1 //===-- ObjectFileMachO.cpp -------------------------------------*- C++ -*-===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 10 // C Includes 11 // C++ Includes 12 // Other libraries and framework includes 13 #include "llvm/ADT/StringRef.h" 14 15 // Project includes 16 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 17 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h" 18 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 19 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 20 #include "lldb/Core/ArchSpec.h" 21 #include "lldb/Core/DataBuffer.h" 22 #include "lldb/Core/Debugger.h" 23 #include "lldb/Core/Error.h" 24 #include "lldb/Core/FileSpecList.h" 25 #include "lldb/Core/Log.h" 26 #include "lldb/Core/Module.h" 27 #include "lldb/Core/ModuleSpec.h" 28 #include "lldb/Core/PluginManager.h" 29 #include "lldb/Core/RangeMap.h" 30 #include "lldb/Core/Section.h" 31 #include "lldb/Core/StreamFile.h" 32 #include "lldb/Core/StreamString.h" 33 #include "lldb/Core/Timer.h" 34 #include "lldb/Core/UUID.h" 35 #include "lldb/Host/FileSpec.h" 36 #include "lldb/Host/Host.h" 37 #include "lldb/Symbol/DWARFCallFrameInfo.h" 38 #include "lldb/Symbol/ObjectFile.h" 39 #include "lldb/Target/DynamicLoader.h" 40 #include "lldb/Target/MemoryRegionInfo.h" 41 #include "lldb/Target/Platform.h" 42 #include "lldb/Target/Process.h" 43 #include "lldb/Target/SectionLoadList.h" 44 #include "lldb/Target/Target.h" 45 #include "lldb/Target/Thread.h" 46 #include "lldb/Target/ThreadList.h" 47 48 #include "lldb/Utility/SafeMachO.h" 49 50 #include "ObjectFileMachO.h" 51 52 #if defined(__APPLE__) && \ 53 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 54 // GetLLDBSharedCacheUUID() needs to call dlsym() 55 #include <dlfcn.h> 56 #endif 57 58 #ifndef __APPLE__ 59 #include "Utility/UuidCompatibility.h" 60 #endif 61 62 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 63 using namespace lldb; 64 using namespace lldb_private; 65 using namespace llvm::MachO; 66 67 // Some structure definitions needed for parsing the dyld shared cache files 68 // found on iOS devices. 69 70 struct lldb_copy_dyld_cache_header_v1 { 71 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 72 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 73 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 74 uint32_t imagesOffset; 75 uint32_t imagesCount; 76 uint64_t dyldBaseAddress; 77 uint64_t codeSignatureOffset; 78 uint64_t codeSignatureSize; 79 uint64_t slideInfoOffset; 80 uint64_t slideInfoSize; 81 uint64_t localSymbolsOffset; 82 uint64_t localSymbolsSize; 83 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 84 // and later 85 }; 86 87 struct lldb_copy_dyld_cache_mapping_info { 88 uint64_t address; 89 uint64_t size; 90 uint64_t fileOffset; 91 uint32_t maxProt; 92 uint32_t initProt; 93 }; 94 95 struct lldb_copy_dyld_cache_local_symbols_info { 96 uint32_t nlistOffset; 97 uint32_t nlistCount; 98 uint32_t stringsOffset; 99 uint32_t stringsSize; 100 uint32_t entriesOffset; 101 uint32_t entriesCount; 102 }; 103 struct lldb_copy_dyld_cache_local_symbols_entry { 104 uint32_t dylibOffset; 105 uint32_t nlistStartIndex; 106 uint32_t nlistCount; 107 }; 108 109 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 110 public: 111 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 112 const DataExtractor &data) 113 : RegisterContextDarwin_x86_64(thread, 0) { 114 SetRegisterDataFrom_LC_THREAD(data); 115 } 116 117 void InvalidateAllRegisters() override { 118 // Do nothing... registers are always valid... 119 } 120 121 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 122 lldb::offset_t offset = 0; 123 SetError(GPRRegSet, Read, -1); 124 SetError(FPURegSet, Read, -1); 125 SetError(EXCRegSet, Read, -1); 126 bool done = false; 127 128 while (!done) { 129 int flavor = data.GetU32(&offset); 130 if (flavor == 0) 131 done = true; 132 else { 133 uint32_t i; 134 uint32_t count = data.GetU32(&offset); 135 switch (flavor) { 136 case GPRRegSet: 137 for (i = 0; i < count; ++i) 138 (&gpr.rax)[i] = data.GetU64(&offset); 139 SetError(GPRRegSet, Read, 0); 140 done = true; 141 142 break; 143 case FPURegSet: 144 // TODO: fill in FPU regs.... 145 // SetError (FPURegSet, Read, -1); 146 done = true; 147 148 break; 149 case EXCRegSet: 150 exc.trapno = data.GetU32(&offset); 151 exc.err = data.GetU32(&offset); 152 exc.faultvaddr = data.GetU64(&offset); 153 SetError(EXCRegSet, Read, 0); 154 done = true; 155 break; 156 case 7: 157 case 8: 158 case 9: 159 // fancy flavors that encapsulate of the above 160 // flavors... 161 break; 162 163 default: 164 done = true; 165 break; 166 } 167 } 168 } 169 } 170 171 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 172 const char *alt_name, size_t reg_byte_size, 173 Stream &data) { 174 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 175 if (reg_info == NULL) 176 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 177 if (reg_info) { 178 lldb_private::RegisterValue reg_value; 179 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 180 if (reg_info->byte_size >= reg_byte_size) 181 data.Write(reg_value.GetBytes(), reg_byte_size); 182 else { 183 data.Write(reg_value.GetBytes(), reg_info->byte_size); 184 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 185 ++i) 186 data.PutChar(0); 187 } 188 return reg_byte_size; 189 } 190 } 191 // Just write zeros if all else fails 192 for (size_t i = 0; i < reg_byte_size; ++i) 193 data.PutChar(0); 194 return reg_byte_size; 195 } 196 197 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 198 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 199 if (reg_ctx_sp) { 200 RegisterContext *reg_ctx = reg_ctx_sp.get(); 201 202 data.PutHex32(GPRRegSet); // Flavor 203 data.PutHex32(GPRWordCount); 204 WriteRegister(reg_ctx, "rax", NULL, 8, data); 205 WriteRegister(reg_ctx, "rbx", NULL, 8, data); 206 WriteRegister(reg_ctx, "rcx", NULL, 8, data); 207 WriteRegister(reg_ctx, "rdx", NULL, 8, data); 208 WriteRegister(reg_ctx, "rdi", NULL, 8, data); 209 WriteRegister(reg_ctx, "rsi", NULL, 8, data); 210 WriteRegister(reg_ctx, "rbp", NULL, 8, data); 211 WriteRegister(reg_ctx, "rsp", NULL, 8, data); 212 WriteRegister(reg_ctx, "r8", NULL, 8, data); 213 WriteRegister(reg_ctx, "r9", NULL, 8, data); 214 WriteRegister(reg_ctx, "r10", NULL, 8, data); 215 WriteRegister(reg_ctx, "r11", NULL, 8, data); 216 WriteRegister(reg_ctx, "r12", NULL, 8, data); 217 WriteRegister(reg_ctx, "r13", NULL, 8, data); 218 WriteRegister(reg_ctx, "r14", NULL, 8, data); 219 WriteRegister(reg_ctx, "r15", NULL, 8, data); 220 WriteRegister(reg_ctx, "rip", NULL, 8, data); 221 WriteRegister(reg_ctx, "rflags", NULL, 