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/Module.h" |
18 | #include "lldb/Core/ModuleSpec.h" |
19 | #include "lldb/Core/PluginManager.h" |
20 | #include "lldb/Core/Progress.h" |
21 | #include "lldb/Core/Section.h" |
22 | #include "lldb/Host/Host.h" |
23 | #include "lldb/Symbol/DWARFCallFrameInfo.h" |
24 | #include "lldb/Symbol/ObjectFile.h" |
25 | #include "lldb/Target/DynamicLoader.h" |
26 | #include "lldb/Target/MemoryRegionInfo.h" |
27 | #include "lldb/Target/Platform.h" |
28 | #include "lldb/Target/Process.h" |
29 | #include "lldb/Target/SectionLoadList.h" |
30 | #include "lldb/Target/Target.h" |
31 | #include "lldb/Target/Thread.h" |
32 | #include "lldb/Target/ThreadList.h" |
33 | #include "lldb/Utility/ArchSpec.h" |
34 | #include "lldb/Utility/DataBuffer.h" |
35 | #include "lldb/Utility/FileSpec.h" |
36 | #include "lldb/Utility/FileSpecList.h" |
37 | #include "lldb/Utility/LLDBLog.h" |
38 | #include "lldb/Utility/Log.h" |
39 | #include "lldb/Utility/RangeMap.h" |
40 | #include "lldb/Utility/RegisterValue.h" |
41 | #include "lldb/Utility/Status.h" |
42 | #include "lldb/Utility/StreamString.h" |
43 | #include "lldb/Utility/Timer.h" |
44 | #include "lldb/Utility/UUID.h" |
45 | |
46 | #include "lldb/Host/SafeMachO.h" |
47 | |
48 | #include "llvm/ADT/DenseSet.h" |
49 | #include "llvm/Support/FormatVariadic.h" |
50 | #include "llvm/Support/MemoryBuffer.h" |
51 | |
52 | #include "ObjectFileMachO.h" |
53 | |
54 | #if defined(__APPLE__) |
55 | #include <TargetConditionals.h> |
56 | // GetLLDBSharedCacheUUID() needs to call dlsym() |
57 | #include <dlfcn.h> |
58 | #include <mach/mach_init.h> |
59 | #include <mach/vm_map.h> |
60 | #include <lldb/Host/SafeMachO.h> |
61 | #endif |
62 | |
63 | #ifndef __APPLE__ |
64 | #include "lldb/Utility/AppleUuidCompatibility.h" |
65 | #else |
66 | #include <uuid/uuid.h> |
67 | #endif |
68 | |
69 | #include <bitset> |
70 | #include <memory> |
71 | #include <optional> |
72 | |
73 | // Unfortunately the signpost header pulls in the system MachO header, too. |
74 | #ifdef CPU_TYPE_ARM |
75 | #undef CPU_TYPE_ARM |
76 | #endif |
77 | #ifdef CPU_TYPE_ARM64 |
78 | #undef CPU_TYPE_ARM64 |
79 | #endif |
80 | #ifdef CPU_TYPE_ARM64_32 |
81 | #undef CPU_TYPE_ARM64_32 |
82 | #endif |
83 | #ifdef CPU_TYPE_I386 |
84 | #undef CPU_TYPE_I386 |
85 | #endif |
86 | #ifdef CPU_TYPE_X86_64 |
87 | #undef CPU_TYPE_X86_64 |
88 | #endif |
89 | #ifdef MH_DYLINKER |
90 | #undef MH_DYLINKER |
91 | #endif |
92 | #ifdef MH_OBJECT |
93 | #undef MH_OBJECT |
94 | #endif |
95 | #ifdef LC_VERSION_MIN_MACOSX |
96 | #undef LC_VERSION_MIN_MACOSX |
97 | #endif |
98 | #ifdef LC_VERSION_MIN_IPHONEOS |
99 | #undef LC_VERSION_MIN_IPHONEOS |
100 | #endif |
101 | #ifdef LC_VERSION_MIN_TVOS |
102 | #undef LC_VERSION_MIN_TVOS |
103 | #endif |
104 | #ifdef LC_VERSION_MIN_WATCHOS |
105 | #undef LC_VERSION_MIN_WATCHOS |
106 | #endif |
107 | #ifdef LC_BUILD_VERSION |
108 | #undef LC_BUILD_VERSION |
109 | #endif |
110 | #ifdef PLATFORM_MACOS |
111 | #undef PLATFORM_MACOS |
112 | #endif |
113 | #ifdef PLATFORM_MACCATALYST |
114 | #undef PLATFORM_MACCATALYST |
115 | #endif |
116 | #ifdef PLATFORM_IOS |
117 | #undef PLATFORM_IOS |
118 | #endif |
119 | #ifdef PLATFORM_IOSSIMULATOR |
120 | #undef PLATFORM_IOSSIMULATOR |
121 | #endif |
122 | #ifdef PLATFORM_TVOS |
123 | #undef PLATFORM_TVOS |
124 | #endif |
125 | #ifdef PLATFORM_TVOSSIMULATOR |
126 | #undef PLATFORM_TVOSSIMULATOR |
127 | #endif |
128 | #ifdef PLATFORM_WATCHOS |
129 | #undef PLATFORM_WATCHOS |
130 | #endif |
131 | #ifdef PLATFORM_WATCHOSSIMULATOR |
132 | #undef PLATFORM_WATCHOSSIMULATOR |
133 | #endif |
134 | |
135 | #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull |
136 | using namespace lldb; |
137 | using namespace lldb_private; |
138 | using namespace llvm::MachO; |
139 | |
140 | LLDB_PLUGIN_DEFINE(ObjectFileMachO) |
141 | |
142 | static void PrintRegisterValue(RegisterContext *reg_ctx, const char *name, |
143 | const char *alt_name, size_t reg_byte_size, |
144 | Stream &data) { |
145 | const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(reg_name: name); |
146 | if (reg_info == nullptr) |
147 | reg_info = reg_ctx->GetRegisterInfoByName(reg_name: alt_name); |
148 | if (reg_info) { |
149 | lldb_private::RegisterValue reg_value; |
150 | if (reg_ctx->ReadRegister(reg_info, reg_value)) { |
151 | if (reg_info->byte_size >= reg_byte_size) |
152 | data.Write(src: reg_value.GetBytes(), src_len: reg_byte_size); |
153 | else { |
154 | data.Write(src: reg_value.GetBytes(), src_len: reg_info->byte_size); |
155 | for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; ++i) |
156 | data.PutChar(ch: 0); |
157 | } |
158 | return; |
159 | } |
160 | } |
161 | // Just write zeros if all else fails |
162 | for (size_t i = 0; i < reg_byte_size; ++i) |
163 | data.PutChar(ch: 0); |
164 | } |
165 | |
166 | class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { |
167 | public: |
168 | (lldb_private::Thread &thread, |
169 | const DataExtractor &data) |
170 | : RegisterContextDarwin_x86_64(thread, 0) { |
171 | SetRegisterDataFrom_LC_THREAD(data); |
172 | } |
173 | |
174 | void InvalidateAllRegisters() override { |
175 | // Do nothing... registers are always valid... |
176 | } |
177 | |
178 | void (const DataExtractor &data) { |
179 | lldb::offset_t offset = 0; |
180 | SetError(flavor: GPRRegSet, err_idx: Read, err: -1); |
181 | SetError(flavor: FPURegSet, err_idx: Read, err: -1); |
182 | SetError(flavor: EXCRegSet, err_idx: Read, err: -1); |
183 | bool done = false; |
184 | |
185 | while (!done) { |
186 | int flavor = data.GetU32(offset_ptr: &offset); |
187 | if (flavor == 0) |
188 | done = true; |
189 | else { |
190 | uint32_t i; |
191 | uint32_t count = data.GetU32(offset_ptr: &offset); |
192 | switch (flavor) { |
193 | case GPRRegSet: |
194 | for (i = 0; i < count; ++i) |
195 | (&gpr.rax)[i] = data.GetU64(offset_ptr: &offset); |
196 | SetError(flavor: GPRRegSet, err_idx: Read, err: 0); |
197 | done = true; |
198 | |
199 | break; |
200 | case FPURegSet: |
201 | // TODO: fill in FPU regs.... |
202 | // SetError (FPURegSet, Read, -1); |
203 | done = true; |
204 | |
205 | break; |
206 | case EXCRegSet: |
207 | exc.trapno = data.GetU32(offset_ptr: &offset); |
208 | exc.err = data.GetU32(offset_ptr: &offset); |
209 | exc.faultvaddr = data.GetU64(offset_ptr: &offset); |
210 | SetError(flavor: EXCRegSet, err_idx: Read, err: 0); |
211 | done = true; |
212 | break; |
213 | case 7: |
214 | case 8: |
215 | case 9: |
216 | // fancy flavors that encapsulate of the above flavors... |
217 | break; |
218 | |
219 | default: |
220 | done = true; |
221 | break; |
222 | } |
223 | } |
224 | } |
225 | } |
226 | |
227 | static bool Create_LC_THREAD(Thread *thread, Stream &data) { |
228 | RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); |
229 | if (reg_ctx_sp) { |
230 | RegisterContext *reg_ctx = reg_ctx_sp.get(); |
231 | |
232 | data.PutHex32(uvalue: GPRRegSet); // Flavor |
233 | data.PutHex32(uvalue: GPRWordCount); |
234 | PrintRegisterValue(reg_ctx, name: "rax" , alt_name: nullptr, reg_byte_size: 8, data); |
235 | PrintRegisterValue(reg_ctx, name: "rbx" , alt_name: nullptr, reg_byte_size: 8, data); |
236 | PrintRegisterValue(reg_ctx, name: "rcx" , alt_name: nullptr, reg_byte_size: 8, data); |
237 | PrintRegisterValue(reg_ctx, name: "rdx" , alt_name: nullptr, reg_byte_size: 8, data); |
238 | PrintRegisterValue(reg_ctx, name: "rdi" , alt_name: nullptr, reg_byte_size: 8, data); |
239 | PrintRegisterValue(reg_ctx, name: "rsi" , alt_name: nullptr, reg_byte_size: 8, data); |
240 | PrintRegisterValue(reg_ctx, name: "rbp" , alt_name: nullptr, reg_byte_size: 8, data); |
241 | PrintRegisterValue(reg_ctx, name: "rsp" , alt_name: nullptr, reg_byte_size: 8, data); |
242 | PrintRegisterValue(reg_ctx, name: "r8" , alt_name: nullptr, reg_byte_size: 8, data); |
243 | PrintRegisterValue(reg_ctx, name: "r9" , alt_name: nullptr, reg_byte_size: 8, data); |
244 | PrintRegisterValue(reg_ctx, name: "r10" , alt_name: nullptr, reg_byte_size: 8, data); |
245 | PrintRegisterValue(reg_ctx, name: "r11" , alt_name: nullptr, reg_byte_size: 8, data); |
246 | PrintRegisterValue(reg_ctx, name: "r12" , alt_name: nullptr, reg_byte_size: 8, data); |
247 | PrintRegisterValue(reg_ctx, name: "r13" , alt_name: nullptr, reg_byte_size: 8, data); |
248 | PrintRegisterValue(reg_ctx, name: "r14" , alt_name: nullptr, reg_byte_size: 8, data); |
249 | PrintRegisterValue(reg_ctx, name: "r15" , alt_name: nullptr, reg_byte_size: 8, data); |
250 | PrintRegisterValue(reg_ctx, name: "rip" , alt_name: nullptr, reg_byte_size: 8, data); |
251 | PrintRegisterValue(reg_ctx, name: "rflags" , alt_name: nullptr, reg_byte_size: 8, data); |
252 | PrintRegisterValue(reg_ctx, name: "cs" , alt_name: nullptr, reg_byte_size: 8, data); |
253 | PrintRegisterValue(reg_ctx, name: "fs" , alt_name: nullptr, reg_byte_size: 8, data); |
254 | PrintRegisterValue(reg_ctx, name: "gs" , alt_name: nullptr, reg_byte_size: 8, data); |
255 | |
256 | // // Write out the FPU registers |
257 | // const size_t fpu_byte_size = sizeof(FPU); |
258 | // size_t bytes_written = 0; |
259 | // data.PutHex32 (FPURegSet); |
260 | // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); |
261 | // bytes_written += data.PutHex32(0); // uint32_t pad[0] |
262 | // bytes_written += data.PutHex32(0); // uint32_t pad[1] |
263 | // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, |
264 | // data); // uint16_t fcw; // "fctrl" |
265 | // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, |
266 | // data); // uint16_t fsw; // "fstat" |
267 | // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, |
268 | // data); // uint8_t ftw; // "ftag" |
269 | // bytes_written += data.PutHex8 (0); // uint8_t pad1; |
270 | // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, |
271 | // data); // uint16_t fop; // "fop" |
272 | // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, |
273 | // data); // uint32_t ip; // "fioff" |
274 | // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, |
275 | // data); // uint16_t cs; // "fiseg" |
276 | // bytes_written += data.PutHex16 (0); // uint16_t pad2; |
277 | // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, |
278 | // data); // uint32_t dp; // "fooff" |
279 | // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, |
280 | // data); // uint16_t ds; // "foseg" |
281 | // bytes_written += data.PutHex16 (0); // uint16_t pad3; |
282 | // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, |
283 | // data); // uint32_t mxcsr; |
284 | // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, |
285 | // 4, data);// uint32_t mxcsrmask; |
286 | // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, |
287 | // sizeof(MMSReg), data); |
288 | // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, |
289 | // sizeof(MMSReg), data); |
290 | // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, |
291 | // sizeof(MMSReg), data); |
292 | // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, |
293 | // sizeof(MMSReg), data); |
294 | // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, |
295 | // sizeof(MMSReg), data); |
296 | // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, |
297 | // sizeof(MMSReg), data); |
298 | // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, |
299 | // sizeof(MMSReg), data); |
300 | // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, |
301 | // sizeof(MMSReg), data); |
302 | // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, |
303 | // sizeof(XMMReg), data); |
304 | // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, |
305 | // sizeof(XMMReg), data); |
306 | // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, |
307 | // sizeof(XMMReg), data); |
308 | // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, |
309 | // sizeof(XMMReg), data); |
310 | // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, |
311 | // sizeof(XMMReg), data); |
312 | // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, |
313 | // sizeof(XMMReg), data); |
314 | // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, |
315 | // sizeof(XMMReg), data); |
316 | // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, |
317 | // sizeof(XMMReg), data); |
318 | // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, |
319 | // sizeof(XMMReg), data); |
320 | // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, |
321 | // sizeof(XMMReg), data); |
322 | // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, |
323 | // sizeof(XMMReg), data); |
324 | // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, |
325 | // sizeof(XMMReg), data); |
326 | // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, |
327 | // sizeof(XMMReg), data); |
328 | // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, |
329 | // sizeof(XMMReg), data); |
330 | // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, |
331 | // sizeof(XMMReg), data); |
332 | // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, |
333 | // sizeof(XMMReg), data); |
334 | // |
335 | // // Fill rest with zeros |
336 | // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ |
337 | // i) |
338 | // data.PutChar(0); |
339 | |
340 | // Write out the EXC registers |
341 | data.PutHex32(uvalue: EXCRegSet); |
342 | data.PutHex32(uvalue: EXCWordCount); |
343 | PrintRegisterValue(reg_ctx, name: "trapno" , alt_name: nullptr, reg_byte_size: 4, data); |
344 | PrintRegisterValue(reg_ctx, name: "err" , alt_name: nullptr, reg_byte_size: 4, data); |
345 | PrintRegisterValue(reg_ctx, name: "faultvaddr" , alt_name: nullptr, reg_byte_size: 8, data); |
346 | return true; |
347 | } |
348 | return false; |
349 | } |
350 | |
351 | protected: |
352 | int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } |
353 | |
354 | int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } |
355 | |
356 | int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } |
357 | |
358 | int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { |
359 | return 0; |
360 | } |
361 | |
362 | int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { |
363 | return 0; |
364 | } |
365 | |
366 | int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { |
367 | return 0; |
368 | } |
369 | }; |
370 | |
371 | class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { |
372 | public: |
373 | (lldb_private::Thread &thread, |
374 | const DataExtractor &data) |
375 | : RegisterContextDarwin_i386(thread, 0) { |
376 | SetRegisterDataFrom_LC_THREAD(data); |
377 | } |
378 | |
379 | void InvalidateAllRegisters() override { |
380 | // Do nothing... registers are always valid... |
381 | } |
382 | |
383 | void (const DataExtractor &data) { |
384 | lldb::offset_t offset = 0; |
385 | SetError(flavor: GPRRegSet, err_idx: Read, err: -1); |
386 | SetError(flavor: FPURegSet, err_idx: Read, err: -1); |
387 | SetError(flavor: EXCRegSet, err_idx: Read, err: -1); |
388 | bool done = false; |
389 | |
390 | while (!done) { |
391 | int flavor = data.GetU32(offset_ptr: &offset); |
392 | if (flavor == 0) |
393 | done = true; |
394 | else { |
395 | uint32_t i; |
396 | uint32_t count = data.GetU32(offset_ptr: &offset); |
397 | switch (flavor) { |
398 | case GPRRegSet: |
399 | for (i = 0; i < count; ++i) |
400 | (&gpr.eax)[i] = data.GetU32(offset_ptr: &offset); |
401 | SetError(flavor: GPRRegSet, err_idx: Read, err: 0); |
402 | done = true; |
403 | |
404 | break; |
405 | case FPURegSet: |
406 | // TODO: fill in FPU regs.... |
407 | // SetError (FPURegSet, Read, -1); |
408 | done = true; |
409 | |
410 | break; |
411 | case EXCRegSet: |
412 | exc.trapno = data.GetU32(offset_ptr: &offset); |
413 | exc.err = data.GetU32(offset_ptr: &offset); |
414 | exc.faultvaddr = data.GetU32(offset_ptr: &offset); |
415 | SetError(flavor: EXCRegSet, err_idx: Read, err: 0); |
416 | done = true; |
417 | break; |
418 | case 7: |
419 | case 8: |
420 | case 9: |
421 | // fancy flavors that encapsulate of the above flavors... |
422 | break; |
423 | |
424 | default: |
425 | done = true; |
426 | break; |
427 | } |
428 | } |
429 | } |
430 | } |
431 | |
432 | static bool Create_LC_THREAD(Thread *thread, Stream &data) { |
433 | RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); |
434 | if (reg_ctx_sp) { |
435 | RegisterContext *reg_ctx = reg_ctx_sp.get(); |
436 | |
437 | data.PutHex32(uvalue: GPRRegSet); // Flavor |
438 | data.PutHex32(uvalue: GPRWordCount); |
439 | PrintRegisterValue(reg_ctx, name: "eax" , alt_name: nullptr, reg_byte_size: 4, data); |
440 | PrintRegisterValue(reg_ctx, name: "ebx" , alt_name: nullptr, reg_byte_size: 4, data); |
441 | PrintRegisterValue(reg_ctx, name: "ecx" , alt_name: nullptr, reg_byte_size: 4, data); |
442 | PrintRegisterValue(reg_ctx, name: "edx" , alt_name: nullptr, reg_byte_size: 4, data); |
443 | PrintRegisterValue(reg_ctx, name: "edi" , alt_name: nullptr, reg_byte_size: 4, data); |
444 | PrintRegisterValue(reg_ctx, name: "esi" , alt_name: nullptr, reg_byte_size: 4, data); |
445 | PrintRegisterValue(reg_ctx, name: "ebp" , alt_name: nullptr, reg_byte_size: 4, data); |
446 | PrintRegisterValue(reg_ctx, name: "esp" , alt_name: nullptr, reg_byte_size: 4, data); |
447 | PrintRegisterValue(reg_ctx, name: "ss" , alt_name: nullptr, reg_byte_size: 4, data); |
448 | PrintRegisterValue(reg_ctx, name: "eflags" , alt_name: nullptr, reg_byte_size: 4, data); |
449 | PrintRegisterValue(reg_ctx, name: "eip" , alt_name: nullptr, reg_byte_size: 4, data); |
450 | PrintRegisterValue(reg_ctx, name: "cs" , alt_name: nullptr, reg_byte_size: 4, data); |
451 | PrintRegisterValue(reg_ctx, name: "ds" , alt_name: nullptr, reg_byte_size: 4, data); |
452 | PrintRegisterValue(reg_ctx, name: "es" , alt_name: nullptr, reg_byte_size: 4, data); |
453 | PrintRegisterValue(reg_ctx, name: "fs" , alt_name: nullptr, reg_byte_size: 4, data); |
454 | PrintRegisterValue(reg_ctx, name: "gs" , alt_name: nullptr, reg_byte_size: 4, data); |
455 | |
456 | // Write out the EXC registers |
457 | data.PutHex32(uvalue: EXCRegSet); |
458 | data.PutHex32(uvalue: EXCWordCount); |
459 | PrintRegisterValue(reg_ctx, name: "trapno" , alt_name: nullptr, reg_byte_size: 4, data); |
460 | PrintRegisterValue(reg_ctx, name: "err" , alt_name: nullptr, reg_byte_size: 4, data); |
461 | PrintRegisterValue(reg_ctx, name: "faultvaddr" , alt_name: nullptr, reg_byte_size: 4, data); |
462 | return true; |
463 | } |
464 | return false; |
465 | } |
466 | |
467 | protected: |
468 | int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } |
469 | |
470 | int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } |
471 | |
472 | int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } |
473 | |
474 | int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { |
475 | return 0; |
476 | } |
477 | |
478 | int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { |
479 | return 0; |
480 | } |
481 | |
482 | int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { |
483 | return 0; |
484 | } |
485 | }; |
486 | |
487 | class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { |
488 | public: |
489 | (lldb_private::Thread &thread, |
490 | const DataExtractor &data) |
491 | : RegisterContextDarwin_arm(thread, 0) { |
492 | SetRegisterDataFrom_LC_THREAD(data); |
493 | } |
494 | |
495 | void InvalidateAllRegisters() override { |
496 | // Do nothing... registers are always valid... |
497 | } |
498 | |
499 | void (const DataExtractor &data) { |
500 | lldb::offset_t offset = 0; |
501 | SetError(flavor: GPRRegSet, err_idx: Read, err: -1); |
502 | SetError(flavor: FPURegSet, err_idx: Read, err: -1); |
503 | SetError(flavor: EXCRegSet, err_idx: Read, err: -1); |
504 | bool done = false; |
505 | |
506 | while (!done) { |
507 | int flavor = data.GetU32(offset_ptr: &offset); |
508 | uint32_t count = data.GetU32(offset_ptr: &offset); |
509 | lldb::offset_t next_thread_state = offset + (count * 4); |
510 | switch (flavor) { |
511 | case GPRAltRegSet: |
512 | case GPRRegSet: { |
513 | // r0-r15, plus CPSR |
514 | uint32_t gpr_buf_count = (sizeof(gpr.r) / sizeof(gpr.r[0])) + 1; |
515 | if (count == gpr_buf_count) { |
516 | for (uint32_t i = 0; i < (count - 1); ++i) { |
517 | gpr.r[i] = data.GetU32(offset_ptr: &offset); |
518 | } |
519 | gpr.cpsr = data.GetU32(offset_ptr: &offset); |
520 | |
521 | SetError(flavor: GPRRegSet, err_idx: Read, err: 0); |
522 | } |
523 | } |
524 | offset = next_thread_state; |
525 | break; |
526 | |
527 | case FPURegSet: { |
528 | uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats; |
529 | const int fpu_reg_buf_size = sizeof(fpu.floats); |
530 | if (data.ExtractBytes(offset, length: fpu_reg_buf_size, dst_byte_order: eByteOrderLittle, |
531 | dst: fpu_reg_buf) == fpu_reg_buf_size) { |
532 | offset += fpu_reg_buf_size; |
533 | fpu.fpscr = data.GetU32(offset_ptr: &offset); |
534 | SetError(flavor: FPURegSet, err_idx: Read, err: 0); |
535 | } else { |
536 | done = true; |
537 | } |
538 | } |
539 | offset = next_thread_state; |
540 | break; |
541 | |
542 | case EXCRegSet: |
543 | if (count == 3) { |
544 | exc.exception = data.GetU32(offset_ptr: &offset); |
545 | exc.fsr = data.GetU32(offset_ptr: &offset); |
546 | exc.far = data.GetU32(offset_ptr: &offset); |
547 | SetError(flavor: EXCRegSet, err_idx: Read, err: 0); |
548 | } |
549 | done = true; |
550 | offset = next_thread_state; |
551 | break; |
552 | |
553 | // Unknown register set flavor, stop trying to parse. |
554 | default: |
555 | done = true; |
556 | } |
557 | } |
558 | } |
559 | |
560 | static bool Create_LC_THREAD(Thread *thread, Stream &data) { |
561 | RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); |
562 | if (reg_ctx_sp) { |
563 | RegisterContext *reg_ctx = reg_ctx_sp.get(); |
564 | |
565 | data.PutHex32(uvalue: GPRRegSet); // Flavor |
566 | data.PutHex32(uvalue: GPRWordCount); |
567 | PrintRegisterValue(reg_ctx, name: "r0" , alt_name: nullptr, reg_byte_size: 4, data); |
568 | PrintRegisterValue(reg_ctx, name: "r1" , alt_name: nullptr, reg_byte_size: 4, data); |
569 | PrintRegisterValue(reg_ctx, name: "r2" , alt_name: nullptr, reg_byte_size: 4, data); |
570 | PrintRegisterValue(reg_ctx, name: "r3" , alt_name: nullptr, reg_byte_size: 4, data); |
571 | PrintRegisterValue(reg_ctx, name: "r4" , alt_name: nullptr, reg_byte_size: 4, data); |
572 | PrintRegisterValue(reg_ctx, name: "r5" , alt_name: nullptr, reg_byte_size: 4, data); |
573 | PrintRegisterValue(reg_ctx, name: "r6" , alt_name: nullptr, reg_byte_size: 4, data); |
574 | PrintRegisterValue(reg_ctx, name: "r7" , alt_name: nullptr, reg_byte_size: 4, data); |
575 | PrintRegisterValue(reg_ctx, name: "r8" , alt_name: nullptr, reg_byte_size: 4, data); |
576 | PrintRegisterValue(reg_ctx, name: "r9" , alt_name: nullptr, reg_byte_size: 4, data); |
577 | PrintRegisterValue(reg_ctx, name: "r10" , alt_name: nullptr, reg_byte_size: 4, data); |
578 | PrintRegisterValue(reg_ctx, name: "r11" , alt_name: nullptr, reg_byte_size: 4, data); |
579 | PrintRegisterValue(reg_ctx, name: "r12" , alt_name: nullptr, reg_byte_size: 4, data); |
580 | PrintRegisterValue(reg_ctx, name: "sp" , alt_name: nullptr, reg_byte_size: 4, data); |
581 | PrintRegisterValue(reg_ctx, name: "lr" , alt_name: nullptr, reg_byte_size: 4, data); |
582 | PrintRegisterValue(reg_ctx, name: "pc" , alt_name: nullptr, reg_byte_size: 4, data); |
583 | PrintRegisterValue(reg_ctx, name: "cpsr" , alt_name: nullptr, reg_byte_size: 4, data); |
584 | |
585 | // Write out the EXC registers |
586 | // data.PutHex32 (EXCRegSet); |
587 | // data.PutHex32 (EXCWordCount); |
588 | // WriteRegister (reg_ctx, "exception", NULL, 4, data); |
589 | // WriteRegister (reg_ctx, "fsr", NULL, 4, data); |
590 | // WriteRegister (reg_ctx, "far", NULL, 4, data); |
591 | return true; |
592 | } |
593 | return false; |
594 | } |
595 | |
596 | protected: |
597 | int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } |
598 | |
599 | int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } |
600 | |
601 | int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } |
602 | |
603 | int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } |
604 | |
605 | int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { |
606 | return 0; |
607 | } |
608 | |
609 | int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { |
610 | return 0; |
611 | } |
612 | |
613 | int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { |
614 | return 0; |
615 | } |
616 | |
617 | int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { |
618 | return -1; |
619 | } |
620 | }; |
621 | |
622 | class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { |
623 | public: |
624 | (lldb_private::Thread &thread, |
625 | const DataExtractor &data) |
626 | : RegisterContextDarwin_arm64(thread, 0) { |
627 | SetRegisterDataFrom_LC_THREAD(data); |
628 | } |
629 | |
630 | void InvalidateAllRegisters() override { |
631 | // Do nothing... registers are always valid... |
632 | } |
633 | |
634 | void (const DataExtractor &data) { |
635 | lldb::offset_t offset = 0; |
636 | SetError(flavor: GPRRegSet, err_idx: Read, err: -1); |
637 | SetError(flavor: FPURegSet, err_idx: Read, err: -1); |
638 | SetError(flavor: EXCRegSet, err_idx: Read, err: -1); |
639 | bool done = false; |
640 | while (!done) { |
641 | int flavor = data.GetU32(offset_ptr: &offset); |
642 | uint32_t count = data.GetU32(offset_ptr: &offset); |
643 | lldb::offset_t next_thread_state = offset + (count * 4); |
644 | switch (flavor) { |
645 | case GPRRegSet: |
646 | // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 |
647 | // 32-bit register) |
648 | if (count >= (33 * 2) + 1) { |
649 | for (uint32_t i = 0; i < 29; ++i) |
650 | gpr.x[i] = data.GetU64(offset_ptr: &offset); |
651 | gpr.fp = data.GetU64(offset_ptr: &offset); |
652 | gpr.lr = data.GetU64(offset_ptr: &offset); |
653 | gpr.sp = data.GetU64(offset_ptr: &offset); |
654 | gpr.pc = data.GetU64(offset_ptr: &offset); |
655 | gpr.cpsr = data.GetU32(offset_ptr: &offset); |
656 | SetError(flavor: GPRRegSet, err_idx: Read, err: 0); |
657 | } |
658 | offset = next_thread_state; |
659 | break; |
660 | case FPURegSet: { |
661 | uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; |
662 | const int fpu_reg_buf_size = sizeof(fpu); |
663 | if (fpu_reg_buf_size == count * sizeof(uint32_t) && |
664 | data.ExtractBytes(offset, length: fpu_reg_buf_size, dst_byte_order: eByteOrderLittle, |
665 | dst: fpu_reg_buf) == fpu_reg_buf_size) { |
666 | SetError(flavor: FPURegSet, err_idx: Read, err: 0); |
667 | } else { |
668 | done = true; |
669 | } |
670 | } |
671 | offset = next_thread_state; |
672 | break; |
673 | case EXCRegSet: |
674 | if (count == 4) { |
675 | exc.far = data.GetU64(offset_ptr: &offset); |
676 | exc.esr = data.GetU32(offset_ptr: &offset); |
677 | exc.exception = data.GetU32(offset_ptr: &offset); |
678 | SetError(flavor: EXCRegSet, err_idx: Read, err: 0); |
679 | } |
680 | offset = next_thread_state; |
681 | break; |
682 | default: |
683 | done = true; |
684 | break; |
685 | } |
686 | } |
687 | } |
688 | |
689 | static bool Create_LC_THREAD(Thread *thread, Stream &data) { |
690 | RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); |
691 | if (reg_ctx_sp) { |
692 | RegisterContext *reg_ctx = reg_ctx_sp.get(); |
693 | |
694 | data.PutHex32(uvalue: GPRRegSet); // Flavor |
695 | data.PutHex32(uvalue: GPRWordCount); |
696 | PrintRegisterValue(reg_ctx, name: "x0" , alt_name: nullptr, reg_byte_size: 8, data); |
697 | PrintRegisterValue(reg_ctx, name: "x1" , alt_name: nullptr, reg_byte_size: 8, data); |
698 | PrintRegisterValue(reg_ctx, name: "x2" , alt_name: nullptr, reg_byte_size: 8, data); |
699 | PrintRegisterValue(reg_ctx, name: "x3" , alt_name: nullptr, reg_byte_size: 8, data); |
700 | PrintRegisterValue(reg_ctx, name: "x4" , alt_name: nullptr, reg_byte_size: 8, data); |
701 | PrintRegisterValue(reg_ctx, name: "x5" , alt_name: nullptr, reg_byte_size: 8, data); |
702 | PrintRegisterValue(reg_ctx, name: "x6" , alt_name: nullptr, reg_byte_size: 8, data); |
703 | PrintRegisterValue(reg_ctx, name: "x7" , alt_name: nullptr, reg_byte_size: 8, data); |
704 | PrintRegisterValue(reg_ctx, name: "x8" , alt_name: nullptr, reg_byte_size: 8, data); |
705 | PrintRegisterValue(reg_ctx, name: "x9" , alt_name: nullptr, reg_byte_size: 8, data); |
706 | PrintRegisterValue(reg_ctx, name: "x10" , alt_name: nullptr, reg_byte_size: 8, data); |
707 | PrintRegisterValue(reg_ctx, name: "x11" , alt_name: nullptr, reg_byte_size: 8, data); |
708 | PrintRegisterValue(reg_ctx, name: "x12" , alt_name: nullptr, reg_byte_size: 8, data); |
709 | PrintRegisterValue(reg_ctx, name: "x13" , alt_name: nullptr, reg_byte_size: 8, data); |
710 | PrintRegisterValue(reg_ctx, name: "x14" , alt_name: nullptr, reg_byte_size: 8, data); |
711 | PrintRegisterValue(reg_ctx, name: "x15" , alt_name: nullptr, reg_byte_size: 8, data); |
712 | PrintRegisterValue(reg_ctx, name: "x16" , alt_name: nullptr, reg_byte_size: 8, data); |
713 | PrintRegisterValue(reg_ctx, name: "x17" , alt_name: nullptr, reg_byte_size: 8, data); |
714 | PrintRegisterValue(reg_ctx, name: "x18" , alt_name: nullptr, reg_byte_size: 8, data); |
715 | PrintRegisterValue(reg_ctx, name: "x19" , alt_name: nullptr, reg_byte_size: 8, data); |
716 | PrintRegisterValue(reg_ctx, name: "x20" , alt_name: nullptr, reg_byte_size: 8, data); |
717 | PrintRegisterValue(reg_ctx, name: "x21" , alt_name: nullptr, reg_byte_size: 8, data); |
718 | PrintRegisterValue(reg_ctx, name: "x22" , alt_name: nullptr, reg_byte_size: 8, data); |
719 | PrintRegisterValue(reg_ctx, name: "x23" , alt_name: nullptr, reg_byte_size: 8, data); |
720 | PrintRegisterValue(reg_ctx, name: "x24" , alt_name: nullptr, reg_byte_size: 8, data); |
721 | PrintRegisterValue(reg_ctx, name: "x25" , alt_name: nullptr, reg_byte_size: 8, data); |
722 | PrintRegisterValue(reg_ctx, name: "x26" , alt_name: nullptr, reg_byte_size: 8, data); |
723 | PrintRegisterValue(reg_ctx, name: "x27" , alt_name: nullptr, reg_byte_size: 8, data); |
724 | PrintRegisterValue(reg_ctx, name: "x28" , alt_name: nullptr, reg_byte_size: 8, data); |
725 | PrintRegisterValue(reg_ctx, name: "fp" , alt_name: nullptr, reg_byte_size: 8, data); |
726 | PrintRegisterValue(reg_ctx, name: "lr" , alt_name: nullptr, reg_byte_size: 8, data); |
727 | PrintRegisterValue(reg_ctx, name: "sp" , alt_name: nullptr, reg_byte_size: 8, data); |
728 | PrintRegisterValue(reg_ctx, name: "pc" , alt_name: nullptr, reg_byte_size: 8, data); |
729 | PrintRegisterValue(reg_ctx, name: "cpsr" , alt_name: nullptr, reg_byte_size: 4, data); |
730 | data.PutHex32(uvalue: 0); // uint32_t pad at the end |
731 | |
732 | // Write out the EXC registers |
733 | data.PutHex32(uvalue: EXCRegSet); |
734 | data.PutHex32(uvalue: EXCWordCount); |
735 | PrintRegisterValue(reg_ctx, name: "far" , alt_name: nullptr, reg_byte_size: 8, data); |
736 | PrintRegisterValue(reg_ctx, name: "esr" , alt_name: nullptr, reg_byte_size: 4, data); |
737 | PrintRegisterValue(reg_ctx, name: "exception" , alt_name: nullptr, reg_byte_size: 4, data); |
738 | return true; |
739 | } |
740 | return false; |
741 | } |
742 | |
743 | protected: |
744 | int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } |
745 | |
746 | int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } |
747 | |
748 | int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } |
749 | |
750 | int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } |
751 | |
752 | int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { |
753 | return 0; |
754 | } |
755 | |
756 | int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { |
757 | return 0; |
758 | } |
759 | |
760 | int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { |
761 | return 0; |
762 | } |
763 | |
764 | int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { |
765 | return -1; |
766 | } |
767 | }; |
768 | |
769 | static uint32_t (uint32_t magic) { |
770 | switch (magic) { |
771 | case MH_MAGIC: |
772 | case MH_CIGAM: |
773 | return sizeof(struct llvm::MachO::mach_header); |
774 | |
775 | case MH_MAGIC_64: |
776 | case MH_CIGAM_64: |
777 | return sizeof(struct llvm::MachO::mach_header_64); |
778 | break; |
779 | |
780 | default: |
781 | break; |
782 | } |
783 | return 0; |
784 | } |
785 | |
786 | #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 |
787 | |
788 | char ObjectFileMachO::ID; |
789 | |
790 | void ObjectFileMachO::Initialize() { |
791 | PluginManager::RegisterPlugin( |
792 | name: GetPluginNameStatic(), description: GetPluginDescriptionStatic(), create_callback: CreateInstance, |
793 | create_memory_callback: CreateMemoryInstance, get_module_specifications: GetModuleSpecifications, save_core: SaveCore); |
794 | } |
795 | |
796 | void ObjectFileMachO::Terminate() { |
797 | PluginManager::UnregisterPlugin(create_callback: CreateInstance); |
798 | } |
799 | |
800 | ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, |
801 | DataBufferSP data_sp, |
802 | lldb::offset_t data_offset, |
803 | const FileSpec *file, |
804 | lldb::offset_t file_offset, |
805 | lldb::offset_t length) { |
806 | if (!data_sp) { |
807 | data_sp = MapFileData(file: *file, Size: length, Offset: file_offset); |
808 | if (!data_sp) |
809 | return nullptr; |
810 | data_offset = 0; |
811 | } |
812 | |
813 | if (!ObjectFileMachO::MagicBytesMatch(data_sp, offset: data_offset, length)) |
814 | return nullptr; |
815 | |
816 | // Update the data to contain the entire file if it doesn't already |
817 | if (data_sp->GetByteSize() < length) { |
818 | data_sp = MapFileData(file: *file, Size: length, Offset: file_offset); |
819 | if (!data_sp) |
820 | return nullptr; |
821 | data_offset = 0; |
822 | } |
823 | auto objfile_up = std::make_unique<ObjectFileMachO>( |
824 | args: module_sp, args&: data_sp, args&: data_offset, args&: file, args&: file_offset, args&: length); |
825 | if (!objfile_up || !objfile_up->ParseHeader()) |
826 | return nullptr; |
827 | |
828 | return objfile_up.release(); |
829 | } |
830 | |
831 | ObjectFile *ObjectFileMachO::CreateMemoryInstance( |
832 | const lldb::ModuleSP &module_sp, WritableDataBufferSP data_sp, |
833 | const ProcessSP &process_sp, lldb::addr_t ) { |
834 | if (ObjectFileMachO::MagicBytesMatch(data_sp, offset: 0, length: data_sp->GetByteSize())) { |
835 | std::unique_ptr<ObjectFile> objfile_up( |
836 | new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); |
837 | if (objfile_up.get() && objfile_up->ParseHeader()) |
838 | return objfile_up.release(); |
839 | } |
840 | return nullptr; |
841 | } |
842 | |
843 | size_t ObjectFileMachO::GetModuleSpecifications( |
844 | const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, |
845 | lldb::offset_t data_offset, lldb::offset_t file_offset, |
846 | lldb::offset_t length, lldb_private::ModuleSpecList &specs) { |
847 | const size_t initial_count = specs.GetSize(); |
848 | |
849 | if (ObjectFileMachO::MagicBytesMatch(data_sp, offset: 0, length: data_sp->GetByteSize())) { |
850 | DataExtractor data; |
851 | data.SetData(data_sp); |
852 | llvm::MachO::mach_header ; |
853 | if (ParseHeader(data, data_offset_ptr: &data_offset, header)) { |
854 | size_t header_and_load_cmds = |
855 | header.sizeofcmds + MachHeaderSizeFromMagic(magic: header.magic); |
856 | if (header_and_load_cmds >= data_sp->GetByteSize()) { |
857 | data_sp = MapFileData(file, Size: header_and_load_cmds, Offset: file_offset); |
858 | data.SetData(data_sp); |
859 | data_offset = MachHeaderSizeFromMagic(magic: header.magic); |
860 | } |
861 | if (data_sp) { |
862 | ModuleSpec base_spec; |
863 | base_spec.GetFileSpec() = file; |
864 | base_spec.SetObjectOffset(file_offset); |
865 | base_spec.SetObjectSize(length); |
866 | GetAllArchSpecs(header, data, lc_offset: data_offset, base_spec, all_specs&: specs); |
867 | } |
868 | } |
869 | } |
870 | return specs.GetSize() - initial_count; |
871 | } |
872 | |
873 | ConstString ObjectFileMachO::GetSegmentNameTEXT() { |
874 | static ConstString g_segment_name_TEXT("__TEXT" ); |
875 | return g_segment_name_TEXT; |
876 | } |
877 | |
878 | ConstString ObjectFileMachO::GetSegmentNameDATA() { |
879 | static ConstString g_segment_name_DATA("__DATA" ); |
880 | return g_segment_name_DATA; |
881 | } |
882 | |
883 | ConstString ObjectFileMachO::GetSegmentNameDATA_DIRTY() { |
884 | static ConstString g_segment_name("__DATA_DIRTY" ); |
885 | return g_segment_name; |
886 | } |
887 | |
888 | ConstString ObjectFileMachO::GetSegmentNameDATA_CONST() { |
889 | static ConstString g_segment_name("__DATA_CONST" ); |
890 | return g_segment_name; |
891 | } |
892 | |
893 | ConstString ObjectFileMachO::GetSegmentNameOBJC() { |
894 | static ConstString g_segment_name_OBJC("__OBJC" ); |
895 | return g_segment_name_OBJC; |
896 | } |
897 | |
898 | ConstString ObjectFileMachO::GetSegmentNameLINKEDIT() { |
899 | static ConstString g_section_name_LINKEDIT("__LINKEDIT" ); |
