "UUID %s with load address 0x%" PRIx64,
7013 image.filename.c_str(), uuidstr.c_str(),
7014 image.load_address);
7015 }
7016 const bool address_is_slide = false;
7017 bool changed = false;
7018 module_sp->SetLoadAddress(process.GetTarget(), image.load_address,
7019 address_is_slide, changed);
7020 } else if (image.slide != 0) {
7021 if (log) {
7022 std::string uuidstr = image.uuid.GetAsString();
7023 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' "
7024 "UUID %s with slide amount 0x%" PRIx64,
7025 image.filename.c_str(), uuidstr.c_str(), image.slide);
7026 }
7027 const bool address_is_slide = true;
7028 bool changed = false;
7029 module_sp->SetLoadAddress(process.GetTarget(), image.slide,
7030 address_is_slide, changed);
7031 } else {
7032 if (log) {
7033 std::string uuidstr = image.uuid.GetAsString();
7034 log->Printf("ObjectFileMachO::LoadCoreFileImages adding binary '%s' "
7035 "UUID %s at its file address, no slide applied",
7036 image.filename.c_str(), uuidstr.c_str());
7037 }
7038 const bool address_is_slide = true;
7039 bool changed = false;
7040 module_sp->SetLoadAddress(process.GetTarget(), 0, address_is_slide,
7041 changed);
7042 }
7043 }
7044 }
7045 if (added_modules.GetSize() > 0) {
7046 process.GetTarget().ModulesDidLoad(added_modules);
7047 process.Flush();
7048 return true;
7049 }
7050 return false;
7051 }
7052