8, data); 222 WriteRegister(reg_ctx, "cs", NULL, 8, data); 223 WriteRegister(reg_ctx, "fs", NULL, 8, data); 224 WriteRegister(reg_ctx, "gs", NULL, 8, data); 225 226 // // Write out the FPU registers 227 // const size_t fpu_byte_size = sizeof(FPU); 228 // size_t bytes_written = 0; 229 // data.PutHex32 (FPURegSet); 230 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 231 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 232 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 233 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 234 // data); // uint16_t fcw; // "fctrl" 235 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 236 // data); // uint16_t fsw; // "fstat" 237 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 238 // data); // uint8_t ftw; // "ftag" 239 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 240 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 241 // data); // uint16_t fop; // "fop" 242 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 243 // data); // uint32_t ip; // "fioff" 244 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 245 // data); // uint16_t cs; // "fiseg" 246 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 247 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 248 // data); // uint32_t dp; // "fooff" 249 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 250 // data); // uint16_t ds; // "foseg" 251 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 252 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 253 // data); // uint32_t mxcsr; 254 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 255 // 4, data);// uint32_t mxcsrmask; 256 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 257 // sizeof(MMSReg), data); 258 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 259 // sizeof(MMSReg), data); 260 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 261 // sizeof(MMSReg), data); 262 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 263 // sizeof(MMSReg), data); 264 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 265 // sizeof(MMSReg), data); 266 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 267 // sizeof(MMSReg), data); 268 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 269 // sizeof(MMSReg), data); 270 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 271 // sizeof(MMSReg), data); 272 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 273 // sizeof(XMMReg), data); 274 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 275 // sizeof(XMMReg), data); 276 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 277 // sizeof(XMMReg), data); 278 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 279 // sizeof(XMMReg), data); 280 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 281 // sizeof(XMMReg), data); 282 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 283 // sizeof(XMMReg), data); 284 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 285 // sizeof(XMMReg), data); 286 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 287 // sizeof(XMMReg), data); 288 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 289 // sizeof(XMMReg), data); 290 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 291 // sizeof(XMMReg), data); 292 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 293 // sizeof(XMMReg), data); 294 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 295 // sizeof(XMMReg), data); 296 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 297 // sizeof(XMMReg), data); 298 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 299 // sizeof(XMMReg), data); 300 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 301 // sizeof(XMMReg), data); 302 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 303 // sizeof(XMMReg), data); 304 // 305 // // Fill rest with zeros 306 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 307 // i) 308 // data.PutChar(0); 309 310 // Write out the EXC registers 311 data.PutHex32(EXCRegSet); 312 data.PutHex32(EXCWordCount); 313 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 314 WriteRegister(reg_ctx, "err", NULL, 4, data); 315 WriteRegister(reg_ctx, "faultvaddr", NULL, 8, data); 316 return true; 317 } 318 return false; 319 } 320 321 protected: 322 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 323 324 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 325 326 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 327 328 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 329 return 0; 330 } 331 332 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 333 return 0; 334 } 335 336 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 337 return 0; 338 } 339 }; 340 341 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 342 public: 343 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 344 const DataExtractor &data) 345 : RegisterContextDarwin_i386(thread, 0) { 346 SetRegisterDataFrom_LC_THREAD(data); 347 } 348 349 void InvalidateAllRegisters() override { 350 // Do nothing... registers are always valid... 351 } 352 353 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 354 lldb::offset_t offset = 0; 355 SetError(GPRRegSet, Read, -1); 356 SetError(FPURegSet, Read, -1); 357 SetError(EXCRegSet, Read, -1); 358 bool done = false; 359 360 while (!done) { 361 int flavor = data.GetU32(&offset); 362 if (flavor == 0) 363 done = true; 364 else { 365 uint32_t i; 366 uint32_t count = data.GetU32(&offset); 367 switch (flavor) { 368 case GPRRegSet: 369 for (i = 0; i < count; ++i) 370 (&gpr.eax)[i] = data.GetU32(&offset); 371 SetError(GPRRegSet, Read, 0); 372 done = true; 373 374 break; 375 case FPURegSet: 376 // TODO: fill in FPU regs.... 377 // SetError (FPURegSet, Read, -1); 378 done = true; 379 380 break; 381 case EXCRegSet: 382 exc.trapno = data.GetU32(&offset); 383 exc.err = data.GetU32(&offset); 384 exc.faultvaddr = data.GetU32(&offset); 385 SetError(EXCRegSet, Read, 0); 386 done = true; 387 break; 388 case 7: 389 case 8: 390 case 9: 391 // fancy flavors that encapsulate of the above 392 // flavors... 