900 | return g_section_name_LINKEDIT; |
901 | } |
902 | |
903 | ConstString ObjectFileMachO::GetSegmentNameDWARF() { |
904 | static ConstString g_section_name("__DWARF" ); |
905 | return g_section_name; |
906 | } |
907 | |
908 | ConstString ObjectFileMachO::GetSegmentNameLLVM_COV() { |
909 | static ConstString g_section_name("__LLVM_COV" ); |
910 | return g_section_name; |
911 | } |
912 | |
913 | ConstString ObjectFileMachO::GetSectionNameEHFrame() { |
914 | static ConstString g_section_name_eh_frame("__eh_frame" ); |
915 | return g_section_name_eh_frame; |
916 | } |
917 | |
918 | bool ObjectFileMachO::MagicBytesMatch(DataBufferSP data_sp, |
919 | lldb::addr_t data_offset, |
920 | lldb::addr_t data_length) { |
921 | DataExtractor data; |
922 | data.SetData(data_sp, offset: data_offset, length: data_length); |
923 | lldb::offset_t offset = 0; |
924 | uint32_t magic = data.GetU32(offset_ptr: &offset); |
925 | |
926 | offset += 4; // cputype |
927 | offset += 4; // cpusubtype |
928 | uint32_t filetype = data.GetU32(offset_ptr: &offset); |
929 | |
930 | // A fileset has a Mach-O header but is not an |
931 | // individual file and must be handled via an |
932 | // ObjectContainer plugin. |
933 | if (filetype == llvm::MachO::MH_FILESET) |
934 | return false; |
935 | |
936 | return MachHeaderSizeFromMagic(magic) != 0; |
937 | } |
938 | |
939 | ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, |
940 | DataBufferSP data_sp, |
941 | lldb::offset_t data_offset, |
942 | const FileSpec *file, |
943 | lldb::offset_t file_offset, |
944 | lldb::offset_t length) |
945 | : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), |
946 | m_mach_sections(), m_entry_point_address(), m_thread_context_offsets(), |
947 | m_thread_context_offsets_valid(false), m_reexported_dylibs(), |
948 | m_allow_assembly_emulation_unwind_plans(true) { |
949 | ::memset(s: &m_header, c: 0, n: sizeof(m_header)); |
950 | ::memset(s: &m_dysymtab, c: 0, n: sizeof(m_dysymtab)); |
951 | } |
952 | |
953 | ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, |
954 | lldb::WritableDataBufferSP , |
955 | const lldb::ProcessSP &process_sp, |
956 | lldb::addr_t ) |
957 | : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), |
958 | m_mach_sections(), m_entry_point_address(), m_thread_context_offsets(), |
959 | m_thread_context_offsets_valid(false), m_reexported_dylibs(), |
960 | m_allow_assembly_emulation_unwind_plans(true) { |
961 | ::memset(s: &m_header, c: 0, n: sizeof(m_header)); |
962 | ::memset(s: &m_dysymtab, c: 0, n: sizeof(m_dysymtab)); |
963 | } |
964 | |
965 | bool ObjectFileMachO::(DataExtractor &data, |
966 | lldb::offset_t *data_offset_ptr, |
967 | llvm::MachO::mach_header &) { |
968 | data.SetByteOrder(endian::InlHostByteOrder()); |
969 | // Leave magic in the original byte order |
970 | header.magic = data.GetU32(offset_ptr: data_offset_ptr); |
971 | bool can_parse = false; |
972 | bool is_64_bit = false; |
973 | switch (header.magic) { |
974 | case MH_MAGIC: |
975 | data.SetByteOrder(endian::InlHostByteOrder()); |
976 | data.SetAddressByteSize(4); |
977 | can_parse = true; |
978 | break; |
979 | |
980 | case MH_MAGIC_64: |
981 | data.SetByteOrder(endian::InlHostByteOrder()); |
982 | data.SetAddressByteSize(8); |
983 | can_parse = true; |
984 | is_64_bit = true; |
985 | break; |
986 | |
987 | case MH_CIGAM: |
988 | data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig |
989 | ? eByteOrderLittle |
990 | : eByteOrderBig); |
991 | data.SetAddressByteSize(4); |
992 | can_parse = true; |
993 | break; |
994 | |
995 | case MH_CIGAM_64: |
996 | data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig |
997 | ? eByteOrderLittle |
998 | : eByteOrderBig); |
999 | data.SetAddressByteSize(8); |
1000 | is_64_bit = true; |
1001 | can_parse = true; |
1002 | break; |
1003 | |
1004 | default: |
1005 | break; |
1006 | } |
1007 | |
1008 | if (can_parse) { |
1009 | data.GetU32(offset_ptr: data_offset_ptr, dst: &header.cputype, count: 6); |
1010 | if (is_64_bit) |
1011 | *data_offset_ptr += 4; |
1012 | return true; |
1013 | } else { |
1014 | memset(s: &header, c: 0, n: sizeof(header)); |
1015 | } |
1016 | return false; |
1017 | } |
1018 | |
1019 | bool ObjectFileMachO::() { |
1020 | ModuleSP module_sp(GetModule()); |
1021 | if (!module_sp) |
1022 | return false; |
1023 | |
1024 | std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); |
1025 | bool can_parse = false; |
1026 | lldb::offset_t offset = 0; |
1027 | m_data.SetByteOrder(endian::InlHostByteOrder()); |
1028 | // Leave magic in the original byte order |
1029 | m_header.magic = m_data.GetU32(offset_ptr: &offset); |
1030 | switch (m_header.magic) { |
1031 | case MH_MAGIC: |
1032 | m_data.SetByteOrder(endian::InlHostByteOrder()); |
1033 | m_data.SetAddressByteSize(4); |
1034 | can_parse = true; |
1035 | break; |
1036 | |
1037 | case MH_MAGIC_64: |
1038 | m_data.SetByteOrder(endian::InlHostByteOrder()); |
1039 | m_data.SetAddressByteSize(8); |
1040 | can_parse = true; |
1041 | break; |
1042 | |
1043 | case MH_CIGAM: |
1044 | m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig |
1045 | ? eByteOrderLittle |
1046 | : eByteOrderBig); |
1047 | m_data.SetAddressByteSize(4); |
1048 | can_parse = true; |
1049 | break; |
1050 | |
1051 | case MH_CIGAM_64: |
1052 | m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig |
1053 | ? eByteOrderLittle |
1054 | : eByteOrderBig); |
1055 | m_data.SetAddressByteSize(8); |
1056 | can_parse = true; |
1057 | break; |
1058 | |
1059 | default: |
1060 | break; |
1061 | } |
1062 | |
1063 | if (can_parse) { |
1064 | m_data.GetU32(offset_ptr: &offset, dst: &m_header.cputype, count: 6); |
1065 | |
1066 | ModuleSpecList all_specs; |
1067 | ModuleSpec base_spec; |
1068 | GetAllArchSpecs(header: m_header, data: m_data, lc_offset: MachHeaderSizeFromMagic(magic: m_header.magic), |
1069 | base_spec, all_specs); |
1070 | |
1071 | for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { |
1072 | ArchSpec mach_arch = |
1073 | all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); |
1074 | |
1075 | // Check if the module has a required architecture |
1076 | const ArchSpec &module_arch = module_sp->GetArchitecture(); |
1077 | if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(rhs: mach_arch)) |
1078 | continue; |
1079 | |
1080 | if (SetModulesArchitecture(mach_arch)) { |
1081 | const size_t header_and_lc_size = |
1082 | m_header.sizeofcmds + MachHeaderSizeFromMagic(magic: m_header.magic); |
1083 | if (m_data.GetByteSize() < header_and_lc_size) { |
1084 | DataBufferSP data_sp; |
1085 | ProcessSP process_sp(m_process_wp.lock()); |
1086 | if (process_sp) { |
1087 | data_sp = ReadMemory(process_sp, addr: m_memory_addr, byte_size: header_and_lc_size); |
1088 | } else { |
1089 | // Read in all only the load command data from the file on disk |
1090 | data_sp = MapFileData(file: m_file, Size: header_and_lc_size, Offset: m_file_offset); |
1091 | if (data_sp->GetByteSize() != header_and_lc_size) |
1092 | continue; |
1093 | } |
1094 | if (data_sp) |
1095 | m_data.SetData(data_sp); |
1096 | } |
1097 | } |
1098 | return true; |
1099 | } |
1100 | // None found. |
1101 | return false; |
1102 | } else { |
1103 | memset(s: &m_header, c: 0, n: sizeof(struct llvm::MachO::mach_header)); |
1104 | } |
1105 | return false; |
1106 | } |
1107 | |
1108 | ByteOrder ObjectFileMachO::GetByteOrder() const { |
1109 | return m_data.GetByteOrder(); |
1110 | } |
1111 | |
1112 | bool ObjectFileMachO::IsExecutable() const { |
1113 | return m_header.filetype == MH_EXECUTE; |
1114 | } |
1115 | |
1116 | bool ObjectFileMachO::IsDynamicLoader() const { |
1117 | return m_header.filetype == MH_DYLINKER; |
1118 | } |
1119 | |
1120 | bool ObjectFileMachO::IsSharedCacheBinary() const { |
1121 | return m_header.flags & MH_DYLIB_IN_CACHE; |
1122 | } |
1123 | |
1124 | bool ObjectFileMachO::IsKext() const { |
1125 | return m_header.filetype == MH_KEXT_BUNDLE; |
1126 | } |
1127 | |
1128 | uint32_t ObjectFileMachO::GetAddressByteSize() const { |
1129 | return m_data.GetAddressByteSize(); |
1130 | } |
1131 | |
1132 | AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { |
1133 | Symtab *symtab = GetSymtab(); |
1134 | if (!symtab) |
1135 | return AddressClass::eUnknown; |
1136 | |
1137 | Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); |
1138 | if (symbol) { |
1139 | if (symbol->ValueIsAddress()) { |
1140 | SectionSP section_sp(symbol->GetAddressRef().GetSection()); |
1141 | if (section_sp) { |
1142 | const lldb::SectionType section_type = section_sp->GetType(); |
1143 | switch (section_type) { |
1144 | case eSectionTypeInvalid: |
1145 | return AddressClass::eUnknown; |
1146 | |
1147 | case eSectionTypeCode: |
1148 | if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { |
1149 | // For ARM we have a bit in the n_desc field of the symbol that |
1150 | // tells us ARM/Thumb which is bit 0x0008. |
1151 | if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) |
1152 | return AddressClass::eCodeAlternateISA; |
1153 | } |
1154 | return AddressClass::eCode; |
1155 | |
1156 | case eSectionTypeContainer: |
1157 | return AddressClass::eUnknown; |
1158 | |
1159 | case eSectionTypeData: |
1160 | case eSectionTypeDataCString: |
1161 | case eSectionTypeDataCStringPointers: |
1162 | case eSectionTypeDataSymbolAddress: |
1163 | case eSectionTypeData4: |
1164 | case eSectionTypeData8: |
1165 | case eSectionTypeData16: |
1166 | case eSectionTypeDataPointers: |
1167 | case eSectionTypeZeroFill: |
1168 | case eSectionTypeDataObjCMessageRefs: |
1169 | case eSectionTypeDataObjCCFStrings: |
1170 | case eSectionTypeGoSymtab: |
1171 | return AddressClass::eData; |
1172 | |
1173 | case eSectionTypeDebug: |
1174 | case eSectionTypeDWARFDebugAbbrev: |
1175 | case eSectionTypeDWARFDebugAbbrevDwo: |
1176 | case eSectionTypeDWARFDebugAddr: |
1177 | case eSectionTypeDWARFDebugAranges: |
1178 | case eSectionTypeDWARFDebugCuIndex: |
1179 | case eSectionTypeDWARFDebugFrame: |
1180 | case eSectionTypeDWARFDebugInfo: |
1181 | case eSectionTypeDWARFDebugInfoDwo: |
1182 | case eSectionTypeDWARFDebugLine: |
1183 | case eSectionTypeDWARFDebugLineStr: |
1184 | case eSectionTypeDWARFDebugLoc: |
1185 | case eSectionTypeDWARFDebugLocDwo: |
1186 | case eSectionTypeDWARFDebugLocLists: |
1187 | case eSectionTypeDWARFDebugLocListsDwo: |
1188 | case eSectionTypeDWARFDebugMacInfo: |
1189 | case eSectionTypeDWARFDebugMacro: |
1190 | case eSectionTypeDWARFDebugNames: |
1191 | case eSectionTypeDWARFDebugPubNames: |
1192 | case eSectionTypeDWARFDebugPubTypes: |
1193 | case eSectionTypeDWARFDebugRanges: |
1194 | case eSectionTypeDWARFDebugRngLists: |
1195 | case eSectionTypeDWARFDebugRngListsDwo: |
1196 | case eSectionTypeDWARFDebugStr: |
1197 | case eSectionTypeDWARFDebugStrDwo: |
1198 | case eSectionTypeDWARFDebugStrOffsets: |
1199 | case eSectionTypeDWARFDebugStrOffsetsDwo: |
1200 | case eSectionTypeDWARFDebugTuIndex: |
1201 | case eSectionTypeDWARFDebugTypes: |
1202 | case eSectionTypeDWARFDebugTypesDwo: |
1203 | case eSectionTypeDWARFAppleNames: |
1204 | case eSectionTypeDWARFAppleTypes: |
1205 | case eSectionTypeDWARFAppleNamespaces: |
1206 | case eSectionTypeDWARFAppleObjC: |
1207 | case eSectionTypeDWARFGNUDebugAltLink: |
1208 | case eSectionTypeCTF: |
1209 | case eSectionTypeSwiftModules: |
1210 | return AddressClass::eDebug; |
1211 | |
1212 | case eSectionTypeEHFrame: |
1213 | case eSectionTypeARMexidx: |
1214 | case eSectionTypeARMextab: |
1215 | case eSectionTypeCompactUnwind: |
1216 | return AddressClass::eRuntime; |
1217 | |
1218 | case eSectionTypeAbsoluteAddress: |
1219 | case eSectionTypeELFSymbolTable: |
1220 | case eSectionTypeELFDynamicSymbols: |
1221 | case eSectionTypeELFRelocationEntries: |
1222 | case eSectionTypeELFDynamicLinkInfo: |
1223 | case eSectionTypeOther: |
1224 | return AddressClass::eUnknown; |
1225 | } |
1226 | } |
1227 | } |
1228 | |
1229 | const SymbolType symbol_type = symbol->GetType(); |
1230 | switch (symbol_type) { |
1231 | case eSymbolTypeAny: |
1232 | return AddressClass::eUnknown; |
1233 | case eSymbolTypeAbsolute: |
1234 | return AddressClass::eUnknown; |
1235 | |
1236 | case eSymbolTypeCode: |
1237 | case eSymbolTypeTrampoline: |
1238 | case eSymbolTypeResolver: |
1239 | if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { |
1240 | // For ARM we have a bit in the n_desc field of the symbol that tells |
1241 | // us ARM/Thumb which is bit 0x0008. |
1242 | if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) |
1243 | return AddressClass::eCodeAlternateISA; |
1244 | } |
1245 | return AddressClass::eCode; |
1246 | |
1247 | case eSymbolTypeData: |
1248 | return AddressClass::eData; |
1249 | case eSymbolTypeRuntime: |
1250 | return AddressClass::eRuntime; |
1251 | case eSymbolTypeException: |
1252 | return AddressClass::eRuntime; |
1253 | case eSymbolTypeSourceFile: |
1254 | return AddressClass::eDebug; |
1255 | case eSymbolTypeHeaderFile: |
1256 | return AddressClass::eDebug; |
1257 | case eSymbolTypeObjectFile: |
1258 | return AddressClass::eDebug; |
1259 | case eSymbolTypeCommonBlock: |
1260 | return AddressClass::eDebug; |
1261 | case eSymbolTypeBlock: |
1262 | return AddressClass::eDebug; |
1263 | case eSymbolTypeLocal: |
1264 | return AddressClass::eData; |
1265 | case eSymbolTypeParam: |
1266 | return AddressClass::eData; |
1267 | case eSymbolTypeVariable: |
1268 | return AddressClass::eData; |
1269 | case eSymbolTypeVariableType: |
1270 | return AddressClass::eDebug; |
1271 | case eSymbolTypeLineEntry: |
1272 | return AddressClass::eDebug; |
1273 | case eSymbolTypeLineHeader: |
1274 | return AddressClass::eDebug; |
1275 | case eSymbolTypeScopeBegin: |
1276 | return AddressClass::eDebug; |
1277 | case eSymbolTypeScopeEnd: |
1278 | return AddressClass::eDebug; |
1279 | case eSymbolTypeAdditional: |
1280 | return AddressClass::eUnknown; |
1281 | case eSymbolTypeCompiler: |
1282 | return AddressClass::eDebug; |
1283 | case eSymbolTypeInstrumentation: |
1284 | return AddressClass::eDebug; |
1285 | case eSymbolTypeUndefined: |
1286 | return AddressClass::eUnknown; |
1287 | case eSymbolTypeObjCClass: |
1288 | return AddressClass::eRuntime; |
1289 | case eSymbolTypeObjCMetaClass: |
1290 | return AddressClass::eRuntime; |
1291 | case eSymbolTypeObjCIVar: |
1292 | return AddressClass::eRuntime; |
1293 | case eSymbolTypeReExported: |
1294 | return AddressClass::eRuntime; |
1295 | } |
1296 | } |
1297 | return AddressClass::eUnknown; |
1298 | } |
1299 | |
1300 | bool ObjectFileMachO::IsStripped() { |
1301 | if (m_dysymtab.cmd == 0) { |
1302 | ModuleSP module_sp(GetModule()); |
1303 | if (module_sp) { |
1304 | lldb::offset_t offset = MachHeaderSizeFromMagic(magic: m_header.magic); |
1305 | for (uint32_t i = 0; i < m_header.ncmds; ++i) { |
1306 | const lldb::offset_t load_cmd_offset = offset; |
1307 | |
1308 | llvm::MachO::load_command lc = {}; |
1309 | if (m_data.GetU32(offset_ptr: &offset, dst: &lc.cmd, count: 2) == nullptr) |
1310 | break; |
1311 | if (lc.cmd == LC_DYSYMTAB) { |
1312 | m_dysymtab.cmd = lc.cmd; |
1313 | m_dysymtab.cmdsize = lc.cmdsize; |
1314 | if (m_data.GetU32(offset_ptr: &offset, dst: &m_dysymtab.ilocalsym, |
1315 | count: (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == |
1316 | nullptr) { |
1317 | // Clear m_dysymtab if we were unable to read all items from the |
1318 | // load command |
1319 | ::memset(s: &m_dysymtab, c: 0, n: sizeof(m_dysymtab)); |
1320 | } |
1321 | } |
1322 | offset = load_cmd_offset + lc.cmdsize; |
1323 | } |
1324 | } |
1325 | } |
1326 | if (m_dysymtab.cmd) |
1327 | return m_dysymtab.nlocalsym <= 1; |
1328 | return false; |
1329 | } |
1330 | |
1331 | ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { |
1332 | EncryptedFileRanges result; |
1333 | lldb::offset_t offset = MachHeaderSizeFromMagic(magic: m_header.magic); |
1334 | |
1335 | llvm::MachO::encryption_info_command encryption_cmd; |
1336 | for (uint32_t i = 0; i < m_header.ncmds; ++i) { |
1337 | const lldb::offset_t load_cmd_offset = offset; |
1338 | if (m_data.GetU32(offset_ptr: &offset, dst: &encryption_cmd, count: 2) == nullptr) |
1339 | break; |
1340 | |
1341 | // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the |
1342 | // 3 fields we care about, so treat them the same. |
1343 | if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || |
1344 | encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { |
1345 | if (m_data.GetU32(offset_ptr: &offset, dst: &encryption_cmd.cryptoff, count: 3)) { |
1346 | if (encryption_cmd.cryptid != 0) { |
1347 | EncryptedFileRanges::Entry entry; |
1348 | entry.SetRangeBase(encryption_cmd.cryptoff); |
1349 | entry.SetByteSize(encryption_cmd.cryptsize); |
1350 | result.Append(entry); |
1351 | } |
1352 | } |
1353 | } |
1354 | offset = load_cmd_offset + encryption_cmd.cmdsize; |
1355 | } |
1356 | |
1357 | return result; |
1358 | } |
1359 | |
1360 | void ObjectFileMachO::SanitizeSegmentCommand( |
1361 | llvm::MachO::segment_command_64 &seg_cmd, uint32_t cmd_idx) { |
1362 | if (m_length == 0 || seg_cmd.filesize == 0) |
1363 | return; |
1364 | |
1365 | if (IsSharedCacheBinary() && !IsInMemory()) { |
1366 | // In shared cache images, the load commands are relative to the |
1367 | // shared cache file, and not the specific image we are |
1368 | // examining. Let's fix this up so that it looks like a normal |
1369 | // image. |
1370 | if (strncmp(s1: seg_cmd.segname, s2: GetSegmentNameTEXT().GetCString(), |
1371 | n: sizeof(seg_cmd.segname)) == 0) |
1372 | m_text_address = seg_cmd.vmaddr; |
1373 | if (strncmp(s1: seg_cmd.segname, s2: GetSegmentNameLINKEDIT().GetCString(), |
1374 | n: sizeof(seg_cmd.segname)) == 0) |
1375 | m_linkedit_original_offset = seg_cmd.fileoff; |
1376 | |
1377 | seg_cmd.fileoff = seg_cmd.vmaddr - m_text_address; |
1378 | } |
1379 | |
1380 | if (seg_cmd.fileoff > m_length) { |
1381 | // We have a load command that says it extends past the end of the file. |
1382 | // This is likely a corrupt file. We don't have any way to return an error |
1383 | // condition here (this method was likely invoked from something like |
1384 | // ObjectFile::GetSectionList()), so we just null out the section contents, |
1385 | // and dump a message to stdout. The most common case here is core file |
1386 | // debugging with a truncated file. |
1387 | const char *lc_segment_name = |
1388 | seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT" ; |
1389 | GetModule()->ReportWarning( |
1390 | format: "load command {0} {1} has a fileoff ({2:x16}) that extends beyond " |
1391 | "the end of the file ({3:x16}), ignoring this section" , |
1392 | args&: cmd_idx, args&: lc_segment_name, args&: seg_cmd.fileoff, args&: m_length); |
1393 | |
1394 | seg_cmd.fileoff = 0; |
1395 | seg_cmd.filesize = 0; |
1396 | } |
1397 | |
1398 | if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { |
1399 | // We have a load command that says it extends past the end of the file. |
1400 | // This is likely a corrupt file. We don't have any way to return an error |
1401 | // condition here (this method was likely invoked from something like |
1402 | // ObjectFile::GetSectionList()), so we just null out the section contents, |
1403 | // and dump a message to stdout. The most common case here is core file |
1404 | // debugging with a truncated file. |
1405 | const char *lc_segment_name = |
1406 | seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT" ; |
1407 | GetModule()->ReportWarning( |
1408 | format: "load command {0} {1} has a fileoff + filesize ({2:x16}) that " |
1409 | "extends beyond the end of the file ({4:x16}), the segment will be " |
1410 | "truncated to match" , |
1411 | args&: cmd_idx, args&: lc_segment_name, args: seg_cmd.fileoff + seg_cmd.filesize, args&: m_length); |
1412 | |
1413 | // Truncate the length |
1414 | seg_cmd.filesize = m_length - seg_cmd.fileoff; |
1415 | } |
1416 | } |
1417 | |
1418 | static uint32_t |
1419 | GetSegmentPermissions(const llvm::MachO::segment_command_64 &seg_cmd) { |
1420 | uint32_t result = 0; |
1421 | if (seg_cmd.initprot & VM_PROT_READ) |
1422 | result |= ePermissionsReadable; |
1423 | if (seg_cmd.initprot & VM_PROT_WRITE) |
1424 | result |= ePermissionsWritable; |
1425 | if (seg_cmd.initprot & VM_PROT_EXECUTE) |
1426 | result |= ePermissionsExecutable; |
1427 | return result; |
1428 | } |
1429 | |
1430 | static lldb::SectionType GetSectionType(uint32_t flags, |
1431 | ConstString section_name) { |
1432 | |
1433 | if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) |
1434 | return eSectionTypeCode; |
1435 | |
1436 | uint32_t mach_sect_type = flags & SECTION_TYPE; |
1437 | static ConstString g_sect_name_objc_data("__objc_data" ); |
1438 | static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs" ); |
1439 | static ConstString g_sect_name_objc_selrefs("__objc_selrefs" ); |
1440 | static ConstString g_sect_name_objc_classrefs("__objc_classrefs" ); |
1441 | static ConstString g_sect_name_objc_superrefs("__objc_superrefs" ); |
1442 | static ConstString g_sect_name_objc_const("__objc_const" ); |
1443 | static ConstString g_sect_name_objc_classlist("__objc_classlist" ); |
1444 | static ConstString g_sect_name_cfstring("__cfstring" ); |
1445 | |
1446 | static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev" ); |
1447 | static ConstString g_sect_name_dwarf_debug_abbrev_dwo("__debug_abbrev.dwo" ); |
1448 | static ConstString g_sect_name_dwarf_debug_addr("__debug_addr" ); |
1449 | static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges" ); |
1450 | static ConstString g_sect_name_dwarf_debug_cu_index("__debug_cu_index" ); |
1451 | static ConstString g_sect_name_dwarf_debug_frame("__debug_frame" ); |
1452 | static ConstString g_sect_name_dwarf_debug_info("__debug_info" ); |
1453 | static ConstString g_sect_name_dwarf_debug_info_dwo("__debug_info.dwo" ); |
1454 | static ConstString g_sect_name_dwarf_debug_line("__debug_line" ); |
1455 | static ConstString g_sect_name_dwarf_debug_line_dwo("__debug_line.dwo" ); |
1456 | static ConstString g_sect_name_dwarf_debug_line_str("__debug_line_str" ); |
1457 | static ConstString g_sect_name_dwarf_debug_loc("__debug_loc" ); |
1458 | static ConstString g_sect_name_dwarf_debug_loclists("__debug_loclists" ); |
1459 | static ConstString g_sect_name_dwarf_debug_loclists_dwo("__debug_loclists.dwo" ); |
1460 | static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo" ); |
1461 | static ConstString g_sect_name_dwarf_debug_macro("__debug_macro" ); |
1462 | static ConstString g_sect_name_dwarf_debug_macro_dwo("__debug_macro.dwo" ); |
1463 | static ConstString g_sect_name_dwarf_debug_names("__debug_names" ); |
1464 | static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames" ); |
1465 | static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes" ); |
1466 | static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges" ); |
1467 | static ConstString g_sect_name_dwarf_debug_rnglists("__debug_rnglists" ); |
1468 | static ConstString g_sect_name_dwarf_debug_str("__debug_str" ); |
1469 | static ConstString g_sect_name_dwarf_debug_str_dwo("__debug_str.dwo" ); |
1470 | static ConstString g_sect_name_dwarf_debug_str_offs("__debug_str_offs" ); |
1471 | static ConstString g_sect_name_dwarf_debug_str_offs_dwo("__debug_str_offs.dwo" ); |
1472 | static ConstString g_sect_name_dwarf_debug_tu_index("__debug_tu_index" ); |
1473 | static ConstString g_sect_name_dwarf_debug_types("__debug_types" ); |
1474 | static ConstString g_sect_name_dwarf_apple_names("__apple_names" ); |
1475 | static ConstString g_sect_name_dwarf_apple_types("__apple_types" ); |
1476 | static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac" ); |
1477 | static ConstString g_sect_name_dwarf_apple_objc("__apple_objc" ); |
1478 | static ConstString g_sect_name_eh_frame("__eh_frame" ); |
1479 | static ConstString g_sect_name_compact_unwind("__unwind_info" ); |
1480 | static ConstString g_sect_name_text("__text" ); |
1481 | static ConstString g_sect_name_data("__data" ); |
1482 | static ConstString g_sect_name_go_symtab("__gosymtab" ); |
1483 | static ConstString g_sect_name_ctf("__ctf" ); |
1484 | static ConstString g_sect_name_swift_ast("__swift_ast" ); |
1485 | |
1486 | if (section_name == g_sect_name_dwarf_debug_abbrev) |
1487 | return eSectionTypeDWARFDebugAbbrev; |
1488 | if (section_name == g_sect_name_dwarf_debug_abbrev_dwo) |
1489 | return eSectionTypeDWARFDebugAbbrevDwo; |
1490 | if (section_name == g_sect_name_dwarf_debug_addr) |
1491 | return eSectionTypeDWARFDebugAddr; |
1492 | if (section_name == g_sect_name_dwarf_debug_aranges) |
1493 | return eSectionTypeDWARFDebugAranges; |
1494 | if (section_name == g_sect_name_dwarf_debug_cu_index) |
1495 | return eSectionTypeDWARFDebugCuIndex; |
1496 | if (section_name == g_sect_name_dwarf_debug_frame) |
1497 | return eSectionTypeDWARFDebugFrame; |
1498 | if (section_name == g_sect_name_dwarf_debug_info) |
1499 | return eSectionTypeDWARFDebugInfo; |
1500 | if (section_name == g_sect_name_dwarf_debug_info_dwo) |
1501 | return eSectionTypeDWARFDebugInfoDwo; |
1502 | if (section_name == g_sect_name_dwarf_debug_line) |
1503 | return eSectionTypeDWARFDebugLine; |
1504 | if (section_name == g_sect_name_dwarf_debug_line_dwo) |
1505 | return eSectionTypeDWARFDebugLine; // Same as debug_line. |
1506 | if (section_name == g_sect_name_dwarf_debug_line_str) |
1507 | return eSectionTypeDWARFDebugLineStr; |
1508 | if (section_name == g_sect_name_dwarf_debug_loc) |
1509 | return eSectionTypeDWARFDebugLoc; |
1510 | if (section_name == g_sect_name_dwarf_debug_loclists) |
1511 | return eSectionTypeDWARFDebugLocLists; |
1512 | if (section_name == g_sect_name_dwarf_debug_loclists_dwo) |
1513 | return eSectionTypeDWARFDebugLocListsDwo; |
1514 | if (section_name == g_sect_name_dwarf_debug_macinfo) |
1515 | return eSectionTypeDWARFDebugMacInfo; |
1516 | if (section_name == g_sect_name_dwarf_debug_macro) |
1517 | return eSectionTypeDWARFDebugMacro; |
1518 | if (section_name == g_sect_name_dwarf_debug_macro_dwo) |
1519 | return eSectionTypeDWARFDebugMacInfo; // Same as debug_macro. |
1520 | if (section_name == g_sect_name_dwarf_debug_names) |
1521 | return eSectionTypeDWARFDebugNames; |
1522 | if (section_name == g_sect_name_dwarf_debug_pubnames) |
1523 | return eSectionTypeDWARFDebugPubNames; |
1524 | if (section_name == g_sect_name_dwarf_debug_pubtypes) |
1525 | return eSectionTypeDWARFDebugPubTypes; |
1526 | if (section_name == g_sect_name_dwarf_debug_ranges) |
1527 | return eSectionTypeDWARFDebugRanges; |
1528 | if (section_name == g_sect_name_dwarf_debug_rnglists) |
1529 | return eSectionTypeDWARFDebugRngLists; |
1530 | if (section_name == g_sect_name_dwarf_debug_str) |
1531 | return eSectionTypeDWARFDebugStr; |
1532 | if (section_name == g_sect_name_dwarf_debug_str_dwo) |
1533 | return eSectionTypeDWARFDebugStrDwo; |
1534 | if (section_name == g_sect_name_dwarf_debug_str_offs) |
1535 | return eSectionTypeDWARFDebugStrOffsets; |
1536 | if (section_name == g_sect_name_dwarf_debug_str_offs_dwo) |
1537 | return eSectionTypeDWARFDebugStrOffsetsDwo; |
1538 | if (section_name == g_sect_name_dwarf_debug_tu_index) |
1539 | return eSectionTypeDWARFDebugTuIndex; |
1540 | if (section_name == g_sect_name_dwarf_debug_types) |
1541 | return eSectionTypeDWARFDebugTypes; |
1542 | if (section_name == g_sect_name_dwarf_apple_names) |
1543 | return eSectionTypeDWARFAppleNames; |
1544 | if (section_name == g_sect_name_dwarf_apple_types) |
1545 | return eSectionTypeDWARFAppleTypes; |
1546 | if (section_name == g_sect_name_dwarf_apple_namespaces) |
1547 | return eSectionTypeDWARFAppleNamespaces; |
1548 | if (section_name == g_sect_name_dwarf_apple_objc) |
1549 | return eSectionTypeDWARFAppleObjC; |
1550 | if (section_name == g_sect_name_objc_selrefs) |
1551 | return eSectionTypeDataCStringPointers; |
1552 | if (section_name == g_sect_name_objc_msgrefs) |
1553 | return eSectionTypeDataObjCMessageRefs; |
1554 | if (section_name == g_sect_name_eh_frame) |
1555 | return eSectionTypeEHFrame; |
1556 | if (section_name == g_sect_name_compact_unwind) |
1557 | return eSectionTypeCompactUnwind; |
1558 | if (section_name == g_sect_name_cfstring) |
1559 | return eSectionTypeDataObjCCFStrings; |
1560 | if (section_name == g_sect_name_go_symtab) |
1561 | return eSectionTypeGoSymtab; |
1562 | if (section_name == g_sect_name_ctf) |
1563 | return eSectionTypeCTF; |
1564 | if (section_name == g_sect_name_swift_ast) |
1565 | return eSectionTypeSwiftModules; |
1566 | if (section_name == g_sect_name_objc_data || |
1567 | section_name == g_sect_name_objc_classrefs || |
1568 | section_name == g_sect_name_objc_superrefs || |
1569 | section_name == g_sect_name_objc_const || |
1570 | section_name == g_sect_name_objc_classlist) { |
1571 | return eSectionTypeDataPointers; |
1572 | } |
1573 | |
1574 | switch (mach_sect_type) { |
1575 | // TODO: categorize sections by other flags for regular sections |
1576 | case S_REGULAR: |
1577 | if (section_name == g_sect_name_text) |
1578 | return eSectionTypeCode; |
1579 | if (section_name == g_sect_name_data) |
1580 | return eSectionTypeData; |
1581 | return eSectionTypeOther; |
1582 | case S_ZEROFILL: |
1583 | return eSectionTypeZeroFill; |
1584 | case S_CSTRING_LITERALS: // section with only literal C strings |
1585 | return eSectionTypeDataCString; |
1586 | case S_4BYTE_LITERALS: // section with only 4 byte literals |
1587 | return eSectionTypeData4; |
1588 | case S_8BYTE_LITERALS: // section with only 8 byte literals |
1589 | return eSectionTypeData8; |
1590 | case S_LITERAL_POINTERS: // section with only pointers to literals |
1591 | return eSectionTypeDataPointers; |
1592 | case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers |
1593 | return eSectionTypeDataPointers; |
1594 | case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers |
1595 | return eSectionTypeDataPointers; |
1596 | case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in |
1597 | // the reserved2 field |
1598 | return eSectionTypeCode; |
1599 | case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for |
1600 | // initialization |
1601 | return eSectionTypeDataPointers; |
1602 | case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for |
1603 | // termination |
1604 | return eSectionTypeDataPointers; |
1605 | case S_COALESCED: |
1606 | return eSectionTypeOther; |
1607 | case S_GB_ZEROFILL: |
1608 | return eSectionTypeZeroFill; |
1609 | case S_INTERPOSING: // section with only pairs of function pointers for |
1610 | // interposing |
1611 | return eSectionTypeCode; |
1612 | case S_16BYTE_LITERALS: // section with only 16 byte literals |
1613 | return eSectionTypeData16; |
1614 | case S_DTRACE_DOF: |
1615 | return eSectionTypeDebug; |
1616 | case S_LAZY_DYLIB_SYMBOL_POINTERS: |
1617 | return eSectionTypeDataPointers; |
1618 | default: |
1619 | return eSectionTypeOther; |
1620 | } |
1621 | } |
1622 | |
1623 | struct ObjectFileMachO::SegmentParsingContext { |
1624 | const EncryptedFileRanges EncryptedRanges; |
1625 | lldb_private::SectionList &UnifiedList; |
1626 | uint32_t NextSegmentIdx = 0; |
1627 | uint32_t NextSectionIdx = 0; |
1628 | bool FileAddressesChanged = false; |
1629 | |
1630 | SegmentParsingContext(EncryptedFileRanges EncryptedRanges, |
1631 | lldb_private::SectionList &UnifiedList) |
1632 | : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} |
1633 | }; |
1634 | |
1635 | void ObjectFileMachO::ProcessSegmentCommand( |
1636 | const llvm::MachO::load_command &load_cmd_, lldb::offset_t offset, |
1637 | uint32_t cmd_idx, SegmentParsingContext &context) { |
1638 | llvm::MachO::segment_command_64 load_cmd; |
1639 | memcpy(dest: &load_cmd, src: &load_cmd_, n: sizeof(load_cmd_)); |
1640 | |
1641 | if (!m_data.GetU8(offset_ptr: &offset, dst: (uint8_t *)load_cmd.segname, count: 16)) |
1642 | return; |
1643 | |
1644 | ModuleSP module_sp = GetModule(); |
1645 | const bool is_core = GetType() == eTypeCoreFile; |
1646 | const bool is_dsym = (m_header.filetype == MH_DSYM); |
1647 | bool add_section = true; |
1648 | bool add_to_unified = true; |
1649 | ConstString const_segname( |
1650 | load_cmd.segname, strnlen(string: load_cmd.segname, maxlen: sizeof(load_cmd.segname))); |
1651 | |
1652 | SectionSP unified_section_sp( |
1653 | context.UnifiedList.FindSectionByName(section_dstr: const_segname)); |
1654 | if (is_dsym && unified_section_sp) { |
1655 | if (const_segname == GetSegmentNameLINKEDIT()) { |
1656 | // We need to keep the __LINKEDIT segment private to this object file |
1657 | // only |
1658 | add_to_unified = false; |
1659 | } else { |
1660 | // This is the dSYM file and this section has already been created by the |
1661 | // object file, no need to create it. |
1662 | add_section = false; |
1663 | } |
1664 | } |
1665 | load_cmd.vmaddr = m_data.GetAddress(offset_ptr: &offset); |
1666 | load_cmd.vmsize = m_data.GetAddress(offset_ptr: &offset); |
1667 | load_cmd.fileoff = m_data.GetAddress(offset_ptr: &offset); |
1668 | load_cmd.filesize = m_data.GetAddress(offset_ptr: &offset); |
1669 | if (!m_data.GetU32(offset_ptr: &offset, dst: &load_cmd.maxprot, count: 4)) |
1670 | return; |
1671 | |
1672 | SanitizeSegmentCommand(seg_cmd&: load_cmd, cmd_idx); |
1673 | |
1674 | const uint32_t segment_permissions = GetSegmentPermissions(seg_cmd: load_cmd); |
1675 | const bool segment_is_encrypted = |
1676 | (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; |
1677 | |
1678 | // Use a segment ID of the segment index shifted left by 8 so they never |
1679 | // conflict with any of the sections. |
1680 | SectionSP segment_sp; |
1681 | if (add_section && (const_segname || is_core)) { |
1682 | segment_sp = std::make_shared<Section>( |
1683 | args&: module_sp, // Module to which this section belongs |
1684 | args: this, // Object file to which this sections belongs |
1685 | args: ++context.NextSegmentIdx |
1686 | << 8, // Section ID is the 1 based segment index |
1687 | // shifted right by 8 bits as not to collide with any of the 256 |
1688 | // section IDs that are possible |
1689 | args&: const_segname, // Name of this section |
1690 | args: eSectionTypeContainer, // This section is a container of other |
1691 | // sections. |
1692 | args&: load_cmd.vmaddr, // File VM address == addresses as they are |
1693 | // found in the object file |
1694 | args&: load_cmd.vmsize, // VM size in bytes of this section |
1695 | args&: load_cmd.fileoff, // Offset to the data for this section in |
1696 | // the file |
1697 | args&: load_cmd.filesize, // Size in bytes of this section as found |
1698 | // in the file |
1699 | args: 0, // Segments have no alignment information |
1700 | args&: load_cmd.flags); // Flags for this section |
1701 | |
1702 | segment_sp->SetIsEncrypted(segment_is_encrypted); |
1703 | m_sections_up->AddSection(section_sp: segment_sp); |
1704 | segment_sp->SetPermissions(segment_permissions); |
1705 | if (add_to_unified) |
1706 | context.UnifiedList.AddSection(section_sp: segment_sp); |
1707 | } else if (unified_section_sp) { |
1708 | // If this is a dSYM and the file addresses in the dSYM differ from the |
1709 | // file addresses in the ObjectFile, we must use the file base address for |
1710 | // the Section from the dSYM for the DWARF to resolve correctly. |
1711 | // This only happens with binaries in the shared cache in practice; |
1712 | // normally a mismatch like this would give a binary & dSYM that do not |
1713 | // match UUIDs. When a binary is included in the shared cache, its |
1714 | // segments are rearranged to optimize the shared cache, so its file |
1715 | // addresses will differ from what the ObjectFile had originally, |
1716 | // and what the dSYM has. |
1717 | if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { |
1718 | Log *log = GetLog(mask: LLDBLog::Symbols); |
1719 | if (log) { |
1720 | log->Printf( |
1721 | format: "Installing dSYM's %s segment file address over ObjectFile's " |
1722 | "so symbol table/debug info resolves correctly for %s" , |
1723 | const_segname.AsCString(), |
1724 | module_sp->GetFileSpec().GetFilename().AsCString()); |
1725 | } |
1726 | |
1727 | // Make sure we've parsed the symbol table from the ObjectFile before |
1728 | // we go around changing its Sections. |
1729 | module_sp->GetObjectFile()->GetSymtab(); |
1730 | // eh_frame would present the same problems but we parse that on a per- |
1731 | // function basis as-needed so it's more difficult to remove its use of |
1732 | // the Sections. Realistically, the environments where this code path |
1733 | // will be taken will not have eh_frame sections. |
1734 | |
1735 | unified_section_sp->SetFileAddress(load_cmd.vmaddr); |
1736 | |
1737 | // Notify the module that the section addresses have been changed once |
1738 | // we're done so any file-address caches can be updated. |
1739 | context.FileAddressesChanged = true; |
1740 | } |
1741 | m_sections_up->AddSection(section_sp: unified_section_sp); |
1742 | } |
1743 | |
1744 | llvm::MachO::section_64 sect64; |
1745 | ::memset(s: §64, c: 0, n: sizeof(sect64)); |
1746 | // Push a section into our mach sections for the section at index zero |
1747 | // (NO_SECT) if we don't have any mach sections yet... |
1748 | if (m_mach_sections.empty()) |
1749 | m_mach_sections.push_back(x: sect64); |
1750 | uint32_t segment_sect_idx; |
1751 | const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; |
1752 | |
1753 | const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; |
1754 | for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; |
1755 | ++segment_sect_idx) { |
1756 | if (m_data.GetU8(offset_ptr: &offset, dst: (uint8_t *)sect64.sectname, |
1757 | count: sizeof(sect64.sectname)) == nullptr) |
1758 | break; |
1759 | if (m_data.GetU8(offset_ptr: &offset, dst: (uint8_t *)sect64.segname, |
1760 | count: sizeof(sect64.segname)) == nullptr) |
1761 | break; |
1762 | sect64.addr = m_data.GetAddress(offset_ptr: &offset); |
1763 | sect64.size = m_data.GetAddress(offset_ptr: &offset); |
1764 | |
1765 | if (m_data.GetU32(offset_ptr: &offset, dst: §64.offset, count: num_u32s) == nullptr) |
1766 | break; |
1767 | |
1768 | if (IsSharedCacheBinary() && !IsInMemory()) { |
1769 | sect64.offset = sect64.addr - m_text_address; |
1770 | } |
1771 | |
1772 | // Keep a list of mach sections around in case we need to get at data that |
1773 | // isn't stored in the abstracted Sections. |
1774 | m_mach_sections.push_back(x: sect64); |
1775 | |
1776 | if (add_section) { |
1777 | ConstString section_name( |
1778 | sect64.sectname, strnlen(string: sect64.sectname, maxlen: sizeof(sect64.sectname))); |
1779 | if (!const_segname) { |
1780 | // We have a segment with no name so we need to conjure up segments |
1781 | // that correspond to the section's segname if there isn't already such |
1782 | // a section. If there is such a section, we resize the section so that |
1783 | // it spans all sections. We also mark these sections as fake so |
1784 | // address matches don't hit if they land in the gaps between the child |
1785 | // sections. |
1786 | const_segname.SetTrimmedCStringWithLength(cstr: sect64.segname, |
1787 | fixed_cstr_len: sizeof(sect64.segname)); |
1788 | segment_sp = context.UnifiedList.FindSectionByName(section_dstr: const_segname); |
1789 | if (segment_sp.get()) { |
1790 | Section *segment = segment_sp.get(); |
1791 | // Grow the section size as needed. |
1792 | const lldb::addr_t sect64_min_addr = sect64.addr; |
1793 | const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; |
1794 | const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); |
1795 | const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); |
1796 | const lldb::addr_t curr_seg_max_addr = |
1797 | curr_seg_min_addr + curr_seg_byte_size; |
1798 | if (sect64_min_addr >= curr_seg_min_addr) { |
1799 | const lldb::addr_t new_seg_byte_size = |
1800 | sect64_max_addr - curr_seg_min_addr; |
1801 | // Only grow the section size if needed |
1802 | if (new_seg_byte_size > curr_seg_byte_size) |
1803 | segment->SetByteSize(new_seg_byte_size); |
1804 | } else { |
1805 | // We need to change the base address of the segment and adjust the |
1806 | // child section offsets for all existing children. |
1807 | const lldb::addr_t slide_amount = |
1808 | sect64_min_addr - curr_seg_min_addr; |
1809 | segment->Slide(slide_amount, slide_children: false); |
1810 | segment->GetChildren().Slide(slide_amount: -slide_amount, slide_children: false); |
1811 | segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); |
1812 | } |
1813 | |
1814 | // Grow the section size as needed. |
1815 | if (sect64.offset) { |
1816 | const lldb::addr_t segment_min_file_offset = |
1817 | segment->GetFileOffset(); |
1818 | const lldb::addr_t segment_max_file_offset = |
1819 | segment_min_file_offset + segment->GetFileSize(); |
1820 | |
1821 | const lldb::addr_t section_min_file_offset = sect64.offset; |
1822 | const lldb::addr_t section_max_file_offset = |
1823 | section_min_file_offset + sect64.size; |
1824 | const lldb::addr_t new_file_offset = |
1825 | std::min(a: section_min_file_offset, b: segment_min_file_offset); |
1826 | const lldb::addr_t new_file_size = |
1827 | std::max(a: section_max_file_offset, b: segment_max_file_offset) - |
1828 | new_file_offset; |
1829 | segment->SetFileOffset(new_file_offset); |
1830 | segment->SetFileSize(new_file_size); |
1831 | } |
1832 | } else { |
1833 | // Create a fake section for the section's named segment |
1834 | segment_sp = std::make_shared<Section>( |
1835 | args&: segment_sp, // Parent section |
1836 | args&: module_sp, // Module to which this section belongs |
1837 | args: this, // Object file to which this section belongs |
1838 | args: ++context.NextSegmentIdx |
1839 | << 8, // Section ID is the 1 based segment index |
1840 | // shifted right by 8 bits as not to |
1841 | // collide with any of the 256 section IDs |
1842 | // that are possible |
1843 | args&: const_segname, // Name of this section |
1844 | args: eSectionTypeContainer, // This section is a container of |
1845 | // other sections. |
1846 | args&: sect64.addr, // File VM address == addresses as they are |
1847 | // found in the object file |
1848 | args&: sect64.size, // VM size in bytes of this section |
1849 | args&: sect64.offset, // Offset to the data for this section in |
1850 | // the file |
1851 | args: sect64.offset ? sect64.size : 0, // Size in bytes of |
1852 | // this section as |
1853 | // found in the file |
1854 | args&: sect64.align, |
1855 | args&: load_cmd.flags); // Flags for this section |
1856 | segment_sp->SetIsFake(true); |
1857 | segment_sp->SetPermissions(segment_permissions); |
1858 | m_sections_up->AddSection(section_sp: segment_sp); |
1859 | if (add_to_unified) |
1860 | context.UnifiedList.AddSection(section_sp: segment_sp); |
1861 | segment_sp->SetIsEncrypted(segment_is_encrypted); |
1862 | } |
1863 | } |
1864 | assert(segment_sp.get()); |
1865 | |
1866 | lldb::SectionType sect_type = GetSectionType(flags: sect64.flags, section_name); |
1867 | |
1868 | SectionSP section_sp(new Section( |
1869 | segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, |
1870 | sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, |
1871 | sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, |
1872 | sect64.flags)); |
1873 | // Set the section to be encrypted to match the segment |
1874 | |
1875 | bool section_is_encrypted = false; |
1876 | if (!segment_is_encrypted && load_cmd.filesize != 0) |
1877 | section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( |
1878 | addr: sect64.offset) != nullptr; |
1879 | |
1880 | section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); |
1881 | section_sp->SetPermissions(segment_permissions); |
1882 | segment_sp->GetChildren().AddSection(section_sp); |
1883 | |
1884 | if (segment_sp->IsFake()) { |
1885 | segment_sp.reset(); |
1886 | const_segname.Clear(); |
1887 | } |
1888 | } |
1889 | } |
1890 | if (segment_sp && is_dsym) { |
1891 | if (first_segment_sectID <= context.NextSectionIdx) { |
1892 | lldb::user_id_t sect_uid; |
1893 | for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; |
1894 | ++sect_uid) { |
1895 | SectionSP curr_section_sp( |
1896 | segment_sp->GetChildren().FindSectionByID(sect_id: sect_uid)); |
1897 | SectionSP next_section_sp; |
1898 | if (sect_uid + 1 <= context.NextSectionIdx) |
1899 | next_section_sp = |
1900 | segment_sp->GetChildren().FindSectionByID(sect_id: sect_uid + 1); |
1901 | |
1902 | if (curr_section_sp.get()) { |
1903 | if (curr_section_sp->GetByteSize() == 0) { |
1904 | if (next_section_sp.get() != nullptr) |
1905 | curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - |
1906 | curr_section_sp->GetFileAddress()); |
1907 | else |
1908 | curr_section_sp->SetByteSize(load_cmd.vmsize); |
1909 | } |
1910 | } |
1911 | } |
1912 | } |
1913 | } |
1914 | } |
1915 | |
1916 | void ObjectFileMachO::ProcessDysymtabCommand( |
1917 | const llvm::MachO::load_command &load_cmd, lldb::offset_t offset) { |
1918 | m_dysymtab.cmd = load_cmd.cmd; |
1919 | m_dysymtab.cmdsize = load_cmd.cmdsize; |
1920 | m_data.GetU32(offset_ptr: &offset, dst: &m_dysymtab.ilocalsym, |
1921 | count: (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); |
1922 | } |
1923 | |
1924 | void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { |
1925 | if (m_sections_up) |
1926 | return; |
1927 | |
1928 | m_sections_up = std::make_unique<SectionList>(); |
1929 | |
1930 | lldb::offset_t offset = MachHeaderSizeFromMagic(magic: m_header.magic); |
1931 | // bool dump_sections = false; |
1932 | ModuleSP module_sp(GetModule()); |
1933 | |
1934 | offset = MachHeaderSizeFromMagic(magic: m_header.magic); |
1935 | |
1936 | SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); |
1937 | llvm::MachO::load_command load_cmd; |
1938 | for (uint32_t i = 0; i < m_header.ncmds; ++i) { |
1939 | const lldb::offset_t load_cmd_offset = offset; |
1940 | if (m_data.GetU32(offset_ptr: &offset, dst: &load_cmd, count: 2) == nullptr) |
1941 | break; |
1942 | |
1943 | if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) |
1944 | ProcessSegmentCommand(load_cmd_: load_cmd, offset, cmd_idx: i, context); |
1945 | else if (load_cmd.cmd == LC_DYSYMTAB) |
1946 | ProcessDysymtabCommand(load_cmd, offset); |
1947 | |
1948 | offset = load_cmd_offset + load_cmd.cmdsize; |
1949 | } |
1950 | |
1951 | if (context.FileAddressesChanged && module_sp) |
1952 | module_sp->SectionFileAddressesChanged(); |
1953 | } |
1954 | |
1955 | class MachSymtabSectionInfo { |
1956 | public: |
1957 | MachSymtabSectionInfo(SectionList *section_list) |
1958 | : m_section_list(section_list), m_section_infos() { |
1959 | // Get the number of sections down to a depth of 1 to include all segments |
1960 | // and their sections, but no other sections that may be added for debug |
1961 | // map or |
1962 | m_section_infos.resize(new_size: section_list->GetNumSections(depth: 1)); |
1963 | } |
1964 | |
1965 | SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { |
1966 | if (n_sect == 0) |
1967 | return SectionSP(); |
1968 | if (n_sect < m_section_infos.size()) { |
1969 | if (!m_section_infos[n_sect].section_sp) { |
1970 | SectionSP section_sp(m_section_list->FindSectionByID(sect_id: n_sect)); |
1971 | m_section_infos[n_sect].section_sp = section_sp; |
1972 | if (section_sp) { |
1973 | m_section_infos[n_sect].vm_range.SetBaseAddress( |
1974 | section_sp->GetFileAddress()); |
1975 | m_section_infos[n_sect].vm_range.SetByteSize( |
1976 | section_sp->GetByteSize()); |
1977 | } else { |
1978 | std::string filename = "<unknown>" ; |
1979 | SectionSP first_section_sp(m_section_list->GetSectionAtIndex(idx: 0)); |
1980 | if (first_section_sp) |
1981 | filename = first_section_sp->GetObjectFile()->GetFileSpec().GetPath(); |
1982 | |
1983 | Debugger::ReportError( |
1984 | message: llvm::formatv(Fmt: "unable to find section {0} for a symbol in " |
1985 | "{1}, corrupt file?" , |
1986 | Vals&: n_sect, Vals&: filename)); |
1987 | } |
1988 | } |
1989 | if (m_section_infos[n_sect].vm_range.Contains(addr: file_addr)) { |
1990 | // Symbol is in section. |
1991 | return m_section_infos[n_sect].section_sp; |
1992 | } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && |
1993 | m_section_infos[n_sect].vm_range.GetBaseAddress() == |
1994 | file_addr) { |
1995 | // Symbol is in section with zero size, but has the same start address |
1996 | // as the section. This can happen with linker symbols (symbols that |
1997 | // start with the letter 'l' or 'L'. |
1998 | return m_section_infos[n_sect].section_sp; |
1999 | } |
2000 | } |
2001 | return m_section_list->FindSectionContainingFileAddress(addr: file_addr); |
2002 | } |
2003 | |
2004 | protected: |
2005 | struct SectionInfo { |
2006 | SectionInfo() : vm_range(), section_sp() {} |
2007 | |
2008 | VMRange vm_range; |
2009 | SectionSP section_sp; |
2010 | }; |
2011 | SectionList *m_section_list; |
2012 | std::vector<SectionInfo> m_section_infos; |
2013 | }; |
2014 | |
2015 | #define TRIE_SYMBOL_IS_THUMB (1ULL << 63) |
2016 | struct TrieEntry { |
2017 | void Dump() const { |
2018 | printf(format: "0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"" , |
2019 | static_cast<unsigned long long>(address), |
2020 | static_cast<unsigned long long>(flags), |
2021 | static_cast<unsigned long long>(other), name.GetCString()); |
2022 | if (import_name) |
2023 | printf(format: " -> \"%s\"\n" , import_name.GetCString()); |
2024 | else |
2025 | printf(format: "\n" ); |
2026 | } |
2027 | ConstString name; |
2028 | uint64_t address = LLDB_INVALID_ADDRESS; |
2029 | uint64_t flags = |
2030 | 0; // EXPORT_SYMBOL_FLAGS_REEXPORT, EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER, |
2031 | // TRIE_SYMBOL_IS_THUMB |
2032 | uint64_t other = 0; |
2033 | ConstString import_name; |
2034 | }; |
2035 | |
2036 | struct TrieEntryWithOffset { |
2037 | lldb::offset_t nodeOffset; |
2038 | TrieEntry entry; |
2039 | |
2040 | TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} |
2041 | |
2042 | void Dump(uint32_t idx) const { |
2043 | printf(format: "[%3u] 0x%16.16llx: " , idx, |
2044 | static_cast<unsigned long long>(nodeOffset)); |
2045 | entry.Dump(); |
2046 | } |
2047 | |
2048 | bool operator<(const TrieEntryWithOffset &other) const { |
2049 | return (nodeOffset < other.nodeOffset); |
2050 | } |
2051 | }; |
2052 | |
2053 | static bool (DataExtractor &data, lldb::offset_t offset, |
2054 | const bool is_arm, addr_t text_seg_base_addr, |
2055 | std::vector<llvm::StringRef> &nameSlices, |
2056 | std::set<lldb::addr_t> &resolver_addresses, |
2057 | std::vector<TrieEntryWithOffset> &reexports, |
2058 | std::vector<TrieEntryWithOffset> &ext_symbols) { |
2059 | if (!data.ValidOffset(offset)) |
2060 | return true; |
2061 | |
2062 | // Terminal node -- end of a branch, possibly add this to |
2063 | // the symbol table or resolver table. |
2064 | const uint64_t terminalSize = data.GetULEB128(offset_ptr: &offset); |
2065 | lldb::offset_t children_offset = offset + terminalSize; |
2066 | if (terminalSize != 0) { |
2067 | TrieEntryWithOffset e(offset); |
2068 | e.entry.flags = data.GetULEB128(offset_ptr: &offset); |
2069 | const char *import_name = nullptr; |
2070 | if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { |
2071 | e.entry.address = 0; |
2072 | e.entry.other = data.GetULEB128(offset_ptr: &offset); // dylib ordinal |
2073 | import_name = data.GetCStr(offset_ptr: &offset); |
2074 | } else { |
2075 | e.entry.address = data.GetULEB128(offset_ptr: &offset); |
2076 | if (text_seg_base_addr != LLDB_INVALID_ADDRESS) |
2077 | e.entry.address += text_seg_base_addr; |
2078 | if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { |
2079 | e.entry.other = data.GetULEB128(offset_ptr: &offset); |
2080 | uint64_t resolver_addr = e.entry.other; |
2081 | if (text_seg_base_addr != LLDB_INVALID_ADDRESS) |
2082 | resolver_addr += text_seg_base_addr; |
2083 | if (is_arm) |
2084 | resolver_addr &= THUMB_ADDRESS_BIT_MASK; |
2085 | resolver_addresses.insert(x: resolver_addr); |
2086 | } else |
2087 | e.entry.other = 0; |
2088 | } |
2089 | bool add_this_entry = false; |
2090 | if (Flags(e.entry.flags).Test(bit: EXPORT_SYMBOL_FLAGS_REEXPORT) && |
2091 | import_name && import_name[0]) { |
2092 | // add symbols that are reexport symbols with a valid import name. |
2093 | add_this_entry = true; |
2094 | } else if (e.entry.flags == 0 && |
2095 | (import_name == nullptr || import_name[0] == '\0')) { |
2096 | // add externally visible symbols, in case the nlist record has |
2097 | // been stripped/omitted. |
2098 | add_this_entry = true; |
2099 | } |
2100 | if (add_this_entry) { |
2101 | std::string name; |
2102 | if (!nameSlices.empty()) { |
2103 | for (auto name_slice : nameSlices) |
2104 | name.append(s: name_slice.data(), n: name_slice.size()); |
2105 | } |
2106 | if (name.size() > 1) { |
2107 | // Skip the leading '_' |
2108 | e.entry.name.SetCStringWithLength(cstr: name.c_str() + 1, cstr_len: name.size() - 1); |
2109 | } |
2110 | if (import_name) { |
2111 | // Skip the leading '_' |
2112 | e.entry.import_name.SetCString(import_name + 1); |
2113 | } |
2114 | if (Flags(e.entry.flags).Test(bit: EXPORT_SYMBOL_FLAGS_REEXPORT)) { |
2115 | reexports.push_back(x: e); |
2116 | } else { |
2117 | if (is_arm && (e.entry.address & 1)) { |
2118 | e.entry.flags |= TRIE_SYMBOL_IS_THUMB; |
2119 | e.entry.address &= THUMB_ADDRESS_BIT_MASK; |
2120 | } |
2121 | ext_symbols.push_back(x: e); |
2122 | } |
2123 | } |
2124 | } |
2125 | |
2126 | const uint8_t childrenCount = data.GetU8(offset_ptr: &children_offset); |
2127 | for (uint8_t i = 0; i < childrenCount; ++i) { |
2128 | const char *cstr = data.GetCStr(offset_ptr: &children_offset); |
2129 | if (cstr) |
2130 | nameSlices.push_back(x: llvm::StringRef(cstr)); |
2131 | else |
2132 | return false; // Corrupt data |
2133 | lldb::offset_t childNodeOffset = data.GetULEB128(offset_ptr: &children_offset); |
2134 | if (childNodeOffset) { |
2135 | if (!ParseTrieEntries(data, offset: childNodeOffset, is_arm, text_seg_base_addr, |
2136 | nameSlices, resolver_addresses, reexports, |
2137 | ext_symbols)) { |
2138 | return false; |
2139 | } |
2140 | } |
2141 | nameSlices.pop_back(); |
2142 | } |
2143 | return true; |
2144 | } |
2145 | |
2146 | static SymbolType GetSymbolType(const char *&symbol_name, |
2147 | bool &demangled_is_synthesized, |
2148 | const SectionSP &text_section_sp, |
2149 | const SectionSP &data_section_sp, |
2150 | const SectionSP &data_dirty_section_sp, |
2151 | const SectionSP &data_const_section_sp, |
2152 | const SectionSP &symbol_section) { |
2153 | SymbolType type = eSymbolTypeInvalid; |
2154 | |
2155 | const char *symbol_sect_name = symbol_section->GetName().AsCString(); |
2156 | if (symbol_section->IsDescendant(section: text_section_sp.get())) { |
2157 | if (symbol_section->IsClear(bit: S_ATTR_PURE_INSTRUCTIONS | |
2158 | S_ATTR_SELF_MODIFYING_CODE | |
2159 | S_ATTR_SOME_INSTRUCTIONS)) |
2160 | type = eSymbolTypeData; |
2161 | else |
2162 | type = eSymbolTypeCode; |
2163 | } else if (symbol_section->IsDescendant(section: data_section_sp.get()) || |
2164 | symbol_section->IsDescendant(section: data_dirty_section_sp.get()) || |
2165 | symbol_section->IsDescendant(section: data_const_section_sp.get())) { |
2166 | if (symbol_sect_name && |
2167 | ::strstr(haystack: symbol_sect_name, needle: "__objc" ) == symbol_sect_name) { |
2168 | type = eSymbolTypeRuntime; |
2169 | |
2170 | if (symbol_name) { |
2171 | llvm::StringRef symbol_name_ref(symbol_name); |
2172 | if (symbol_name_ref.starts_with(Prefix: "OBJC_" )) { |
2173 | static const llvm::StringRef g_objc_v2_prefix_class("OBJC_CLASS_$_" ); |
2174 | static const llvm::StringRef g_objc_v2_prefix_metaclass( |
2175 | "OBJC_METACLASS_$_" ); |
2176 | static const llvm::StringRef g_objc_v2_prefix_ivar("OBJC_IVAR_$_" ); |
2177 | if (symbol_name_ref.starts_with(Prefix: g_objc_v2_prefix_class)) { |
2178 | symbol_name = symbol_name + g_objc_v2_prefix_class.size(); |
2179 | type = eSymbolTypeObjCClass; |
2180 | demangled_is_synthesized = true; |
2181 | } else if (symbol_name_ref.starts_with(Prefix: g_objc_v2_prefix_metaclass)) { |
2182 | symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); |
2183 | type = eSymbolTypeObjCMetaClass; |
2184 | demangled_is_synthesized = true; |
2185 | } else if (symbol_name_ref.starts_with(Prefix: g_objc_v2_prefix_ivar)) { |
2186 | symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); |
2187 | type = eSymbolTypeObjCIVar; |
2188 | demangled_is_synthesized = true; |
2189 | } |
2190 | } |
2191 | } |
2192 | } else if (symbol_sect_name && |
2193 | ::strstr(haystack: symbol_sect_name, needle: "__gcc_except_tab" ) == |
2194 | symbol_sect_name) { |
2195 | type = eSymbolTypeException; |
2196 | } else { |
2197 | type = eSymbolTypeData; |
2198 | } |
2199 | } else if (symbol_sect_name && |
2200 | ::strstr(haystack: symbol_sect_name, needle: "__IMPORT" ) == symbol_sect_name) { |
2201 | type = eSymbolTypeTrampoline; |
2202 | } |
2203 | return type; |
2204 | } |
2205 | |
2206 | static std::optional<struct nlist_64> |
2207 | (DataExtractor &nlist_data, lldb::offset_t &nlist_data_offset, |
2208 | size_t nlist_byte_size) { |
2209 | struct nlist_64 nlist; |
2210 | if (!nlist_data.ValidOffsetForDataOfSize(offset: nlist_data_offset, length: nlist_byte_size)) |
2211 | return {}; |
2212 | nlist.n_strx = nlist_data.GetU32_unchecked(offset_ptr: &nlist_data_offset); |
2213 | nlist.n_type = nlist_data.GetU8_unchecked(offset_ptr: &nlist_data_offset); |
2214 | nlist.n_sect = nlist_data.GetU8_unchecked(offset_ptr: &nlist_data_offset); |
2215 | nlist.n_desc = nlist_data.GetU16_unchecked(offset_ptr: &nlist_data_offset); |
2216 | nlist.n_value = nlist_data.GetAddress_unchecked(offset_ptr: &nlist_data_offset); |
2217 | return nlist; |
2218 | } |
2219 | |
2220 | enum { DebugSymbols = true, NonDebugSymbols = false }; |
2221 | |
2222 | void ObjectFileMachO::ParseSymtab(Symtab &symtab) { |
2223 | ModuleSP module_sp(GetModule()); |
2224 | if (!module_sp) |
2225 | return; |
2226 | |
2227 | Log *log = GetLog(mask: LLDBLog::Symbols); |
2228 | |
2229 | const FileSpec &file = m_file ? m_file : module_sp->GetFileSpec(); |
2230 | const char *file_name = file.GetFilename().AsCString(value_if_empty: "<Unknown>" ); |
2231 | LLDB_SCOPED_TIMERF("ObjectFileMachO::ParseSymtab () module = %s" , file_name); |
2232 | LLDB_LOG(log, "Parsing symbol table for {0}" , file_name); |
2233 | Progress progress("Parsing symbol table" , file_name); |
2234 | |
2235 | llvm::MachO::symtab_command symtab_load_command = {.cmd: 0, .cmdsize: 0, .symoff: 0, .nsyms: 0, .stroff: 0, .strsize: 0}; |
2236 | llvm::MachO::linkedit_data_command function_starts_load_command = {.cmd: 0, .cmdsize: 0, .dataoff: 0, .datasize: 0}; |
2237 | llvm::MachO::linkedit_data_command exports_trie_load_command = {.cmd: 0, .cmdsize: 0, .dataoff: 0, .datasize: 0}; |
2238 | llvm::MachO::dyld_info_command dyld_info = {.cmd: 0, .cmdsize: 0, .rebase_off: 0, .rebase_size: 0, .bind_off: 0, .bind_size: 0, .weak_bind_off: 0, .weak_bind_size: 0, .lazy_bind_off: 0, .lazy_bind_size: 0, .export_off: 0, .export_size: 0}; |
2239 | llvm::MachO::dysymtab_command dysymtab = m_dysymtab; |
2240 | // The data element of type bool indicates that this entry is thumb |
2241 | // code. |
2242 | typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; |
2243 | |
2244 | // Record the address of every function/data that we add to the symtab. |
2245 | // We add symbols to the table in the order of most information (nlist |
2246 | // records) to least (function starts), and avoid duplicating symbols |
2247 | // via this set. |
2248 | llvm::DenseSet<addr_t> symbols_added; |
2249 | |
2250 | // We are using a llvm::DenseSet for "symbols_added" so we must be sure we |
2251 | // do not add the tombstone or empty keys to the set. |
2252 | auto add_symbol_addr = [&symbols_added](lldb::addr_t file_addr) { |
2253 | // Don't add the tombstone or empty keys. |
2254 | if (file_addr == UINT64_MAX || file_addr == UINT64_MAX - 1) |
2255 | return; |
2256 | symbols_added.insert(V: file_addr); |
2257 | }; |
2258 | FunctionStarts function_starts; |
2259 | lldb::offset_t offset = MachHeaderSizeFromMagic(magic: m_header.magic); |
2260 | uint32_t i; |
2261 | FileSpecList dylib_files; |
2262 | llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_" ); |
2263 | llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_" ); |
2264 | llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_" ); |
2265 | UUID image_uuid; |
2266 | |
2267 | for (i = 0; i < m_header.ncmds; ++i) { |
2268 | const lldb::offset_t cmd_offset = offset; |
2269 | // Read in the load command and load command size |
2270 | llvm::MachO::load_command lc; |
2271 | if (m_data.GetU32(offset_ptr: &offset, dst: &lc, count: 2) == nullptr) |
2272 | break; |
2273 | // Watch for the symbol table load command |
2274 | switch (lc.cmd) { |
2275 | case LC_SYMTAB: |
2276 | symtab_load_command.cmd = lc.cmd; |
2277 | symtab_load_command.cmdsize = lc.cmdsize; |
2278 | // Read in the rest of the symtab load command |
2279 | if (m_data.GetU32(offset_ptr: &offset, dst: &symtab_load_command.symoff, count: 4) == |
2280 | nullptr) // fill in symoff, nsyms, stroff, strsize fields |
2281 | return; |
2282 | break; |
2283 | |
2284 | case LC_DYLD_INFO: |
2285 | case LC_DYLD_INFO_ONLY: |
2286 | if (m_data.GetU32(offset_ptr: &offset, dst: &dyld_info.rebase_off, count: 10)) { |
2287 | dyld_info.cmd = lc.cmd; |
2288 | dyld_info.cmdsize = lc.cmdsize; |
2289 | } else { |
2290 | memset(s: &dyld_info, c: 0, n: sizeof(dyld_info)); |
2291 | } |
2292 | break; |
2293 | |
2294 | case LC_LOAD_DYLIB: |
2295 | case LC_LOAD_WEAK_DYLIB: |
2296 | case LC_REEXPORT_DYLIB: |
2297 | case LC_LOADFVMLIB: |
2298 | case LC_LOAD_UPWARD_DYLIB: { |
2299 | uint32_t name_offset = cmd_offset + m_data.GetU32(offset_ptr: &offset); |
2300 | const char *path = m_data.PeekCStr(offset: name_offset); |
2301 | if (path) { |
2302 | FileSpec file_spec(path); |
2303 | // Strip the path if there is @rpath, @executable, etc so we just use |
2304 | // the basename |
2305 | if (path[0] == '@') |
2306 | file_spec.ClearDirectory(); |
2307 | |
2308 | if (lc.cmd == LC_REEXPORT_DYLIB) { |
2309 | m_reexported_dylibs.AppendIfUnique(file: file_spec); |
2310 | } |
2311 | |
2312 | dylib_files.Append(file: file_spec); |
2313 | } |
2314 | } break; |
2315 | |
2316 | case LC_DYLD_EXPORTS_TRIE: |
2317 | exports_trie_load_command.cmd = lc.cmd; |
2318 | exports_trie_load_command.cmdsize = lc.cmdsize; |
2319 | if (m_data.GetU32(offset_ptr: &offset, dst: &exports_trie_load_command.dataoff, count: 2) == |
2320 | nullptr) // fill in offset and size fields |
2321 | memset(s: &exports_trie_load_command, c: 0, |
2322 | n: sizeof(exports_trie_load_command)); |
2323 | break; |
2324 | case LC_FUNCTION_STARTS: |
2325 | function_starts_load_command.cmd = lc.cmd; |
2326 | function_starts_load_command.cmdsize = lc.cmdsize; |
2327 | if (m_data.GetU32(offset_ptr: &offset, dst: &function_starts_load_command.dataoff, count: 2) == |
2328 | nullptr) // fill in data offset and size fields |
2329 | memset(s: &function_starts_load_command, c: 0, |
2330 | n: sizeof(function_starts_load_command)); |
2331 | break; |
2332 | |
2333 | case LC_UUID: { |
2334 | const uint8_t *uuid_bytes = m_data.PeekData(offset, length: 16); |
2335 | |
2336 | if (uuid_bytes) |
2337 | image_uuid = UUID(uuid_bytes, 16); |
2338 | break; |
2339 | } |
2340 | |
2341 | default: |
2342 | break; |
2343 | } |
2344 | offset = cmd_offset + lc.cmdsize; |
2345 | } |
2346 | |
2347 | if (!symtab_load_command.cmd) |
2348 | return; |
2349 | |
2350 | SectionList *section_list = GetSectionList(); |
2351 | if (section_list == nullptr) |
2352 | return; |
2353 | |
2354 | const uint32_t addr_byte_size = m_data.GetAddressByteSize(); |
2355 | const ByteOrder byte_order = m_data.GetByteOrder(); |
2356 | bool bit_width_32 = addr_byte_size == 4; |
2357 | const size_t nlist_byte_size = |
2358 | bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); |
2359 | |
2360 | DataExtractor nlist_data(nullptr, 0, byte_order, addr_byte_size); |
2361 | DataExtractor strtab_data(nullptr, 0, byte_order, addr_byte_size); |
2362 | DataExtractor function_starts_data(nullptr, 0, byte_order, addr_byte_size); |
2363 | DataExtractor indirect_symbol_index_data(nullptr, 0, byte_order, |
2364 | addr_byte_size); |
2365 | DataExtractor dyld_trie_data(nullptr, 0, byte_order, addr_byte_size); |
2366 | |
2367 | const addr_t nlist_data_byte_size = |
2368 | symtab_load_command.nsyms * nlist_byte_size; |
2369 | const addr_t strtab_data_byte_size = symtab_load_command.strsize; |
2370 | addr_t strtab_addr = LLDB_INVALID_ADDRESS; |
2371 | |
2372 | ProcessSP process_sp(m_process_wp.lock()); |
2373 | Process *process = process_sp.get(); |
2374 | |
2375 | uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; |
2376 | bool is_shared_cache_image = IsSharedCacheBinary(); |
2377 | bool is_local_shared_cache_image = is_shared_cache_image && !IsInMemory(); |
2378 | SectionSP linkedit_section_sp( |
2379 | section_list->FindSectionByName(section_dstr: GetSegmentNameLINKEDIT())); |
2380 | |
2381 | if (process && m_header.filetype != llvm::MachO::MH_OBJECT && |
2382 | !is_local_shared_cache_image) { |
2383 | Target &target = process->GetTarget(); |
2384 | |
2385 | memory_module_load_level = target.GetMemoryModuleLoadLevel(); |
2386 | |
2387 | // Reading mach file from memory in a process or core file... |
2388 | |
2389 | if (linkedit_section_sp) { |
2390 | addr_t linkedit_load_addr = |
2391 | linkedit_section_sp->GetLoadBaseAddress(target: &target); |
2392 | if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { |
2393 | // We might be trying to access the symbol table before the |
2394 | // __LINKEDIT's load address has been set in the target. We can't |
2395 | // fail to read the symbol table, so calculate the right address |
2396 | // manually |
2397 | linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( |
2398 | mach_header_load_address: m_memory_addr, mach_header_section: GetMachHeaderSection(), section: linkedit_section_sp.get()); |
2399 | } |
2400 | |
2401 | const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); |
2402 | const addr_t symoff_addr = linkedit_load_addr + |
2403 | symtab_load_command.symoff - |
2404 | linkedit_file_offset; |
2405 | strtab_addr = linkedit_load_addr + symtab_load_command.stroff - |
2406 | linkedit_file_offset; |
2407 | |
2408 | // Always load dyld - the dynamic linker - from memory if we didn't |
2409 | // find a binary anywhere else. lldb will not register |
2410 | // dylib/framework/bundle loads/unloads if we don't have the dyld |
2411 | // symbols, we force dyld to load from memory despite the user's |
2412 | // target.memory-module-load-level setting. |
2413 | if (memory_module_load_level == eMemoryModuleLoadLevelComplete || |
2414 | m_header.filetype == llvm::MachO::MH_DYLINKER) { |
2415 | DataBufferSP nlist_data_sp( |
2416 | ReadMemory(process_sp, addr: symoff_addr, byte_size: nlist_data_byte_size)); |
2417 | if (nlist_data_sp) |
2418 | nlist_data.SetData(data_sp: nlist_data_sp, offset: 0, length: nlist_data_sp->GetByteSize()); |
2419 | if (dysymtab.nindirectsyms != 0) { |
2420 | const addr_t indirect_syms_addr = linkedit_load_addr + |
2421 | dysymtab.indirectsymoff - |
2422 | linkedit_file_offset; |
2423 | DataBufferSP indirect_syms_data_sp(ReadMemory( |
2424 | process_sp, addr: indirect_syms_addr, byte_size: dysymtab.nindirectsyms * 4)); |
2425 | if (indirect_syms_data_sp) |
2426 | indirect_symbol_index_data.SetData( |
2427 | data_sp: indirect_syms_data_sp, offset: 0, |
2428 | length: indirect_syms_data_sp->GetByteSize()); |
2429 | // If this binary is outside the shared cache, |
2430 | // cache the string table. |
2431 | // Binaries in the shared cache all share a giant string table, |
2432 | // and we can't share the string tables across multiple |
2433 | // ObjectFileMachO's, so we'd end up re-reading this mega-strtab |
2434 | // for every binary in the shared cache - it would be a big perf |
2435 | // problem. For binaries outside the shared cache, it's faster to |
2436 | // read the entire strtab at once instead of piece-by-piece as we |
2437 | // process the nlist records. |
2438 | if (!is_shared_cache_image) { |
2439 | DataBufferSP strtab_data_sp( |
2440 | ReadMemory(process_sp, addr: strtab_addr, byte_size: strtab_data_byte_size)); |
2441 | if (strtab_data_sp) { |
2442 | strtab_data.SetData(data_sp: strtab_data_sp, offset: 0, |
2443 | length: strtab_data_sp->GetByteSize()); |
2444 | } |
2445 | } |
2446 | } |
2447 | if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) { |
2448 | if (function_starts_load_command.cmd) { |
2449 | const addr_t func_start_addr = |
2450 | linkedit_load_addr + function_starts_load_command.dataoff - |
2451 | linkedit_file_offset; |
2452 | DataBufferSP func_start_data_sp( |
2453 | ReadMemory(process_sp, addr: func_start_addr, |
2454 | byte_size: function_starts_load_command.datasize)); |
2455 | if (func_start_data_sp) |
2456 | function_starts_data.SetData(data_sp: func_start_data_sp, offset: 0, |
2457 | length: func_start_data_sp->GetByteSize()); |
2458 | } |
2459 | } |
2460 | } |
2461 | } |
2462 | } else { |
2463 | if (is_local_shared_cache_image) { |
2464 | // The load commands in shared cache images are relative to the |
2465 | // beginning of the shared cache, not the library image. The |
2466 | // data we get handed when creating the ObjectFileMachO starts |
2467 | // at the beginning of a specific library and spans to the end |
2468 | // of the cache to be able to reach the shared LINKEDIT |
2469 | // segments. We need to convert the load command offsets to be |
2470 | // relative to the beginning of our specific image. |
2471 | lldb::addr_t linkedit_offset = linkedit_section_sp->GetFileOffset(); |
2472 | lldb::offset_t linkedit_slide = |
2473 | linkedit_offset - m_linkedit_original_offset; |
2474 | symtab_load_command.symoff += linkedit_slide; |
2475 | symtab_load_command.stroff += linkedit_slide; |
2476 | dyld_info.export_off += linkedit_slide; |
2477 | dysymtab.indirectsymoff += linkedit_slide; |
2478 | function_starts_load_command.dataoff += linkedit_slide; |
2479 | exports_trie_load_command.dataoff += linkedit_slide; |
2480 | } |
2481 | |
2482 | nlist_data.SetData(data: m_data, offset: symtab_load_command.symoff, |
2483 | length: nlist_data_byte_size); |
2484 | strtab_data.SetData(data: m_data, offset: symtab_load_command.stroff, |
2485 | length: strtab_data_byte_size); |
2486 | |
2487 | // We shouldn't have exports data from both the LC_DYLD_INFO command |
2488 | // AND the LC_DYLD_EXPORTS_TRIE command in the same binary: |
2489 | lldbassert(!((dyld_info.export_size > 0) |
2490 | && (exports_trie_load_command.datasize > 0))); |
2491 | if (dyld_info.export_size > 0) { |
2492 | dyld_trie_data.SetData(data: m_data, offset: dyld_info.export_off, |
2493 | length: dyld_info.export_size); |
2494 | } else if (exports_trie_load_command.datasize > 0) { |
2495 | dyld_trie_data.SetData(data: m_data, offset: exports_trie_load_command.dataoff, |
2496 | length: exports_trie_load_command.datasize); |
2497 | } |
2498 | |
2499 | if (dysymtab.nindirectsyms != 0) { |
2500 | indirect_symbol_index_data.SetData(data: m_data, offset: dysymtab.indirectsymoff, |
2501 | length: dysymtab.nindirectsyms * 4); |
2502 | } |
2503 | if (function_starts_load_command.cmd) { |
2504 | function_starts_data.SetData(data: m_data, offset: function_starts_load_command.dataoff, |
2505 | length: function_starts_load_command.datasize); |
2506 | } |
2507 | } |
2508 | |
2509 | const bool have_strtab_data = strtab_data.GetByteSize() > 0; |
2510 | |
2511 | ConstString g_segment_name_TEXT = GetSegmentNameTEXT(); |
2512 | ConstString g_segment_name_DATA = GetSegmentNameDATA(); |
2513 | ConstString g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); |
2514 | ConstString g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); |
2515 | ConstString g_segment_name_OBJC = GetSegmentNameOBJC(); |
2516 | ConstString g_section_name_eh_frame = GetSectionNameEHFrame(); |
2517 | SectionSP text_section_sp( |
2518 | section_list->FindSectionByName(section_dstr: g_segment_name_TEXT)); |
2519 | SectionSP data_section_sp( |
2520 | section_list->FindSectionByName(section_dstr: g_segment_name_DATA)); |
2521 | SectionSP data_dirty_section_sp( |
2522 | section_list->FindSectionByName(section_dstr: g_segment_name_DATA_DIRTY)); |
2523 | SectionSP data_const_section_sp( |
2524 | section_list->FindSectionByName(section_dstr: g_segment_name_DATA_CONST)); |
2525 | SectionSP objc_section_sp( |
2526 | section_list->FindSectionByName(section_dstr: g_segment_name_OBJC)); |
2527 | SectionSP eh_frame_section_sp; |
2528 | if (text_section_sp.get()) |
2529 | eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( |
2530 | section_dstr: g_section_name_eh_frame); |
2531 | else |
2532 | eh_frame_section_sp = |
2533 | section_list->FindSectionByName(section_dstr: g_section_name_eh_frame); |
2534 | |
2535 | const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); |
2536 | const bool always_thumb = GetArchitecture().IsAlwaysThumbInstructions(); |
2537 | |
2538 | // lldb works best if it knows the start address of all functions in a |
2539 | // module. Linker symbols or debug info are normally the best source of |
2540 | // information for start addr / size but they may be stripped in a released |
2541 | // binary. Two additional sources of information exist in Mach-O binaries: |
2542 | // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each |