393 break; 394 395 default: 396 done = true; 397 break; 398 } 399 } 400 } 401 } 402 403 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 404 const char *alt_name, size_t reg_byte_size, 405 Stream &data) { 406 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 407 if (reg_info == NULL) 408 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 409 if (reg_info) { 410 lldb_private::RegisterValue reg_value; 411 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 412 if (reg_info->byte_size >= reg_byte_size) 413 data.Write(reg_value.GetBytes(), reg_byte_size); 414 else { 415 data.Write(reg_value.GetBytes(), reg_info->byte_size); 416 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 417 ++i) 418 data.PutChar(0); 419 } 420 return reg_byte_size; 421 } 422 } 423 // Just write zeros if all else fails 424 for (size_t i = 0; i < reg_byte_size; ++i) 425 data.PutChar(0); 426 return reg_byte_size; 427 } 428 429 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 430 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 431 if (reg_ctx_sp) { 432 RegisterContext *reg_ctx = reg_ctx_sp.get(); 433 434 data.PutHex32(GPRRegSet); // Flavor 435 data.PutHex32(GPRWordCount); 436 WriteRegister(reg_ctx, "eax", NULL, 4, data); 437 WriteRegister(reg_ctx, "ebx", NULL, 4, data); 438 WriteRegister(reg_ctx, "ecx", NULL, 4, data); 439 WriteRegister(reg_ctx, "edx", NULL, 4, data); 440 WriteRegister(reg_ctx, "edi", NULL, 4, data); 441 WriteRegister(reg_ctx, "esi", NULL, 4, data); 442 WriteRegister(reg_ctx, "ebp", NULL, 4, data); 443 WriteRegister(reg_ctx, "esp", NULL, 4, data); 444 WriteRegister(reg_ctx, "ss", NULL, 4, data); 445 WriteRegister(reg_ctx, "eflags", NULL, 4, data); 446 WriteRegister(reg_ctx, "eip", NULL, 4, data); 447 WriteRegister(reg_ctx, "cs", NULL, 4, data); 448 WriteRegister(reg_ctx, "ds", NULL, 4, data); 449 WriteRegister(reg_ctx, "es", NULL, 4, data); 450 WriteRegister(reg_ctx, "fs", NULL, 4, data); 451 WriteRegister(reg_ctx, "gs", NULL, 4, data); 452 453 // Write out the EXC registers 454 data.PutHex32(EXCRegSet); 455 data.PutHex32(EXCWordCount); 456 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 457 WriteRegister(reg_ctx, "err", NULL, 4, data); 458 WriteRegister(reg_ctx, "faultvaddr", NULL, 4, data); 459 return true; 460 } 461 return false; 462 } 463 464 protected: 465 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 466 467 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 468 469 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 470 471 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 472 return 0; 473 } 474 475 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 476 return 0; 477 } 478 479 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 480 return 0; 481 } 482 }; 483 484 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 485 public: 486 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 487 const DataExtractor &data) 488 : RegisterContextDarwin_arm(thread, 0) { 489 SetRegisterDataFrom_LC_THREAD(data); 490 } 491 492 void InvalidateAllRegisters() override { 493 // Do nothing... registers are always valid... 494 } 495 496 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 497 lldb::offset_t offset = 0; 498 SetError(GPRRegSet, Read, -1); 499 SetError(FPURegSet, Read, -1); 500 SetError(EXCRegSet, Read, -1); 501 bool done = false; 502 503 while (!done) { 504 int flavor = data.GetU32(&offset); 505 uint32_t count = data.GetU32(&offset); 506 lldb::offset_t next_thread_state = offset + (count * 4); 507 switch (flavor) { 508 case GPRAltRegSet: 509 case GPRRegSet: 510 for (uint32_t i = 0; i < count; ++i) { 511 gpr.r[i] = data.GetU32(&offset); 512 } 513 514 // Note that gpr.cpsr is also copied by the above loop; this loop 515 // technically extends 516 // one element past the end of the gpr.r[] array. 517 518 SetError(GPRRegSet, Read, 0); 519 offset = next_thread_state; 520 break; 521 522 case FPURegSet: { 523 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 524 const int fpu_reg_buf_size = sizeof(fpu.floats); 525 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 526 fpu_reg_buf) == fpu_reg_buf_size) { 527 offset += fpu_reg_buf_size; 528 fpu.fpscr = data.GetU32(&offset); 529 SetError(FPURegSet, Read, 0); 530 } else { 531 done = true; 532 } 533 } 534 offset = next_thread_state; 535 break; 536 537 case EXCRegSet: 538 if (count == 3) { 539 exc.exception = data.GetU32(&offset); 540 exc.fsr = data.GetU32(&offset); 541 exc.far = data.GetU32(&offset); 542 SetError(EXCRegSet, Read, 0); 543 } 544 done = true; 545 offset = next_thread_state; 546 break; 547 548 // Unknown register set flavor, stop trying to parse. 549 default: 550 done = true; 551 } 552 } 553 } 554 555 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 556 const char *alt_name, size_t reg_byte_size, 557 Stream &data) { 558 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 559 if (reg_info == NULL) 560 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 561 if (reg_info) { 562 lldb_private::RegisterValue reg_value; 563 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 564 if (reg_info->byte_size >= reg_byte_size) 565 data.Write(reg_value.GetBytes(), reg_byte_size); 566 else { 567 data.Write(reg_value.GetBytes(), reg_info->byte_size); 568 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 569 ++i) 570 data.PutChar(0); 571 } 572 return reg_byte_size; 573 } 574 } 575 // Just write zeros if all else fails 576 for (size_t i = 0; i < reg_byte_size; ++i) 577 data.PutChar(0); 578 return reg_byte_size; 579 } 580 581 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 582 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 583 if (reg_ctx_sp) { 584 RegisterContext *reg_ctx = reg_ctx_sp.get(); 585 586 data.PutHex32(GPRRegSet); // Flavor 587 data.PutHex32(GPRWordCount); 588 WriteRegister(reg_ctx, "r0", NULL, 4, data); 589 WriteRegister(reg_ctx, "r1", NULL, 4, data); 590 WriteRegister(reg_ctx, "r2", NULL, 4, data); 591 WriteRegister(reg_ctx, "r3", NULL, 4, data); 592 WriteRegister(reg_ctx, "r4", NULL, 4, data); 593 WriteRegister(reg_ctx, "r5", NULL, 4, data); 594 WriteRegister(reg_ctx, "r6", NULL, 4, data); 595 WriteRegister(reg_ctx, "r7", NULL, 4, data); 596 WriteRegister(reg_ctx, "r8", NULL, 4, data); 597 WriteRegister(reg_ctx, "r9", NULL, 4, data); 598 WriteRegister(reg_ctx, "r10", NULL, 4, data); 599 WriteRegister(reg_ctx, "r11", NULL, 4, data); 600 WriteRegister(reg_ctx, "r12", NULL, 4, data); 601 WriteRegister(reg_ctx, "sp", NULL, 4, data); 602 WriteRegister(reg_ctx, "lr", NULL, 4, data); 603 WriteRegister(reg_ctx, "pc", NULL, 4, data); 604 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 605 606 // Write out the EXC registers 607 // data.PutHex32 (EXCRegSet); 608 // data.PutHex32 (EXCWordCount); 609 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 610 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 611 // WriteRegister (reg_ctx, "far", NULL, 4, data); 612 return true; 613 } 614 return false; 615 } 616 617 protected: 618 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 619 620 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 621 622 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 623 624 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 625 626 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 627 return 0; 628 } 629 630 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 631 return 0; 632 } 633 634 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 635 return 0; 636 } 637 638 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 639 return -1; 640 } 641 }; 642 643 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 644 public: 645 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 646 const DataExtractor &data) 647 : RegisterContextDarwin_arm64(thread, 0) { 648 SetRegisterDataFrom_LC_THREAD(data); 649 } 650 651 void InvalidateAllRegisters() override { 652 // Do nothing... registers are always valid... 