2543 | // function's start address in the |
2544 | // binary, relative to the text section. |
2545 | // eh_frame - the eh_frame FDEs have the start addr & size of |
2546 | // each function |
2547 | // LC_FUNCTION_STARTS is the fastest source to read in, and is present on |
2548 | // all modern binaries. |
2549 | // Binaries built to run on older releases may need to use eh_frame |
2550 | // information. |
2551 | |
2552 | if (text_section_sp && function_starts_data.GetByteSize()) { |
2553 | FunctionStarts::Entry function_start_entry; |
2554 | function_start_entry.data = false; |
2555 | lldb::offset_t function_start_offset = 0; |
2556 | function_start_entry.addr = text_section_sp->GetFileAddress(); |
2557 | uint64_t delta; |
2558 | while ((delta = function_starts_data.GetULEB128(offset_ptr: &function_start_offset)) > |
2559 | 0) { |
2560 | // Now append the current entry |
2561 | function_start_entry.addr += delta; |
2562 | if (is_arm) { |
2563 | if (function_start_entry.addr & 1) { |
2564 | function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; |
2565 | function_start_entry.data = true; |
2566 | } else if (always_thumb) { |
2567 | function_start_entry.data = true; |
2568 | } |
2569 | } |
2570 | function_starts.Append(entry: function_start_entry); |
2571 | } |
2572 | } else { |
2573 | // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the |
2574 | // load command claiming an eh_frame but it doesn't actually have the |
2575 | // eh_frame content. And if we have a dSYM, we don't need to do any of |
2576 | // this fill-in-the-missing-symbols works anyway - the debug info should |
2577 | // give us all the functions in the module. |
2578 | if (text_section_sp.get() && eh_frame_section_sp.get() && |
2579 | m_type != eTypeDebugInfo) { |
2580 | DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, |
2581 | DWARFCallFrameInfo::EH); |
2582 | DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; |
2583 | eh_frame.GetFunctionAddressAndSizeVector(function_info&: functions); |
2584 | addr_t text_base_addr = text_section_sp->GetFileAddress(); |
2585 | size_t count = functions.GetSize(); |
2586 | for (size_t i = 0; i < count; ++i) { |
2587 | const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = |
2588 | functions.GetEntryAtIndex(i); |
2589 | if (func) { |
2590 | FunctionStarts::Entry function_start_entry; |
2591 | function_start_entry.addr = func->base - text_base_addr; |
2592 | if (is_arm) { |
2593 | if (function_start_entry.addr & 1) { |
2594 | function_start_entry.addr &= THUMB_ADDRESS_BIT_MASK; |
2595 | function_start_entry.data = true; |
2596 | } else if (always_thumb) { |
2597 | function_start_entry.data = true; |
2598 | } |
2599 | } |
2600 | function_starts.Append(entry: function_start_entry); |
2601 | } |
2602 | } |
2603 | } |
2604 | } |
2605 | |
2606 | const size_t function_starts_count = function_starts.GetSize(); |
2607 | |
2608 | // For user process binaries (executables, dylibs, frameworks, bundles), if |
2609 | // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're |
2610 | // going to assume the binary has been stripped. Don't allow assembly |
2611 | // language instruction emulation because we don't know proper function |
2612 | // start boundaries. |
2613 | // |
2614 | // For all other types of binaries (kernels, stand-alone bare board |
2615 | // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame |
2616 | // sections - we should not make any assumptions about them based on that. |
2617 | if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { |
2618 | m_allow_assembly_emulation_unwind_plans = false; |
2619 | Log *unwind_or_symbol_log(GetLog(mask: LLDBLog::Symbols | LLDBLog::Unwind)); |
2620 | |
2621 | if (unwind_or_symbol_log) |
2622 | module_sp->LogMessage( |
2623 | log: unwind_or_symbol_log, |
2624 | format: "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds" ); |
2625 | } |
2626 | |
2627 | const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() |
2628 | ? eh_frame_section_sp->GetID() |
2629 | : static_cast<user_id_t>(NO_SECT); |
2630 | |
2631 | uint32_t N_SO_index = UINT32_MAX; |
2632 | |
2633 | MachSymtabSectionInfo section_info(section_list); |
2634 | std::vector<uint32_t> N_FUN_indexes; |
2635 | std::vector<uint32_t> N_NSYM_indexes; |
2636 | std::vector<uint32_t> N_INCL_indexes; |
2637 | std::vector<uint32_t> N_BRAC_indexes; |
2638 | std::vector<uint32_t> N_COMM_indexes; |
2639 | typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; |
2640 | typedef llvm::DenseMap<uint32_t, uint32_t> NListIndexToSymbolIndexMap; |
2641 | typedef llvm::DenseMap<const char *, uint32_t> ConstNameToSymbolIndexMap; |
2642 | ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; |
2643 | ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; |
2644 | ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; |
2645 | // Any symbols that get merged into another will get an entry in this map |
2646 | // so we know |
2647 | NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; |
2648 | uint32_t nlist_idx = 0; |
2649 | Symbol *symbol_ptr = nullptr; |
2650 | |
2651 | uint32_t sym_idx = 0; |
2652 | Symbol *sym = nullptr; |
2653 | size_t num_syms = 0; |
2654 | std::string memory_symbol_name; |
2655 | uint32_t unmapped_local_symbols_found = 0; |
2656 | |
2657 | std::vector<TrieEntryWithOffset> reexport_trie_entries; |
2658 | std::vector<TrieEntryWithOffset> external_sym_trie_entries; |
2659 | std::set<lldb::addr_t> resolver_addresses; |
2660 | |
2661 | const size_t dyld_trie_data_size = dyld_trie_data.GetByteSize(); |
2662 | if (dyld_trie_data_size > 0) { |
2663 | LLDB_LOG(log, "Parsing {0} bytes of dyld trie data" , dyld_trie_data_size); |
2664 | SectionSP text_segment_sp = |
2665 | GetSectionList()->FindSectionByName(section_dstr: GetSegmentNameTEXT()); |
2666 | lldb::addr_t text_segment_file_addr = LLDB_INVALID_ADDRESS; |
2667 | if (text_segment_sp) |
2668 | text_segment_file_addr = text_segment_sp->GetFileAddress(); |
2669 | std::vector<llvm::StringRef> nameSlices; |
2670 | ParseTrieEntries(data&: dyld_trie_data, offset: 0, is_arm, text_seg_base_addr: text_segment_file_addr, |
2671 | nameSlices, resolver_addresses, reexports&: reexport_trie_entries, |
2672 | ext_symbols&: external_sym_trie_entries); |
2673 | } |
2674 | |
2675 | typedef std::set<ConstString> IndirectSymbols; |
2676 | IndirectSymbols indirect_symbol_names; |
2677 | |
2678 | #if TARGET_OS_IPHONE |
2679 | |
2680 | // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been |
2681 | // optimized by moving LOCAL symbols out of the memory mapped portion of |
2682 | // the DSC. The symbol information has all been retained, but it isn't |
2683 | // available in the normal nlist data. However, there *are* duplicate |
2684 | // entries of *some* |
2685 | // LOCAL symbols in the normal nlist data. To handle this situation |
2686 | // correctly, we must first attempt |
2687 | // to parse any DSC unmapped symbol information. If we find any, we set a |
2688 | // flag that tells the normal nlist parser to ignore all LOCAL symbols. |
2689 | |
2690 | if (IsSharedCacheBinary()) { |
2691 | // Before we can start mapping the DSC, we need to make certain the |
2692 | // target process is actually using the cache we can find. |
2693 | |
2694 | // Next we need to determine the correct path for the dyld shared cache. |
2695 | |
2696 | ArchSpec header_arch = GetArchitecture(); |
2697 | |
2698 | UUID dsc_uuid; |
2699 | UUID process_shared_cache_uuid; |
2700 | addr_t process_shared_cache_base_addr; |
2701 | |
2702 | if (process) { |
2703 | GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, |
2704 | process_shared_cache_uuid); |
2705 | } |
2706 | |
2707 | __block bool found_image = false; |
2708 | __block void *nlist_buffer = nullptr; |
2709 | __block unsigned nlist_count = 0; |
2710 | __block char *string_table = nullptr; |
2711 | __block vm_offset_t vm_nlist_memory = 0; |
2712 | __block mach_msg_type_number_t vm_nlist_bytes_read = 0; |
2713 | __block vm_offset_t vm_string_memory = 0; |
2714 | __block mach_msg_type_number_t vm_string_bytes_read = 0; |
2715 | |
2716 | auto _ = llvm::make_scope_exit(^{ |
2717 | if (vm_nlist_memory) |
2718 | vm_deallocate(mach_task_self(), vm_nlist_memory, vm_nlist_bytes_read); |
2719 | if (vm_string_memory) |
2720 | vm_deallocate(mach_task_self(), vm_string_memory, vm_string_bytes_read); |
2721 | }); |
2722 | |
2723 | typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; |
2724 | typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; |
2725 | UndefinedNameToDescMap undefined_name_to_desc; |
2726 | SymbolIndexToName reexport_shlib_needs_fixup; |
2727 | |
2728 | dyld_for_each_installed_shared_cache(^(dyld_shared_cache_t shared_cache) { |
2729 | uuid_t cache_uuid; |
2730 | dyld_shared_cache_copy_uuid(shared_cache, &cache_uuid); |
2731 | if (found_image) |
2732 | return; |
2733 | |
2734 | if (process_shared_cache_uuid.IsValid() && |
2735 | process_shared_cache_uuid != UUID::fromData(&cache_uuid, 16)) |
2736 | return; |
2737 | |
2738 | dyld_shared_cache_for_each_image(shared_cache, ^(dyld_image_t image) { |
2739 | uuid_t dsc_image_uuid; |
2740 | if (found_image) |
2741 | return; |
2742 | |
2743 | dyld_image_copy_uuid(image, &dsc_image_uuid); |
2744 | if (image_uuid != UUID::fromData(dsc_image_uuid, 16)) |
2745 | return; |
2746 | |
2747 | found_image = true; |
2748 | |
2749 | // Compute the size of the string table. We need to ask dyld for a |
2750 | // new SPI to avoid this step. |
2751 | dyld_image_local_nlist_content_4Symbolication( |
2752 | image, ^(const void *nlistStart, uint64_t nlistCount, |
2753 | const char *stringTable) { |
2754 | if (!nlistStart || !nlistCount) |
2755 | return; |
2756 | |
2757 | // The buffers passed here are valid only inside the block. |
2758 | // Use vm_read to make a cheap copy of them available for our |
2759 | // processing later. |
2760 | kern_return_t ret = |
2761 | vm_read(mach_task_self(), (vm_address_t)nlistStart, |
2762 | nlist_byte_size * nlistCount, &vm_nlist_memory, |
2763 | &vm_nlist_bytes_read); |
2764 | if (ret != KERN_SUCCESS) |
2765 | return; |
2766 | assert(vm_nlist_bytes_read == nlist_byte_size * nlistCount); |
2767 | |
2768 | // We don't know the size of the string table. It's cheaper |
2769 | // to map the whole VM region than to determine the size by |
2770 | // parsing all the nlist entries. |
2771 | vm_address_t string_address = (vm_address_t)stringTable; |
2772 | vm_size_t region_size; |
2773 | mach_msg_type_number_t info_count = VM_REGION_BASIC_INFO_COUNT_64; |
2774 | vm_region_basic_info_data_t info; |
2775 | memory_object_name_t object; |
2776 | ret = vm_region_64(mach_task_self(), &string_address, |
2777 | ®ion_size, VM_REGION_BASIC_INFO_64, |
2778 | (vm_region_info_t)&info, &info_count, &object); |
2779 | if (ret != KERN_SUCCESS) |
2780 | return; |
2781 | |
2782 | ret = vm_read(mach_task_self(), (vm_address_t)stringTable, |
2783 | region_size - |
2784 | ((vm_address_t)stringTable - string_address), |
2785 | &vm_string_memory, &vm_string_bytes_read); |
2786 | if (ret != KERN_SUCCESS) |
2787 | return; |
2788 | |
2789 | nlist_buffer = (void *)vm_nlist_memory; |
2790 | string_table = (char *)vm_string_memory; |
2791 | nlist_count = nlistCount; |
2792 | }); |
2793 | }); |
2794 | }); |
2795 | if (nlist_buffer) { |
2796 | DataExtractor dsc_local_symbols_data(nlist_buffer, |
2797 | nlist_count * nlist_byte_size, |
2798 | byte_order, addr_byte_size); |
2799 | unmapped_local_symbols_found = nlist_count; |
2800 | |
2801 | // The normal nlist code cannot correctly size the Symbols |
2802 | // array, we need to allocate it here. |
2803 | sym = symtab.Resize( |
2804 | symtab_load_command.nsyms + m_dysymtab.nindirectsyms + |
2805 | unmapped_local_symbols_found - m_dysymtab.nlocalsym); |
2806 | num_syms = symtab.GetNumSymbols(); |
2807 | |
2808 | lldb::offset_t nlist_data_offset = 0; |
2809 | |
2810 | for (uint32_t nlist_index = 0; |
2811 | nlist_index < nlist_count; |
2812 | nlist_index++) { |
2813 | ///////////////////////////// |
2814 | { |
2815 | std::optional<struct nlist_64> nlist_maybe = |
2816 | ParseNList(dsc_local_symbols_data, nlist_data_offset, |
2817 | nlist_byte_size); |
2818 | if (!nlist_maybe) |
2819 | break; |
2820 | struct nlist_64 nlist = *nlist_maybe; |
2821 | |
2822 | SymbolType type = eSymbolTypeInvalid; |
2823 | const char *symbol_name = string_table + nlist.n_strx; |
2824 | |
2825 | if (symbol_name == NULL) { |
2826 | // No symbol should be NULL, even the symbols with no |
2827 | // string values should have an offset zero which |
2828 | // points to an empty C-string |
2829 | Debugger::ReportError(llvm::formatv( |
2830 | "DSC unmapped local symbol[{0}] has invalid " |
2831 | "string table offset {1:x} in {2}, ignoring symbol" , |
2832 | nlist_index, nlist.n_strx, |
2833 | module_sp->GetFileSpec().GetPath()); |
2834 | continue; |
2835 | } |
2836 | if (symbol_name[0] == '\0') |
2837 | symbol_name = NULL; |
2838 | |
2839 | const char *symbol_name_non_abi_mangled = NULL; |
2840 | |
2841 | SectionSP symbol_section; |
2842 | uint32_t symbol_byte_size = 0; |
2843 | bool add_nlist = true; |
2844 | bool is_debug = ((nlist.n_type & N_STAB) != 0); |
2845 | bool demangled_is_synthesized = false; |
2846 | bool is_gsym = false; |
2847 | bool set_value = true; |
2848 | |
2849 | assert(sym_idx < num_syms); |
2850 | |
2851 | sym[sym_idx].SetDebug(is_debug); |
2852 | |
2853 | if (is_debug) { |
2854 | switch (nlist.n_type) { |
2855 | case N_GSYM: |
2856 | // global symbol: name,,NO_SECT,type,0 |
2857 | // Sometimes the N_GSYM value contains the address. |
2858 | |
2859 | // FIXME: In the .o files, we have a GSYM and a debug |
2860 | // symbol for all the ObjC data. They |
2861 | // have the same address, but we want to ensure that |
2862 | // we always find only the real symbol, 'cause we |
2863 | // don't currently correctly attribute the |
2864 | // GSYM one to the ObjCClass/Ivar/MetaClass |
2865 | // symbol type. This is a temporary hack to make |
2866 | // sure the ObjectiveC symbols get treated correctly. |
2867 | // To do this right, we should coalesce all the GSYM |
2868 | // & global symbols that have the same address. |
2869 | |
2870 | is_gsym = true; |
2871 | sym[sym_idx].SetExternal(true); |
2872 | |
2873 | if (symbol_name && symbol_name[0] == '_' && |
2874 | symbol_name[1] == 'O') { |
2875 | llvm::StringRef symbol_name_ref(symbol_name); |
2876 | if (symbol_name_ref.starts_with( |
2877 | g_objc_v2_prefix_class)) { |
2878 | symbol_name_non_abi_mangled = symbol_name + 1; |
2879 | symbol_name = |
2880 | symbol_name + g_objc_v2_prefix_class.size(); |
2881 | type = eSymbolTypeObjCClass; |
2882 | demangled_is_synthesized = true; |
2883 | |
2884 | } else if (symbol_name_ref.starts_with( |
2885 | g_objc_v2_prefix_metaclass)) { |
2886 | symbol_name_non_abi_mangled = symbol_name + 1; |
2887 | symbol_name = |
2888 | symbol_name + g_objc_v2_prefix_metaclass.size(); |
2889 | type = eSymbolTypeObjCMetaClass; |
2890 | demangled_is_synthesized = true; |
2891 | } else if (symbol_name_ref.starts_with( |
2892 | g_objc_v2_prefix_ivar)) { |
2893 | symbol_name_non_abi_mangled = symbol_name + 1; |
2894 | symbol_name = |
2895 | symbol_name + g_objc_v2_prefix_ivar.size(); |
2896 | type = eSymbolTypeObjCIVar; |
2897 | demangled_is_synthesized = true; |
2898 | } |
2899 | } else { |
2900 | if (nlist.n_value != 0) |
2901 | symbol_section = section_info.GetSection( |
2902 | nlist.n_sect, nlist.n_value); |
2903 | type = eSymbolTypeData; |
2904 | } |
2905 | break; |
2906 | |
2907 | case N_FNAME: |
2908 | // procedure name (f77 kludge): name,,NO_SECT,0,0 |
2909 | type = eSymbolTypeCompiler; |
2910 | break; |
2911 | |
2912 | case N_FUN: |
2913 | // procedure: name,,n_sect,linenumber,address |
2914 | if (symbol_name) { |
2915 | type = eSymbolTypeCode; |
2916 | symbol_section = section_info.GetSection( |
2917 | nlist.n_sect, nlist.n_value); |
2918 | |
2919 | N_FUN_addr_to_sym_idx.insert( |
2920 | std::make_pair(nlist.n_value, sym_idx)); |
2921 | // We use the current number of symbols in the |
2922 | // symbol table in lieu of using nlist_idx in case |
2923 | // we ever start trimming entries out |
2924 | N_FUN_indexes.push_back(sym_idx); |
2925 | } else { |
2926 | type = eSymbolTypeCompiler; |
2927 | |
2928 | if (!N_FUN_indexes.empty()) { |
2929 | // Copy the size of the function into the |
2930 | // original |
2931 | // STAB entry so we don't have |
2932 | // to hunt for it later |
2933 | symtab.SymbolAtIndex(N_FUN_indexes.back()) |
2934 | ->SetByteSize(nlist.n_value); |
2935 | N_FUN_indexes.pop_back(); |
2936 | // We don't really need the end function STAB as |
2937 | // it contains the size which we already placed |
2938 | // with the original symbol, so don't add it if |
2939 | // we want a minimal symbol table |
2940 | add_nlist = false; |
2941 | } |
2942 | } |
2943 | break; |
2944 | |
2945 | case N_STSYM: |
2946 | // static symbol: name,,n_sect,type,address |
2947 | N_STSYM_addr_to_sym_idx.insert( |
2948 | std::make_pair(nlist.n_value, sym_idx)); |
2949 | symbol_section = section_info.GetSection(nlist.n_sect, |
2950 | nlist.n_value); |
2951 | if (symbol_name && symbol_name[0]) { |
2952 | type = ObjectFile::GetSymbolTypeFromName( |
2953 | symbol_name + 1, eSymbolTypeData); |
2954 | } |
2955 | break; |
2956 | |
2957 | case N_LCSYM: |
2958 | // .lcomm symbol: name,,n_sect,type,address |
2959 | symbol_section = section_info.GetSection(nlist.n_sect, |
2960 | nlist.n_value); |
2961 | type = eSymbolTypeCommonBlock; |
2962 | break; |
2963 | |
2964 | case N_BNSYM: |
2965 | // We use the current number of symbols in the symbol |
2966 | // table in lieu of using nlist_idx in case we ever |
2967 | // start trimming entries out Skip these if we want |
2968 | // minimal symbol tables |
2969 | add_nlist = false; |
2970 | break; |
2971 | |
2972 | case N_ENSYM: |
2973 | // Set the size of the N_BNSYM to the terminating |
2974 | // index of this N_ENSYM so that we can always skip |
2975 | // the entire symbol if we need to navigate more |
2976 | // quickly at the source level when parsing STABS |
2977 | // Skip these if we want minimal symbol tables |
2978 | add_nlist = false; |
2979 | break; |
2980 | |
2981 | case N_OPT: |
2982 | // emitted with gcc2_compiled and in gcc source |
2983 | type = eSymbolTypeCompiler; |
2984 | break; |
2985 | |
2986 | case N_RSYM: |
2987 | // register sym: name,,NO_SECT,type,register |
2988 | type = eSymbolTypeVariable; |
2989 | break; |
2990 | |
2991 | case N_SLINE: |
2992 | // src line: 0,,n_sect,linenumber,address |
2993 | symbol_section = section_info.GetSection(nlist.n_sect, |
2994 | nlist.n_value); |
2995 | type = eSymbolTypeLineEntry; |
2996 | break; |
2997 | |
2998 | case N_SSYM: |
2999 | // structure elt: name,,NO_SECT,type,struct_offset |
3000 | type = eSymbolTypeVariableType; |
3001 | break; |
3002 | |
3003 | case N_SO: |
3004 | // source file name |
3005 | type = eSymbolTypeSourceFile; |
3006 | if (symbol_name == NULL) { |
3007 | add_nlist = false; |
3008 | if (N_SO_index != UINT32_MAX) { |
3009 | // Set the size of the N_SO to the terminating |
3010 | // index of this N_SO so that we can always skip |
3011 | // the entire N_SO if we need to navigate more |
3012 | // quickly at the source level when parsing STABS |
3013 | symbol_ptr = symtab.SymbolAtIndex(N_SO_index); |
3014 | symbol_ptr->SetByteSize(sym_idx); |
3015 | symbol_ptr->SetSizeIsSibling(true); |
3016 | } |
3017 | N_NSYM_indexes.clear(); |
3018 | N_INCL_indexes.clear(); |
3019 | N_BRAC_indexes.clear(); |
3020 | N_COMM_indexes.clear(); |
3021 | N_FUN_indexes.clear(); |
3022 | N_SO_index = UINT32_MAX; |
3023 | } else { |
3024 | // We use the current number of symbols in the |
3025 | // symbol table in lieu of using nlist_idx in case |
3026 | // we ever start trimming entries out |
3027 | const bool N_SO_has_full_path = symbol_name[0] == '/'; |
3028 | if (N_SO_has_full_path) { |
3029 | if ((N_SO_index == sym_idx - 1) && |
3030 | ((sym_idx - 1) < num_syms)) { |
3031 | // We have two consecutive N_SO entries where |
3032 | // the first contains a directory and the |
3033 | // second contains a full path. |
3034 | sym[sym_idx - 1].GetMangled().SetValue( |
3035 | ConstString(symbol_name)); |
3036 | m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; |
3037 | add_nlist = false; |
3038 | } else { |
3039 | // This is the first entry in a N_SO that |
3040 | // contains a directory or |
3041 | // a full path to the source file |
3042 | N_SO_index = sym_idx; |
3043 | } |
3044 | } else if ((N_SO_index == sym_idx - 1) && |
3045 | ((sym_idx - 1) < num_syms)) { |
3046 | // This is usually the second N_SO entry that |
3047 | // contains just the filename, so here we combine |
3048 | // it with the first one if we are minimizing the |
3049 | // symbol table |
3050 | const char *so_path = sym[sym_idx - 1] |
3051 | .GetMangled() |
3052 | .GetDemangledName() |
3053 | .AsCString(); |
3054 | if (so_path && so_path[0]) { |
3055 | std::string full_so_path(so_path); |
3056 | const size_t double_slash_pos = |
3057 | full_so_path.find("//" ); |
3058 | if (double_slash_pos != std::string::npos) { |
3059 | // The linker has been generating bad N_SO |
3060 | // entries with doubled up paths |
3061 | // in the format "%s%s" where the first |
3062 | // string in the DW_AT_comp_dir, and the |
3063 | // second is the directory for the source |
3064 | // file so you end up with a path that looks |
3065 | // like "/tmp/src//tmp/src/" |
3066 | FileSpec so_dir(so_path); |
3067 | if (!FileSystem::Instance().Exists(so_dir)) { |
3068 | so_dir.SetFile( |
3069 | &full_so_path[double_slash_pos + 1], |
3070 | FileSpec::Style::native); |
3071 | if (FileSystem::Instance().Exists(so_dir)) { |
3072 | // Trim off the incorrect path |
3073 | full_so_path.erase(0, double_slash_pos + 1); |
3074 | } |
3075 | } |
3076 | } |
3077 | if (*full_so_path.rbegin() != '/') |
3078 | full_so_path += '/'; |
3079 | full_so_path += symbol_name; |
3080 | sym[sym_idx - 1].GetMangled().SetValue( |
3081 | ConstString(full_so_path.c_str())); |
3082 | add_nlist = false; |
3083 | m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; |
3084 | } |
3085 | } else { |
3086 | // This could be a relative path to a N_SO |
3087 | N_SO_index = sym_idx; |
3088 | } |
3089 | } |
3090 | break; |
3091 | |
3092 | case N_OSO: |
3093 | // object file name: name,,0,0,st_mtime |
3094 | type = eSymbolTypeObjectFile; |
3095 | break; |
3096 | |
3097 | case N_LSYM: |
3098 | // local sym: name,,NO_SECT,type,offset |
3099 | type = eSymbolTypeLocal; |
3100 | break; |
3101 | |
3102 | // INCL scopes |
3103 | case N_BINCL: |
3104 | // include file beginning: name,,NO_SECT,0,sum We use |
3105 | // the current number of symbols in the symbol table |
3106 | // in lieu of using nlist_idx in case we ever start |
3107 | // trimming entries out |
3108 | N_INCL_indexes.push_back(sym_idx); |
3109 | type = eSymbolTypeScopeBegin; |
3110 | break; |
3111 | |
3112 | case N_EINCL: |
3113 | // include file end: name,,NO_SECT,0,0 |
3114 | // Set the size of the N_BINCL to the terminating |
3115 | // index of this N_EINCL so that we can always skip |
3116 | // the entire symbol if we need to navigate more |
3117 | // quickly at the source level when parsing STABS |
3118 | if (!N_INCL_indexes.empty()) { |
3119 | symbol_ptr = |
3120 | symtab.SymbolAtIndex(N_INCL_indexes.back()); |
3121 | symbol_ptr->SetByteSize(sym_idx + 1); |
3122 | symbol_ptr->SetSizeIsSibling(true); |
3123 | N_INCL_indexes.pop_back(); |
3124 | } |
3125 | type = eSymbolTypeScopeEnd; |
3126 | break; |
3127 | |
3128 | case N_SOL: |
3129 | // #included file name: name,,n_sect,0,address |
3130 | type = eSymbolTypeHeaderFile; |
3131 | |
3132 | // We currently don't use the header files on darwin |
3133 | add_nlist = false; |
3134 | break; |
3135 | |
3136 | case N_PARAMS: |
3137 | // compiler parameters: name,,NO_SECT,0,0 |
3138 | type = eSymbolTypeCompiler; |
3139 | break; |
3140 | |
3141 | case N_VERSION: |
3142 | // compiler version: name,,NO_SECT,0,0 |
3143 | type = eSymbolTypeCompiler; |
3144 | break; |
3145 | |
3146 | case N_OLEVEL: |
3147 | // compiler -O level: name,,NO_SECT,0,0 |
3148 | type = eSymbolTypeCompiler; |
3149 | break; |
3150 | |
3151 | case N_PSYM: |
3152 | // parameter: name,,NO_SECT,type,offset |
3153 | type = eSymbolTypeVariable; |
3154 | break; |
3155 | |
3156 | case N_ENTRY: |
3157 | // alternate entry: name,,n_sect,linenumber,address |
3158 | symbol_section = section_info.GetSection(nlist.n_sect, |
3159 | nlist.n_value); |
3160 | type = eSymbolTypeLineEntry; |
3161 | break; |
3162 | |
3163 | // Left and Right Braces |
3164 | case N_LBRAC: |
3165 | // left bracket: 0,,NO_SECT,nesting level,address We |
3166 | // use the current number of symbols in the symbol |
3167 | // table in lieu of using nlist_idx in case we ever |
3168 | // start trimming entries out |
3169 | symbol_section = section_info.GetSection(nlist.n_sect, |
3170 | nlist.n_value); |
3171 | N_BRAC_indexes.push_back(sym_idx); |
3172 | type = eSymbolTypeScopeBegin; |
3173 | break; |
3174 | |
3175 | case N_RBRAC: |
3176 | // right bracket: 0,,NO_SECT,nesting level,address |
3177 | // Set the size of the N_LBRAC to the terminating |
3178 | // index of this N_RBRAC so that we can always skip |
3179 | // the entire symbol if we need to navigate more |
3180 | // quickly at the source level when parsing STABS |
3181 | symbol_section = section_info.GetSection(nlist.n_sect, |
3182 | nlist.n_value); |
3183 | if (!N_BRAC_indexes.empty()) { |
3184 | symbol_ptr = |
3185 | symtab.SymbolAtIndex(N_BRAC_indexes.back()); |
3186 | symbol_ptr->SetByteSize(sym_idx + 1); |
3187 | symbol_ptr->SetSizeIsSibling(true); |
3188 | N_BRAC_indexes.pop_back(); |
3189 | } |
3190 | type = eSymbolTypeScopeEnd; |
3191 | break; |
3192 | |
3193 | case N_EXCL: |
3194 | // deleted include file: name,,NO_SECT,0,sum |
3195 | type = eSymbolTypeHeaderFile; |
3196 | break; |
3197 | |
3198 | // COMM scopes |
3199 | case N_BCOMM: |
3200 | // begin common: name,,NO_SECT,0,0 |
3201 | // We use the current number of symbols in the symbol |
3202 | // table in lieu of using nlist_idx in case we ever |
3203 | // start trimming entries out |
3204 | type = eSymbolTypeScopeBegin; |
3205 | N_COMM_indexes.push_back(sym_idx); |
3206 | break; |
3207 | |
3208 | case N_ECOML: |
3209 | // end common (local name): 0,,n_sect,0,address |
3210 | symbol_section = section_info.GetSection(nlist.n_sect, |
3211 | nlist.n_value); |
3212 | // Fall through |
3213 | |
3214 | case N_ECOMM: |
3215 | // end common: name,,n_sect,0,0 |
3216 | // Set the size of the N_BCOMM to the terminating |
3217 | // index of this N_ECOMM/N_ECOML so that we can |
3218 | // always skip the entire symbol if we need to |
3219 | // navigate more quickly at the source level when |
3220 | // parsing STABS |
3221 | if (!N_COMM_indexes.empty()) { |
3222 | symbol_ptr = |
3223 | symtab.SymbolAtIndex(N_COMM_indexes.back()); |
3224 | symbol_ptr->SetByteSize(sym_idx + 1); |
3225 | symbol_ptr->SetSizeIsSibling(true); |
3226 | N_COMM_indexes.pop_back(); |
3227 | } |
3228 | type = eSymbolTypeScopeEnd; |
3229 | break; |
3230 | |
3231 | case N_LENG: |
3232 | // second stab entry with length information |
3233 | type = eSymbolTypeAdditional; |
3234 | break; |
3235 | |
3236 | default: |
3237 | break; |
3238 | } |
3239 | } else { |
3240 | // uint8_t n_pext = N_PEXT & nlist.n_type; |
3241 | uint8_t n_type = N_TYPE & nlist.n_type; |
3242 | sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); |
3243 | |
3244 | switch (n_type) { |
3245 | case N_INDR: { |
3246 | const char *reexport_name_cstr = |
3247 | strtab_data.PeekCStr(nlist.n_value); |
3248 | if (reexport_name_cstr && reexport_name_cstr[0]) { |
3249 | type = eSymbolTypeReExported; |
3250 | ConstString reexport_name( |
3251 | reexport_name_cstr + |
3252 | ((reexport_name_cstr[0] == '_') ? 1 : 0)); |
3253 | sym[sym_idx].SetReExportedSymbolName(reexport_name); |
3254 | set_value = false; |
3255 | reexport_shlib_needs_fixup[sym_idx] = reexport_name; |
3256 | indirect_symbol_names.insert(ConstString( |
3257 | symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); |
3258 | } else |
3259 | type = eSymbolTypeUndefined; |
3260 | } break; |
3261 | |
3262 | case N_UNDF: |
3263 | if (symbol_name && symbol_name[0]) { |
3264 | ConstString undefined_name( |
3265 | symbol_name + ((symbol_name[0] == '_') ? 1 : 0)); |
3266 | undefined_name_to_desc[undefined_name] = nlist.n_desc; |
3267 | } |
3268 | // Fall through |
3269 | case N_PBUD: |
3270 | type = eSymbolTypeUndefined; |
3271 | break; |
3272 | |
3273 | case N_ABS: |
3274 | type = eSymbolTypeAbsolute; |
3275 | break; |
3276 | |
3277 | case N_SECT: { |
3278 | symbol_section = section_info.GetSection(nlist.n_sect, |
3279 | nlist.n_value); |
3280 | |
3281 | if (symbol_section == NULL) { |
3282 | // TODO: warn about this? |
3283 | add_nlist = false; |
3284 | break; |
3285 | } |
3286 | |
3287 | if (TEXT_eh_frame_sectID == nlist.n_sect) { |
3288 | type = eSymbolTypeException; |
3289 | } else { |
3290 | uint32_t section_type = |
3291 | symbol_section->Get() & SECTION_TYPE; |
3292 | |
3293 | switch (section_type) { |
3294 | case S_CSTRING_LITERALS: |
3295 | type = eSymbolTypeData; |
3296 | break; // section with only literal C strings |
3297 | case S_4BYTE_LITERALS: |
3298 | type = eSymbolTypeData; |
3299 | break; // section with only 4 byte literals |
3300 | case S_8BYTE_LITERALS: |
3301 | type = eSymbolTypeData; |
3302 | break; // section with only 8 byte literals |
3303 | case S_LITERAL_POINTERS: |
3304 | type = eSymbolTypeTrampoline; |
3305 | break; // section with only pointers to literals |
3306 | case S_NON_LAZY_SYMBOL_POINTERS: |
3307 | type = eSymbolTypeTrampoline; |
3308 | break; // section with only non-lazy symbol |
3309 | // pointers |
3310 | case S_LAZY_SYMBOL_POINTERS: |
3311 | type = eSymbolTypeTrampoline; |
3312 | break; // section with only lazy symbol pointers |
3313 | case S_SYMBOL_STUBS: |
3314 | type = eSymbolTypeTrampoline; |
3315 | break; // section with only symbol stubs, byte |
3316 | // size of stub in the reserved2 field |
3317 | case S_MOD_INIT_FUNC_POINTERS: |
3318 | type = eSymbolTypeCode; |
3319 | break; // section with only function pointers for |
3320 | // initialization |
3321 | case S_MOD_TERM_FUNC_POINTERS: |
3322 | type = eSymbolTypeCode; |
3323 | break; // section with only function pointers for |
3324 | // termination |
3325 | case S_INTERPOSING: |
3326 | type = eSymbolTypeTrampoline; |
3327 | break; // section with only pairs of function |
3328 | // pointers for interposing |
3329 | case S_16BYTE_LITERALS: |
3330 | type = eSymbolTypeData; |
3331 | break; // section with only 16 byte literals |
3332 | case S_DTRACE_DOF: |
3333 | type = eSymbolTypeInstrumentation; |
3334 | break; |
3335 | case S_LAZY_DYLIB_SYMBOL_POINTERS: |
3336 | type = eSymbolTypeTrampoline; |
3337 | break; |
3338 | default: |
3339 | switch (symbol_section->GetType()) { |
3340 | case lldb::eSectionTypeCode: |
3341 | type = eSymbolTypeCode; |
3342 | break; |
3343 | case eSectionTypeData: |
3344 | case eSectionTypeDataCString: // Inlined C string |
3345 | // data |
3346 | case eSectionTypeDataCStringPointers: // Pointers |
3347 | // to C |
3348 | // string |
3349 | // data |
3350 | case eSectionTypeDataSymbolAddress: // Address of |
3351 | // a symbol in |
3352 | // the symbol |
3353 | // table |
3354 | case eSectionTypeData4: |
3355 | case eSectionTypeData8: |
3356 | case eSectionTypeData16: |
3357 | type = eSymbolTypeData; |
3358 | break; |
3359 | default: |
3360 | break; |
3361 | } |
3362 | break; |
3363 | } |
3364 | |
3365 | if (type == eSymbolTypeInvalid) { |
3366 | const char *symbol_sect_name = |
3367 | symbol_section->GetName().AsCString(); |
3368 | if (symbol_section->IsDescendant( |
3369 | text_section_sp.get())) { |
3370 | if (symbol_section->IsClear( |
3371 | S_ATTR_PURE_INSTRUCTIONS | |
3372 | S_ATTR_SELF_MODIFYING_CODE | |
3373 | S_ATTR_SOME_INSTRUCTIONS)) |
3374 | type = eSymbolTypeData; |
3375 | else |
3376 | type = eSymbolTypeCode; |
3377 | } else if (symbol_section->IsDescendant( |
3378 | data_section_sp.get()) || |
3379 | symbol_section->IsDescendant( |
3380 | data_dirty_section_sp.get()) || |
3381 | symbol_section->IsDescendant( |
3382 | data_const_section_sp.get())) { |
3383 | if (symbol_sect_name && |
3384 | ::strstr(symbol_sect_name, "__objc" ) == |
3385 | symbol_sect_name) { |
3386 | type = eSymbolTypeRuntime; |
3387 | |
3388 | if (symbol_name) { |
3389 | llvm::StringRef symbol_name_ref(symbol_name); |
3390 | if (symbol_name_ref.starts_with("_OBJC_" )) { |
3391 | llvm::StringRef |
3392 | g_objc_v2_prefix_class( |
3393 | "_OBJC_CLASS_$_" ); |
3394 | llvm::StringRef |
3395 | g_objc_v2_prefix_metaclass( |
3396 | "_OBJC_METACLASS_$_" ); |
3397 | llvm::StringRef |
3398 | g_objc_v2_prefix_ivar("_OBJC_IVAR_$_" ); |
3399 | if (symbol_name_ref.starts_with( |
3400 | g_objc_v2_prefix_class)) { |
3401 | symbol_name_non_abi_mangled = |
3402 | symbol_name + 1; |
3403 | symbol_name = |
3404 | symbol_name + |
3405 | g_objc_v2_prefix_class.size(); |
3406 | type = eSymbolTypeObjCClass; |
3407 | demangled_is_synthesized = true; |
3408 | } else if ( |
3409 | symbol_name_ref.starts_with( |
3410 | g_objc_v2_prefix_metaclass)) { |
3411 | symbol_name_non_abi_mangled = |
3412 | symbol_name + 1; |
3413 | symbol_name = |
3414 | symbol_name + |
3415 | g_objc_v2_prefix_metaclass.size(); |
3416 | type = eSymbolTypeObjCMetaClass; |
3417 | demangled_is_synthesized = true; |
3418 | } else if (symbol_name_ref.starts_with( |
3419 | g_objc_v2_prefix_ivar)) { |
3420 | symbol_name_non_abi_mangled = |
3421 | symbol_name + 1; |
3422 | symbol_name = |
3423 | symbol_name + |
3424 | g_objc_v2_prefix_ivar.size(); |
3425 | type = eSymbolTypeObjCIVar; |
3426 | demangled_is_synthesized = true; |
3427 | } |
3428 | } |
3429 | } |
3430 | } else if (symbol_sect_name && |
3431 | ::strstr(symbol_sect_name, |
3432 | "__gcc_except_tab" ) == |
3433 | symbol_sect_name) { |
3434 | type = eSymbolTypeException; |
3435 | } else { |
3436 | type = eSymbolTypeData; |
3437 | } |
3438 | } else if (symbol_sect_name && |
3439 | ::strstr(symbol_sect_name, "__IMPORT" ) == |
3440 | symbol_sect_name) { |
3441 | type = eSymbolTypeTrampoline; |
3442 | } else if (symbol_section->IsDescendant( |
3443 | objc_section_sp.get())) { |
3444 | type = eSymbolTypeRuntime; |
3445 | if (symbol_name && symbol_name[0] == '.') { |
3446 | llvm::StringRef symbol_name_ref(symbol_name); |
3447 | llvm::StringRef |
3448 | g_objc_v1_prefix_class(".objc_class_name_" ); |
3449 | if (symbol_name_ref.starts_with( |
3450 | g_objc_v1_prefix_class)) { |
3451 | symbol_name_non_abi_mangled = symbol_name; |
3452 | symbol_name = symbol_name + |
3453 | g_objc_v1_prefix_class.size(); |
3454 | type = eSymbolTypeObjCClass; |
3455 | demangled_is_synthesized = true; |
3456 | } |
3457 | } |
3458 | } |
3459 | } |
3460 | } |
3461 | } break; |
3462 | } |
3463 | } |
3464 | |
3465 | if (add_nlist) { |
3466 | uint64_t symbol_value = nlist.n_value; |
3467 | if (symbol_name_non_abi_mangled) { |
3468 | sym[sym_idx].GetMangled().SetMangledName( |
3469 | ConstString(symbol_name_non_abi_mangled)); |
3470 | sym[sym_idx].GetMangled().SetDemangledName( |
3471 | ConstString(symbol_name)); |
3472 | } else { |
3473 | if (symbol_name && symbol_name[0] == '_') { |
3474 | symbol_name++; // Skip the leading underscore |
3475 | } |
3476 | |
3477 | if (symbol_name) { |
3478 | ConstString const_symbol_name(symbol_name); |
3479 | sym[sym_idx].GetMangled().SetValue(const_symbol_name); |
3480 | if (is_gsym && is_debug) { |
3481 | const char *gsym_name = |
3482 | sym[sym_idx] |
3483 | .GetMangled() |
3484 | .GetName(Mangled::ePreferMangled) |
3485 | .GetCString(); |
3486 | if (gsym_name) |
3487 | N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; |
3488 | } |
3489 | } |
3490 | } |
3491 | if (symbol_section) { |
3492 | const addr_t section_file_addr = |
3493 | symbol_section->GetFileAddress(); |
3494 | if (symbol_byte_size == 0 && |
3495 | function_starts_count > 0) { |
3496 | addr_t symbol_lookup_file_addr = nlist.n_value; |
3497 | // Do an exact address match for non-ARM addresses, |
3498 | // else get the closest since the symbol might be a |
3499 | // thumb symbol which has an address with bit zero |
3500 | // set |
3501 | FunctionStarts::Entry *func_start_entry = |
3502 | function_starts.FindEntry(symbol_lookup_file_addr, |
3503 | !is_arm); |
3504 | if (is_arm && func_start_entry) { |
3505 | // Verify that the function start address is the |
3506 | // symbol address (ARM) or the symbol address + 1 |
3507 | // (thumb) |
3508 | if (func_start_entry->addr != |
3509 | symbol_lookup_file_addr && |
3510 | func_start_entry->addr != |
3511 | (symbol_lookup_file_addr + 1)) { |
3512 | // Not the right entry, NULL it out... |
3513 | func_start_entry = NULL; |
3514 | } |
3515 | } |
3516 | if (func_start_entry) { |
3517 | func_start_entry->data = true; |
3518 | |
3519 | addr_t symbol_file_addr = func_start_entry->addr; |
3520 | uint32_t symbol_flags = 0; |
3521 | if (is_arm) { |
3522 | if (symbol_file_addr & 1) |
3523 | symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; |
3524 | symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; |
3525 | } |
3526 | |
3527 | const FunctionStarts::Entry *next_func_start_entry = |
3528 | function_starts.FindNextEntry(func_start_entry); |
3529 | const addr_t section_end_file_addr = |
3530 | section_file_addr + |
3531 | symbol_section->GetByteSize(); |
3532 | if (next_func_start_entry) { |
3533 | addr_t next_symbol_file_addr = |
3534 | next_func_start_entry->addr; |
3535 | // Be sure the clear the Thumb address bit when |
3536 | // we calculate the size from the current and |
3537 | // next address |
3538 | if (is_arm) |
3539 | next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; |
3540 | symbol_byte_size = std::min<lldb::addr_t>( |
3541 | next_symbol_file_addr - symbol_file_addr, |
3542 | section_end_file_addr - symbol_file_addr); |
3543 | } else { |
3544 | symbol_byte_size = |
3545 | section_end_file_addr - symbol_file_addr; |
3546 | } |
3547 | } |
3548 | } |
3549 | symbol_value -= section_file_addr; |
3550 | } |
3551 | |
3552 | if (is_debug == false) { |
3553 | if (type == eSymbolTypeCode) { |
3554 | // See if we can find a N_FUN entry for any code |
3555 | // symbols. If we do find a match, and the name |
3556 | // matches, then we can merge the two into just the |
3557 | // function symbol to avoid duplicate entries in |
3558 | // the symbol table |
3559 | auto range = |
3560 | N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); |
3561 | if (range.first != range.second) { |
3562 | bool found_it = false; |
3563 | for (auto pos = range.first; pos != range.second; |
3564 | ++pos) { |
3565 | if (sym[sym_idx].GetMangled().GetName( |
3566 | Mangled::ePreferMangled) == |
3567 | sym[pos->second].GetMangled().GetName( |
3568 | Mangled::ePreferMangled)) { |
3569 | m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; |
3570 | // We just need the flags from the linker |
3571 | // symbol, so put these flags |
3572 | // into the N_FUN flags to avoid duplicate |
3573 | // symbols in the symbol table |
3574 | sym[pos->second].SetExternal( |
3575 | sym[sym_idx].IsExternal()); |
3576 | sym[pos->second].SetFlags(nlist.n_type << 16 | |
3577 | nlist.n_desc); |
3578 | if (resolver_addresses.find(nlist.n_value) != |
3579 | resolver_addresses.end()) |
3580 | sym[pos->second].SetType(eSymbolTypeResolver); |
3581 | sym[sym_idx].Clear(); |
3582 | found_it = true; |
3583 | break; |
3584 | } |
3585 | } |
3586 | if (found_it) |
3587 | continue; |
3588 | } else { |
3589 | if (resolver_addresses.find(nlist.n_value) != |
3590 | resolver_addresses.end()) |
3591 | type = eSymbolTypeResolver; |
3592 | } |
3593 | } else if (type == eSymbolTypeData || |
3594 | type == eSymbolTypeObjCClass || |
3595 | type == eSymbolTypeObjCMetaClass || |
3596 | type == eSymbolTypeObjCIVar) { |
3597 | // See if we can find a N_STSYM entry for any data |
3598 | // symbols. If we do find a match, and the name |
3599 | // matches, then we can merge the two into just the |
3600 | // Static symbol to avoid duplicate entries in the |
3601 | // symbol table |
3602 | auto range = N_STSYM_addr_to_sym_idx.equal_range( |
3603 | nlist.n_value); |
3604 | if (range.first != range.second) { |
3605 | bool found_it = false; |
3606 | for (auto pos = range.first; pos != range.second; |
3607 | ++pos) { |
3608 | if (sym[sym_idx].GetMangled().GetName( |
3609 | Mangled::ePreferMangled) == |
3610 | sym[pos->second].GetMangled().GetName( |
3611 | Mangled::ePreferMangled)) { |
3612 | m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; |
3613 | // We just need the flags from the linker |
3614 | // symbol, so put these flags |
3615 | // into the N_STSYM flags to avoid duplicate |
3616 | // symbols in the symbol table |
3617 | sym[pos->second].SetExternal( |
3618 | sym[sym_idx].IsExternal()); |
3619 | sym[pos->second].SetFlags(nlist.n_type << 16 | |
3620 | nlist.n_desc); |
3621 | sym[sym_idx].Clear(); |
3622 | found_it = true; |
3623 | break; |
3624 | } |
3625 | } |
3626 | if (found_it) |
3627 | continue; |
3628 | } else { |
3629 | const char *gsym_name = |
3630 | sym[sym_idx] |
3631 | .GetMangled() |
3632 | .GetName(Mangled::ePreferMangled) |
3633 | .GetCString(); |
3634 | if (gsym_name) { |
3635 | // Combine N_GSYM stab entries with the non |
3636 | // stab symbol |
3637 | ConstNameToSymbolIndexMap::const_iterator pos = |
3638 | N_GSYM_name_to_sym_idx.find(gsym_name); |
3639 | if (pos != N_GSYM_name_to_sym_idx.end()) { |
3640 | const uint32_t GSYM_sym_idx = pos->second; |
3641 | m_nlist_idx_to_sym_idx[nlist_idx] = |
3642 | GSYM_sym_idx; |
3643 | // Copy the address, because often the N_GSYM |
3644 | // address has an invalid address of zero |
3645 | // when the global is a common symbol |
3646 | sym[GSYM_sym_idx].GetAddressRef().SetSection( |
3647 | symbol_section); |
3648 | sym[GSYM_sym_idx].GetAddressRef().SetOffset( |
3649 | symbol_value); |
3650 | add_symbol_addr(sym[GSYM_sym_idx] |
3651 | .GetAddress() |
3652 | .GetFileAddress()); |
3653 | // We just need the flags from the linker |
3654 | // symbol, so put these flags |
3655 | // into the N_GSYM flags to avoid duplicate |
3656 | // symbols in the symbol table |
3657 | sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | |
3658 | nlist.n_desc); |
3659 | sym[sym_idx].Clear(); |
3660 | continue; |
3661 | } |
3662 | } |
3663 | } |
3664 | } |
3665 | } |
3666 | |
3667 | sym[sym_idx].SetID(nlist_idx); |
3668 | sym[sym_idx].SetType(type); |
3669 | if (set_value) { |
3670 | sym[sym_idx].GetAddressRef().SetSection(symbol_section); |
3671 | sym[sym_idx].GetAddressRef().SetOffset(symbol_value); |
3672 | add_symbol_addr( |
3673 | sym[sym_idx].GetAddress().GetFileAddress()); |
3674 | } |
3675 | sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); |
3676 | |
3677 | if (symbol_byte_size > 0) |
3678 | sym[sym_idx].SetByteSize(symbol_byte_size); |
3679 | |
3680 | if (demangled_is_synthesized) |
3681 | sym[sym_idx].SetDemangledNameIsSynthesized(true); |
3682 | ++sym_idx; |
3683 | } else { |
3684 | sym[sym_idx].Clear(); |
3685 | } |
3686 | } |
3687 | ///////////////////////////// |
3688 | } |
3689 | } |
3690 | |
3691 | for (const auto &pos : reexport_shlib_needs_fixup) { |
3692 | const auto undef_pos = undefined_name_to_desc.find(pos.second); |
3693 | if (undef_pos != undefined_name_to_desc.end()) { |
3694 | const uint8_t dylib_ordinal = |
3695 | llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); |
3696 | if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) |
3697 | sym[pos.first].SetReExportedSymbolSharedLibrary( |
3698 | dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); |
3699 | } |
3700 | } |
3701 | } |
3702 | |
3703 | #endif |
3704 | lldb::offset_t nlist_data_offset = 0; |
3705 | |
3706 | if (nlist_data.GetByteSize() > 0) { |
3707 | |
3708 | // If the sym array was not created while parsing the DSC unmapped |
3709 | // symbols, create it now. |
3710 | if (sym == nullptr) { |
3711 | sym = |
3712 | symtab.Resize(count: symtab_load_command.nsyms + m_dysymtab.nindirectsyms); |
3713 | num_syms = symtab.GetNumSymbols(); |
3714 | } |
3715 | |
3716 | if (unmapped_local_symbols_found) { |
3717 | assert(m_dysymtab.ilocalsym == 0); |
3718 | nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); |
3719 | nlist_idx = m_dysymtab.nlocalsym; |
3720 | } else { |
3721 | nlist_idx = 0; |
3722 | } |
3723 | |
3724 | typedef llvm::DenseMap<ConstString, uint16_t> UndefinedNameToDescMap; |
3725 | typedef llvm::DenseMap<uint32_t, ConstString> SymbolIndexToName; |
3726 | UndefinedNameToDescMap undefined_name_to_desc; |
3727 | SymbolIndexToName reexport_shlib_needs_fixup; |
3728 | |
3729 | // Symtab parsing is a huge mess. Everything is entangled and the code |
3730 | // requires access to a ridiculous amount of variables. LLDB depends |
3731 | // heavily on the proper merging of symbols and to get that right we need |
3732 | // to make sure we have parsed all the debug symbols first. Therefore we |
3733 | // invoke the lambda twice, once to parse only the debug symbols and then |
3734 | // once more to parse the remaining symbols. |
3735 | auto ParseSymbolLambda = [&](struct nlist_64 &nlist, uint32_t nlist_idx, |
3736 | bool debug_only) { |
3737 | const bool is_debug = ((nlist.n_type & N_STAB) != 0); |
3738 | if (is_debug != debug_only) |
3739 | return true; |
3740 | |
3741 | const char *symbol_name_non_abi_mangled = nullptr; |
3742 | const char *symbol_name = nullptr; |
3743 | |
3744 | if (have_strtab_data) { |
3745 | symbol_name = strtab_data.PeekCStr(offset: nlist.n_strx); |
3746 | |
3747 | if (symbol_name == nullptr) { |
3748 | // No symbol should be NULL, even the symbols with no string values |
3749 | // should have an offset zero which points to an empty C-string |
3750 | Debugger::ReportError(message: llvm::formatv( |
3751 | Fmt: "symbol[{0}] has invalid string table offset {1:x} in {2}, " |
3752 | "ignoring symbol" , |
3753 | Vals&: nlist_idx, Vals&: nlist.n_strx, Vals: module_sp->GetFileSpec().GetPath())); |
3754 | return true; |
3755 | } |
3756 | if (symbol_name[0] == '\0') |
3757 | symbol_name = nullptr; |
3758 | } else { |
3759 | const addr_t str_addr = strtab_addr + nlist.n_strx; |
3760 | Status str_error; |
3761 | if (process->ReadCStringFromMemory(vm_addr: str_addr, out_str&: memory_symbol_name, |
3762 | error&: str_error)) |
3763 | symbol_name = memory_symbol_name.c_str(); |
3764 | } |
3765 | |
3766 | SymbolType type = eSymbolTypeInvalid; |
3767 | SectionSP symbol_section; |
3768 | lldb::addr_t symbol_byte_size = 0; |
3769 | bool add_nlist = true; |
3770 | bool is_gsym = false; |
3771 | bool demangled_is_synthesized = false; |
3772 | bool set_value = true; |
3773 | |
3774 | assert(sym_idx < num_syms); |
3775 | sym[sym_idx].SetDebug(is_debug); |
3776 | |
3777 | if (is_debug) { |
3778 | switch (nlist.n_type) { |
3779 | case N_GSYM: |
3780 | // global symbol: name,,NO_SECT,type,0 |
3781 | // Sometimes the N_GSYM value contains the address. |
3782 | |
3783 | // FIXME: In the .o files, we have a GSYM and a debug symbol for all |
3784 | // the ObjC data. They |
3785 | // have the same address, but we want to ensure that we always find |
3786 | // only the real symbol, 'cause we don't currently correctly |
3787 | // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol |
3788 | // type. This is a temporary hack to make sure the ObjectiveC |
3789 | // symbols get treated correctly. To do this right, we should |
3790 | // coalesce all the GSYM & global symbols that have the same |
3791 | // address. |
3792 | is_gsym = true; |
3793 | sym[sym_idx].SetExternal(true); |
3794 | |
3795 | if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { |
3796 | llvm::StringRef symbol_name_ref(symbol_name); |
3797 | if (symbol_name_ref.starts_with(Prefix: g_objc_v2_prefix_class)) { |
3798 | symbol_name_non_abi_mangled = symbol_name + 1; |
3799 | symbol_name = symbol_name + g_objc_v2_prefix_class.size(); |
3800 | type = eSymbolTypeObjCClass; |
3801 | demangled_is_synthesized = true; |
3802 | |
3803 | } else if (symbol_name_ref.starts_with( |
3804 | Prefix: g_objc_v2_prefix_metaclass)) { |
3805 | symbol_name_non_abi_mangled = symbol_name + 1; |
3806 | symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); |
3807 | type = eSymbolTypeObjCMetaClass; |
3808 | demangled_is_synthesized = true; |
3809 | } else if (symbol_name_ref.starts_with(Prefix: g_objc_v2_prefix_ivar)) { |
3810 | symbol_name_non_abi_mangled = symbol_name + 1; |
3811 | symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); |
3812 | type = eSymbolTypeObjCIVar; |
3813 | demangled_is_synthesized = true; |
3814 | } |
3815 | } else { |
3816 | if (nlist.n_value != 0) |
3817 | symbol_section = |
3818 | section_info.GetSection(n_sect: nlist.n_sect, file_addr: nlist.n_value); |
3819 | type = eSymbolTypeData; |
3820 | } |
3821 | break; |
3822 | |
3823 | case N_FNAME: |
3824 | // procedure name (f77 kludge): name,,NO_SECT,0,0 |
3825 | type = eSymbolTypeCompiler; |
3826 | break; |
3827 | |
3828 | case N_FUN: |
3829 | // procedure: name,,n_sect,linenumber,address |
3830 | if (symbol_name) { |
3831 | type = eSymbolTypeCode; |
3832 | symbol_section = |
3833 | section_info.GetSection(n_sect: nlist.n_sect, file_addr: nlist.n_value); |
3834 | |
3835 | N_FUN_addr_to_sym_idx.insert( |
3836 | x: std::make_pair(x&: nlist.n_value, y&: sym_idx)); |
3837 | // We use the current number of symbols in the symbol table in |
3838 | // lieu of using nlist_idx in case we ever start trimming entries |
3839 | // out |
3840 | N_FUN_indexes.push_back(x: sym_idx); |
3841 | } else { |
3842 | type = eSymbolTypeCompiler; |
3843 | |
3844 | if (!N_FUN_indexes.empty()) { |
3845 | // Copy the size of the function into the original STAB entry |
3846 | // so we don't have to hunt for it later |
3847 | symtab.SymbolAtIndex(idx: N_FUN_indexes.back()) |
3848 | ->SetByteSize(nlist.n_value); |
3849 | N_FUN_indexes.pop_back(); |
3850 | // We don't really need the end function STAB as it contains |
3851 | // the size which we already placed with the original symbol, |
3852 | // so don't add it if we want a minimal symbol table |
3853 | add_nlist = false; |
3854 | } |
3855 | } |
3856 | break; |
3857 | |
3858 | case N_STSYM: |
3859 | // static symbol: name,,n_sect,type,address |
3860 | N_STSYM_addr_to_sym_idx.insert( |
3861 | x: std::make_pair(x&: nlist.n_value, y&: sym_idx)); |
3862 | symbol_section = section_info.GetSection(n_sect: nlist.n_sect, file_addr: nlist.n_value); |
3863 | if (symbol_name && symbol_name[0]) { |
3864 | type = ObjectFile::GetSymbolTypeFromName(name: symbol_name + 1, |
3865 | symbol_type_hint: eSymbolTypeData); |
3866 | } |
3867 | break; |
3868 | |
3869 | case N_LCSYM: |
3870 | // .lcomm symbol: name,,n_sect,type,address |
3871 | symbol_section = section_info.GetSection(n_sect: nlist.n_sect, file_addr: nlist.n_value); |
3872 | type = eSymbolTypeCommonBlock; |
3873 | break; |
3874 | |
3875 | case N_BNSYM: |
3876 | // We use the current number of symbols in the symbol table in lieu |
3877 | // of using nlist_idx in case we ever start trimming entries out |
3878 | // Skip these if we want minimal symbol tables |
3879 | add_nlist = false; |
3880 | break; |
3881 | |
3882 | case N_ENSYM: |
3883 | // Set the size of the N_BNSYM to the terminating index of this |
3884 | // N_ENSYM so that we can always skip the entire symbol if we need |
3885 | // to navigate more quickly at the source level when parsing STABS |
3886 | // Skip these if we want minimal symbol tables |
3887 | add_nlist = false; |
3888 | break; |
3889 | |
3890 | case N_OPT: |
3891 | // emitted with gcc2_compiled and in gcc source |
3892 | type = eSymbolTypeCompiler; |
3893 | break; |
3894 | |
3895 | case N_RSYM: |
3896 | // register sym: name,,NO_SECT,type,register |
3897 | type = eSymbolTypeVariable; |
3898 | break; |
3899 | |
3900 | case N_SLINE: |
3901 | // src line: 0,,n_sect,linenumber,address |
3902 | symbol_section = section_info.GetSection(n_sect: nlist.n_sect, file_addr: nlist.n_value); |
3903 | type = eSymbolTypeLineEntry; |
3904 | break; |
3905 | |
3906 | case N_SSYM: |
3907 | // structure elt: name,,NO_SECT,type,struct_offset |
3908 | type = eSymbolTypeVariableType; |
3909 | break; |
3910 | |
3911 | case N_SO: |
3912 | // source file name |
3913 | type = eSymbolTypeSourceFile; |
3914 | if (symbol_name == nullptr) { |
3915 | add_nlist = false; |
3916 | if (N_SO_index != UINT32_MAX) { |
3917 | // Set the size of the N_SO to the terminating index of this |
3918 | // N_SO so that we can always skip the entire N_SO if we need |
3919 | // to navigate more quickly at the source level when parsing |
3920 | // STABS |
3921 | symbol_ptr = symtab.SymbolAtIndex(idx: N_SO_index); |
3922 | symbol_ptr->SetByteSize(sym_idx); |
3923 | symbol_ptr->SetSizeIsSibling(true); |
3924 | } |
3925 | N_NSYM_indexes.clear(); |
3926 | N_INCL_indexes.clear(); |
3927 | N_BRAC_indexes.clear(); |
3928 | N_COMM_indexes.clear(); |
3929 | N_FUN_indexes.clear(); |
3930 | N_SO_index = UINT32_MAX; |
3931 | } else { |
3932 | // We use the current number of symbols in the symbol table in |
3933 | // lieu of using nlist_idx in case we ever start trimming entries |
3934 | // out |
3935 | const bool N_SO_has_full_path = symbol_name[0] == '/'; |
3936 | if (N_SO_has_full_path) { |
3937 | if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { |
3938 | // We have two consecutive N_SO entries where the first |
3939 | // contains a directory and the second contains a full path. |
3940 | sym[sym_idx - 1].GetMangled().SetValue( |
3941 | ConstString(symbol_name)); |
3942 | m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; |
3943 | add_nlist = false; |
3944 | } else { |
3945 | // This is the first entry in a N_SO that contains a |
3946 | // directory or a full path to the source file |
3947 | N_SO_index = sym_idx; |
3948 | } |
3949 | } else if ((N_SO_index == sym_idx - 1) && |
3950 | ((sym_idx - 1) < num_syms)) { |
3951 | // This is usually the second N_SO entry that contains just the |
3952 | // filename, so here we combine it with the first one if we are |
3953 | // minimizing the symbol table |
3954 | const char *so_path = |
3955 | sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); |
3956 | if (so_path && so_path[0]) { |
3957 | std::string full_so_path(so_path); |
3958 | const size_t double_slash_pos = full_so_path.find(s: "//" ); |
3959 | if (double_slash_pos != std::string::npos) { |
3960 | // The linker has been generating bad N_SO entries with |
3961 | // doubled up paths in the format "%s%s" where the first |
3962 | // string in the DW_AT_comp_dir, and the second is the |
3963 | // directory for the source file so you end up with a path |
3964 | // that looks like "/tmp/src//tmp/src/" |
3965 | FileSpec so_dir(so_path); |
3966 | if (!FileSystem::Instance().Exists(file_spec: so_dir)) { |
3967 | so_dir.SetFile(path: &full_so_path[double_slash_pos + 1], |
3968 | style: FileSpec::Style::native); |
3969 | if (FileSystem::Instance().Exists(file_spec: so_dir)) { |
3970 | // Trim off the incorrect path |
3971 | full_so_path.erase(pos: 0, n: double_slash_pos + 1); |
3972 | } |
3973 | } |
3974 | } |
3975 | if (*full_so_path.rbegin() != '/') |
3976 | full_so_path += '/'; |
3977 | full_so_path += symbol_name; |
3978 | sym[sym_idx - 1].GetMangled().SetValue( |
3979 | ConstString(full_so_path.c_str())); |
3980 | add_nlist = false; |
3981 | m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; |
3982 | } |
3983 | } else { |
3984 | // This could be a relative path to a N_SO |
3985 | N_SO_index = sym_idx; |
3986 | } |
3987 | } |
3988 | break; |
3989 | |
3990 | case N_OSO: |
3991 | // object file name: name,,0,0,st_mtime |
3992 | type = eSymbolTypeObjectFile; |
3993 | break; |
3994 | |
3995 | case N_LSYM: |
3996 | // local sym: name,,NO_SECT,type,offset |
3997 | type = eSymbolTypeLocal; |
3998 | break; |
3999 | |
4000 | // INCL scopes |
4001 | case N_BINCL: |
4002 | // include file beginning: name,,NO_SECT,0,sum We use the current |
4003 | // number of symbols in the symbol table in lieu of using nlist_idx |
4004 | // in case we ever start trimming entries out |
4005 | N_INCL_indexes.push_back(x: sym_idx); |
4006 | type = eSymbolTypeScopeBegin; |
4007 | break; |
4008 | |
4009 | case N_EINCL: |
4010 | // include file end: name,,NO_SECT,0,0 |
4011 | // Set the size of the N_BINCL to the terminating index of this |
4012 | // N_EINCL so that we can always skip the entire symbol if we need |
4013 | // to navigate more quickly at the source level when parsing STABS |
4014 | if (!N_INCL_indexes.empty()) { |
4015 | symbol_ptr = symtab.SymbolAtIndex(idx: N_INCL_indexes.back()); |
4016 | symbol_ptr->SetByteSize(sym_idx + 1); |
4017 | symbol_ptr->SetSizeIsSibling(true); |
4018 | N_INCL_indexes.pop_back(); |
4019 | } |
4020 | type = eSymbolTypeScopeEnd; |
4021 | break; |
4022 | |
4023 | case N_SOL: |
4024 | // #included file name: name,,n_sect,0,address |
4025 | type = eSymbolTypeHeaderFile; |
4026 | |
4027 | // We currently don't use the header files on darwin |
4028 | add_nlist = false; |
4029 | break; |
4030 | |
4031 | case N_PARAMS: |
4032 | // compiler parameters: name,,NO_SECT,0,0 |
4033 | type = eSymbolTypeCompiler; |
4034 | break; |
4035 | |
4036 | case N_VERSION: |
4037 | // compiler version: name,,NO_SECT,0,0 |
4038 | type = eSymbolTypeCompiler; |
4039 | break; |
4040 | |
4041 | case N_OLEVEL: |
4042 | // compiler -O level: name,,NO_SECT,0,0 |
4043 | type = eSymbolTypeCompiler; |
4044 | break; |
4045 | |
4046 | case N_PSYM: |
4047 | // parameter: name,,NO_SECT,type,offset |
4048 | type = eSymbolTypeVariable; |
4049 | break; |
4050 | |
4051 | case N_ENTRY: |
4052 | // alternate entry: name,,n_sect,linenumber,address |
4053 | symbol_section = section_info.GetSection(n_sect: nlist.n_sect, file_addr: nlist.n_value); |
4054 | type = eSymbolTypeLineEntry; |
4055 | break; |
4056 | |
4057 | // Left and Right Braces |
4058 | case N_LBRAC: |
4059 | // left bracket: 0,,NO_SECT,nesting level,address We use the |
4060 | // current number of symbols in the symbol table in lieu of using |
4061 | // nlist_idx in case we ever start trimming entries out |
4062 | symbol_section = section_info.GetSection(n_sect: nlist.n_sect, file_addr: nlist.n_value); |
4063 | N_BRAC_indexes.push_back(x: sym_idx); |
4064 | type = eSymbolTypeScopeBegin; |
4065 | break; |
4066 | |
4067 | case N_RBRAC: |
4068 | // right bracket: 0,,NO_SECT,nesting level,address Set the size of |
4069 | // the N_LBRAC to the terminating index of this N_RBRAC so that we |
4070 | // can always skip the entire symbol if we need to navigate more |
4071 | // quickly at the source level when parsing STABS |
4072 | symbol_section = section_info.GetSection(n_sect: nlist.n_sect, file_addr: nlist.n_value); |
4073 | if (!N_BRAC_indexes.empty()) { |
4074 | symbol_ptr = symtab.SymbolAtIndex(idx: N_BRAC_indexes.back()); |
4075 | symbol_ptr->SetByteSize(sym_idx + 1); |
4076 | symbol_ptr->SetSizeIsSibling(true); |
4077 | N_BRAC_indexes.pop_back(); |
4078 | } |
4079 | type = eSymbolTypeScopeEnd; |
4080 | break; |
4081 | |
4082 | case N_EXCL: |
4083 | // deleted include file: name,,NO_SECT,0,sum |
4084 | type = eSymbolTypeHeaderFile; |
4085 | break; |
4086 | |
4087 | // COMM scopes |
4088 | case N_BCOMM: |
4089 | // begin common: name,,NO_SECT,0,0 |
4090 | // We use the current number of symbols in the symbol table in lieu |
4091 | // of using nlist_idx in case we ever start trimming entries out |
4092 | type = eSymbolTypeScopeBegin; |
4093 | N_COMM_indexes.push_back(x: sym_idx); |
4094 | break; |
4095 | |
4096 | case N_ECOML: |
4097 | // end common (local name): 0,,n_sect,0,address |
4098 | symbol_section = section_info.GetSection(n_sect: nlist.n_sect, file_addr: nlist.n_value); |
4099 | [[fallthrough]]; |
4100 | |
4101 | case N_ECOMM: |
4102 | // end common: name,,n_sect,0,0 |
4103 | // Set the size of the N_BCOMM to the terminating index of this |
4104 | // N_ECOMM/N_ECOML so that we can always skip the entire symbol if |
4105 | // we need to navigate more quickly at the source level when |
4106 | // parsing STABS |
4107 | if (!N_COMM_indexes.empty()) { |
4108 | symbol_ptr = symtab.SymbolAtIndex(idx: N_COMM_indexes.back()); |
4109 | symbol_ptr->SetByteSize(sym_idx + 1); |
4110 | symbol_ptr->SetSizeIsSibling(true); |
4111 | N_COMM_indexes.pop_back(); |
4112 | } |
4113 | type = eSymbolTypeScopeEnd; |
4114 | break; |
4115 | |
4116 | case N_LENG: |
4117 | // second stab entry with length information |
4118 | type = eSymbolTypeAdditional; |
4119 | break; |
4120 | |
4121 | default: |
4122 | break; |
4123 | } |
4124 | } else { |
4125 | uint8_t n_type = N_TYPE & nlist.n_type; |
4126 | sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); |
4127 | |
4128 | switch (n_type) { |
4129 | case N_INDR: { |
4130 | const char *reexport_name_cstr = strtab_data.PeekCStr(offset: nlist.n_value); |
4131 | if (reexport_name_cstr && reexport_name_cstr[0] && symbol_name) { |
4132 | type = eSymbolTypeReExported; |
4133 | ConstString reexport_name(reexport_name_cstr + |
4134 | ((reexport_name_cstr[0] == '_') ? 1 : 0)); |
4135 | sym[sym_idx].SetReExportedSymbolName(reexport_name); |
4136 | set_value = false; |
4137 | reexport_shlib_needs_fixup[sym_idx] = reexport_name; |
4138 | indirect_symbol_names.insert( |
4139 | x: ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); |
4140 | } else |
4141 | type = eSymbolTypeUndefined; |
4142 | } break; |
4143 | |
4144 | case N_UNDF: |
4145 | if (symbol_name && symbol_name[0]) { |
4146 | ConstString undefined_name(symbol_name + |
4147 | ((symbol_name[0] == '_') ? 1 : 0)); |
4148 | undefined_name_to_desc[undefined_name] = nlist.n_desc; |
4149 | } |
4150 | [[fallthrough]]; |
4151 | |
4152 | case N_PBUD: |
4153 | type = eSymbolTypeUndefined; |
4154 | break; |
4155 | |
4156 | case N_ABS: |
4157 | type = eSymbolTypeAbsolute; |
4158 | break; |
4159 | |
4160 | case N_SECT: { |
4161 | symbol_section = section_info.GetSection(n_sect: nlist.n_sect, file_addr: nlist.n_value); |
4162 | |
4163 | if (!symbol_section) { |
4164 | // TODO: warn about this? |
4165 | add_nlist = false; |
4166 | break; |
4167 | } |
4168 | |
4169 | if (TEXT_eh_frame_sectID == nlist.n_sect) { |
4170 | type = eSymbolTypeException; |
4171 | } else { |
4172 | uint32_t section_type = symbol_section->Get() & SECTION_TYPE; |
4173 | |
4174 | switch (section_type) { |
4175 | case S_CSTRING_LITERALS: |
4176 | type = eSymbolTypeData; |
4177 | break; // section with only literal C strings |
4178 | case S_4BYTE_LITERALS: |
4179 | type = eSymbolTypeData; |
4180 | break; // section with only 4 byte literals |
4181 | case S_8BYTE_LITERALS: |
4182 | type = eSymbolTypeData; |
4183 | break; // section with only 8 byte literals |
4184 | case S_LITERAL_POINTERS: |
4185 | type = eSymbolTypeTrampoline; |
4186 | break; // section with only pointers to literals |
4187 | case S_NON_LAZY_SYMBOL_POINTERS: |
4188 | type = eSymbolTypeTrampoline; |
4189 | break; // section with only non-lazy symbol pointers |
4190 | case S_LAZY_SYMBOL_POINTERS: |
4191 | type = eSymbolTypeTrampoline; |
4192 | break; // section with only lazy symbol pointers |
4193 | case S_SYMBOL_STUBS: |
4194 | type = eSymbolTypeTrampoline; |
4195 | break; // section with only symbol stubs, byte size of stub in |
4196 | // the reserved2 field |
4197 | case S_MOD_INIT_FUNC_POINTERS: |
4198 | type = eSymbolTypeCode; |
4199 | break; // section with only function pointers for initialization |
4200 | case S_MOD_TERM_FUNC_POINTERS: |
4201 | type = eSymbolTypeCode; |
4202 | break; // section with only function pointers for termination |
4203 | case S_INTERPOSING: |
4204 | type = eSymbolTypeTrampoline; |
4205 | break; // section with only pairs of function pointers for |
4206 | // interposing |
4207 | case S_16BYTE_LITERALS: |
4208 | type = eSymbolTypeData; |
4209 | break; // section with only 16 byte literals |
4210 | case S_DTRACE_DOF: |
4211 | type = eSymbolTypeInstrumentation; |
4212 | break; |
4213 | case S_LAZY_DYLIB_SYMBOL_POINTERS: |
4214 | type = eSymbolTypeTrampoline; |
4215 | break; |
4216 | default: |
4217 | switch (symbol_section->GetType()) { |
4218 | case lldb::eSectionTypeCode: |
4219 | type = eSymbolTypeCode; |
4220 | break; |
4221 | case eSectionTypeData: |
4222 | case eSectionTypeDataCString: // Inlined C string data |
4223 | case eSectionTypeDataCStringPointers: // Pointers to C string |
4224 | // data |
4225 | case eSectionTypeDataSymbolAddress: // Address of a symbol in |
4226 | // the symbol table |
4227 | case eSectionTypeData4: |
4228 | case eSectionTypeData8: |
4229 | case eSectionTypeData16: |
4230 | type = eSymbolTypeData; |
4231 | break; |
4232 | default: |
4233 | break; |
4234 | } |
4235 | break; |
4236 | } |
4237 | |
4238 | if (type == eSymbolTypeInvalid) { |
4239 | const char *symbol_sect_name = |
4240 | symbol_section->GetName().AsCString(); |
4241 | if (symbol_section->IsDescendant(section: text_section_sp.get())) { |
4242 | if (symbol_section->IsClear(bit: S_ATTR_PURE_INSTRUCTIONS | |
4243 | S_ATTR_SELF_MODIFYING_CODE | |
4244 | S_ATTR_SOME_INSTRUCTIONS)) |
4245 | type = eSymbolTypeData; |
4246 | else |
4247 | type = eSymbolTypeCode; |
4248 | } else if (symbol_section->IsDescendant(section: data_section_sp.get()) || |
4249 | symbol_section->IsDescendant( |
4250 | section: data_dirty_section_sp.get()) || |
4251 | symbol_section->IsDescendant( |
4252 | section: data_const_section_sp.get())) { |
4253 | if (symbol_sect_name && |
4254 | ::strstr(haystack: symbol_sect_name, needle: "__objc" ) == symbol_sect_name) { |
4255 | type = eSymbolTypeRuntime; |
4256 | |
4257 | if (symbol_name) { |
4258 | llvm::StringRef symbol_name_ref(symbol_name); |
4259 | if (symbol_name_ref.starts_with(Prefix: "_OBJC_" )) { |
4260 | llvm::StringRef g_objc_v2_prefix_class( |
4261 | "_OBJC_CLASS_$_" ); |
4262 | llvm::StringRef g_objc_v2_prefix_metaclass( |
4263 | "_OBJC_METACLASS_$_" ); |
4264 | llvm::StringRef g_objc_v2_prefix_ivar( |
4265 | "_OBJC_IVAR_$_" ); |
4266 | if (symbol_name_ref.starts_with(Prefix: g_objc_v2_prefix_class)) { |
4267 | symbol_name_non_abi_mangled = symbol_name + 1; |
4268 | symbol_name = |
4269 | symbol_name + g_objc_v2_prefix_class.size(); |
4270 | type = eSymbolTypeObjCClass; |
4271 | demangled_is_synthesized = true; |
4272 | } else if (symbol_name_ref.starts_with( |
4273 | Prefix: g_objc_v2_prefix_metaclass)) { |
4274 | symbol_name_non_abi_mangled = symbol_name + 1; |
4275 | symbol_name = |
4276 | symbol_name + g_objc_v2_prefix_metaclass.size(); |
4277 | type = eSymbolTypeObjCMetaClass; |
4278 | demangled_is_synthesized = true; |
4279 | } else if (symbol_name_ref.starts_with( |
4280 | Prefix: g_objc_v2_prefix_ivar)) { |
4281 | symbol_name_non_abi_mangled = symbol_name + 1; |
4282 | symbol_name = |
4283 | symbol_name + g_objc_v2_prefix_ivar.size(); |
4284 | type = eSymbolTypeObjCIVar; |
4285 | demangled_is_synthesized = true; |
4286 | } |
4287 | } |
4288 | } |
4289 | } else if (symbol_sect_name && |
4290 | ::strstr(haystack: symbol_sect_name, needle: "__gcc_except_tab" ) == |
4291 | symbol_sect_name) { |
4292 | type = eSymbolTypeException; |
4293 | } else { |
4294 | type = eSymbolTypeData; |
4295 | } |
4296 | } else if (symbol_sect_name && |
4297 | ::strstr(haystack: symbol_sect_name, needle: "__IMPORT" ) == |
4298 | symbol_sect_name) { |
4299 | type = eSymbolTypeTrampoline; |
4300 | } else if (symbol_section->IsDescendant(section: objc_section_sp.get())) { |
4301 | type = eSymbolTypeRuntime; |
4302 | if (symbol_name && symbol_name[0] == '.') { |
4303 | llvm::StringRef symbol_name_ref(symbol_name); |
4304 | llvm::StringRef g_objc_v1_prefix_class( |
4305 | ".objc_class_name_" ); |
4306 | if (symbol_name_ref.starts_with(Prefix: g_objc_v1_prefix_class)) { |
4307 | symbol_name_non_abi_mangled = symbol_name; |
4308 | symbol_name = symbol_name + g_objc_v1_prefix_class.size(); |
4309 | type = eSymbolTypeObjCClass; |
4310 | demangled_is_synthesized = true; |
4311 | } |
4312 | } |
4313 | } |
4314 | } |
4315 | } |
4316 | } break; |
4317 | } |
4318 | } |
4319 | |
4320 | if (!add_nlist) { |
4321 | sym[sym_idx].Clear(); |
4322 | return true; |
4323 | } |
4324 | |
4325 | uint64_t symbol_value = nlist.n_value; |
4326 | |
4327 | if (symbol_name_non_abi_mangled) { |
4328 | sym[sym_idx].GetMangled().SetMangledName( |
4329 | ConstString(symbol_name_non_abi_mangled)); |
4330 | sym[sym_idx].GetMangled().SetDemangledName(ConstString(symbol_name)); |
4331 | } else { |
4332 | |
4333 | if (symbol_name && symbol_name[0] == '_') { |
4334 | symbol_name++; // Skip the leading underscore |
4335 | } |
4336 | |
4337 | if (symbol_name) { |
4338 | ConstString const_symbol_name(symbol_name); |
4339 | sym[sym_idx].GetMangled().SetValue(const_symbol_name); |
4340 | } |
4341 | } |
4342 | |
4343 | if (is_gsym) { |
4344 | const char *gsym_name = sym[sym_idx] |
4345 | .GetMangled() |
4346 | .GetName(preference: Mangled::ePreferMangled) |
4347 | .GetCString(); |
4348 | if (gsym_name) |
4349 | N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; |
4350 | } |
4351 | |
4352 | if (symbol_section) { |
4353 | const addr_t section_file_addr = symbol_section->GetFileAddress(); |
4354 | if (symbol_byte_size == 0 && function_starts_count > 0) { |
4355 | addr_t symbol_lookup_file_addr = nlist.n_value; |
4356 | // Do an exact address match for non-ARM addresses, else get the |
4357 | // closest since the symbol might be a thumb symbol which has an |
4358 | // address with bit zero set. |
4359 | FunctionStarts::Entry *func_start_entry = |
4360 | function_starts.FindEntry(addr: symbol_lookup_file_addr, exact_match_only: !is_arm); |
4361 | if (is_arm && func_start_entry) { |
4362 | // Verify that the function start address is the symbol address |
4363 | // (ARM) or the symbol address + 1 (thumb). |
4364 | if (func_start_entry->addr != symbol_lookup_file_addr && |
4365 | func_start_entry->addr != (symbol_lookup_file_addr + 1)) { |
4366 | // Not the right entry, NULL it out... |
4367 | func_start_entry = nullptr; |
4368 | } |
4369 | } |
4370 | if (func_start_entry) { |
4371 | func_start_entry->data = true; |
4372 | |
4373 | addr_t symbol_file_addr = func_start_entry->addr; |
4374 | if (is_arm) |
4375 | symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; |
4376 | |
4377 | const FunctionStarts::Entry *next_func_start_entry = |
4378 | function_starts.FindNextEntry(entry: func_start_entry); |
4379 | const addr_t section_end_file_addr = |
4380 | section_file_addr + symbol_section->GetByteSize(); |
4381 | if (next_func_start_entry) { |
4382 | addr_t next_symbol_file_addr = next_func_start_entry->addr; |
4383 | // Be sure the clear the Thumb address bit when we calculate the |
4384 | // size from the current and next address |
4385 | if (is_arm) |
4386 | next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; |
4387 | symbol_byte_size = std::min<lldb::addr_t>( |
4388 | a: next_symbol_file_addr - symbol_file_addr, |
4389 | b: section_end_file_addr - symbol_file_addr); |
4390 | } else { |
4391 | symbol_byte_size = section_end_file_addr - symbol_file_addr; |
4392 | } |
4393 | } |
4394 | } |
4395 | symbol_value -= section_file_addr; |
4396 | } |
4397 | |
4398 | if (!is_debug) { |
4399 | if (type == eSymbolTypeCode) { |
4400 | // See if we can find a N_FUN entry for any code symbols. If we do |
4401 | // find a match, and the name matches, then we can merge the two into |
4402 | // just the function symbol to avoid duplicate entries in the symbol |
4403 | // table. |
4404 | std::pair<ValueToSymbolIndexMap::const_iterator, |
4405 | ValueToSymbolIndexMap::const_iterator> |
4406 | range; |
4407 | range = N_FUN_addr_to_sym_idx.equal_range(x: nlist.n_value); |
4408 | if (range.first != range.second) { |
4409 | for (ValueToSymbolIndexMap::const_iterator pos = range.first; |
4410 | pos != range.second; ++pos) { |
4411 | if (sym[sym_idx].GetMangled().GetName(preference: Mangled::ePreferMangled) == |
4412 | sym[pos->second].GetMangled().GetName( |
4413 | preference: Mangled::ePreferMangled)) { |
4414 | m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; |
4415 | // We just need the flags from the linker symbol, so put these |
4416 | // flags into the N_FUN flags to avoid duplicate symbols in the |
4417 | // symbol table. |
4418 | sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); |
4419 | sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); |
4420 | if (resolver_addresses.find(x: nlist.n_value) != |
4421 | resolver_addresses.end()) |
4422 | sym[pos->second].SetType(eSymbolTypeResolver); |
4423 | sym[sym_idx].Clear(); |
4424 | return true; |
4425 | } |
4426 | } |
4427 | } else { |
4428 | if (resolver_addresses.find(x: nlist.n_value) != |
4429 | resolver_addresses.end()) |
4430 | type = eSymbolTypeResolver; |
4431 | } |
4432 | } else if (type == eSymbolTypeData || type == eSymbolTypeObjCClass || |
4433 | type == eSymbolTypeObjCMetaClass || |
4434 | type == eSymbolTypeObjCIVar) { |
4435 | // See if we can find a N_STSYM entry for any data symbols. If we do |
4436 | // find a match, and the name matches, then we can merge the two into |
4437 | // just the Static symbol to avoid duplicate entries in the symbol |
4438 | // table. |
4439 | std::pair<ValueToSymbolIndexMap::const_iterator, |
4440 | ValueToSymbolIndexMap::const_iterator> |
4441 | range; |
4442 | range = N_STSYM_addr_to_sym_idx.equal_range(x: nlist.n_value); |