653 } 654 655 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 656 lldb::offset_t offset = 0; 657 SetError(GPRRegSet, Read, -1); 658 SetError(FPURegSet, Read, -1); 659 SetError(EXCRegSet, Read, -1); 660 bool done = false; 661 while (!done) { 662 int flavor = data.GetU32(&offset); 663 uint32_t count = data.GetU32(&offset); 664 lldb::offset_t next_thread_state = offset + (count * 4); 665 switch (flavor) { 666 case GPRRegSet: 667 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 668 // 32-bit register) 669 if (count >= (33 * 2) + 1) { 670 for (uint32_t i = 0; i < 33; ++i) 671 gpr.x[i] = data.GetU64(&offset); 672 gpr.cpsr = data.GetU32(&offset); 673 SetError(GPRRegSet, Read, 0); 674 } 675 offset = next_thread_state; 676 break; 677 case FPURegSet: { 678 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 679 const int fpu_reg_buf_size = sizeof(fpu); 680 if (fpu_reg_buf_size == count && 681 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 682 fpu_reg_buf) == fpu_reg_buf_size) { 683 SetError(FPURegSet, Read, 0); 684 } else { 685 done = true; 686 } 687 } 688 offset = next_thread_state; 689 break; 690 case EXCRegSet: 691 if (count == 4) { 692 exc.far = data.GetU64(&offset); 693 exc.esr = data.GetU32(&offset); 694 exc.exception = data.GetU32(&offset); 695 SetError(EXCRegSet, Read, 0); 696 } 697 offset = next_thread_state; 698 break; 699 default: 700 done = true; 701 break; 702 } 703 } 704 } 705 706 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 707 const char *alt_name, size_t reg_byte_size, 708 Stream &data) { 709 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 710 if (reg_info == NULL) 711 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 712 if (reg_info) { 713 lldb_private::RegisterValue reg_value; 714 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 715 if (reg_info->byte_size >= reg_byte_size) 716 data.Write(reg_value.GetBytes(), reg_byte_size); 717 else { 718 data.Write(reg_value.GetBytes(), reg_info->byte_size); 719 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 720 ++i) 721 data.PutChar(0); 722 } 723 return reg_byte_size; 724 } 725 } 726 // Just write zeros if all else fails 727 for (size_t i = 0; i < reg_byte_size; ++i) 728 data.PutChar(0); 729 return reg_byte_size; 730 } 731 732 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 733 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 734 if (reg_ctx_sp) { 735 RegisterContext *reg_ctx = reg_ctx_sp.get(); 736 737 data.PutHex32(GPRRegSet); // Flavor 738 data.PutHex32(GPRWordCount); 739 WriteRegister(reg_ctx, "x0", NULL, 8, data); 740 WriteRegister(reg_ctx, "x1", NULL, 8, data); 741 WriteRegister(reg_ctx, "x2", NULL, 8, data); 742 WriteRegister(reg_ctx, "x3", NULL, 8, data); 743 WriteRegister(reg_ctx, "x4", NULL, 8, data); 744 WriteRegister(reg_ctx, "x5", NULL, 8, data); 745 WriteRegister(reg_ctx, "x6", NULL, 8, data); 746 WriteRegister(reg_ctx, "x7", NULL, 8, data); 747 WriteRegister(reg_ctx, "x8", NULL, 8, data); 748 WriteRegister(reg_ctx, "x9", NULL, 8, data); 749 WriteRegister(reg_ctx, "x10", NULL, 8, data); 750 WriteRegister(reg_ctx, "x11", NULL, 8, data); 751 WriteRegister(reg_ctx, "x12", NULL, 8, data); 752 WriteRegister(reg_ctx, "x13", NULL, 8, data); 753 WriteRegister(reg_ctx, "x14", NULL, 8, data); 754 WriteRegister(reg_ctx, "x15", NULL, 8, data); 755 WriteRegister(reg_ctx, "x16", NULL, 8, data); 756 WriteRegister(reg_ctx, "x17", NULL, 8, data); 757 WriteRegister(reg_ctx, "x18", NULL, 8, data); 758 WriteRegister(reg_ctx, "x19", NULL, 8, data); 759 WriteRegister(reg_ctx, "x20", NULL, 8, data); 760 WriteRegister(reg_ctx, "x21", NULL, 8, data); 761 WriteRegister(reg_ctx, "x22", NULL, 8, data); 762 WriteRegister(reg_ctx, "x23", NULL, 8, data); 763 WriteRegister(reg_ctx, "x24", NULL, 8, data); 764 WriteRegister(reg_ctx, "x25", NULL, 8, data); 765 WriteRegister(reg_ctx, "x26", NULL, 8, data); 766 WriteRegister(reg_ctx, "x27", NULL, 8, data); 767 WriteRegister(reg_ctx, "x28", NULL, 8, data); 768 WriteRegister(reg_ctx, "fp", NULL, 8, data); 769 WriteRegister(reg_ctx, "lr", NULL, 8, data); 770 WriteRegister(reg_ctx, "sp", NULL, 8, data); 771 WriteRegister(reg_ctx, "pc", NULL, 8, data); 772 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 773 774 // Write out the EXC registers 775 // data.PutHex32 (EXCRegSet); 776 // data.PutHex32 (EXCWordCount); 777 // WriteRegister (reg_ctx, "far", NULL, 8, data); 778 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 779 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 780 return true; 781 } 782 return false; 783 } 784 785 protected: 786 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 787 788 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 789 790 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 791 792 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 793 794 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 795 return 0; 796 } 797 798 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 799 return 0; 800 } 801 802 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 803 return 0; 804 } 805 806 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 807 return -1; 808 } 809 }; 810 811 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 812 switch (magic) { 813 case MH_MAGIC: 814 case MH_CIGAM: 815 return sizeof(struct mach_header); 816 817 case MH_MAGIC_64: 818 case MH_CIGAM_64: 819 return sizeof(struct mach_header_64); 820 break; 821 822 default: 823 break; 824 } 825 return 0; 826 } 827 828 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 829 830 void ObjectFileMachO::Initialize() { 831 PluginManager::RegisterPlugin( 832 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 833 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 834 } 835 836 void ObjectFileMachO::Terminate() { 837 PluginManager::UnregisterPlugin(CreateInstance); 838 } 