4443 | if (range.first != range.second) { |
4444 | for (ValueToSymbolIndexMap::const_iterator pos = range.first; |
4445 | pos != range.second; ++pos) { |
4446 | if (sym[sym_idx].GetMangled().GetName(preference: Mangled::ePreferMangled) == |
4447 | sym[pos->second].GetMangled().GetName( |
4448 | preference: Mangled::ePreferMangled)) { |
4449 | m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; |
4450 | // We just need the flags from the linker symbol, so put these |
4451 | // flags into the N_STSYM flags to avoid duplicate symbols in |
4452 | // the symbol table. |
4453 | sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); |
4454 | sym[pos->second].SetFlags(nlist.n_type << 16 | nlist.n_desc); |
4455 | sym[sym_idx].Clear(); |
4456 | return true; |
4457 | } |
4458 | } |
4459 | } else { |
4460 | // Combine N_GSYM stab entries with the non stab symbol. |
4461 | const char *gsym_name = sym[sym_idx] |
4462 | .GetMangled() |
4463 | .GetName(preference: Mangled::ePreferMangled) |
4464 | .GetCString(); |
4465 | if (gsym_name) { |
4466 | ConstNameToSymbolIndexMap::const_iterator pos = |
4467 | N_GSYM_name_to_sym_idx.find(Val: gsym_name); |
4468 | if (pos != N_GSYM_name_to_sym_idx.end()) { |
4469 | const uint32_t GSYM_sym_idx = pos->second; |
4470 | m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; |
4471 | // Copy the address, because often the N_GSYM address has an |
4472 | // invalid address of zero when the global is a common symbol. |
4473 | sym[GSYM_sym_idx].GetAddressRef().SetSection(symbol_section); |
4474 | sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); |
4475 | add_symbol_addr( |
4476 | sym[GSYM_sym_idx].GetAddress().GetFileAddress()); |
4477 | // We just need the flags from the linker symbol, so put these |
4478 | // flags into the N_GSYM flags to avoid duplicate symbols in |
4479 | // the symbol table. |
4480 | sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); |
4481 | sym[sym_idx].Clear(); |
4482 | return true; |
4483 | } |
4484 | } |
4485 | } |
4486 | } |
4487 | } |
4488 | |
4489 | sym[sym_idx].SetID(nlist_idx); |
4490 | sym[sym_idx].SetType(type); |
4491 | if (set_value) { |
4492 | sym[sym_idx].GetAddressRef().SetSection(symbol_section); |
4493 | sym[sym_idx].GetAddressRef().SetOffset(symbol_value); |
4494 | if (symbol_section) |
4495 | add_symbol_addr(sym[sym_idx].GetAddress().GetFileAddress()); |
4496 | } |
4497 | sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); |
4498 | if (nlist.n_desc & N_WEAK_REF) |
4499 | sym[sym_idx].SetIsWeak(true); |
4500 | |
4501 | if (symbol_byte_size > 0) |
4502 | sym[sym_idx].SetByteSize(symbol_byte_size); |
4503 | |
4504 | if (demangled_is_synthesized) |
4505 | sym[sym_idx].SetDemangledNameIsSynthesized(true); |
4506 | |
4507 | ++sym_idx; |
4508 | return true; |
4509 | }; |
4510 | |
4511 | // First parse all the nlists but don't process them yet. See the next |
4512 | // comment for an explanation why. |
4513 | std::vector<struct nlist_64> nlists; |
4514 | nlists.reserve(n: symtab_load_command.nsyms); |
4515 | for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { |
4516 | if (auto nlist = |
4517 | ParseNList(nlist_data, nlist_data_offset, nlist_byte_size)) |
4518 | nlists.push_back(x: *nlist); |
4519 | else |
4520 | break; |
4521 | } |
4522 | |
4523 | // Now parse all the debug symbols. This is needed to merge non-debug |
4524 | // symbols in the next step. Non-debug symbols are always coalesced into |
4525 | // the debug symbol. Doing this in one step would mean that some symbols |
4526 | // won't be merged. |
4527 | nlist_idx = 0; |
4528 | for (auto &nlist : nlists) { |
4529 | if (!ParseSymbolLambda(nlist, nlist_idx++, DebugSymbols)) |
4530 | break; |
4531 | } |
4532 | |
4533 | // Finally parse all the non debug symbols. |
4534 | nlist_idx = 0; |
4535 | for (auto &nlist : nlists) { |
4536 | if (!ParseSymbolLambda(nlist, nlist_idx++, NonDebugSymbols)) |
4537 | break; |
4538 | } |
4539 | |
4540 | for (const auto &pos : reexport_shlib_needs_fixup) { |
4541 | const auto undef_pos = undefined_name_to_desc.find(Val: pos.second); |
4542 | if (undef_pos != undefined_name_to_desc.end()) { |
4543 | const uint8_t dylib_ordinal = |
4544 | llvm::MachO::GET_LIBRARY_ORDINAL(n_desc: undef_pos->second); |
4545 | if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) |
4546 | sym[pos.first].SetReExportedSymbolSharedLibrary( |
4547 | dylib_files.GetFileSpecAtIndex(idx: dylib_ordinal - 1)); |
4548 | } |
4549 | } |
4550 | } |
4551 | |
4552 | // Count how many trie symbols we'll add to the symbol table |
4553 | int trie_symbol_table_augment_count = 0; |
4554 | for (auto &e : external_sym_trie_entries) { |
4555 | if (!symbols_added.contains(V: e.entry.address)) |
4556 | trie_symbol_table_augment_count++; |
4557 | } |
4558 | |
4559 | if (num_syms < sym_idx + trie_symbol_table_augment_count) { |
4560 | num_syms = sym_idx + trie_symbol_table_augment_count; |
4561 | sym = symtab.Resize(count: num_syms); |
4562 | } |
4563 | uint32_t synthetic_sym_id = symtab_load_command.nsyms; |
4564 | |
4565 | // Add symbols from the trie to the symbol table. |
4566 | for (auto &e : external_sym_trie_entries) { |
4567 | if (symbols_added.contains(V: e.entry.address)) |
4568 | continue; |
4569 | |
4570 | // Find the section that this trie address is in, use that to annotate |
4571 | // symbol type as we add the trie address and name to the symbol table. |
4572 | Address symbol_addr; |
4573 | if (module_sp->ResolveFileAddress(vm_addr: e.entry.address, so_addr&: symbol_addr)) { |
4574 | SectionSP symbol_section(symbol_addr.GetSection()); |
4575 | const char *symbol_name = e.entry.name.GetCString(); |
4576 | bool demangled_is_synthesized = false; |
4577 | SymbolType type = |
4578 | GetSymbolType(symbol_name, demangled_is_synthesized, text_section_sp, |
4579 | data_section_sp, data_dirty_section_sp, |
4580 | data_const_section_sp, symbol_section); |
4581 | |
4582 | sym[sym_idx].SetType(type); |
4583 | if (symbol_section) { |
4584 | sym[sym_idx].SetID(synthetic_sym_id++); |
4585 | sym[sym_idx].GetMangled().SetMangledName(ConstString(symbol_name)); |
4586 | if (demangled_is_synthesized) |
4587 | sym[sym_idx].SetDemangledNameIsSynthesized(true); |
4588 | sym[sym_idx].SetIsSynthetic(true); |
4589 | sym[sym_idx].SetExternal(true); |
4590 | sym[sym_idx].GetAddressRef() = symbol_addr; |
4591 | add_symbol_addr(symbol_addr.GetFileAddress()); |
4592 | if (e.entry.flags & TRIE_SYMBOL_IS_THUMB) |
4593 | sym[sym_idx].SetFlags(MACHO_NLIST_ARM_SYMBOL_IS_THUMB); |
4594 | ++sym_idx; |
4595 | } |
4596 | } |
4597 | } |
4598 | |
4599 | if (function_starts_count > 0) { |
4600 | uint32_t num_synthetic_function_symbols = 0; |
4601 | for (i = 0; i < function_starts_count; ++i) { |
4602 | if (!symbols_added.contains(V: function_starts.GetEntryRef(i).addr)) |
4603 | ++num_synthetic_function_symbols; |
4604 | } |
4605 | |
4606 | if (num_synthetic_function_symbols > 0) { |
4607 | if (num_syms < sym_idx + num_synthetic_function_symbols) { |
4608 | num_syms = sym_idx + num_synthetic_function_symbols; |
4609 | sym = symtab.Resize(count: num_syms); |
4610 | } |
4611 | for (i = 0; i < function_starts_count; ++i) { |
4612 | const FunctionStarts::Entry *func_start_entry = |
4613 | function_starts.GetEntryAtIndex(i); |
4614 | if (!symbols_added.contains(V: func_start_entry->addr)) { |
4615 | addr_t symbol_file_addr = func_start_entry->addr; |
4616 | uint32_t symbol_flags = 0; |
4617 | if (func_start_entry->data) |
4618 | symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; |
4619 | Address symbol_addr; |
4620 | if (module_sp->ResolveFileAddress(vm_addr: symbol_file_addr, so_addr&: symbol_addr)) { |
4621 | SectionSP symbol_section(symbol_addr.GetSection()); |
4622 | uint32_t symbol_byte_size = 0; |
4623 | if (symbol_section) { |
4624 | const addr_t section_file_addr = symbol_section->GetFileAddress(); |
4625 | const FunctionStarts::Entry *next_func_start_entry = |
4626 | function_starts.FindNextEntry(entry: func_start_entry); |
4627 | const addr_t section_end_file_addr = |
4628 | section_file_addr + symbol_section->GetByteSize(); |
4629 | if (next_func_start_entry) { |
4630 | addr_t next_symbol_file_addr = next_func_start_entry->addr; |
4631 | if (is_arm) |
4632 | next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; |
4633 | symbol_byte_size = std::min<lldb::addr_t>( |
4634 | a: next_symbol_file_addr - symbol_file_addr, |
4635 | b: section_end_file_addr - symbol_file_addr); |
4636 | } else { |
4637 | symbol_byte_size = section_end_file_addr - symbol_file_addr; |
4638 | } |
4639 | sym[sym_idx].SetID(synthetic_sym_id++); |
4640 | // Don't set the name for any synthetic symbols, the Symbol |
4641 | // object will generate one if needed when the name is accessed |
4642 | // via accessors. |
4643 | sym[sym_idx].GetMangled().SetDemangledName(ConstString()); |
4644 | sym[sym_idx].SetType(eSymbolTypeCode); |
4645 | sym[sym_idx].SetIsSynthetic(true); |
4646 | sym[sym_idx].GetAddressRef() = symbol_addr; |
4647 | add_symbol_addr(symbol_addr.GetFileAddress()); |
4648 | if (symbol_flags) |
4649 | sym[sym_idx].SetFlags(symbol_flags); |
4650 | if (symbol_byte_size) |
4651 | sym[sym_idx].SetByteSize(symbol_byte_size); |
4652 | ++sym_idx; |
4653 | } |
4654 | } |
4655 | } |
4656 | } |
4657 | } |
4658 | } |
4659 | |
4660 | // Trim our symbols down to just what we ended up with after removing any |
4661 | // symbols. |
4662 | if (sym_idx < num_syms) { |
4663 | num_syms = sym_idx; |
4664 | sym = symtab.Resize(count: num_syms); |
4665 | } |
4666 | |
4667 | // Now synthesize indirect symbols |
4668 | if (m_dysymtab.nindirectsyms != 0) { |
4669 | if (indirect_symbol_index_data.GetByteSize()) { |
4670 | NListIndexToSymbolIndexMap::const_iterator end_index_pos = |
4671 | m_nlist_idx_to_sym_idx.end(); |
4672 | |
4673 | for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); |
4674 | ++sect_idx) { |
4675 | if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == |
4676 | S_SYMBOL_STUBS) { |
4677 | uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; |
4678 | if (symbol_stub_byte_size == 0) |
4679 | continue; |
4680 | |
4681 | const uint32_t num_symbol_stubs = |
4682 | m_mach_sections[sect_idx].size / symbol_stub_byte_size; |
4683 | |
4684 | if (num_symbol_stubs == 0) |
4685 | continue; |
4686 | |
4687 | const uint32_t symbol_stub_index_offset = |
4688 | m_mach_sections[sect_idx].reserved1; |
4689 | for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) { |
4690 | const uint32_t symbol_stub_index = |
4691 | symbol_stub_index_offset + stub_idx; |
4692 | const lldb::addr_t symbol_stub_addr = |
4693 | m_mach_sections[sect_idx].addr + |
4694 | (stub_idx * symbol_stub_byte_size); |
4695 | lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; |
4696 | if (indirect_symbol_index_data.ValidOffsetForDataOfSize( |
4697 | offset: symbol_stub_offset, length: 4)) { |
4698 | const uint32_t stub_sym_id = |
4699 | indirect_symbol_index_data.GetU32(offset_ptr: &symbol_stub_offset); |
4700 | if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) |
4701 | continue; |
4702 | |
4703 | NListIndexToSymbolIndexMap::const_iterator index_pos = |
4704 | m_nlist_idx_to_sym_idx.find(Val: stub_sym_id); |
4705 | Symbol *stub_symbol = nullptr; |
4706 | if (index_pos != end_index_pos) { |
4707 | // We have a remapping from the original nlist index to a |
4708 | // current symbol index, so just look this up by index |
4709 | stub_symbol = symtab.SymbolAtIndex(idx: index_pos->second); |
4710 | } else { |
4711 | // We need to lookup a symbol using the original nlist symbol |
4712 | // index since this index is coming from the S_SYMBOL_STUBS |
4713 | stub_symbol = symtab.FindSymbolByID(uid: stub_sym_id); |
4714 | } |
4715 | |
4716 | if (stub_symbol) { |
4717 | Address so_addr(symbol_stub_addr, section_list); |
4718 | |
4719 | if (stub_symbol->GetType() == eSymbolTypeUndefined) { |
4720 | // Change the external symbol into a trampoline that makes |
4721 | // sense These symbols were N_UNDF N_EXT, and are useless |
4722 | // to us, so we can re-use them so we don't have to make up |
4723 | // a synthetic symbol for no good reason. |
4724 | if (resolver_addresses.find(x: symbol_stub_addr) == |
4725 | resolver_addresses.end()) |
4726 | stub_symbol->SetType(eSymbolTypeTrampoline); |
4727 | else |
4728 | stub_symbol->SetType(eSymbolTypeResolver); |
4729 | stub_symbol->SetExternal(false); |
4730 | stub_symbol->GetAddressRef() = so_addr; |
4731 | stub_symbol->SetByteSize(symbol_stub_byte_size); |
4732 | } else { |
4733 | // Make a synthetic symbol to describe the trampoline stub |
4734 | Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); |
4735 | if (sym_idx >= num_syms) { |
4736 | sym = symtab.Resize(count: ++num_syms); |
4737 | stub_symbol = nullptr; // this pointer no longer valid |
4738 | } |
4739 | sym[sym_idx].SetID(synthetic_sym_id++); |
4740 | sym[sym_idx].GetMangled() = stub_symbol_mangled_name; |
4741 | if (resolver_addresses.find(x: symbol_stub_addr) == |
4742 | resolver_addresses.end()) |
4743 | sym[sym_idx].SetType(eSymbolTypeTrampoline); |
4744 | else |
4745 | sym[sym_idx].SetType(eSymbolTypeResolver); |
4746 | sym[sym_idx].SetIsSynthetic(true); |
4747 | sym[sym_idx].GetAddressRef() = so_addr; |
4748 | add_symbol_addr(so_addr.GetFileAddress()); |
4749 | sym[sym_idx].SetByteSize(symbol_stub_byte_size); |
4750 | ++sym_idx; |
4751 | } |
4752 | } else { |
4753 | if (log) |
4754 | log->Warning(fmt: "symbol stub referencing symbol table symbol " |
4755 | "%u that isn't in our minimal symbol table, " |
4756 | "fix this!!!" , |
4757 | stub_sym_id); |
4758 | } |
4759 | } |
4760 | } |
4761 | } |
4762 | } |
4763 | } |
4764 | } |
4765 | |
4766 | if (!reexport_trie_entries.empty()) { |
4767 | for (const auto &e : reexport_trie_entries) { |
4768 | if (e.entry.import_name) { |
4769 | // Only add indirect symbols from the Trie entries if we didn't have |
4770 | // a N_INDR nlist entry for this already |
4771 | if (indirect_symbol_names.find(x: e.entry.name) == |
4772 | indirect_symbol_names.end()) { |
4773 | // Make a synthetic symbol to describe re-exported symbol. |
4774 | if (sym_idx >= num_syms) |
4775 | sym = symtab.Resize(count: ++num_syms); |
4776 | sym[sym_idx].SetID(synthetic_sym_id++); |
4777 | sym[sym_idx].GetMangled() = Mangled(e.entry.name); |
4778 | sym[sym_idx].SetType(eSymbolTypeReExported); |
4779 | sym[sym_idx].SetIsSynthetic(true); |
4780 | sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); |
4781 | if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { |
4782 | sym[sym_idx].SetReExportedSymbolSharedLibrary( |
4783 | dylib_files.GetFileSpecAtIndex(idx: e.entry.other - 1)); |
4784 | } |
4785 | ++sym_idx; |
4786 | } |
4787 | } |
4788 | } |
4789 | } |
4790 | } |
4791 | |
4792 | void ObjectFileMachO::Dump(Stream *s) { |
4793 | ModuleSP module_sp(GetModule()); |
4794 | if (module_sp) { |
4795 | std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); |
4796 | s->Printf(format: "%p: " , static_cast<void *>(this)); |
4797 | s->Indent(); |
4798 | if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) |
4799 | s->PutCString(cstr: "ObjectFileMachO64" ); |
4800 | else |
4801 | s->PutCString(cstr: "ObjectFileMachO32" ); |
4802 | |
4803 | *s << ", file = '" << m_file; |
4804 | ModuleSpecList all_specs; |
4805 | ModuleSpec base_spec; |
4806 | GetAllArchSpecs(header: m_header, data: m_data, lc_offset: MachHeaderSizeFromMagic(magic: m_header.magic), |
4807 | base_spec, all_specs); |
4808 | for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { |
4809 | *s << "', triple" ; |
4810 | if (e) |
4811 | s->Printf(format: "[%d]" , i); |
4812 | *s << " = " ; |
4813 | *s << all_specs.GetModuleSpecRefAtIndex(i) |
4814 | .GetArchitecture() |
4815 | .GetTriple() |
4816 | .getTriple(); |
4817 | } |
4818 | *s << "\n" ; |
4819 | SectionList *sections = GetSectionList(); |
4820 | if (sections) |
4821 | sections->Dump(s&: s->AsRawOstream(), indent: s->GetIndentLevel(), target: nullptr, show_header: true, |
4822 | UINT32_MAX); |
4823 | |
4824 | if (m_symtab_up) |
4825 | m_symtab_up->Dump(s, target: nullptr, sort_type: eSortOrderNone); |
4826 | } |
4827 | } |
4828 | |
4829 | UUID ObjectFileMachO::(const llvm::MachO::mach_header &, |
4830 | const lldb_private::DataExtractor &data, |
4831 | lldb::offset_t lc_offset) { |
4832 | uint32_t i; |
4833 | llvm::MachO::uuid_command load_cmd; |
4834 | |
4835 | lldb::offset_t offset = lc_offset; |
4836 | for (i = 0; i < header.ncmds; ++i) { |
4837 | const lldb::offset_t cmd_offset = offset; |
4838 | if (data.GetU32(offset_ptr: &offset, dst: &load_cmd, count: 2) == nullptr) |
4839 | break; |
4840 | |
4841 | if (load_cmd.cmd == LC_UUID) { |
4842 | const uint8_t *uuid_bytes = data.PeekData(offset, length: 16); |
4843 | |
4844 | if (uuid_bytes) { |
4845 | // OpenCL on Mac OS X uses the same UUID for each of its object files. |
4846 | // We pretend these object files have no UUID to prevent crashing. |
4847 | |
4848 | const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, |
4849 | 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, |
4850 | 0xbb, 0x14, 0xf0, 0x0d}; |
4851 | |
4852 | if (!memcmp(s1: uuid_bytes, s2: opencl_uuid, n: 16)) |
4853 | return UUID(); |
4854 | |
4855 | return UUID(uuid_bytes, 16); |
4856 | } |
4857 | return UUID(); |
4858 | } |
4859 | offset = cmd_offset + load_cmd.cmdsize; |
4860 | } |
4861 | return UUID(); |
4862 | } |
4863 | |
4864 | static llvm::StringRef GetOSName(uint32_t cmd) { |
4865 | switch (cmd) { |
4866 | case llvm::MachO::LC_VERSION_MIN_IPHONEOS: |
4867 | return llvm::Triple::getOSTypeName(Kind: llvm::Triple::IOS); |
4868 | case llvm::MachO::LC_VERSION_MIN_MACOSX: |
4869 | return llvm::Triple::getOSTypeName(Kind: llvm::Triple::MacOSX); |
4870 | case llvm::MachO::LC_VERSION_MIN_TVOS: |
4871 | return llvm::Triple::getOSTypeName(Kind: llvm::Triple::TvOS); |
4872 | case llvm::MachO::LC_VERSION_MIN_WATCHOS: |
4873 | return llvm::Triple::getOSTypeName(Kind: llvm::Triple::WatchOS); |
4874 | default: |
4875 | llvm_unreachable("unexpected LC_VERSION load command" ); |
4876 | } |
4877 | } |
4878 | |
4879 | namespace { |
4880 | struct OSEnv { |
4881 | llvm::StringRef os_type; |
4882 | llvm::StringRef environment; |
4883 | OSEnv(uint32_t cmd) { |
4884 | switch (cmd) { |
4885 | case llvm::MachO::PLATFORM_MACOS: |
4886 | os_type = llvm::Triple::getOSTypeName(Kind: llvm::Triple::MacOSX); |
4887 | return; |
4888 | case llvm::MachO::PLATFORM_IOS: |
4889 | os_type = llvm::Triple::getOSTypeName(Kind: llvm::Triple::IOS); |
4890 | return; |
4891 | case llvm::MachO::PLATFORM_TVOS: |
4892 | os_type = llvm::Triple::getOSTypeName(Kind: llvm::Triple::TvOS); |
4893 | return; |
4894 | case llvm::MachO::PLATFORM_WATCHOS: |
4895 | os_type = llvm::Triple::getOSTypeName(Kind: llvm::Triple::WatchOS); |
4896 | return; |
4897 | case llvm::MachO::PLATFORM_BRIDGEOS: |
4898 | os_type = llvm::Triple::getOSTypeName(Kind: llvm::Triple::BridgeOS); |
4899 | return; |
4900 | case llvm::MachO::PLATFORM_DRIVERKIT: |
4901 | os_type = llvm::Triple::getOSTypeName(Kind: llvm::Triple::DriverKit); |
4902 | return; |
4903 | case llvm::MachO::PLATFORM_MACCATALYST: |
4904 | os_type = llvm::Triple::getOSTypeName(Kind: llvm::Triple::IOS); |
4905 | environment = llvm::Triple::getEnvironmentTypeName(Kind: llvm::Triple::MacABI); |
4906 | return; |
4907 | case llvm::MachO::PLATFORM_IOSSIMULATOR: |
4908 | os_type = llvm::Triple::getOSTypeName(Kind: llvm::Triple::IOS); |
4909 | environment = |
4910 | llvm::Triple::getEnvironmentTypeName(Kind: llvm::Triple::Simulator); |
4911 | return; |
4912 | case llvm::MachO::PLATFORM_TVOSSIMULATOR: |
4913 | os_type = llvm::Triple::getOSTypeName(Kind: llvm::Triple::TvOS); |
4914 | environment = |
4915 | llvm::Triple::getEnvironmentTypeName(Kind: llvm::Triple::Simulator); |
4916 | return; |
4917 | case llvm::MachO::PLATFORM_WATCHOSSIMULATOR: |
4918 | os_type = llvm::Triple::getOSTypeName(Kind: llvm::Triple::WatchOS); |
4919 | environment = |
4920 | llvm::Triple::getEnvironmentTypeName(Kind: llvm::Triple::Simulator); |
4921 | return; |
4922 | case llvm::MachO::PLATFORM_XROS: |
4923 | os_type = llvm::Triple::getOSTypeName(Kind: llvm::Triple::XROS); |
4924 | return; |
4925 | case llvm::MachO::PLATFORM_XROS_SIMULATOR: |
4926 | os_type = llvm::Triple::getOSTypeName(Kind: llvm::Triple::XROS); |
4927 | environment = |
4928 | llvm::Triple::getEnvironmentTypeName(Kind: llvm::Triple::Simulator); |
4929 | return; |
4930 | default: { |
4931 | Log *log(GetLog(mask: LLDBLog::Symbols | LLDBLog::Process)); |
4932 | LLDB_LOGF(log, "unsupported platform in LC_BUILD_VERSION" ); |
4933 | } |
4934 | } |
4935 | } |
4936 | }; |
4937 | |
4938 | struct MinOS { |
4939 | uint32_t major_version, minor_version, patch_version; |
4940 | MinOS(uint32_t version) |
4941 | : major_version(version >> 16), minor_version((version >> 8) & 0xffu), |
4942 | patch_version(version & 0xffu) {} |
4943 | }; |
4944 | } // namespace |
4945 | |
4946 | void ObjectFileMachO::(const llvm::MachO::mach_header &, |
4947 | const lldb_private::DataExtractor &data, |
4948 | lldb::offset_t lc_offset, |
4949 | ModuleSpec &base_spec, |
4950 | lldb_private::ModuleSpecList &all_specs) { |
4951 | auto &base_arch = base_spec.GetArchitecture(); |
4952 | base_arch.SetArchitecture(arch_type: eArchTypeMachO, cpu: header.cputype, sub: header.cpusubtype); |
4953 | if (!base_arch.IsValid()) |
4954 | return; |
4955 | |
4956 | bool found_any = false; |
4957 | auto add_triple = [&](const llvm::Triple &triple) { |
4958 | auto spec = base_spec; |
4959 | spec.GetArchitecture().GetTriple() = triple; |
4960 | if (spec.GetArchitecture().IsValid()) { |
4961 | spec.GetUUID() = ObjectFileMachO::GetUUID(header, data, lc_offset); |
4962 | all_specs.Append(spec); |
4963 | found_any = true; |
4964 | } |
4965 | }; |
4966 | |
4967 | // Set OS to an unspecified unknown or a "*" so it can match any OS |
4968 | llvm::Triple base_triple = base_arch.GetTriple(); |
4969 | base_triple.setOS(llvm::Triple::UnknownOS); |
4970 | base_triple.setOSName(llvm::StringRef()); |
4971 | |
4972 | if (header.filetype == MH_PRELOAD) { |
4973 | if (header.cputype == CPU_TYPE_ARM) { |
4974 | // If this is a 32-bit arm binary, and it's a standalone binary, force |
4975 | // the Vendor to Apple so we don't accidentally pick up the generic |
4976 | // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the |
4977 | // frame pointer register; most other armv7 ABIs use a combination of |
4978 | // r7 and r11. |
4979 | base_triple.setVendor(llvm::Triple::Apple); |
4980 | } else { |
4981 | // Set vendor to an unspecified unknown or a "*" so it can match any |
4982 | // vendor This is required for correct behavior of EFI debugging on |
4983 | // x86_64 |
4984 | base_triple.setVendor(llvm::Triple::UnknownVendor); |
4985 | base_triple.setVendorName(llvm::StringRef()); |
4986 | } |
4987 | return add_triple(base_triple); |
4988 | } |
4989 | |
4990 | llvm::MachO::load_command load_cmd; |
4991 | |
4992 | // See if there is an LC_VERSION_MIN_* load command that can give |
4993 | // us the OS type. |
4994 | lldb::offset_t offset = lc_offset; |
4995 | for (uint32_t i = 0; i < header.ncmds; ++i) { |
4996 | const lldb::offset_t cmd_offset = offset; |
4997 | if (data.GetU32(offset_ptr: &offset, dst: &load_cmd, count: 2) == nullptr) |
4998 | break; |
4999 | |
5000 | llvm::MachO::version_min_command version_min; |
5001 | switch (load_cmd.cmd) { |
5002 | case llvm::MachO::LC_VERSION_MIN_MACOSX: |
5003 | case llvm::MachO::LC_VERSION_MIN_IPHONEOS: |
5004 | case llvm::MachO::LC_VERSION_MIN_TVOS: |
5005 | case llvm::MachO::LC_VERSION_MIN_WATCHOS: { |
5006 | if (load_cmd.cmdsize != sizeof(version_min)) |
5007 | break; |
5008 | if (data.ExtractBytes(offset: cmd_offset, length: sizeof(version_min), |
5009 | dst_byte_order: data.GetByteOrder(), dst: &version_min) == 0) |
5010 | break; |
5011 | MinOS min_os(version_min.version); |
5012 | llvm::SmallString<32> os_name; |
5013 | llvm::raw_svector_ostream os(os_name); |
5014 | os << GetOSName(cmd: load_cmd.cmd) << min_os.major_version << '.' |
5015 | << min_os.minor_version << '.' << min_os.patch_version; |
5016 | |
5017 | auto triple = base_triple; |
5018 | triple.setOSName(os.str()); |
5019 | |
5020 | // Disambiguate legacy simulator platforms. |
5021 | if (load_cmd.cmd != llvm::MachO::LC_VERSION_MIN_MACOSX && |
5022 | (base_triple.getArch() == llvm::Triple::x86_64 || |
5023 | base_triple.getArch() == llvm::Triple::x86)) { |
5024 | // The combination of legacy LC_VERSION_MIN load command and |
5025 | // x86 architecture always indicates a simulator environment. |
5026 | // The combination of LC_VERSION_MIN and arm architecture only |
5027 | // appears for native binaries. Back-deploying simulator |
5028 | // binaries on Apple Silicon Macs use the modern unambigous |
5029 | // LC_BUILD_VERSION load commands; no special handling required. |
5030 | triple.setEnvironment(llvm::Triple::Simulator); |
5031 | } |
5032 | add_triple(triple); |
5033 | break; |
5034 | } |
5035 | default: |
5036 | break; |
5037 | } |
5038 | |
5039 | offset = cmd_offset + load_cmd.cmdsize; |
5040 | } |
5041 | |
5042 | // See if there are LC_BUILD_VERSION load commands that can give |
5043 | // us the OS type. |
5044 | offset = lc_offset; |
5045 | for (uint32_t i = 0; i < header.ncmds; ++i) { |
5046 | const lldb::offset_t cmd_offset = offset; |
5047 | if (data.GetU32(offset_ptr: &offset, dst: &load_cmd, count: 2) == nullptr) |
5048 | break; |
5049 | |
5050 | do { |
5051 | if (load_cmd.cmd == llvm::MachO::LC_BUILD_VERSION) { |
5052 | llvm::MachO::build_version_command build_version; |
5053 | if (load_cmd.cmdsize < sizeof(build_version)) { |
5054 | // Malformed load command. |
5055 | break; |
5056 | } |
5057 | if (data.ExtractBytes(offset: cmd_offset, length: sizeof(build_version), |
5058 | dst_byte_order: data.GetByteOrder(), dst: &build_version) == 0) |
5059 | break; |
5060 | MinOS min_os(build_version.minos); |
5061 | OSEnv os_env(build_version.platform); |
5062 | llvm::SmallString<16> os_name; |
5063 | llvm::raw_svector_ostream os(os_name); |
5064 | os << os_env.os_type << min_os.major_version << '.' |
5065 | << min_os.minor_version << '.' << min_os.patch_version; |
5066 | auto triple = base_triple; |
5067 | triple.setOSName(os.str()); |
5068 | os_name.clear(); |
5069 | if (!os_env.environment.empty()) |
5070 | triple.setEnvironmentName(os_env.environment); |
5071 | add_triple(triple); |
5072 | } |
5073 | } while (false); |
5074 | offset = cmd_offset + load_cmd.cmdsize; |
5075 | } |
5076 | |
5077 | if (!found_any) { |
5078 | add_triple(base_triple); |
5079 | } |
5080 | } |
5081 | |
5082 | ArchSpec ObjectFileMachO::( |
5083 | ModuleSP module_sp, const llvm::MachO::mach_header &, |
5084 | const lldb_private::DataExtractor &data, lldb::offset_t lc_offset) { |
5085 | ModuleSpecList all_specs; |
5086 | ModuleSpec base_spec; |
5087 | GetAllArchSpecs(header, data, lc_offset: MachHeaderSizeFromMagic(magic: header.magic), |
5088 | base_spec, all_specs); |
5089 | |
5090 | // If the object file offers multiple alternative load commands, |
5091 | // pick the one that matches the module. |
5092 | if (module_sp) { |
5093 | const ArchSpec &module_arch = module_sp->GetArchitecture(); |
5094 | for (unsigned i = 0, e = all_specs.GetSize(); i != e; ++i) { |
5095 | ArchSpec mach_arch = |
5096 | all_specs.GetModuleSpecRefAtIndex(i).GetArchitecture(); |
5097 | if (module_arch.IsCompatibleMatch(rhs: mach_arch)) |
5098 | return mach_arch; |
5099 | } |
5100 | } |
5101 | |
5102 | // Return the first arch we found. |
5103 | if (all_specs.GetSize() == 0) |
5104 | return {}; |
5105 | return all_specs.GetModuleSpecRefAtIndex(i: 0).GetArchitecture(); |
5106 | } |
5107 | |
5108 | UUID ObjectFileMachO::GetUUID() { |
5109 | ModuleSP module_sp(GetModule()); |
5110 | if (module_sp) { |
5111 | std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); |
5112 | lldb::offset_t offset = MachHeaderSizeFromMagic(magic: m_header.magic); |
5113 | return GetUUID(header: m_header, data: m_data, lc_offset: offset); |
5114 | } |
5115 | return UUID(); |
5116 | } |
5117 | |
5118 | uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { |
5119 | uint32_t count = 0; |
5120 | ModuleSP module_sp(GetModule()); |
5121 | if (module_sp) { |
5122 | std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); |
5123 | llvm::MachO::load_command load_cmd; |
5124 | lldb::offset_t offset = MachHeaderSizeFromMagic(magic: m_header.magic); |
5125 | std::vector<std::string> rpath_paths; |
5126 | std::vector<std::string> rpath_relative_paths; |
5127 | std::vector<std::string> at_exec_relative_paths; |
5128 | uint32_t i; |
5129 | for (i = 0; i < m_header.ncmds; ++i) { |
5130 | const uint32_t cmd_offset = offset; |
5131 | if (m_data.GetU32(offset_ptr: &offset, dst: &load_cmd, count: 2) == nullptr) |
5132 | break; |
5133 | |
5134 | switch (load_cmd.cmd) { |
5135 | case LC_RPATH: |
5136 | case LC_LOAD_DYLIB: |
5137 | case LC_LOAD_WEAK_DYLIB: |
5138 | case LC_REEXPORT_DYLIB: |
5139 | case LC_LOAD_DYLINKER: |
5140 | case LC_LOADFVMLIB: |
5141 | case LC_LOAD_UPWARD_DYLIB: { |
5142 | uint32_t name_offset = cmd_offset + m_data.GetU32(offset_ptr: &offset); |
5143 | const char *path = m_data.PeekCStr(offset: name_offset); |
5144 | if (path) { |
5145 | if (load_cmd.cmd == LC_RPATH) |
5146 | rpath_paths.push_back(x: path); |
5147 | else { |
5148 | if (path[0] == '@') { |
5149 | if (strncmp(s1: path, s2: "@rpath" , n: strlen(s: "@rpath" )) == 0) |
5150 | rpath_relative_paths.push_back(x: path + strlen(s: "@rpath" )); |
5151 | else if (strncmp(s1: path, s2: "@executable_path" , |
5152 | n: strlen(s: "@executable_path" )) == 0) |
5153 | at_exec_relative_paths.push_back(x: path + |
5154 | strlen(s: "@executable_path" )); |
5155 | } else { |
5156 | FileSpec file_spec(path); |
5157 | if (files.AppendIfUnique(file: file_spec)) |
5158 | count++; |
5159 | } |
5160 | } |
5161 | } |
5162 | } break; |
5163 | |
5164 | default: |
5165 | break; |
5166 | } |
5167 | offset = cmd_offset + load_cmd.cmdsize; |
5168 | } |
5169 | |
5170 | FileSpec this_file_spec(m_file); |
5171 | FileSystem::Instance().Resolve(file_spec&: this_file_spec); |
5172 | |
5173 | if (!rpath_paths.empty()) { |
5174 | // Fixup all LC_RPATH values to be absolute paths |
5175 | std::string loader_path("@loader_path" ); |
5176 | std::string executable_path("@executable_path" ); |
5177 | for (auto &rpath : rpath_paths) { |
5178 | if (llvm::StringRef(rpath).starts_with(Prefix: loader_path)) { |
5179 | rpath.erase(pos: 0, n: loader_path.size()); |
5180 | rpath.insert(pos: 0, s: this_file_spec.GetDirectory().GetCString()); |
5181 | } else if (llvm::StringRef(rpath).starts_with(Prefix: executable_path)) { |
5182 | rpath.erase(pos: 0, n: executable_path.size()); |
5183 | rpath.insert(pos: 0, s: this_file_spec.GetDirectory().GetCString()); |
5184 | } |
5185 | } |
5186 | |
5187 | for (const auto &rpath_relative_path : rpath_relative_paths) { |
5188 | for (const auto &rpath : rpath_paths) { |
5189 | std::string path = rpath; |
5190 | path += rpath_relative_path; |
5191 | // It is OK to resolve this path because we must find a file on disk |
5192 | // for us to accept it anyway if it is rpath relative. |
5193 | FileSpec file_spec(path); |
5194 | FileSystem::Instance().Resolve(file_spec); |
5195 | if (FileSystem::Instance().Exists(file_spec) && |
5196 | files.AppendIfUnique(file: file_spec)) { |
5197 | count++; |
5198 | break; |
5199 | } |
5200 | } |
5201 | } |
5202 | } |
5203 | |
5204 | // We may have @executable_paths but no RPATHS. Figure those out here. |
5205 | // Only do this if this object file is the executable. We have no way to |
5206 | // get back to the actual executable otherwise, so we won't get the right |
5207 | // path. |
5208 | if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { |
5209 | FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); |
5210 | for (const auto &at_exec_relative_path : at_exec_relative_paths) { |
5211 | FileSpec file_spec = |
5212 | exec_dir.CopyByAppendingPathComponent(component: at_exec_relative_path); |
5213 | if (FileSystem::Instance().Exists(file_spec) && |
5214 | files.AppendIfUnique(file: file_spec)) |
5215 | count++; |
5216 | } |
5217 | } |
5218 | } |
5219 | return count; |
5220 | } |
5221 | |
5222 | lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { |
5223 | // If the object file is not an executable it can't hold the entry point. |
5224 | // m_entry_point_address is initialized to an invalid address, so we can just |
5225 | // return that. If m_entry_point_address is valid it means we've found it |
5226 | // already, so return the cached value. |
5227 | |
5228 | if ((!IsExecutable() && !IsDynamicLoader()) || |
5229 | m_entry_point_address.IsValid()) { |
5230 | return m_entry_point_address; |
5231 | } |
5232 | |
5233 | // Otherwise, look for the UnixThread or Thread command. The data for the |
5234 | // Thread command is given in /usr/include/mach-o.h, but it is basically: |
5235 | // |
5236 | // uint32_t flavor - this is the flavor argument you would pass to |
5237 | // thread_get_state |
5238 | // uint32_t count - this is the count of longs in the thread state data |
5239 | // struct XXX_thread_state state - this is the structure from |
5240 | // <machine/thread_status.h> corresponding to the flavor. |
5241 | // <repeat this trio> |
5242 | // |
5243 | // So we just keep reading the various register flavors till we find the GPR |
5244 | // one, then read the PC out of there. |
5245 | // FIXME: We will need to have a "RegisterContext data provider" class at some |
5246 | // point that can get all the registers |
5247 | // out of data in this form & attach them to a given thread. That should |
5248 | // underlie the MacOS X User process plugin, and we'll also need it for the |
5249 | // MacOS X Core File process plugin. When we have that we can also use it |
5250 | // here. |
5251 | // |
5252 | // For now we hard-code the offsets and flavors we need: |
5253 | // |
5254 | // |
5255 | |
5256 | ModuleSP module_sp(GetModule()); |
5257 | if (module_sp) { |
5258 | std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); |
5259 | llvm::MachO::load_command load_cmd; |
5260 | lldb::offset_t offset = MachHeaderSizeFromMagic(magic: m_header.magic); |
5261 | uint32_t i; |
5262 | lldb::addr_t start_address = LLDB_INVALID_ADDRESS; |
5263 | bool done = false; |
5264 | |
5265 | for (i = 0; i < m_header.ncmds; ++i) { |
5266 | const lldb::offset_t cmd_offset = offset; |
5267 | if (m_data.GetU32(offset_ptr: &offset, dst: &load_cmd, count: 2) == nullptr) |
5268 | break; |
5269 | |
5270 | switch (load_cmd.cmd) { |
5271 | case LC_UNIXTHREAD: |
5272 | case LC_THREAD: { |
5273 | while (offset < cmd_offset + load_cmd.cmdsize) { |
5274 | uint32_t flavor = m_data.GetU32(offset_ptr: &offset); |
5275 | uint32_t count = m_data.GetU32(offset_ptr: &offset); |
5276 | if (count == 0) { |
5277 | // We've gotten off somehow, log and exit; |
5278 | return m_entry_point_address; |
5279 | } |
5280 | |
5281 | switch (m_header.cputype) { |
5282 | case llvm::MachO::CPU_TYPE_ARM: |
5283 | if (flavor == 1 || |
5284 | flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 |
5285 | // from mach/arm/thread_status.h |
5286 | { |
5287 | offset += 60; // This is the offset of pc in the GPR thread state |
5288 | // data structure. |
5289 | start_address = m_data.GetU32(offset_ptr: &offset); |
5290 | done = true; |
5291 | } |
5292 | break; |
5293 | case llvm::MachO::CPU_TYPE_ARM64: |
5294 | case llvm::MachO::CPU_TYPE_ARM64_32: |
5295 | if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h |
5296 | { |
5297 | offset += 256; // This is the offset of pc in the GPR thread state |
5298 | // data structure. |
5299 | start_address = m_data.GetU64(offset_ptr: &offset); |
5300 | done = true; |
5301 | } |
5302 | break; |
5303 | case llvm::MachO::CPU_TYPE_I386: |
5304 | if (flavor == |
5305 | 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h |
5306 | { |
5307 | offset += 40; // This is the offset of eip in the GPR thread state |
5308 | // data structure. |
5309 | start_address = m_data.GetU32(offset_ptr: &offset); |
5310 | done = true; |
5311 | } |
5312 | break; |
5313 | case llvm::MachO::CPU_TYPE_X86_64: |
5314 | if (flavor == |
5315 | 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h |
5316 | { |
5317 | offset += 16 * 8; // This is the offset of rip in the GPR thread |
5318 | // state data structure. |
5319 | start_address = m_data.GetU64(offset_ptr: &offset); |
5320 | done = true; |
5321 | } |
5322 | break; |
5323 | default: |
5324 | return m_entry_point_address; |
5325 | } |
5326 | // Haven't found the GPR flavor yet, skip over the data for this |
5327 | // flavor: |
5328 | if (done) |
5329 | break; |