839 840 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 841 static ConstString g_name("mach-o"); 842 return g_name; 843 } 844 845 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 846 return "Mach-o object file reader (32 and 64 bit)"; 847 } 848 849 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 850 DataBufferSP &data_sp, 851 lldb::offset_t data_offset, 852 const FileSpec *file, 853 lldb::offset_t file_offset, 854 lldb::offset_t length) { 855 if (!data_sp) { 856 data_sp = file->MemoryMapFileContentsIfLocal(file_offset, length); 857 data_offset = 0; 858 } 859 860 if (ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) { 861 // Update the data to contain the entire file if it doesn't already 862 if (data_sp->GetByteSize() < length) { 863 data_sp = file->MemoryMapFileContentsIfLocal(file_offset, length); 864 data_offset = 0; 865 } 866 std::unique_ptr<ObjectFile> objfile_ap(new ObjectFileMachO( 867 module_sp, data_sp, data_offset, file, file_offset, length)); 868 if (objfile_ap.get() && objfile_ap->ParseHeader()) 869 return objfile_ap.release(); 870 } 871 return NULL; 872 } 873 874 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 875 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 876 const ProcessSP &process_sp, lldb::addr_t header_addr) { 877 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 878 std::unique_ptr<ObjectFile> objfile_ap( 879 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 880 if (objfile_ap.get() && objfile_ap->ParseHeader()) 881 return objfile_ap.release(); 882 } 883 return NULL; 884 } 885 886 size_t ObjectFileMachO::GetModuleSpecifications( 887 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 888 lldb::offset_t data_offset, lldb::offset_t file_offset, 889 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 890 const size_t initial_count = specs.GetSize(); 891 892 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 893 DataExtractor data; 894 data.SetData(data_sp); 895 llvm::MachO::mach_header header; 896 if (ParseHeader(data, &data_offset, header)) { 897 size_t header_and_load_cmds = 898 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 899 if (header_and_load_cmds >= data_sp->GetByteSize()) { 900 data_sp = file.ReadFileContents(file_offset, header_and_load_cmds); 901 data.SetData(data_sp); 902 data_offset = MachHeaderSizeFromMagic(header.magic); 903 } 904 if (data_sp) { 905 ModuleSpec spec; 906 spec.GetFileSpec() = file; 907 spec.SetObjectOffset(file_offset); 908 spec.SetObjectSize(length); 909 910 if (GetArchitecture(header, data, data_offset, 911 spec.GetArchitecture())) { 912 if (spec.GetArchitecture().IsValid()) { 913 GetUUID(header, data, data_offset, spec.GetUUID()); 914 specs.Append(spec); 915 } 916 } 917 } 918 } 919 } 920 return specs.GetSize() - initial_count; 921 } 922 923 const ConstString &ObjectFileMachO::GetSegmentNameTEXT() { 924 static ConstString g_segment_name_TEXT("__TEXT"); 925 return g_segment_name_TEXT; 926 } 927 928 const ConstString &ObjectFileMachO::GetSegmentNameDATA() { 929 static ConstString g_segment_name_DATA("__DATA"); 930 return g_segment_name_DATA; 931 } 932 933 const ConstString &ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 934 static ConstString g_segment_name("__DATA_DIRTY"); 935 return g_segment_name; 936 } 937 938 const ConstString &ObjectFileMachO::GetSegmentNameDATA_CONST() { 939 static ConstString g_segment_name("__DATA_CONST"); 940 return g_segment_name; 941 } 942 943 const ConstString &ObjectFileMachO::GetSegmentNameOBJC() { 944 static ConstString g_segment_name_OBJC("__OBJC"); 945 return g_segment_name_OBJC; 946 } 947 948 const ConstString &ObjectFileMachO::GetSegmentNameLINKEDIT() { 949 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 950 return g_section_name_LINKEDIT; 951 } 952 953 const ConstString &ObjectFileMachO::GetSectionNameEHFrame() { 954 static ConstString g_section_name_eh_frame("__eh_frame"); 955 return g_section_name_eh_frame; 956 } 957 958 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 959 lldb::addr_t data_offset, 960 lldb::addr_t data_length) { 961 DataExtractor data; 962 data.SetData(data_sp, data_offset, data_length); 963 lldb::offset_t offset = 0; 964 uint32_t magic = data.GetU32(&offset); 965 return MachHeaderSizeFromMagic(magic) != 0; 966 } 967 968 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 969 DataBufferSP &data_sp, 970 lldb::offset_t data_offset, 971 const FileSpec *file, 972 lldb::offset_t file_offset, 973 lldb::offset_t length) 974 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 975 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 976 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 977 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 978 ::memset(&m_header, 0, sizeof(m_header)); 979 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 980 } 981 982 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 983 lldb::DataBufferSP &header_data_sp, 984 const lldb::ProcessSP &process_sp, 985 lldb::addr_t header_addr) 986 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 987 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 988 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 989 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 990 ::memset(&m_header, 0, sizeof(m_header)); 991 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 992 } 993 994 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 995 lldb::offset_t *data_offset_ptr, 996 llvm::MachO::mach_header &header) { 997 data.SetByteOrder(endian::InlHostByteOrder()); 998 // Leave magic in the original byte order 999 header.magic = data.GetU32(data_offset_ptr); 1000 bool can_parse = false; 1001 bool is_64_bit = false; 1002 switch (header.magic) { 1003 case MH_MAGIC: 1004 data.SetByteOrder(endian::InlHostByteOrder()); 1005 data.SetAddressByteSize(4); 1006 can_parse = true; 1007 break; 1008 1009 case MH_MAGIC_64: 1010 data.SetByteOrder(endian::InlHostByteOrder()); 1011 data.SetAddressByteSize(8); 1012 can_parse = true; 1013 is_64_bit = true; 1014 break; 1015 1016 case MH_CIGAM: 1017 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1018 ? eByteOrderLittle 1019 : eByteOrderBig); 1020 data.SetAddressByteSize(4); 1021 can_parse = true; 1022 break; 1023 1024 case MH_CIGAM_64: 1025 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1026 ? eByteOrderLittle 1027 : eByteOrderBig); 1028 data.SetAddressByteSize(8); 1029 is_64_bit = true; 1030 can_parse = true; 1031 break; 1032 1033 default: 1034 break; 1035 } 1036 1037 if (can_parse) { 1038 data.GetU32(data_offset_ptr, &header.cputype, 6); 1039 if (is_64_bit) 1040 *data_offset_ptr += 4; 1041 return true; 1042 } else { 1043 memset(&header, 0, sizeof(header)); 1044 } 1045 return false; 1046 } 1047 1048 bool ObjectFileMachO::ParseHeader() { 1049 ModuleSP module_sp(GetModule()); 1050 if (module_sp) { 1051 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1052 bool can_parse = false; 1053 lldb::offset_t offset = 0; 1054 m_data.SetByteOrder(endian::InlHostByteOrder()); 1055 // Leave magic in the original byte order 1056 m_header.magic = m_data.GetU32(&offset); 1057 switch (m_header.magic) { 1058 case MH_MAGIC: 1059 m_data.SetByteOrder(endian::InlHostByteOrder()); 1060 m_data.SetAddressByteSize(4); 1061 can_parse = true; 1062 break; 1063 1064 case MH_MAGIC_64: 1065 m_data.SetByteOrder(endian::InlHostByteOrder()); 1066 m_data.SetAddressByteSize(8); 1067 can_parse = true; 1068 break; 1069 1070 case MH_CIGAM: 1071 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1072 ? eByteOrderLittle 1073 : eByteOrderBig); 1074 m_data.SetAddressByteSize(4); 1075 can_parse = true; 1076 break; 1077 1078 case MH_CIGAM_64: 1079 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1080 ? eByteOrderLittle 1081 : eByteOrderBig); 1082 m_data.SetAddressByteSize(8); 1083 can_parse = true; 1084 break; 1085 1086 default: 1087 break; 1088 } 1089 1090 if (can_parse) { 1091 m_data.GetU32(&offset, &m_header.cputype, 6); 1092 1093 ArchSpec mach_arch; 1094 1095 if (GetArchitecture(mach_arch)) { 1096 // Check if the module has a required architecture 1097 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1098 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1099 return false; 1100 1101 if (SetModulesArchitecture(mach_arch)) { 1102 const size_t header_and_lc_size = 1103 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1104 if (m_data.GetByteSize() < header_and_lc_size) { 1105 DataBufferSP data_sp; 1106 ProcessSP process_sp(m_process_wp.lock()); 1107 if (process_sp) { 1108 data_sp = 1109 ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1110 } else { 1111 // Read in all only the load command data from the file on disk 1112 data_sp = 1113 m_file.ReadFileContents(m_file_offset, header_and_lc_size); 1114 if (data_sp->GetByteSize() != header_and_lc_size) 1115 return false; 1116 } 1117 if (data_sp) 1118 m_data.SetData(data_sp); 1119 } 1120 } 1121 return true; 1122 } 1123 } else { 1124 memset(&m_header, 0, sizeof(struct mach_header)); 1125 } 1126 } 1127 return false; 1128 } 1129 1130 ByteOrder ObjectFileMachO::GetByteOrder() const { 1131 return m_data.GetByteOrder(); 1132 } 1133 1134 bool ObjectFileMachO::IsExecutable() const { 1135 return m_header.filetype == MH_EXECUTE; 1136 } 1137 1138 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1139 return m_data.GetAddressByteSize(); 1140 } 1141 1142 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1143 Symtab *symtab = GetSymtab(); 1144 if (symtab) { 1145 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1146 if (symbol) { 1147 if (symbol->ValueIsAddress()) { 1148 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1149 if (section_sp) { 1150 const lldb::SectionType section_type = section_sp->GetType(); 1151 switch (section_type) { 1152 case eSectionTypeInvalid: 1153 return eAddressClassUnknown; 1154 1155 case eSectionTypeCode: 1156 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1157 // For ARM we have a bit in the n_desc field of the symbol 1158 // that tells us ARM/Thumb which is bit 0x0008. 1159 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1160 return eAddressClassCodeAlternateISA; 1161 } 1162 return eAddressClassCode; 1163 1164 case eSectionTypeContainer: 1165 return eAddressClassUnknown; 1166 1167 case eSectionTypeData: 1168 case eSectionTypeDataCString: 1169 case eSectionTypeDataCStringPointers: 1170 case eSectionTypeDataSymbolAddress: 1171 case eSectionTypeData4: 1172 case eSectionTypeData8: 1173 case eSectionTypeData16: 1174 case eSectionTypeDataPointers: 1175 case eSectionTypeZeroFill: 1176 case eSectionTypeDataObjCMessageRefs: 1177 case eSectionTypeDataObjCCFStrings: 1178 case eSectionTypeGoSymtab: 1179 return eAddressClassData; 1180 1181 case eSectionTypeDebug: 1182 case eSectionTypeDWARFDebugAbbrev: 1183 case eSectionTypeDWARFDebugAddr: 1184 case eSectionTypeDWARFDebugAranges: 1185 case eSectionTypeDWARFDebugFrame: 1186 case eSectionTypeDWARFDebugInfo: 1187 case eSectionTypeDWARFDebugLine: 1188 case eSectionTypeDWARFDebugLoc: 1189 case eSectionTypeDWARFDebugMacInfo: 1190 case eSectionTypeDWARFDebugMacro: 1191 case eSectionTypeDWARFDebugPubNames: 1192 case eSectionTypeDWARFDebugPubTypes: 1193 case eSectionTypeDWARFDebugRanges: 1194 case eSectionTypeDWARFDebugStr: 1195 case eSectionTypeDWARFDebugStrOffsets: 1196 case eSectionTypeDWARFAppleNames: 1197 case eSectionTypeDWARFAppleTypes: 1198 case eSectionTypeDWARFAppleNamespaces: 1199 case eSectionTypeDWARFAppleObjC: 1200 return eAddressClassDebug; 1201 1202 case eSectionTypeEHFrame: 1203 case eSectionTypeARMexidx: 1204 case eSectionTypeARMextab: 1205 case eSectionTypeCompactUnwind: 1206 return eAddressClassRuntime; 1207 1208 case eSectionTypeAbsoluteAddress: 1209 case eSectionTypeELFSymbolTable: 1210 case eSectionTypeELFDynamicSymbols: 1211 case eSectionTypeELFRelocationEntries: 1212 case eSectionTypeELFDynamicLinkInfo: 1213 case eSectionTypeOther: 1214 return eAddressClassUnknown; 1215 } 1216 } 1217 } 1218 1219 const SymbolType symbol_type = symbol->GetType(); 1220 switch (symbol_type) { 1221 case eSymbolTypeAny: 1222 return eAddressClassUnknown; 1223 case eSymbolTypeAbsolute: 1224 return eAddressClassUnknown; 1225 1226 case eSymbolTypeCode: 1227 case eSymbolTypeTrampoline: 1228 case eSymbolTypeResolver: 1229 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1230 // For ARM we have a bit in the n_desc field of the symbol 1231 // that tells us ARM/Thumb which is bit 0x0008. 