5330 | offset += count * 4; |
5331 | } |
5332 | } break; |
5333 | case LC_MAIN: { |
5334 | uint64_t entryoffset = m_data.GetU64(offset_ptr: &offset); |
5335 | SectionSP text_segment_sp = |
5336 | GetSectionList()->FindSectionByName(section_dstr: GetSegmentNameTEXT()); |
5337 | if (text_segment_sp) { |
5338 | done = true; |
5339 | start_address = text_segment_sp->GetFileAddress() + entryoffset; |
5340 | } |
5341 | } break; |
5342 | |
5343 | default: |
5344 | break; |
5345 | } |
5346 | if (done) |
5347 | break; |
5348 | |
5349 | // Go to the next load command: |
5350 | offset = cmd_offset + load_cmd.cmdsize; |
5351 | } |
5352 | |
5353 | if (start_address == LLDB_INVALID_ADDRESS && IsDynamicLoader()) { |
5354 | if (GetSymtab()) { |
5355 | Symbol *dyld_start_sym = GetSymtab()->FindFirstSymbolWithNameAndType( |
5356 | name: ConstString("_dyld_start" ), symbol_type: SymbolType::eSymbolTypeCode, |
5357 | symbol_debug_type: Symtab::eDebugAny, symbol_visibility: Symtab::eVisibilityAny); |
5358 | if (dyld_start_sym && dyld_start_sym->GetAddress().IsValid()) { |
5359 | start_address = dyld_start_sym->GetAddress().GetFileAddress(); |
5360 | } |
5361 | } |
5362 | } |
5363 | |
5364 | if (start_address != LLDB_INVALID_ADDRESS) { |
5365 | // We got the start address from the load commands, so now resolve that |
5366 | // address in the sections of this ObjectFile: |
5367 | if (!m_entry_point_address.ResolveAddressUsingFileSections( |
5368 | addr: start_address, sections: GetSectionList())) { |
5369 | m_entry_point_address.Clear(); |
5370 | } |
5371 | } else { |
5372 | // We couldn't read the UnixThread load command - maybe it wasn't there. |
5373 | // As a fallback look for the "start" symbol in the main executable. |
5374 | |
5375 | ModuleSP module_sp(GetModule()); |
5376 | |
5377 | if (module_sp) { |
5378 | SymbolContextList contexts; |
5379 | SymbolContext context; |
5380 | module_sp->FindSymbolsWithNameAndType(name: ConstString("start" ), |
5381 | symbol_type: eSymbolTypeCode, sc_list&: contexts); |
5382 | if (contexts.GetSize()) { |
5383 | if (contexts.GetContextAtIndex(idx: 0, sc&: context)) |
5384 | m_entry_point_address = context.symbol->GetAddress(); |
5385 | } |
5386 | } |
5387 | } |
5388 | } |
5389 | |
5390 | return m_entry_point_address; |
5391 | } |
5392 | |
5393 | lldb_private::Address ObjectFileMachO::GetBaseAddress() { |
5394 | lldb_private::Address ; |
5395 | SectionList *section_list = GetSectionList(); |
5396 | if (section_list) { |
5397 | SectionSP text_segment_sp( |
5398 | section_list->FindSectionByName(section_dstr: GetSegmentNameTEXT())); |
5399 | if (text_segment_sp) { |
5400 | header_addr.SetSection(text_segment_sp); |
5401 | header_addr.SetOffset(0); |
5402 | } |
5403 | } |
5404 | return header_addr; |
5405 | } |
5406 | |
5407 | uint32_t ObjectFileMachO::GetNumThreadContexts() { |
5408 | ModuleSP module_sp(GetModule()); |
5409 | if (module_sp) { |
5410 | std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); |
5411 | if (!m_thread_context_offsets_valid) { |
5412 | m_thread_context_offsets_valid = true; |
5413 | lldb::offset_t offset = MachHeaderSizeFromMagic(magic: m_header.magic); |
5414 | FileRangeArray::Entry file_range; |
5415 | llvm::MachO::thread_command thread_cmd; |
5416 | for (uint32_t i = 0; i < m_header.ncmds; ++i) { |
5417 | const uint32_t cmd_offset = offset; |
5418 | if (m_data.GetU32(offset_ptr: &offset, dst: &thread_cmd, count: 2) == nullptr) |
5419 | break; |
5420 | |
5421 | if (thread_cmd.cmd == LC_THREAD) { |
5422 | file_range.SetRangeBase(offset); |
5423 | file_range.SetByteSize(thread_cmd.cmdsize - 8); |
5424 | m_thread_context_offsets.Append(entry: file_range); |
5425 | } |
5426 | offset = cmd_offset + thread_cmd.cmdsize; |
5427 | } |
5428 | } |
5429 | } |
5430 | return m_thread_context_offsets.GetSize(); |
5431 | } |
5432 | |
5433 | std::vector<std::tuple<offset_t, offset_t>> |
5434 | ObjectFileMachO::FindLC_NOTEByName(std::string name) { |
5435 | std::vector<std::tuple<offset_t, offset_t>> results; |
5436 | ModuleSP module_sp(GetModule()); |
5437 | if (module_sp) { |
5438 | std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); |
5439 | |
5440 | offset_t offset = MachHeaderSizeFromMagic(magic: m_header.magic); |
5441 | for (uint32_t i = 0; i < m_header.ncmds; ++i) { |
5442 | const uint32_t cmd_offset = offset; |
5443 | llvm::MachO::load_command lc = {}; |
5444 | if (m_data.GetU32(offset_ptr: &offset, dst: &lc.cmd, count: 2) == nullptr) |
5445 | break; |
5446 | if (lc.cmd == LC_NOTE) { |
5447 | char data_owner[17]; |
5448 | m_data.CopyData(offset, length: 16, dst: data_owner); |
5449 | data_owner[16] = '\0'; |
5450 | offset += 16; |
5451 | |
5452 | if (name == data_owner) { |
5453 | offset_t payload_offset = m_data.GetU64_unchecked(offset_ptr: &offset); |
5454 | offset_t payload_size = m_data.GetU64_unchecked(offset_ptr: &offset); |
5455 | results.push_back(x: {payload_offset, payload_size}); |
5456 | } |
5457 | } |
5458 | offset = cmd_offset + lc.cmdsize; |
5459 | } |
5460 | } |
5461 | return results; |
5462 | } |
5463 | |
5464 | std::string ObjectFileMachO::GetIdentifierString() { |
5465 | Log *log( |
5466 | GetLog(mask: LLDBLog::Symbols | LLDBLog::Process | LLDBLog::DynamicLoader)); |
5467 | ModuleSP module_sp(GetModule()); |
5468 | if (module_sp) { |
5469 | std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); |
5470 | |
5471 | auto lc_notes = FindLC_NOTEByName(name: "kern ver str" ); |
5472 | for (auto lc_note : lc_notes) { |
5473 | offset_t payload_offset = std::get<0>(t&: lc_note); |
5474 | offset_t payload_size = std::get<1>(t&: lc_note); |
5475 | uint32_t version; |
5476 | if (m_data.GetU32(offset_ptr: &payload_offset, dst: &version, count: 1) != nullptr) { |
5477 | if (version == 1) { |
5478 | uint32_t strsize = payload_size - sizeof(uint32_t); |
5479 | std::string result(strsize, '\0'); |
5480 | m_data.CopyData(offset: payload_offset, length: strsize, dst: result.data()); |
5481 | LLDB_LOGF(log, "LC_NOTE 'kern ver str' found with text '%s'" , |
5482 | result.c_str()); |
5483 | return result; |
5484 | } |
5485 | } |
5486 | } |
5487 | |
5488 | // Second, make a pass over the load commands looking for an obsolete |
5489 | // LC_IDENT load command. |
5490 | offset_t offset = MachHeaderSizeFromMagic(magic: m_header.magic); |
5491 | for (uint32_t i = 0; i < m_header.ncmds; ++i) { |
5492 | const uint32_t cmd_offset = offset; |
5493 | llvm::MachO::ident_command ident_command; |
5494 | if (m_data.GetU32(offset_ptr: &offset, dst: &ident_command, count: 2) == nullptr) |
5495 | break; |
5496 | if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { |
5497 | std::string result(ident_command.cmdsize, '\0'); |
5498 | if (m_data.CopyData(offset, length: ident_command.cmdsize, dst: result.data()) == |
5499 | ident_command.cmdsize) { |
5500 | LLDB_LOGF(log, "LC_IDENT found with text '%s'" , result.c_str()); |
5501 | return result; |
5502 | } |
5503 | } |
5504 | offset = cmd_offset + ident_command.cmdsize; |
5505 | } |
5506 | } |
5507 | return {}; |
5508 | } |
5509 | |
5510 | AddressableBits ObjectFileMachO::GetAddressableBits() { |
5511 | AddressableBits addressable_bits; |
5512 | |
5513 | Log *log(GetLog(mask: LLDBLog::Process)); |
5514 | ModuleSP module_sp(GetModule()); |
5515 | if (module_sp) { |
5516 | std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); |
5517 | auto lc_notes = FindLC_NOTEByName(name: "addrable bits" ); |
5518 | for (auto lc_note : lc_notes) { |
5519 | offset_t payload_offset = std::get<0>(t&: lc_note); |
5520 | uint32_t version; |
5521 | if (m_data.GetU32(offset_ptr: &payload_offset, dst: &version, count: 1) != nullptr) { |
5522 | if (version == 3) { |
5523 | uint32_t num_addr_bits = m_data.GetU32_unchecked(offset_ptr: &payload_offset); |
5524 | addressable_bits.SetAddressableBits(num_addr_bits); |
5525 | LLDB_LOGF(log, |
5526 | "LC_NOTE 'addrable bits' v3 found, value %d " |
5527 | "bits" , |
5528 | num_addr_bits); |
5529 | } |
5530 | if (version == 4) { |
5531 | uint32_t lo_addr_bits = m_data.GetU32_unchecked(offset_ptr: &payload_offset); |
5532 | uint32_t hi_addr_bits = m_data.GetU32_unchecked(offset_ptr: &payload_offset); |
5533 | |
5534 | if (lo_addr_bits == hi_addr_bits) |
5535 | addressable_bits.SetAddressableBits(lo_addr_bits); |
5536 | else |
5537 | addressable_bits.SetAddressableBits(lowmem_addressing_bits: lo_addr_bits, highmem_addressing_bits: hi_addr_bits); |
5538 | LLDB_LOGF(log, "LC_NOTE 'addrable bits' v4 found, value %d & %d bits" , |
5539 | lo_addr_bits, hi_addr_bits); |
5540 | } |
5541 | } |
5542 | } |
5543 | } |
5544 | return addressable_bits; |
5545 | } |
5546 | |
5547 | bool ObjectFileMachO::GetCorefileMainBinaryInfo(addr_t &value, |
5548 | bool &value_is_offset, |
5549 | UUID &uuid, |
5550 | ObjectFile::BinaryType &type) { |
5551 | Log *log( |
5552 | GetLog(mask: LLDBLog::Symbols | LLDBLog::Process | LLDBLog::DynamicLoader)); |
5553 | value = LLDB_INVALID_ADDRESS; |
5554 | value_is_offset = false; |
5555 | uuid.Clear(); |
5556 | uint32_t log2_pagesize = 0; // not currently passed up to caller |
5557 | uint32_t platform = 0; // not currently passed up to caller |
5558 | ModuleSP module_sp(GetModule()); |
5559 | if (module_sp) { |
5560 | std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); |
5561 | |
5562 | auto lc_notes = FindLC_NOTEByName(name: "main bin spec" ); |
5563 | for (auto lc_note : lc_notes) { |
5564 | offset_t payload_offset = std::get<0>(t&: lc_note); |
5565 | |
5566 | // struct main_bin_spec |
5567 | // { |
5568 | // uint32_t version; // currently 2 |
5569 | // uint32_t type; // 0 == unspecified, 1 == kernel, |
5570 | // // 2 == user process, |
5571 | // // 3 == standalone binary |
5572 | // uint64_t address; // UINT64_MAX if address not specified |
5573 | // uint64_t slide; // slide, UINT64_MAX if unspecified |
5574 | // // 0 if no slide needs to be applied to |
5575 | // // file address |
5576 | // uuid_t uuid; // all zero's if uuid not specified |
5577 | // uint32_t log2_pagesize; // process page size in log base 2, |
5578 | // // e.g. 4k pages are 12. |
5579 | // // 0 for unspecified |
5580 | // uint32_t platform; // The Mach-O platform for this corefile. |
5581 | // // 0 for unspecified. |
5582 | // // The values are defined in |
5583 | // // <mach-o/loader.h>, PLATFORM_*. |
5584 | // } __attribute((packed)); |
5585 | |
5586 | // "main bin spec" (main binary specification) data payload is |
5587 | // formatted: |
5588 | // uint32_t version [currently 1] |
5589 | // uint32_t type [0 == unspecified, 1 == kernel, |
5590 | // 2 == user process, 3 == firmware ] |
5591 | // uint64_t address [ UINT64_MAX if address not specified ] |
5592 | // uuid_t uuid [ all zero's if uuid not specified ] |
5593 | // uint32_t log2_pagesize [ process page size in log base |
5594 | // 2, e.g. 4k pages are 12. |
5595 | // 0 for unspecified ] |
5596 | // uint32_t unused [ for alignment ] |
5597 | |
5598 | uint32_t version; |
5599 | if (m_data.GetU32(offset_ptr: &payload_offset, dst: &version, count: 1) != nullptr && |
5600 | version <= 2) { |
5601 | uint32_t binspec_type = 0; |
5602 | uuid_t raw_uuid; |
5603 | memset(s: raw_uuid, c: 0, n: sizeof(uuid_t)); |
5604 | |
5605 | if (!m_data.GetU32(offset_ptr: &payload_offset, dst: &binspec_type, count: 1)) |
5606 | return false; |
5607 | if (!m_data.GetU64(offset_ptr: &payload_offset, dst: &value, count: 1)) |
5608 | return false; |
5609 | uint64_t slide = LLDB_INVALID_ADDRESS; |
5610 | if (version > 1 && !m_data.GetU64(offset_ptr: &payload_offset, dst: &slide, count: 1)) |
5611 | return false; |
5612 | if (value == LLDB_INVALID_ADDRESS && slide != LLDB_INVALID_ADDRESS) { |
5613 | value = slide; |
5614 | value_is_offset = true; |
5615 | } |
5616 | |
5617 | if (m_data.CopyData(offset: payload_offset, length: sizeof(uuid_t), dst: raw_uuid) != 0) { |
5618 | uuid = UUID(raw_uuid, sizeof(uuid_t)); |
5619 | // convert the "main bin spec" type into our |
5620 | // ObjectFile::BinaryType enum |
5621 | const char *typestr = "unrecognized type" ; |
5622 | switch (binspec_type) { |
5623 | case 0: |
5624 | type = eBinaryTypeUnknown; |
5625 | typestr = "uknown" ; |
5626 | break; |
5627 | case 1: |
5628 | type = eBinaryTypeKernel; |
5629 | typestr = "xnu kernel" ; |
5630 | break; |
5631 | case 2: |
5632 | type = eBinaryTypeUser; |
5633 | typestr = "userland dyld" ; |
5634 | break; |
5635 | case 3: |
5636 | type = eBinaryTypeStandalone; |
5637 | typestr = "standalone" ; |
5638 | break; |
5639 | } |
5640 | LLDB_LOGF(log, |
5641 | "LC_NOTE 'main bin spec' found, version %d type %d " |
5642 | "(%s), value 0x%" PRIx64 " value-is-slide==%s uuid %s" , |
5643 | version, type, typestr, value, |
5644 | value_is_offset ? "true" : "false" , |
5645 | uuid.GetAsString().c_str()); |
5646 | if (!m_data.GetU32(offset_ptr: &payload_offset, dst: &log2_pagesize, count: 1)) |
5647 | return false; |
5648 | if (version > 1 && !m_data.GetU32(offset_ptr: &payload_offset, dst: &platform, count: 1)) |
5649 | return false; |
5650 | return true; |
5651 | } |
5652 | } |
5653 | } |
5654 | } |
5655 | return false; |
5656 | } |
5657 | |
5658 | bool ObjectFileMachO::(std::vector<tid_t> &tids) { |
5659 | tids.clear(); |
5660 | ModuleSP module_sp(GetModule()); |
5661 | if (module_sp) { |
5662 | std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); |
5663 | |
5664 | Log *log(GetLog(mask: LLDBLog::Object | LLDBLog::Process | LLDBLog::Thread)); |
5665 | auto lc_notes = FindLC_NOTEByName(name: "process metadata" ); |
5666 | for (auto lc_note : lc_notes) { |
5667 | offset_t payload_offset = std::get<0>(t&: lc_note); |
5668 | offset_t strsize = std::get<1>(t&: lc_note); |
5669 | std::string buf(strsize, '\0'); |
5670 | if (m_data.CopyData(offset: payload_offset, length: strsize, dst: buf.data()) != strsize) { |
5671 | LLDB_LOGF(log, |
5672 | "Unable to read %" PRIu64 |
5673 | " bytes of 'process metadata' LC_NOTE JSON contents" , |
5674 | strsize); |
5675 | return false; |
5676 | } |
5677 | while (buf.back() == '\0') |
5678 | buf.resize(n: buf.size() - 1); |
5679 | StructuredData::ObjectSP object_sp = StructuredData::ParseJSON(json_text: buf); |
5680 | StructuredData::Dictionary *dict = object_sp->GetAsDictionary(); |
5681 | if (!dict) { |
5682 | LLDB_LOGF(log, "Unable to read 'process metadata' LC_NOTE, did not " |
5683 | "get a dictionary." ); |
5684 | return false; |
5685 | } |
5686 | StructuredData::Array *threads; |
5687 | if (!dict->GetValueForKeyAsArray(key: "threads" , result&: threads) || !threads) { |
5688 | LLDB_LOGF(log, |
5689 | "'process metadata' LC_NOTE does not have a 'threads' key" ); |
5690 | return false; |
5691 | } |
5692 | if (threads->GetSize() != GetNumThreadContexts()) { |
5693 | LLDB_LOGF(log, "Unable to read 'process metadata' LC_NOTE, number of " |
5694 | "threads does not match number of LC_THREADS." ); |
5695 | return false; |
5696 | } |
5697 | const size_t num_threads = threads->GetSize(); |
5698 | for (size_t i = 0; i < num_threads; i++) { |
5699 | std::optional<StructuredData::Dictionary *> maybe_thread = |
5700 | threads->GetItemAtIndexAsDictionary(idx: i); |
5701 | if (!maybe_thread) { |
5702 | LLDB_LOGF(log, |
5703 | "Unable to read 'process metadata' LC_NOTE, threads " |
5704 | "array does not have a dictionary at index %zu." , |
5705 | i); |
5706 | return false; |
5707 | } |
5708 | StructuredData::Dictionary *thread = *maybe_thread; |
5709 | tid_t tid = LLDB_INVALID_THREAD_ID; |
5710 | if (thread->GetValueForKeyAsInteger<tid_t>(key: "thread_id" , result&: tid)) |
5711 | if (tid == 0) |
5712 | tid = LLDB_INVALID_THREAD_ID; |
5713 | tids.push_back(x: tid); |
5714 | } |
5715 | |
5716 | if (log) { |
5717 | StreamString logmsg; |
5718 | logmsg.Printf(format: "LC_NOTE 'process metadata' found: " ); |
5719 | dict->Dump(s&: logmsg, /* pretty_print */ false); |
5720 | LLDB_LOGF(log, "%s" , logmsg.GetData()); |
5721 | } |
5722 | return true; |
5723 | } |
5724 | } |
5725 | return false; |
5726 | } |
5727 | |
5728 | lldb::RegisterContextSP |
5729 | ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, |
5730 | lldb_private::Thread &thread) { |
5731 | lldb::RegisterContextSP reg_ctx_sp; |
5732 | |
5733 | ModuleSP module_sp(GetModule()); |
5734 | if (module_sp) { |
5735 | std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); |
5736 | if (!m_thread_context_offsets_valid) |
5737 | GetNumThreadContexts(); |
5738 | |
5739 | const FileRangeArray::Entry *thread_context_file_range = |
5740 | m_thread_context_offsets.GetEntryAtIndex(i: idx); |
5741 | if (thread_context_file_range) { |
5742 | |
5743 | DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), |
5744 | thread_context_file_range->GetByteSize()); |
5745 | |
5746 | switch (m_header.cputype) { |
5747 | case llvm::MachO::CPU_TYPE_ARM64: |
5748 | case llvm::MachO::CPU_TYPE_ARM64_32: |
5749 | reg_ctx_sp = |
5750 | std::make_shared<RegisterContextDarwin_arm64_Mach>(args&: thread, args&: data); |
5751 | break; |
5752 | |
5753 | case llvm::MachO::CPU_TYPE_ARM: |
5754 | reg_ctx_sp = |
5755 | std::make_shared<RegisterContextDarwin_arm_Mach>(args&: thread, args&: data); |
5756 | break; |
5757 | |
5758 | case llvm::MachO::CPU_TYPE_I386: |
5759 | reg_ctx_sp = |
5760 | std::make_shared<RegisterContextDarwin_i386_Mach>(args&: thread, args&: data); |
5761 | break; |
5762 | |
5763 | case llvm::MachO::CPU_TYPE_X86_64: |
5764 | reg_ctx_sp = |
5765 | std::make_shared<RegisterContextDarwin_x86_64_Mach>(args&: thread, args&: data); |
5766 | break; |
5767 | } |
5768 | } |
5769 | } |
5770 | return reg_ctx_sp; |
5771 | } |
5772 | |
5773 | ObjectFile::Type ObjectFileMachO::CalculateType() { |
5774 | switch (m_header.filetype) { |
5775 | case MH_OBJECT: // 0x1u |
5776 | if (GetAddressByteSize() == 4) { |
5777 | // 32 bit kexts are just object files, but they do have a valid |
5778 | // UUID load command. |
5779 | if (GetUUID()) { |
5780 | // this checking for the UUID load command is not enough we could |
5781 | // eventually look for the symbol named "OSKextGetCurrentIdentifier" as |
5782 | // this is required of kexts |
5783 | if (m_strata == eStrataInvalid) |
5784 | m_strata = eStrataKernel; |
5785 | return eTypeSharedLibrary; |
5786 | } |
5787 | } |
5788 | return eTypeObjectFile; |
5789 | |
5790 | case MH_EXECUTE: |
5791 | return eTypeExecutable; // 0x2u |
5792 | case MH_FVMLIB: |
5793 | return eTypeSharedLibrary; // 0x3u |
5794 | case MH_CORE: |
5795 | return eTypeCoreFile; // 0x4u |
5796 | case MH_PRELOAD: |
5797 | return eTypeSharedLibrary; // 0x5u |
5798 | case MH_DYLIB: |
5799 | return eTypeSharedLibrary; // 0x6u |
5800 | case MH_DYLINKER: |
5801 | return eTypeDynamicLinker; // 0x7u |
5802 | case MH_BUNDLE: |
5803 | return eTypeSharedLibrary; // 0x8u |
5804 | case MH_DYLIB_STUB: |
5805 | return eTypeStubLibrary; // 0x9u |
5806 | case MH_DSYM: |
5807 | return eTypeDebugInfo; // 0xAu |
5808 | case MH_KEXT_BUNDLE: |
5809 | return eTypeSharedLibrary; // 0xBu |
5810 | default: |
5811 | break; |
5812 | } |
5813 | return eTypeUnknown; |
5814 | } |
5815 | |
5816 | ObjectFile::Strata ObjectFileMachO::CalculateStrata() { |
5817 | switch (m_header.filetype) { |
5818 | case MH_OBJECT: // 0x1u |
5819 | { |
5820 | // 32 bit kexts are just object files, but they do have a valid |
5821 | // UUID load command. |
5822 | if (GetUUID()) { |
5823 | // this checking for the UUID load command is not enough we could |
5824 | // eventually look for the symbol named "OSKextGetCurrentIdentifier" as |
5825 | // this is required of kexts |
5826 | if (m_type == eTypeInvalid) |
5827 | m_type = eTypeSharedLibrary; |
5828 | |
5829 | return eStrataKernel; |
5830 | } |
5831 | } |
5832 | return eStrataUnknown; |
5833 | |
5834 | case MH_EXECUTE: // 0x2u |
5835 | // Check for the MH_DYLDLINK bit in the flags |
5836 | if (m_header.flags & MH_DYLDLINK) { |
5837 | return eStrataUser; |
5838 | } else { |
5839 | SectionList *section_list = GetSectionList(); |
5840 | if (section_list) { |
5841 | static ConstString g_kld_section_name("__KLD" ); |
5842 | if (section_list->FindSectionByName(section_dstr: g_kld_section_name)) |
5843 | return eStrataKernel; |
5844 | } |
5845 | } |
5846 | return eStrataRawImage; |
5847 | |
5848 | case MH_FVMLIB: |
5849 | return eStrataUser; // 0x3u |
5850 | case MH_CORE: |
5851 | return eStrataUnknown; // 0x4u |
5852 | case MH_PRELOAD: |
5853 | return eStrataRawImage; // 0x5u |
5854 | case MH_DYLIB: |
5855 | return eStrataUser; // 0x6u |
5856 | case MH_DYLINKER: |
5857 | return eStrataUser; // 0x7u |
5858 | case MH_BUNDLE: |
5859 | return eStrataUser; // 0x8u |
5860 | case MH_DYLIB_STUB: |
5861 | return eStrataUser; // 0x9u |
5862 | case MH_DSYM: |
5863 | return eStrataUnknown; // 0xAu |
5864 | case MH_KEXT_BUNDLE: |
5865 | return eStrataKernel; // 0xBu |
5866 | default: |
5867 | break; |
5868 | } |
5869 | return eStrataUnknown; |
5870 | } |
5871 | |
5872 | llvm::VersionTuple ObjectFileMachO::GetVersion() { |
5873 | ModuleSP module_sp(GetModule()); |
5874 | if (module_sp) { |
5875 | std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); |
5876 | llvm::MachO::dylib_command load_cmd; |
5877 | lldb::offset_t offset = MachHeaderSizeFromMagic(magic: m_header.magic); |
5878 | uint32_t version_cmd = 0; |
5879 | uint64_t version = 0; |
5880 | uint32_t i; |
5881 | for (i = 0; i < m_header.ncmds; ++i) { |
5882 | const lldb::offset_t cmd_offset = offset; |
5883 | if (m_data.GetU32(offset_ptr: &offset, dst: &load_cmd, count: 2) == nullptr) |
5884 | break; |
5885 | |
5886 | if (load_cmd.cmd == LC_ID_DYLIB) { |
5887 | if (version_cmd == 0) { |
5888 | version_cmd = load_cmd.cmd; |
5889 | if (m_data.GetU32(offset_ptr: &offset, dst: &load_cmd.dylib, count: 4) == nullptr) |
5890 | break; |
5891 | version = load_cmd.dylib.current_version; |
5892 | } |
5893 | break; // Break for now unless there is another more complete version |
5894 | // number load command in the future. |
5895 | } |
5896 | offset = cmd_offset + load_cmd.cmdsize; |
5897 | } |
5898 | |
5899 | if (version_cmd == LC_ID_DYLIB) { |
5900 | unsigned major = (version & 0xFFFF0000ull) >> 16; |
5901 | unsigned minor = (version & 0x0000FF00ull) >> 8; |
5902 | unsigned subminor = (version & 0x000000FFull); |
5903 | return llvm::VersionTuple(major, minor, subminor); |
5904 | } |
5905 | } |
5906 | return llvm::VersionTuple(); |
5907 | } |
5908 | |
5909 | ArchSpec ObjectFileMachO::GetArchitecture() { |
5910 | ModuleSP module_sp(GetModule()); |
5911 | ArchSpec arch; |
5912 | if (module_sp) { |
5913 | std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); |
5914 | |
5915 | return GetArchitecture(module_sp, header: m_header, data: m_data, |
5916 | lc_offset: MachHeaderSizeFromMagic(magic: m_header.magic)); |
5917 | } |
5918 | return arch; |
5919 | } |
5920 | |
5921 | void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, |
5922 | addr_t &base_addr, UUID &uuid) { |
5923 | uuid.Clear(); |
5924 | base_addr = LLDB_INVALID_ADDRESS; |
5925 | if (process && process->GetDynamicLoader()) { |
5926 | DynamicLoader *dl = process->GetDynamicLoader(); |
5927 | LazyBool using_shared_cache; |
5928 | LazyBool private_shared_cache; |
5929 | dl->GetSharedCacheInformation(base_address&: base_addr, uuid, using_shared_cache, |
5930 | private_shared_cache); |
5931 | } |
5932 | Log *log(GetLog(mask: LLDBLog::Symbols | LLDBLog::Process)); |
5933 | LLDB_LOGF( |
5934 | log, |
5935 | "inferior process shared cache has a UUID of %s, base address 0x%" PRIx64, |
5936 | uuid.GetAsString().c_str(), base_addr); |
5937 | } |
5938 | |
5939 | // From dyld SPI header dyld_process_info.h |
5940 | typedef void *dyld_process_info; |
5941 | struct lldb_copy__dyld_process_cache_info { |
5942 | uuid_t cacheUUID; // UUID of cache used by process |
5943 | uint64_t cacheBaseAddress; // load address of dyld shared cache |
5944 | bool noCache; // process is running without a dyld cache |
5945 | bool privateCache; // process is using a private copy of its dyld cache |
5946 | }; |
5947 | |
5948 | // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with |
5949 | // llvm enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile |
5950 | // errors. So we need to use the actual underlying types of task_t and |
5951 | // kern_return_t below. |
5952 | extern "C" unsigned int /*task_t*/ mach_task_self(); |
5953 | |
5954 | void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { |
5955 | uuid.Clear(); |
5956 | base_addr = LLDB_INVALID_ADDRESS; |
5957 | |
5958 | #if defined(__APPLE__) |
5959 | uint8_t *(*dyld_get_all_image_infos)(void); |
5960 | dyld_get_all_image_infos = |
5961 | (uint8_t * (*)()) dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos" ); |
5962 | if (dyld_get_all_image_infos) { |
5963 | uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); |
5964 | if (dyld_all_image_infos_address) { |
5965 | uint32_t *version = (uint32_t *) |
5966 | dyld_all_image_infos_address; // version <mach-o/dyld_images.h> |
5967 | if (*version >= 13) { |
5968 | uuid_t *sharedCacheUUID_address = 0; |
5969 | int wordsize = sizeof(uint8_t *); |
5970 | if (wordsize == 8) { |
5971 | sharedCacheUUID_address = |
5972 | (uuid_t *)((uint8_t *)dyld_all_image_infos_address + |
5973 | 160); // sharedCacheUUID <mach-o/dyld_images.h> |
5974 | if (*version >= 15) |
5975 | base_addr = |
5976 | *(uint64_t |
5977 | *)((uint8_t *)dyld_all_image_infos_address + |
5978 | 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> |
5979 | } else { |
5980 | sharedCacheUUID_address = |
5981 | (uuid_t *)((uint8_t *)dyld_all_image_infos_address + |
5982 | 84); // sharedCacheUUID <mach-o/dyld_images.h> |
5983 | if (*version >= 15) { |
5984 | base_addr = 0; |
5985 | base_addr = |
5986 | *(uint32_t |
5987 | *)((uint8_t *)dyld_all_image_infos_address + |
5988 | 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> |
5989 | } |
5990 | } |
5991 | uuid = UUID(sharedCacheUUID_address, sizeof(uuid_t)); |
5992 | } |
5993 | } |
5994 | } else { |
5995 | // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI |
5996 | dyld_process_info (*dyld_process_info_create)( |
5997 | unsigned int /* task_t */ task, uint64_t timestamp, |
5998 | unsigned int /*kern_return_t*/ *kernelError); |
5999 | void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); |
6000 | void (*dyld_process_info_release)(dyld_process_info info); |
6001 | |
6002 | dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, |
6003 | unsigned int /*kern_return_t*/ *)) |
6004 | dlsym(RTLD_DEFAULT, "_dyld_process_info_create" ); |
6005 | dyld_process_info_get_cache = (void (*)(void *, void *))dlsym( |
6006 | RTLD_DEFAULT, "_dyld_process_info_get_cache" ); |
6007 | dyld_process_info_release = |
6008 | (void (*)(void *))dlsym(RTLD_DEFAULT, "_dyld_process_info_release" ); |
6009 | |
6010 | if (dyld_process_info_create && dyld_process_info_get_cache) { |
6011 | unsigned int /*kern_return_t */ kern_ret; |
6012 | dyld_process_info process_info = |
6013 | dyld_process_info_create(::mach_task_self(), 0, &kern_ret); |
6014 | if (process_info) { |
6015 | struct lldb_copy__dyld_process_cache_info sc_info; |
6016 | memset(&sc_info, 0, sizeof(struct lldb_copy__dyld_process_cache_info)); |
6017 | dyld_process_info_get_cache(process_info, &sc_info); |
6018 | if (sc_info.cacheBaseAddress != 0) { |
6019 | base_addr = sc_info.cacheBaseAddress; |
6020 | uuid = UUID(sc_info.cacheUUID, sizeof(uuid_t)); |
6021 | } |
6022 | dyld_process_info_release(process_info); |
6023 | } |
6024 | } |
6025 | } |
6026 | Log *log(GetLog(LLDBLog::Symbols | LLDBLog::Process)); |
6027 | if (log && uuid.IsValid()) |
6028 | LLDB_LOGF(log, |
6029 | "lldb's in-memory shared cache has a UUID of %s base address of " |
6030 | "0x%" PRIx64, |
6031 | uuid.GetAsString().c_str(), base_addr); |
6032 | #endif |
6033 | } |
6034 | |
6035 | static llvm::VersionTuple (DataExtractor &data, |
6036 | lldb::offset_t offset, |
6037 | size_t ncmds) { |
6038 | for (size_t i = 0; i < ncmds; i++) { |
6039 | const lldb::offset_t load_cmd_offset = offset; |
6040 | llvm::MachO::load_command lc = {}; |
6041 | if (data.GetU32(offset_ptr: &offset, dst: &lc.cmd, count: 2) == nullptr) |
6042 | break; |
6043 | |
6044 | uint32_t version = 0; |
6045 | if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || |
6046 | lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || |
6047 | lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || |
6048 | lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { |
6049 | // struct version_min_command { |
6050 | // uint32_t cmd; // LC_VERSION_MIN_* |
6051 | // uint32_t cmdsize; |
6052 | // uint32_t version; // X.Y.Z encoded in nibbles xxxx.yy.zz |
6053 | // uint32_t sdk; |
6054 | // }; |
6055 | // We want to read version. |
6056 | version = data.GetU32(offset_ptr: &offset); |
6057 | } else if (lc.cmd == llvm::MachO::LC_BUILD_VERSION) { |
6058 | // struct build_version_command { |
6059 | // uint32_t cmd; // LC_BUILD_VERSION |
6060 | // uint32_t cmdsize; |
6061 | // uint32_t platform; |
6062 | // uint32_t minos; // X.Y.Z encoded in nibbles xxxx.yy.zz |
6063 | // uint32_t sdk; |
6064 | // uint32_t ntools; |
6065 | // }; |
6066 | // We want to read minos. |
6067 | offset += sizeof(uint32_t); // Skip over platform |
6068 | version = data.GetU32(offset_ptr: &offset); // Extract minos |
6069 | } |
6070 | |
6071 | if (version) { |
6072 | const uint32_t xxxx = version >> 16; |
6073 | const uint32_t yy = (version >> 8) & 0xffu; |
6074 | const uint32_t zz = version & 0xffu; |
6075 | if (xxxx) |
6076 | return llvm::VersionTuple(xxxx, yy, zz); |
6077 | } |
6078 | offset = load_cmd_offset + lc.cmdsize; |
6079 | } |
6080 | return llvm::VersionTuple(); |
6081 | } |
6082 | |
6083 | llvm::VersionTuple ObjectFileMachO::GetMinimumOSVersion() { |
6084 | if (!m_min_os_version) |
6085 | m_min_os_version = FindMinimumVersionInfo( |
6086 | data&: m_data, offset: MachHeaderSizeFromMagic(magic: m_header.magic), ncmds: m_header.ncmds); |
6087 | return *m_min_os_version; |
6088 | } |
6089 | |
6090 | llvm::VersionTuple ObjectFileMachO::GetSDKVersion() { |
6091 | if (!m_sdk_versions) |
6092 | m_sdk_versions = FindMinimumVersionInfo( |
6093 | data&: m_data, offset: MachHeaderSizeFromMagic(magic: m_header.magic), ncmds: m_header.ncmds); |
6094 | return *m_sdk_versions; |
6095 | } |
6096 | |
6097 | bool ObjectFileMachO::GetIsDynamicLinkEditor() { |
6098 | return m_header.filetype == llvm::MachO::MH_DYLINKER; |
6099 | } |
6100 | |
6101 | bool ObjectFileMachO::CanTrustAddressRanges() { |
6102 | // Dsymutil guarantees that the .debug_aranges accelerator is complete and can |
6103 | // be trusted by LLDB. |
6104 | return m_header.filetype == llvm::MachO::MH_DSYM; |
6105 | } |
6106 | |
6107 | bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { |
6108 | return m_allow_assembly_emulation_unwind_plans; |
6109 | } |
6110 | |
6111 | Section *ObjectFileMachO::() { |
6112 | // Find the first address of the mach header which is the first non-zero file |
6113 | // sized section whose file offset is zero. This is the base file address of |
6114 | // the mach-o file which can be subtracted from the vmaddr of the other |
6115 | // segments found in memory and added to the load address |
6116 | ModuleSP module_sp = GetModule(); |
6117 | if (!module_sp) |
6118 | return nullptr; |
6119 | SectionList *section_list = GetSectionList(); |
6120 | if (!section_list) |
6121 | return nullptr; |
6122 | |
6123 | // Some binaries can have a TEXT segment with a non-zero file offset. |
6124 | // Binaries in the shared cache are one example. Some hand-generated |
6125 | // binaries may not be laid out in the normal TEXT,DATA,LC_SYMTAB order |
6126 | // in the file, even though they're laid out correctly in vmaddr terms. |
6127 | SectionSP text_segment_sp = |
6128 | section_list->FindSectionByName(section_dstr: GetSegmentNameTEXT()); |
6129 | if (text_segment_sp.get() && SectionIsLoadable(section: text_segment_sp.get())) |
6130 | return text_segment_sp.get(); |
6131 | |
6132 | const size_t num_sections = section_list->GetSize(); |
6133 | for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { |
6134 | Section *section = section_list->GetSectionAtIndex(idx: sect_idx).get(); |
6135 | if (section->GetFileOffset() == 0 && SectionIsLoadable(section)) |
6136 | return section; |
6137 | } |
6138 | |
6139 | return nullptr; |
6140 | } |
6141 | |
6142 | bool ObjectFileMachO::SectionIsLoadable(const Section *section) { |
6143 | if (!section) |
6144 | return false; |
6145 | const bool is_dsym = (m_header.filetype == MH_DSYM); |
6146 | if (section->GetFileSize() == 0 && !is_dsym && |
6147 | section->GetName() != GetSegmentNameDATA()) |
6148 | return false; |
6149 | if (section->IsThreadSpecific()) |
6150 | return false; |
6151 | if (GetModule().get() != section->GetModule().get()) |
6152 | return false; |
6153 | // firmware style binaries with llvm gcov segment do |
6154 | // not have that segment mapped into memory. |
6155 | if (section->GetName() == GetSegmentNameLLVM_COV()) { |
6156 | const Strata strata = GetStrata(); |
6157 | if (strata == eStrataKernel || strata == eStrataRawImage) |
6158 | return false; |
6159 | } |
6160 | // Be careful with __LINKEDIT and __DWARF segments |
6161 | if (section->GetName() == GetSegmentNameLINKEDIT() || |
6162 | section->GetName() == GetSegmentNameDWARF()) { |
6163 | // Only map __LINKEDIT and __DWARF if we have an in memory image and |
6164 | // this isn't a kernel binary like a kext or mach_kernel. |
6165 | const bool is_memory_image = (bool)m_process_wp.lock(); |
6166 | const Strata strata = GetStrata(); |
6167 | if (is_memory_image == false || strata == eStrataKernel) |
6168 | return false; |
6169 | } |
6170 | return true; |
6171 | } |
6172 | |
6173 | lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( |
6174 | lldb::addr_t , const Section *, |
6175 | const Section *section) { |
6176 | ModuleSP module_sp = GetModule(); |
6177 | if (module_sp && header_section && section && |
6178 | header_load_address != LLDB_INVALID_ADDRESS) { |
6179 | lldb::addr_t file_addr = header_section->GetFileAddress(); |
6180 | if (file_addr != LLDB_INVALID_ADDRESS && SectionIsLoadable(section)) |
6181 | return section->GetFileAddress() - file_addr + header_load_address; |
6182 | } |
6183 | return LLDB_INVALID_ADDRESS; |
6184 | } |
6185 | |
6186 | bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, |
6187 | bool value_is_offset) { |
6188 | ModuleSP module_sp = GetModule(); |
6189 | if (!module_sp) |
6190 | return false; |
6191 | |
6192 | SectionList *section_list = GetSectionList(); |
6193 | if (!section_list) |
6194 | return false; |
6195 | |
6196 | size_t num_loaded_sections = 0; |
6197 | const size_t num_sections = section_list->GetSize(); |
6198 | |
6199 | // Warn if some top-level segments map to the same address. The binary may be |
6200 | // malformed. |
6201 | const bool warn_multiple = true; |
6202 | |
6203 | if (value_is_offset) { |
6204 | // "value" is an offset to apply to each top level segment |
6205 | for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { |
6206 | // Iterate through the object file sections to find all of the |
6207 | // sections that size on disk (to avoid __PAGEZERO) and load them |
6208 | SectionSP section_sp(section_list->GetSectionAtIndex(idx: sect_idx)); |
6209 | if (SectionIsLoadable(section: section_sp.get())) |
6210 | if (target.GetSectionLoadList().SetSectionLoadAddress( |
6211 | section_sp, load_addr: section_sp->GetFileAddress() + value, |
6212 | warn_multiple)) |
6213 | ++num_loaded_sections; |
6214 | } |
6215 | } else { |
6216 | // "value" is the new base address of the mach_header, adjust each |
6217 | // section accordingly |
6218 | |
6219 | Section * = GetMachHeaderSection(); |
6220 | if (mach_header_section) { |
6221 | for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { |
6222 | SectionSP section_sp(section_list->GetSectionAtIndex(idx: sect_idx)); |
6223 | |
6224 | lldb::addr_t section_load_addr = |
6225 | CalculateSectionLoadAddressForMemoryImage( |
6226 | header_load_address: value, header_section: mach_header_section, section: section_sp.get()); |
6227 | if (section_load_addr != LLDB_INVALID_ADDRESS) { |
6228 | if (target.GetSectionLoadList().SetSectionLoadAddress( |
6229 | section_sp, load_addr: section_load_addr, warn_multiple)) |
6230 | ++num_loaded_sections; |
6231 | } |
6232 | } |
6233 | } |
6234 | } |
6235 | return num_loaded_sections > 0; |
6236 | } |
6237 | |
6238 | struct { |
6239 | uint32_t ; // currently 1 |
6240 | uint32_t ; // number of binary images |
6241 | uint64_t ; // file offset in the corefile of where the array of |
6242 | // struct entry's begin. |
6243 | uint32_t ; // size of 'struct entry'. |
6244 | uint32_t ; |
6245 | }; |
6246 | |
6247 | struct image_entry { |
6248 | uint64_t filepath_offset; // offset in corefile to c-string of the file path, |
6249 | // UINT64_MAX if unavailable. |
6250 | uuid_t uuid; // uint8_t[16]. should be set to all zeroes if |
6251 | // uuid is unknown. |
6252 | uint64_t load_address; // UINT64_MAX if unknown. |
6253 | uint64_t seg_addrs_offset; // offset to the array of struct segment_vmaddr's. |
6254 | uint32_t segment_count; // The number of segments for this binary. |
6255 | uint32_t unused; |
6256 | |
6257 | image_entry() { |
6258 | filepath_offset = UINT64_MAX; |
6259 | memset(s: &uuid, c: 0, n: sizeof(uuid_t)); |
6260 | segment_count = 0; |
6261 | load_address = UINT64_MAX; |
6262 | seg_addrs_offset = UINT64_MAX; |
6263 | unused = 0; |
6264 | } |
6265 | image_entry(const image_entry &rhs) { |
6266 | filepath_offset = rhs.filepath_offset; |
6267 | memcpy(dest: &uuid, src: &rhs.uuid, n: sizeof(uuid_t)); |
6268 | segment_count = rhs.segment_count; |
6269 | seg_addrs_offset = rhs.seg_addrs_offset; |
6270 | load_address = rhs.load_address; |
6271 | unused = rhs.unused; |
6272 | } |
6273 | }; |
6274 | |
6275 | struct segment_vmaddr { |
6276 | char segname[16]; |
6277 | uint64_t vmaddr; |
6278 | uint64_t unused; |
6279 | |
6280 | segment_vmaddr() { |
6281 | memset(s: &segname, c: 0, n: 16); |
6282 | vmaddr = UINT64_MAX; |
6283 | unused = 0; |
6284 | } |
6285 | segment_vmaddr(const segment_vmaddr &rhs) { |
6286 | memcpy(dest: &segname, src: &rhs.segname, n: 16); |
6287 | vmaddr = rhs.vmaddr; |
6288 | unused = rhs.unused; |
6289 | } |
6290 | }; |
6291 | |
6292 | // Write the payload for the "all image infos" LC_NOTE into |
6293 | // the supplied all_image_infos_payload, assuming that this |
6294 | // will be written into the corefile starting at |
6295 | // initial_file_offset. |
6296 | // |
6297 | // The placement of this payload is a little tricky. We're |
6298 | // laying this out as |
6299 | // |
6300 | // 1. header (struct all_image_info_header) |
6301 | // 2. Array of fixed-size (struct image_entry)'s, one |
6302 | // per binary image present in the process. |
6303 | // 3. Arrays of (struct segment_vmaddr)'s, a varying number |
6304 | // for each binary image. |
6305 | // 4. Variable length c-strings of binary image filepaths, |
6306 | // one per binary. |
6307 | // |
6308 | // To compute where everything will be laid out in the |
6309 | // payload, we need to iterate over the images and calculate |
6310 | // how many segment_vmaddr structures each image will need, |
6311 | // and how long each image's filepath c-string is. There |
6312 | // are some multiple passes over the image list while calculating |
6313 | // everything. |
6314 | |
6315 | static offset_t CreateAllImageInfosPayload( |
6316 | const lldb::ProcessSP &process_sp, offset_t initial_file_offset, |
6317 | StreamString &all_image_infos_payload, SaveCoreStyle core_style) { |
6318 | Target &target = process_sp->GetTarget(); |
6319 | ModuleList modules = target.GetImages(); |
6320 | |
6321 | // stack-only corefiles have no reason to include binaries that |
6322 | // are not executing; we're trying to make the smallest corefile |
6323 | // we can, so leave the rest out. |
6324 | if (core_style == SaveCoreStyle::eSaveCoreStackOnly) |
6325 | modules.Clear(); |
6326 | |
6327 | std::set<std::string> executing_uuids; |
6328 | ThreadList &thread_list(process_sp->GetThreadList()); |
6329 | for (uint32_t i = 0; i < thread_list.GetSize(); i++) { |
6330 | ThreadSP thread_sp = thread_list.GetThreadAtIndex(idx: i); |
6331 | uint32_t stack_frame_count = thread_sp->GetStackFrameCount(); |
6332 | for (uint32_t j = 0; j < stack_frame_count; j++) { |
6333 | StackFrameSP stack_frame_sp = thread_sp->GetStackFrameAtIndex(idx: j); |
6334 | Address pc = stack_frame_sp->GetFrameCodeAddress(); |
6335 | ModuleSP module_sp = pc.GetModule(); |
6336 | if (module_sp) { |
6337 | UUID uuid = module_sp->GetUUID(); |
6338 | if (uuid.IsValid()) { |
6339 | executing_uuids.insert(x: uuid.GetAsString()); |
6340 | modules.AppendIfNeeded(new_module: module_sp); |
6341 | } |
6342 | } |
6343 | } |
6344 | } |
6345 | size_t modules_count = modules.GetSize(); |
6346 | |
6347 | struct all_image_infos_header infos; |
6348 | infos.version = 1; |
6349 | infos.imgcount = modules_count; |
6350 | infos.entries_size = sizeof(image_entry); |
6351 | infos.entries_fileoff = initial_file_offset + sizeof(all_image_infos_header); |
6352 | infos.unused = 0; |
6353 | |
6354 | all_image_infos_payload.PutHex32(uvalue: infos.version); |
6355 | all_image_infos_payload.PutHex32(uvalue: infos.imgcount); |
6356 | all_image_infos_payload.PutHex64(uvalue: infos.entries_fileoff); |
6357 | all_image_infos_payload.PutHex32(uvalue: infos.entries_size); |
6358 | all_image_infos_payload.PutHex32(uvalue: infos.unused); |
6359 | |
6360 | // First create the structures for all of the segment name+vmaddr vectors |
6361 | // for each module, so we will know the size of them as we add the |
6362 | // module entries. |
6363 | std::vector<std::vector<segment_vmaddr>> modules_segment_vmaddrs; |
6364 | for (size_t i = 0; i < modules_count; i++) { |
6365 | ModuleSP module = modules.GetModuleAtIndex(idx: i); |
6366 | |
6367 | SectionList *sections = module->GetSectionList(); |
6368 | size_t sections_count = sections->GetSize(); |
6369 | std::vector<segment_vmaddr> segment_vmaddrs; |
6370 | for (size_t j = 0; j < sections_count; j++) { |
6371 | SectionSP section = sections->GetSectionAtIndex(idx: j); |
6372 | if (!section->GetParent().get()) { |
6373 | addr_t vmaddr = section->GetLoadBaseAddress(target: &target); |
6374 | if (vmaddr == LLDB_INVALID_ADDRESS) |
6375 | continue; |
6376 | ConstString name = section->GetName(); |
6377 | segment_vmaddr seg_vmaddr; |
6378 | // This is the uncommon case where strncpy is exactly |
6379 | // the right one, doesn't need to be nul terminated. |
6380 | // The segment name in a Mach-O LC_SEGMENT/LC_SEGMENT_64 is char[16] and |
6381 | // is not guaranteed to be nul-terminated if all 16 characters are |
6382 | // used. |
6383 | // coverity[buffer_size_warning] |
6384 | strncpy(dest: seg_vmaddr.segname, src: name.AsCString(), |
6385 | n: sizeof(seg_vmaddr.segname)); |
6386 | seg_vmaddr.vmaddr = vmaddr; |
6387 | seg_vmaddr.unused = 0; |
6388 | segment_vmaddrs.push_back(x: seg_vmaddr); |
6389 | } |
6390 | } |
6391 | modules_segment_vmaddrs.push_back(x: segment_vmaddrs); |
6392 | } |
6393 | |
6394 | offset_t size_of_vmaddr_structs = 0; |
6395 | for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) { |
6396 | size_of_vmaddr_structs += |
6397 | modules_segment_vmaddrs[i].size() * sizeof(segment_vmaddr); |
6398 | } |
6399 | |
6400 | offset_t size_of_filepath_cstrings = 0; |
6401 | for (size_t i = 0; i < modules_count; i++) { |
6402 | ModuleSP module_sp = modules.GetModuleAtIndex(idx: i); |
6403 | size_of_filepath_cstrings += module_sp->GetFileSpec().GetPath().size() + 1; |
6404 | } |
6405 | |
6406 | // Calculate the file offsets of our "all image infos" payload in the |
6407 | // corefile. initial_file_offset the original value passed in to this method. |
6408 | |
6409 | offset_t start_of_entries = |
6410 | initial_file_offset + sizeof(all_image_infos_header); |
6411 | offset_t start_of_seg_vmaddrs = |
6412 | start_of_entries + sizeof(image_entry) * modules_count; |
6413 | offset_t start_of_filenames = start_of_seg_vmaddrs + size_of_vmaddr_structs; |
6414 | |
6415 | offset_t final_file_offset = start_of_filenames + size_of_filepath_cstrings; |
6416 | |
6417 | // Now write the one-per-module 'struct image_entry' into the |
6418 | // StringStream; keep track of where the struct segment_vmaddr |
6419 | // entries for each module will end up in the corefile. |
6420 | |
6421 | offset_t current_string_offset = start_of_filenames; |
6422 | offset_t current_segaddrs_offset = start_of_seg_vmaddrs; |
6423 | std::vector<struct image_entry> image_entries; |
6424 | for (size_t i = 0; i < modules_count; i++) { |
6425 | ModuleSP module_sp = modules.GetModuleAtIndex(idx: i); |
6426 | |
6427 | struct image_entry ent; |
6428 | memcpy(dest: &ent.uuid, src: module_sp->GetUUID().GetBytes().data(), n: sizeof(ent.uuid)); |
6429 | if (modules_segment_vmaddrs[i].size() > 0) { |
6430 | ent.segment_count = modules_segment_vmaddrs[i].size(); |
6431 | ent.seg_addrs_offset = current_segaddrs_offset; |
6432 | } |
6433 | ent.filepath_offset = current_string_offset; |
6434 | ObjectFile *objfile = module_sp->GetObjectFile(); |
6435 | if (objfile) { |
6436 | Address base_addr(objfile->GetBaseAddress()); |
6437 | if (base_addr.IsValid()) { |
6438 | ent.load_address = base_addr.GetLoadAddress(target: &target); |
6439 | } |
6440 | } |
6441 | |
6442 | all_image_infos_payload.PutHex64(uvalue: ent.filepath_offset); |
6443 | all_image_infos_payload.PutRawBytes(s: ent.uuid, src_len: sizeof(ent.uuid)); |
6444 | all_image_infos_payload.PutHex64(uvalue: ent.load_address); |
6445 | all_image_infos_payload.PutHex64(uvalue: ent.seg_addrs_offset); |
6446 | all_image_infos_payload.PutHex32(uvalue: ent.segment_count); |
6447 | |
6448 | if (executing_uuids.find(x: module_sp->GetUUID().GetAsString()) != |
6449 | executing_uuids.end()) |
6450 | all_image_infos_payload.PutHex32(uvalue: 1); |
6451 | else |
6452 | all_image_infos_payload.PutHex32(uvalue: 0); |
6453 | |
6454 | current_segaddrs_offset += ent.segment_count * sizeof(segment_vmaddr); |
6455 | current_string_offset += module_sp->GetFileSpec().GetPath().size() + 1; |
6456 | } |
6457 | |
6458 | // Now write the struct segment_vmaddr entries into the StringStream. |
6459 | |
6460 | for (size_t i = 0; i < modules_segment_vmaddrs.size(); i++) { |
6461 | if (modules_segment_vmaddrs[i].size() == 0) |
6462 | continue; |
6463 | for (struct segment_vmaddr segvm : modules_segment_vmaddrs[i]) { |
6464 | all_image_infos_payload.PutRawBytes(s: segvm.segname, src_len: sizeof(segvm.segname)); |
6465 | all_image_infos_payload.PutHex64(uvalue: segvm.vmaddr); |
6466 | all_image_infos_payload.PutHex64(uvalue: segvm.unused); |
6467 | } |
6468 | } |
6469 | |
6470 | for (size_t i = 0; i < modules_count; i++) { |
6471 | ModuleSP module_sp = modules.GetModuleAtIndex(idx: i); |
6472 | std::string filepath = module_sp->GetFileSpec().GetPath(); |
6473 | all_image_infos_payload.PutRawBytes(s: filepath.data(), src_len: filepath.size() + 1); |
6474 | } |
6475 | |
6476 | return final_file_offset; |
6477 | } |
6478 | |
6479 | // Temp struct used to combine contiguous memory regions with |
6480 | // identical permissions. |
6481 | struct page_object { |
6482 | addr_t addr; |
6483 | addr_t size; |
6484 | uint32_t prot; |
6485 | }; |
6486 | |
6487 | bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, |
6488 | const FileSpec &outfile, |
6489 | lldb::SaveCoreStyle &core_style, Status &error) { |
6490 | if (!process_sp) |
6491 | return false; |
6492 | |
6493 | // Default on macOS is to create a dirty-memory-only corefile. |
6494 | if (core_style == SaveCoreStyle::eSaveCoreUnspecified) |
6495 | core_style = SaveCoreStyle::eSaveCoreDirtyOnly; |
6496 | |
6497 | Target &target = process_sp->GetTarget(); |
6498 | const ArchSpec target_arch = target.GetArchitecture(); |
6499 | const llvm::Triple &target_triple = target_arch.GetTriple(); |
6500 | if (target_triple.getVendor() == llvm::Triple::Apple && |
6501 | (target_triple.getOS() == llvm::Triple::MacOSX || |
6502 | target_triple.getOS() == llvm::Triple::IOS || |
6503 | target_triple.getOS() == llvm::Triple::WatchOS || |
6504 | target_triple.getOS() == llvm::Triple::TvOS || |
6505 | target_triple.getOS() == llvm::Triple::XROS)) { |
6506 | // NEED_BRIDGEOS_TRIPLE target_triple.getOS() == llvm::Triple::BridgeOS)) |
6507 | // { |
6508 | bool make_core = false; |
6509 | switch (target_arch.GetMachine()) { |
6510 | case llvm::Triple::aarch64: |
6511 | case llvm::Triple::aarch64_32: |
6512 | case llvm::Triple::arm: |
6513 | case llvm::Triple::thumb: |
6514 | case llvm::Triple::x86: |
6515 | case llvm::Triple::x86_64: |
6516 | make_core = true; |
6517 | break; |
6518 | default: |
6519 | error.SetErrorStringWithFormat("unsupported core architecture: %s" , |
6520 | target_triple.str().c_str()); |
6521 | break; |
6522 | } |
6523 | |
6524 | if (make_core) { |
6525 | Process::CoreFileMemoryRanges core_ranges; |
6526 | error = process_sp->CalculateCoreFileSaveRanges(core_style, ranges&: core_ranges); |
6527 | if (error.Success()) { |
6528 | const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); |
6529 | const ByteOrder byte_order = target_arch.GetByteOrder(); |
6530 | std::vector<llvm::MachO::segment_command_64> segment_load_commands; |
6531 | for (const auto &core_range : core_ranges) { |
6532 | uint32_t cmd_type = LC_SEGMENT_64; |
6533 | uint32_t segment_size = sizeof(llvm::MachO::segment_command_64); |
6534 | if (addr_byte_size == 4) { |
6535 | cmd_type = LC_SEGMENT; |
6536 | segment_size = sizeof(llvm::MachO::segment_command); |
6537 | } |
6538 | // Skip any ranges with no read/write/execute permissions and empty |
6539 | // ranges. |
6540 | if (core_range.lldb_permissions == 0 || core_range.range.size() == 0) |
6541 | continue; |
6542 | uint32_t vm_prot = 0; |
6543 | if (core_range.lldb_permissions & ePermissionsReadable) |
6544 | vm_prot |= VM_PROT_READ; |
6545 | if (core_range.lldb_permissions & ePermissionsWritable) |
6546 | vm_prot |= VM_PROT_WRITE; |
6547 | if (core_range.lldb_permissions & ePermissionsExecutable) |
6548 | vm_prot |= VM_PROT_EXECUTE; |
6549 | const addr_t vm_addr = core_range.range.start(); |
6550 | const addr_t vm_size = core_range.range.size(); |
6551 | llvm::MachO::segment_command_64 segment = { |
6552 | .cmd: cmd_type, // uint32_t cmd; |
6553 | .cmdsize: segment_size, // uint32_t cmdsize; |
6554 | .segname: {0}, // char segname[16]; |
6555 | .vmaddr: vm_addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O |
6556 | .vmsize: vm_size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O |
6557 | .fileoff: 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O |
6558 | .filesize: vm_size, // uint64_t filesize; // uint32_t for 32-bit Mach-O |
6559 | .maxprot: vm_prot, // uint32_t maxprot; |
6560 | .initprot: vm_prot, // uint32_t initprot; |
6561 | .nsects: 0, // uint32_t nsects; |
6562 | .flags: 0}; // uint32_t flags; |
6563 | segment_load_commands.push_back(x: segment); |
6564 | } |
6565 | |
6566 | StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); |
6567 | |
6568 | llvm::MachO::mach_header_64 ; |
6569 | mach_header.magic = addr_byte_size == 8 ? MH_MAGIC_64 : MH_MAGIC; |
6570 | mach_header.cputype = target_arch.GetMachOCPUType(); |
6571 | mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); |
6572 | mach_header.filetype = MH_CORE; |
6573 | mach_header.ncmds = segment_load_commands.size(); |
6574 | mach_header.flags = 0; |
6575 | mach_header.reserved = 0; |
6576 | ThreadList &thread_list = process_sp->GetThreadList(); |
6577 | const uint32_t num_threads = thread_list.GetSize(); |
6578 | |
6579 | // Make an array of LC_THREAD data items. Each one contains the |
6580 | // contents of the LC_THREAD load command. The data doesn't contain |
6581 | // the load command + load command size, we will add the load command |
6582 | // and load command size as we emit the data. |
6583 | std::vector<StreamString> LC_THREAD_datas(num_threads); |
6584 | for (auto &LC_THREAD_data : LC_THREAD_datas) { |
6585 | LC_THREAD_data.GetFlags().Set(Stream::eBinary); |
6586 | LC_THREAD_data.SetAddressByteSize(addr_byte_size); |
6587 | LC_THREAD_data.SetByteOrder(byte_order); |
6588 | } |
6589 | for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) { |
6590 | ThreadSP thread_sp(thread_list.GetThreadAtIndex(idx: thread_idx)); |
6591 | if (thread_sp) { |
6592 | switch (mach_header.cputype) { |
6593 | case llvm::MachO::CPU_TYPE_ARM64: |
6594 | case llvm::MachO::CPU_TYPE_ARM64_32: |
6595 | RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( |
6596 | thread: thread_sp.get(), data&: LC_THREAD_datas[thread_idx]); |
6597 | break; |
6598 | |
6599 | case llvm::MachO::CPU_TYPE_ARM: |
6600 | RegisterContextDarwin_arm_Mach::Create_LC_THREAD( |
6601 | thread: thread_sp.get(), data&: LC_THREAD_datas[thread_idx]); |
6602 | break; |
6603 | |
6604 | case llvm::MachO::CPU_TYPE_I386: |
6605 | RegisterContextDarwin_i386_Mach::Create_LC_THREAD( |
6606 | thread: thread_sp.get(), data&: LC_THREAD_datas[thread_idx]); |
6607 | break; |
6608 | |
6609 | case llvm::MachO::CPU_TYPE_X86_64: |
6610 | RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( |
6611 | thread: thread_sp.get(), data&: LC_THREAD_datas[thread_idx]); |
6612 | break; |
6613 | } |
6614 | } |
6615 | } |
6616 | |
6617 | // The size of the load command is the size of the segments... |
6618 | if (addr_byte_size == 8) { |
6619 | mach_header.sizeofcmds = segment_load_commands.size() * |
6620 | sizeof(llvm::MachO::segment_command_64); |
6621 | } else { |
6622 | mach_header.sizeofcmds = segment_load_commands.size() * |
6623 | sizeof(llvm::MachO::segment_command); |
6624 | } |
6625 | |
6626 | // and the size of all LC_THREAD load command |
6627 | for (const auto &LC_THREAD_data : LC_THREAD_datas) { |
6628 | ++mach_header.ncmds; |
6629 | mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); |
6630 | } |
6631 | |
6632 | // Bits will be set to indicate which bits are NOT used in |
6633 | // addressing in this process or 0 for unknown. |
6634 | uint64_t address_mask = process_sp->GetCodeAddressMask(); |
6635 | if (address_mask != LLDB_INVALID_ADDRESS_MASK) { |
6636 | // LC_NOTE "addrable bits" |
6637 | mach_header.ncmds++; |
6638 | mach_header.sizeofcmds += sizeof(llvm::MachO::note_command); |
6639 | } |
6640 | |
6641 | // LC_NOTE "process metadata" |
6642 | mach_header.ncmds++; |
6643 | mach_header.sizeofcmds += sizeof(llvm::MachO::note_command); |
6644 | |
6645 | // LC_NOTE "all image infos" |
6646 | mach_header.ncmds++; |
6647 | mach_header.sizeofcmds += sizeof(llvm::MachO::note_command); |
6648 | |
6649 | // Write the mach header |
6650 | buffer.PutHex32(uvalue: mach_header.magic); |
6651 | buffer.PutHex32(uvalue: mach_header.cputype); |
6652 | buffer.PutHex32(uvalue: mach_header.cpusubtype); |
6653 | buffer.PutHex32(uvalue: mach_header.filetype); |
6654 | buffer.PutHex32(uvalue: mach_header.ncmds); |
6655 | buffer.PutHex32(uvalue: mach_header.sizeofcmds); |
6656 | buffer.PutHex32(uvalue: mach_header.flags); |
6657 | if (addr_byte_size == 8) { |
6658 | buffer.PutHex32(uvalue: mach_header.reserved); |
6659 | } |
6660 | |
6661 | // Skip the mach header and all load commands and align to the next |
6662 | // 0x1000 byte boundary |
6663 | addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; |
6664 | |
6665 | file_offset = llvm::alignTo(Value: file_offset, Align: 16); |
6666 | std::vector<std::unique_ptr<LCNoteEntry>> lc_notes; |
6667 | |
6668 | // Add "addrable bits" LC_NOTE when an address mask is available |
6669 | if (address_mask != LLDB_INVALID_ADDRESS_MASK) { |
6670 | std::unique_ptr<LCNoteEntry> addrable_bits_lcnote_up( |
6671 | new LCNoteEntry(addr_byte_size, byte_order)); |
6672 | addrable_bits_lcnote_up->name = "addrable bits" ; |
6673 | addrable_bits_lcnote_up->payload_file_offset = file_offset; |
6674 | int bits = std::bitset<64>(~address_mask).count(); |
6675 | addrable_bits_lcnote_up->payload.PutHex32(uvalue: 4); // version |
6676 | addrable_bits_lcnote_up->payload.PutHex32( |
6677 | uvalue: bits); // # of bits used for low addresses |
6678 | addrable_bits_lcnote_up->payload.PutHex32( |
6679 | uvalue: bits); // # of bits used for high addresses |
6680 | addrable_bits_lcnote_up->payload.PutHex32(uvalue: 0); // reserved |
6681 | |
6682 | file_offset += addrable_bits_lcnote_up->payload.GetSize(); |
6683 | |
6684 | lc_notes.push_back(x: std::move(addrable_bits_lcnote_up)); |
6685 | } |
6686 | |
6687 | // Add "process metadata" LC_NOTE |
6688 | std::unique_ptr<LCNoteEntry> ( |
6689 | new LCNoteEntry(addr_byte_size, byte_order)); |
6690 | thread_extrainfo_lcnote_up->name = "process metadata" ; |
6691 | thread_extrainfo_lcnote_up->payload_file_offset = file_offset; |
6692 | |
6693 | StructuredData::DictionarySP dict( |
6694 | std::make_shared<StructuredData::Dictionary>()); |
6695 | StructuredData::ArraySP threads( |
6696 | std::make_shared<StructuredData::Array>()); |
6697 | for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) { |
6698 | ThreadSP thread_sp(thread_list.GetThreadAtIndex(idx: thread_idx)); |
6699 | StructuredData::DictionarySP thread( |
6700 | std::make_shared<StructuredData::Dictionary>()); |
6701 | thread->AddIntegerItem(key: "thread_id" , value: thread_sp->GetID()); |
6702 | threads->AddItem(item: thread); |
6703 | } |
6704 | dict->AddItem(key: "threads" , value_sp: threads); |
6705 | StreamString strm; |
6706 | dict->Dump(s&: strm, /* pretty */ pretty_print: false); |
6707 | thread_extrainfo_lcnote_up->payload.PutRawBytes(s: strm.GetData(), |
6708 | src_len: strm.GetSize()); |
6709 | |
6710 | file_offset += thread_extrainfo_lcnote_up->payload.GetSize(); |
6711 | file_offset = llvm::alignTo(Value: file_offset, Align: 16); |
6712 | lc_notes.push_back(x: std::move(thread_extrainfo_lcnote_up)); |
6713 | |
6714 | // Add "all image infos" LC_NOTE |
6715 | std::unique_ptr<LCNoteEntry> all_image_infos_lcnote_up( |
6716 | new LCNoteEntry(addr_byte_size, byte_order)); |
6717 | all_image_infos_lcnote_up->name = "all image infos" ; |
6718 | all_image_infos_lcnote_up->payload_file_offset = file_offset; |
6719 | file_offset = CreateAllImageInfosPayload( |
6720 | process_sp, initial_file_offset: file_offset, all_image_infos_payload&: all_image_infos_lcnote_up->payload, |
6721 | core_style); |
6722 | lc_notes.push_back(x: std::move(all_image_infos_lcnote_up)); |
6723 | |
6724 | // Add LC_NOTE load commands |
6725 | for (auto &lcnote : lc_notes) { |
6726 | // Add the LC_NOTE load command to the file. |
6727 | buffer.PutHex32(uvalue: LC_NOTE); |
6728 | buffer.PutHex32(uvalue: sizeof(llvm::MachO::note_command)); |
6729 | char namebuf[16]; |
6730 | memset(s: namebuf, c: 0, n: sizeof(namebuf)); |
6731 | // This is the uncommon case where strncpy is exactly |
6732 | // the right one, doesn't need to be nul terminated. |
6733 | // LC_NOTE name field is char[16] and is not guaranteed to be |
6734 | // nul-terminated. |
6735 | // coverity[buffer_size_warning] |
6736 | strncpy(dest: namebuf, src: lcnote->name.c_str(), n: sizeof(namebuf)); |
6737 | buffer.PutRawBytes(s: namebuf, src_len: sizeof(namebuf)); |
6738 | buffer.PutHex64(uvalue: lcnote->payload_file_offset); |
6739 | buffer.PutHex64(uvalue: lcnote->payload.GetSize()); |
6740 | } |
6741 | |
6742 | // Align to 4096-byte page boundary for the LC_SEGMENTs. |
6743 | file_offset = llvm::alignTo(Value: file_offset, Align: 4096); |
6744 | |
6745 | for (auto &segment : segment_load_commands) { |
6746 | segment.fileoff = file_offset; |
6747 | file_offset += segment.filesize; |
6748 | } |
6749 | |
6750 | // Write out all of the LC_THREAD load commands |
6751 | for (const auto &LC_THREAD_data : LC_THREAD_datas) { |
6752 | const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); |
6753 | buffer.PutHex32(uvalue: LC_THREAD); |
6754 | buffer.PutHex32(uvalue: 8 + LC_THREAD_data_size); // cmd + cmdsize + data |
6755 | buffer.Write(src: LC_THREAD_data.GetString().data(), src_len: LC_THREAD_data_size); |
6756 | } |
6757 | |
6758 | // Write out all of the segment load commands |
6759 | for (const auto &segment : segment_load_commands) { |
6760 | buffer.PutHex32(uvalue: segment.cmd); |
6761 | buffer.PutHex32(uvalue: segment.cmdsize); |
6762 | buffer.PutRawBytes(s: segment.segname, src_len: sizeof(segment.segname)); |
6763 | if (addr_byte_size == 8) { |
6764 | buffer.PutHex64(uvalue: segment.vmaddr); |
6765 | buffer.PutHex64(uvalue: segment.vmsize); |
6766 | buffer.PutHex64(uvalue: segment.fileoff); |
6767 | buffer.PutHex64(uvalue: segment.filesize); |
6768 | } else { |
6769 | buffer.PutHex32(uvalue: static_cast<uint32_t>(segment.vmaddr)); |
6770 | buffer.PutHex32(uvalue: static_cast<uint32_t>(segment.vmsize)); |
6771 | buffer.PutHex32(uvalue: static_cast<uint32_t>(segment.fileoff)); |
6772 | buffer.PutHex32(uvalue: static_cast<uint32_t>(segment.filesize)); |
6773 | } |
6774 | buffer.PutHex32(uvalue: segment.maxprot); |
6775 | buffer.PutHex32(uvalue: segment.initprot); |
6776 | buffer.PutHex32(uvalue: segment.nsects); |
6777 | buffer.PutHex32(uvalue: segment.flags); |
6778 | } |
6779 | |
6780 | std::string core_file_path(outfile.GetPath()); |
6781 | auto core_file = FileSystem::Instance().Open( |
6782 | file_spec: outfile, options: File::eOpenOptionWriteOnly | File::eOpenOptionTruncate | |
6783 | File::eOpenOptionCanCreate); |
6784 | if (!core_file) { |
6785 | error = core_file.takeError(); |
6786 | } else { |
6787 | // Read 1 page at a time |
6788 | uint8_t bytes[0x1000]; |
6789 | // Write the mach header and load commands out to the core file |
6790 | size_t bytes_written = buffer.GetString().size(); |
6791 | error = |
6792 | core_file.get()->Write(buf: buffer.GetString().data(), num_bytes&: bytes_written); |
6793 | if (error.Success()) { |
6794 | |
6795 | for (auto &lcnote : lc_notes) { |
6796 | if (core_file.get()->SeekFromStart(offset: lcnote->payload_file_offset) == |
6797 | -1) { |
6798 | error.SetErrorStringWithFormat("Unable to seek to corefile pos " |
6799 | "to write '%s' LC_NOTE payload" , |
6800 | lcnote->name.c_str()); |
6801 | return false; |
6802 | } |
6803 | bytes_written = lcnote->payload.GetSize(); |
6804 | error = core_file.get()->Write(buf: lcnote->payload.GetData(), |
6805 | num_bytes&: bytes_written); |
6806 | if (!error.Success()) |
6807 | return false; |
6808 | } |
6809 | |
6810 | // Now write the file data for all memory segments in the process |
6811 | for (const auto &segment : segment_load_commands) { |
6812 | if (core_file.get()->SeekFromStart(offset: segment.fileoff) == -1) { |
6813 | error.SetErrorStringWithFormat( |
6814 | "unable to seek to offset 0x%" PRIx64 " in '%s'" , |
6815 | segment.fileoff, core_file_path.c_str()); |
6816 | break; |
6817 | } |
6818 | |
6819 | target.GetDebugger().GetAsyncOutputStream()->Printf( |
6820 | format: "Saving %" PRId64 |
6821 | " bytes of data for memory region at 0x%" PRIx64 "\n" , |
6822 | segment.vmsize, segment.vmaddr); |
6823 | addr_t bytes_left = segment.vmsize; |
6824 | addr_t addr = segment.vmaddr; |
6825 | Status memory_read_error; |
6826 | while (bytes_left > 0 && error.Success()) { |
6827 | const size_t bytes_to_read = |
6828 | bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; |
6829 | |
6830 | // In a savecore setting, we don't really care about caching, |
6831 | // as the data is dumped and very likely never read again, |
6832 | // so we call ReadMemoryFromInferior to bypass it. |
6833 | const size_t bytes_read = process_sp->ReadMemoryFromInferior( |
6834 | vm_addr: addr, buf: bytes, size: bytes_to_read, error&: memory_read_error); |
6835 | |
6836 | if (bytes_read == bytes_to_read) { |
6837 | size_t bytes_written = bytes_read; |
6838 | error = core_file.get()->Write(buf: bytes, num_bytes&: bytes_written); |
6839 | bytes_left -= bytes_read; |
6840 | addr += bytes_read; |
6841 | } else { |
6842 | // Some pages within regions are not readable, those should |
6843 | // be zero filled |
6844 | memset(s: bytes, c: 0, n: bytes_to_read); |
6845 | size_t bytes_written = bytes_to_read; |
6846 | error = core_file.get()->Write(buf: bytes, num_bytes&: bytes_written); |
6847 | bytes_left -= bytes_to_read; |
6848 | addr += bytes_to_read; |
6849 | } |
6850 | } |
6851 | } |
6852 | } |
6853 | } |
6854 | } |
6855 | } |
6856 | return true; // This is the right plug to handle saving core files for |
6857 | // this process |
6858 | } |
6859 | return false; |
6860 | } |
6861 | |
6862 | ObjectFileMachO::MachOCorefileAllImageInfos |
6863 | ObjectFileMachO::GetCorefileAllImageInfos() { |
6864 | MachOCorefileAllImageInfos image_infos; |
6865 | Log *log(GetLog(mask: LLDBLog::Object | LLDBLog::Symbols | LLDBLog::Process | |
6866 | LLDBLog::DynamicLoader)); |
6867 | |
6868 | auto lc_notes = FindLC_NOTEByName(name: "all image infos" ); |
6869 | for (auto lc_note : lc_notes) { |
6870 | offset_t payload_offset = std::get<0>(t&: lc_note); |
6871 | // Read the struct all_image_infos_header. |
6872 | uint32_t version = m_data.GetU32(offset_ptr: &payload_offset); |
6873 | if (version != 1) { |
6874 | return image_infos; |
6875 | } |
6876 | uint32_t imgcount = m_data.GetU32(offset_ptr: &payload_offset); |
6877 | uint64_t entries_fileoff = m_data.GetU64(offset_ptr: &payload_offset); |
6878 | // 'entries_size' is not used, nor is the 'unused' entry. |
6879 | // offset += 4; // uint32_t entries_size; |
6880 | // offset += 4; // uint32_t unused; |
6881 | |
6882 | LLDB_LOGF(log, "LC_NOTE 'all image infos' found version %d with %d images" , |
6883 | version, imgcount); |
6884 | payload_offset = entries_fileoff; |
6885 | for (uint32_t i = 0; i < imgcount; i++) { |
6886 | // Read the struct image_entry. |
6887 | offset_t filepath_offset = m_data.GetU64(offset_ptr: &payload_offset); |
6888 | uuid_t uuid; |
6889 | memcpy(dest: &uuid, src: m_data.GetData(offset_ptr: &payload_offset, length: sizeof(uuid_t)), |
6890 | n: sizeof(uuid_t)); |
6891 | uint64_t load_address = m_data.GetU64(offset_ptr: &payload_offset); |
6892 | offset_t seg_addrs_offset = m_data.GetU64(offset_ptr: &payload_offset); |
6893 | uint32_t segment_count = m_data.GetU32(offset_ptr: &payload_offset); |
6894 | uint32_t currently_executing = m_data.GetU32(offset_ptr: &payload_offset); |
6895 | |
6896 | MachOCorefileImageEntry image_entry; |
6897 | image_entry.filename = (const char *)m_data.GetCStr(offset_ptr: &filepath_offset); |
6898 | image_entry.uuid = UUID(uuid, sizeof(uuid_t)); |
6899 | image_entry.load_address = load_address; |
6900 | image_entry.currently_executing = currently_executing; |
6901 | |
6902 | offset_t seg_vmaddrs_offset = seg_addrs_offset; |
6903 | for (uint32_t j = 0; j < segment_count; j++) { |
6904 | char segname[17]; |
6905 | m_data.CopyData(offset: seg_vmaddrs_offset, length: 16, dst: segname); |
6906 | segname[16] = '\0'; |
6907 | seg_vmaddrs_offset += 16; |
6908 | uint64_t vmaddr = m_data.GetU64(offset_ptr: &seg_vmaddrs_offset); |
6909 | seg_vmaddrs_offset += 8; /* unused */ |
6910 | |
6911 | std::tuple<ConstString, addr_t> new_seg{ConstString(segname), vmaddr}; |
6912 | image_entry.segment_load_addresses.push_back(x: new_seg); |
6913 | } |
6914 | LLDB_LOGF(log, " image entry: %s %s 0x%" PRIx64 " %s" , |
6915 | image_entry.filename.c_str(), |
6916 | image_entry.uuid.GetAsString().c_str(), |
6917 | image_entry.load_address, |
6918 | image_entry.currently_executing ? "currently executing" |
6919 | : "not currently executing" ); |
6920 | image_infos.all_image_infos.push_back(x: image_entry); |
6921 | } |
6922 | } |
6923 | |
6924 | lc_notes = FindLC_NOTEByName(name: "load binary" ); |
6925 | for (auto lc_note : lc_notes) { |
6926 | offset_t payload_offset = std::get<0>(t&: lc_note); |
6927 | uint32_t version = m_data.GetU32(offset_ptr: &payload_offset); |
6928 | if (version == 1) { |
6929 | uuid_t uuid; |
6930 | memcpy(dest: &uuid, src: m_data.GetData(offset_ptr: &payload_offset, length: sizeof(uuid_t)), |
6931 | n: sizeof(uuid_t)); |
6932 | uint64_t load_address = m_data.GetU64(offset_ptr: &payload_offset); |
6933 | uint64_t slide = m_data.GetU64(offset_ptr: &payload_offset); |
6934 | std::string filename = m_data.GetCStr(offset_ptr: &payload_offset); |
6935 | |
6936 | MachOCorefileImageEntry image_entry; |
6937 | image_entry.filename = filename; |
6938 | image_entry.uuid = UUID(uuid, sizeof(uuid_t)); |
6939 | image_entry.load_address = load_address; |
6940 | image_entry.slide = slide; |
6941 | image_entry.currently_executing = true; |
6942 | image_infos.all_image_infos.push_back(x: image_entry); |
6943 | LLDB_LOGF(log, |
6944 | "LC_NOTE 'load binary' found, filename %s uuid %s load " |
6945 | "address 0x%" PRIx64 " slide 0x%" PRIx64, |
6946 | filename.c_str(), |
6947 | image_entry.uuid.IsValid() |
6948 | ? image_entry.uuid.GetAsString().c_str() |
6949 | : "00000000-0000-0000-0000-000000000000" , |
6950 | load_address, slide); |
6951 | } |
6952 | } |
6953 | |
6954 | return image_infos; |
6955 | } |
6956 | |
6957 | bool ObjectFileMachO::LoadCoreFileImages(lldb_private::Process &process) { |
6958 | MachOCorefileAllImageInfos image_infos = GetCorefileAllImageInfos(); |
6959 | Log *log = GetLog(mask: LLDBLog::Object | LLDBLog::DynamicLoader); |
6960 | Status error; |
6961 | |
6962 | bool found_platform_binary = false; |
6963 | ModuleList added_modules; |
6964 | for (MachOCorefileImageEntry &image : image_infos.all_image_infos) { |
6965 | ModuleSP module_sp, local_filesystem_module_sp; |
6966 | |
6967 | // If this is a platform binary, it has been loaded (or registered with |
6968 | // the DynamicLoader to be loaded), we don't need to do any further |
6969 | // processing. We're not going to call ModulesDidLoad on this in this |
6970 | // method, so notify==true. |
6971 | if (process.GetTarget() |
6972 | .GetDebugger() |
6973 | .GetPlatformList() |
6974 | .LoadPlatformBinaryAndSetup(process: &process, addr: image.load_address, |
6975 | notify: true /* notify */)) { |
6976 | LLDB_LOGF(log, |
6977 | "ObjectFileMachO::%s binary at 0x%" PRIx64 |
6978 | " is a platform binary, has been handled by a Platform plugin." , |
6979 | __FUNCTION__, image.load_address); |
6980 | continue; |
6981 | } |
6982 | |
6983 | bool value_is_offset = image.load_address == LLDB_INVALID_ADDRESS; |
6984 | uint64_t value = value_is_offset ? image.slide : image.load_address; |
6985 | if (value_is_offset && value == LLDB_INVALID_ADDRESS) { |
6986 | // We have neither address nor slide; so we will find the binary |
6987 | // by UUID and load it at slide/offset 0. |
6988 | value = 0; |
6989 | } |
6990 | |
6991 | // We have either a UUID, or we have a load address which |
6992 | // and can try to read load commands and find a UUID. |
6993 | if (image.uuid.IsValid() || |
6994 | (!value_is_offset && value != LLDB_INVALID_ADDRESS)) { |
6995 | const bool set_load_address = image.segment_load_addresses.size() == 0; |
6996 | const bool notify = false; |
6997 | // Userland Darwin binaries will have segment load addresses via |
6998 | // the `all image infos` LC_NOTE. |
6999 | const bool allow_memory_image_last_resort = |
7000 | image.segment_load_addresses.size(); |
7001 | module_sp = DynamicLoader::LoadBinaryWithUUIDAndAddress( |
7002 | process: &process, name: image.filename, uuid: image.uuid, value, value_is_offset, |
7003 | force_symbol_search: image.currently_executing, notify, set_address_in_target: set_load_address, |
7004 | allow_memory_image_last_resort); |
7005 | } |
7006 | |
7007 | // We have a ModuleSP to load in the Target. Load it at the |
7008 | // correct address/slide and notify/load scripting resources. |
7009 | if (module_sp) { |
7010 | added_modules.Append(module_sp, notify: false /* notify */); |
7011 | |
7012 | // We have a list of segment load address |
7013 | if (image.segment_load_addresses.size() > 0) { |
7014 | if (log) { |
7015 | std::string uuidstr = image.uuid.GetAsString(); |
7016 | log->Printf(format: "ObjectFileMachO::LoadCoreFileImages adding binary '%s' " |
7017 | "UUID %s with section load addresses" , |
7018 | module_sp->GetFileSpec().GetPath().c_str(), |
7019 | uuidstr.c_str()); |
7020 | } |
7021 | for (auto name_vmaddr_tuple : image.segment_load_addresses) { |
7022 | SectionList *sectlist = module_sp->GetObjectFile()->GetSectionList(); |
7023 | if (sectlist) { |
7024 | SectionSP sect_sp = |
7025 | sectlist->FindSectionByName(section_dstr: std::get<0>(t&: name_vmaddr_tuple)); |
7026 | if (sect_sp) { |
7027 | process.GetTarget().SetSectionLoadAddress( |
7028 | section: sect_sp, load_addr: std::get<1>(t&: name_vmaddr_tuple)); |
7029 | } |
7030 | } |
7031 | } |
7032 | } else { |
7033 | if (log) { |
7034 | std::string uuidstr = image.uuid.GetAsString(); |
7035 | log->Printf(format: "ObjectFileMachO::LoadCoreFileImages adding binary '%s' " |
7036 | "UUID %s with %s 0x%" PRIx64, |
7037 | module_sp->GetFileSpec().GetPath().c_str(), |
7038 | uuidstr.c_str(), |
7039 | value_is_offset ? "slide" : "load address" , value); |
7040 | } |
7041 | bool changed; |
7042 | module_sp->SetLoadAddress(target&: process.GetTarget(), value, value_is_offset, |
7043 | changed); |
7044 | } |
7045 | } |
7046 | } |
7047 | if (added_modules.GetSize() > 0) { |
7048 | process.GetTarget().ModulesDidLoad(module_list&: added_modules); |
7049 | process.Flush(); |
7050 | return true; |
7051 | } |
7052 | // Return true if the only binary we found was the platform binary, |
7053 | // and it was loaded outside the scope of this method. |
7054 | if (found_platform_binary) |
7055 | return true; |
7056 | |
7057 | // No binaries. |
7058 | return false; |
7059 | } |
7060 | |