1232 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1233 return eAddressClassCodeAlternateISA; 1234 } 1235 return eAddressClassCode; 1236 1237 case eSymbolTypeData: 1238 return eAddressClassData; 1239 case eSymbolTypeRuntime: 1240 return eAddressClassRuntime; 1241 case eSymbolTypeException: 1242 return eAddressClassRuntime; 1243 case eSymbolTypeSourceFile: 1244 return eAddressClassDebug; 1245 case eSymbolTypeHeaderFile: 1246 return eAddressClassDebug; 1247 case eSymbolTypeObjectFile: 1248 return eAddressClassDebug; 1249 case eSymbolTypeCommonBlock: 1250 return eAddressClassDebug; 1251 case eSymbolTypeBlock: 1252 return eAddressClassDebug; 1253 case eSymbolTypeLocal: 1254 return eAddressClassData; 1255 case eSymbolTypeParam: 1256 return eAddressClassData; 1257 case eSymbolTypeVariable: 1258 return eAddressClassData; 1259 case eSymbolTypeVariableType: 1260 return eAddressClassDebug; 1261 case eSymbolTypeLineEntry: 1262 return eAddressClassDebug; 1263 case eSymbolTypeLineHeader: 1264 return eAddressClassDebug; 1265 case eSymbolTypeScopeBegin: 1266 return eAddressClassDebug; 1267 case eSymbolTypeScopeEnd: 1268 return eAddressClassDebug; 1269 case eSymbolTypeAdditional: 1270 return eAddressClassUnknown; 1271 case eSymbolTypeCompiler: 1272 return eAddressClassDebug; 1273 case eSymbolTypeInstrumentation: 1274 return eAddressClassDebug; 1275 case eSymbolTypeUndefined: 1276 return eAddressClassUnknown; 1277 case eSymbolTypeObjCClass: 1278 return eAddressClassRuntime; 1279 case eSymbolTypeObjCMetaClass: 1280 return eAddressClassRuntime; 1281 case eSymbolTypeObjCIVar: 1282 return eAddressClassRuntime; 1283 case eSymbolTypeReExported: 1284 return eAddressClassRuntime; 1285 } 1286 } 1287 } 1288 return eAddressClassUnknown; 1289 } 1290 1291 Symtab *ObjectFileMachO::GetSymtab() { 1292 ModuleSP module_sp(GetModule()); 1293 if (module_sp) { 1294 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1295 if (m_symtab_ap.get() == NULL) { 1296 m_symtab_ap.reset(new Symtab(this)); 1297 std::lock_guard<std::recursive_mutex> symtab_guard( 1298 m_symtab_ap->GetMutex()); 1299 ParseSymtab(); 1300 m_symtab_ap->Finalize(); 1301 } 1302 } 1303 return m_symtab_ap.get(); 1304 } 1305 1306 bool ObjectFileMachO::IsStripped() { 1307 if (m_dysymtab.cmd == 0) { 1308 ModuleSP module_sp(GetModule()); 1309 if (module_sp) { 1310 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1311 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1312 const lldb::offset_t load_cmd_offset = offset; 1313 1314 load_command lc; 1315 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 1316 break; 1317 if (lc.cmd == LC_DYSYMTAB) { 1318 m_dysymtab.cmd = lc.cmd; 1319 m_dysymtab.cmdsize = lc.cmdsize; 1320 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1321 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1322 NULL) { 1323 // Clear m_dysymtab if we were unable to read all items from the 1324 // load command 1325 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1326 } 1327 } 1328 offset = load_cmd_offset + lc.cmdsize; 1329 } 1330 } 1331 } 1332 if (m_dysymtab.cmd) 1333 return m_dysymtab.nlocalsym <= 1; 1334 return false; 1335 } 1336 1337 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1338 if (!m_sections_ap.get()) { 1339 m_sections_ap.reset(new SectionList()); 1340 1341 const bool is_dsym = (m_header.filetype == MH_DSYM); 1342 lldb::user_id_t segID = 0; 1343 lldb::user_id_t sectID = 0; 1344 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1345 uint32_t i; 1346 const bool is_core = GetType() == eTypeCoreFile; 1347 // bool dump_sections = false; 1348 ModuleSP module_sp(GetModule()); 1349 // First look up any LC_ENCRYPTION_INFO load commands 1350 typedef RangeArray<uint32_t, uint32_t, 8> EncryptedFileRanges; 1351 EncryptedFileRanges encrypted_file_ranges; 1352 encryption_info_command encryption_cmd; 1353 for (i = 0; i < m_header.ncmds; ++i) { 1354 const lldb::offset_t load_cmd_offset = offset; 1355 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1356 break; 1357 1358 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for 1359 // the 3 fields we care about, so treat them the same. 1360 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1361 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1362 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1363 if (encryption_cmd.cryptid != 0) { 1364 EncryptedFileRanges::Entry entry; 1365 entry.SetRangeBase(encryption_cmd.cryptoff); 1366 entry.SetByteSize(encryption_cmd.cryptsize); 1367 encrypted_file_ranges.Append(entry); 1368 } 1369 } 1370 } 1371 offset = load_cmd_offset + encryption_cmd.cmdsize; 1372 } 1373 1374 bool section_file_addresses_changed = false; 1375 1376 offset = MachHeaderSizeFromMagic(m_header.magic); 1377 1378 struct segment_command_64 load_cmd; 1379 for (i = 0; i < m_header.ncmds; ++i) { 1380 const lldb::offset_t load_cmd_offset = offset; 1381 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1382 break; 1383 1384 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) { 1385 if (m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) { 1386 bool add_section = true; 1387 bool add_to_unified = true; 1388 ConstString const_segname(load_cmd.segname, 1389 std::min<size_t>(strlen(load_cmd.segname), 1390 sizeof(load_cmd.segname))); 1391 1392 SectionSP unified_section_sp( 1393 unified_section_list.FindSectionByName(const_segname)); 1394 if (is_dsym && unified_section_sp) { 1395 if (const_segname == GetSegmentNameLINKEDIT()) { 1396 // We need to keep the __LINKEDIT segment private to this object 1397 // file only 1398 add_to_unified = false; 1399 } else { 1400 // This is the dSYM file and this section has already been created 1401 // by 1402 // the object file, no need to create it. 1403 add_section = false; 1404 } 1405 } 1406 load_cmd.vmaddr = m_data.GetAddress(&offset); 1407 load_cmd.vmsize = m_data.GetAddress(&offset); 1408 load_cmd.fileoff = m_data.GetAddress(&offset); 1409 load_cmd.filesize = m_data.GetAddress(&offset); 1410 if (m_length != 0 && load_cmd.filesize != 0) { 1411 if (load_cmd.fileoff > m_length) { 1412 // We have a load command that says it extends past the end of the 1413 // file. This is likely 1414 // a corrupt file. We don't have any way to return an error 1415 // condition here (this method 1416 // was likely invoked from something like 1417 // ObjectFile::GetSectionList()) -- all we can do 1418 // is null out the SectionList vector and if a process has been 1419 // set up, dump a message 1420 // to stdout. The most common case here is core file debugging 1421 // with a truncated file. 1422 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 1423 ? "LC_SEGMENT_64" 1424 : "LC_SEGMENT"; 1425 module_sp->ReportWarning( 1426 "load command %u %s has a fileoff (0x%" PRIx64 1427 ") that extends beyond the end of the file (0x%" PRIx64 1428 "), ignoring this section", 1429 i, lc_segment_name, load_cmd.fileoff, m_length); 1430 1431 load_cmd.fileoff = 0; 1432 load_cmd.filesize = 0; 1433 } 1434 1435 if (load_cmd.fileoff + load_cmd.filesize > m_length) { 1436 // We have a load command that says it extends past the end of the 1437 // file. This is likely 1438 // a corrupt file. We don't have any way to return an error 1439 // condition here (this method 1440 // was likely invoked from something like 1441 // ObjectFile::GetSectionList()) -- all we can do 1442 // is null out the SectionList vector and if a process has been 1443 // set up, dump a message 1444 // to stdout. The most common case here is core file debugging 1445 // with a truncated file. 1446 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 1447 ? "LC_SEGMENT_64" 1448 : "LC_SEGMENT"; 1449 GetModule()->ReportWarning( 1450 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1451 ") that extends beyond the end of the file (0x%" PRIx64 1452 "), the segment will be truncated to match", 1453 i, lc_segment_name, load_cmd.fileoff + load_cmd.filesize, 1454 m_length); 1455 1456 // Tuncase the length 1457 load_cmd.filesize = m_length - load_cmd.fileoff; 1458 } 1459 } 1460 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) { 1461 uint32_t segment_permissions = 0; 1462 if (load_cmd.initprot & VM_PROT_READ) 1463 segment_permissions |= ePermissionsReadable; 1464 if (load_cmd.initprot & VM_PROT_WRITE) 1465 segment_permissions |= ePermissionsWritable; 1466 if (load_cmd.initprot & VM_PROT_EXECUTE) 1467 segment_permissions |= ePermissionsExecutable; 1468 1469 const bool segment_is_encrypted = 1470 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1471 1472 // Keep a list of mach segments around in case we need to 1473 // get at data that isn't stored in the abstracted Sections. 1474 m_mach_segments.push_back(load_cmd); 1475 1476 // Use a segment ID of the segment index shifted left by 8 so they 1477 // never conflict with any of the sections. 1478 SectionSP segment_sp; 1479 if (add_section && (const_segname || is_core)) { 1480 segment_sp.reset(new Section( 1481 module_sp, // Module to which this section belongs 1482 this, // Object file to which this sections belongs 1483 ++segID << 8, // Section ID is the 1 based segment index 1484 // shifted right by 8 bits as not to collide 1485 // with any of the 256 section IDs that are 1486 // possible 1487 const_segname, // Name of this section 1488 eSectionTypeContainer, // This section is a container of other 1489 // sections. 1490 load_cmd.vmaddr, // File VM address == addresses as they are 1491 // found in the object file 1492 load_cmd.vmsize, // VM size in bytes of this section 1493 load_cmd.fileoff, // Offset to the data for this section in 1494 // the file 1495 load_cmd.filesize, // Size in bytes of this section as found 1496 // in the file 1497 0, // Segments have no alignment information 1498 load_cmd.flags)); // Flags for this section 1499 1500 segment_sp->SetIsEncrypted(segment_is_encrypted); 1501 m_sections_ap->AddSection(segment_sp); 1502 segment_sp->SetPermissions(segment_permissions); 1503 if (add_to_unified) 1504 unified_section_list.AddSection(segment_sp); 1505 } else if (unified_section_sp) { 1506 if (is_dsym && 1507 unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1508 // Check to see if the module was read from memory? 1509 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) { 1510 // We have a module that is in memory and needs to have its 1511 // file address adjusted. We need to do this because when we 1512 // load a file from memory, its addresses will be slid 1513 // already, 1514 // yet the addresses in the new symbol file will still be 1515 // unslid. 1516 // Since everything is stored as section offset, this 1517 // shouldn't 1518 // cause any problems. 1519 1520 // Make sure we've parsed the symbol table from the 1521 // ObjectFile before we go around changing its Sections. 1522 module_sp->GetObjectFile()->GetSymtab(); 1523 // eh_frame would present the same problems but we parse that 1524 // on 1525 // a per-function basis as-needed so it's more difficult to 1526 // remove its use of the Sections. Realistically, the 1527 // environments 1528 // where this code path will be taken will not have eh_frame 1529 // sections. 1530 1531 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1532 1533 // Notify the module that the section addresses have been 1534 // changed once 1535 // we're done so any file-address caches can be updated. 1536 section_file_addresses_changed = true; 1537 } 1538 } 1539 m_sections_ap->AddSection(unified_section_sp); 1540 } 1541 1542 struct section_64 sect64; 1543 ::memset(§64, 0, sizeof(sect64)); 1544 // Push a section into our mach sections for the section at 1545 // index zero (NO_SECT) if we don't have any mach sections yet... 1546 if (m_mach_sections.empty()) 1547 m_mach_sections.push_back(sect64); 1548 uint32_t segment_sect_idx; 1549 const lldb::user_id_t first_segment_sectID = sectID + 1; 1550 1551 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1552 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1553 ++segment_sect_idx) { 1554 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1555 sizeof(sect64.sectname)) == NULL) 1556 break; 1557 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1558 sizeof(sect64.segname)) == NULL) 1559 break; 1560 sect64.addr = m_data.GetAddress(&offset); 1561 sect64.size = m_data.GetAddress(&offset); 1562 1563 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1564 break; 1565 1566 // Keep a list of mach sections around in case we need to 1567 // get at data that isn't stored in the abstracted Sections. 1568 m_mach_sections.push_back(sect64); 1569 1570 if (add_section) { 1571 ConstString section_name( 1572 sect64.sectname, std::min<size_t>(strlen(sect64.sectname), 1573 sizeof(sect64.sectname))); 1574 if (!const_segname) { 1575 // We have a segment with no name so we need to conjure up 1576 // segments that correspond to the section's segname if there 1577 // isn't already such a section. If there is such a section, 1578 // we resize the section so that it spans all sections. 1579 // We also mark these sections as fake so address matches 1580 // don't 1581 // hit if they land in the gaps between the child sections. 1582 const_segname.SetTrimmedCStringWithLength( 1583 sect64.segname, sizeof(sect64.segname)); 1584 segment_sp = 1585 unified_section_list.FindSectionByName(const_segname); 1586 if (segment_sp.get()) { 1587 Section *segment = segment_sp.get(); 1588 // Grow the section size as needed. 1589 const lldb::addr_t sect64_min_addr = sect64.addr; 1590 const lldb::addr_t sect64_max_addr = 1591 sect64_min_addr + sect64.size; 1592 const lldb::addr_t curr_seg_byte_size = 1593 segment->GetByteSize(); 1594 const lldb::addr_t curr_seg_min_addr = 1595 segment->GetFileAddress(); 1596 const lldb::addr_t curr_seg_max_addr = 1597 curr_seg_min_addr + curr_seg_byte_size; 1598 if (sect64_min_addr >= curr_seg_min_addr) { 1599 const lldb::addr_t new_seg_byte_size = 1600 sect64_max_addr - curr_seg_min_addr; 1601 // Only grow the section size if needed 1602 if (new_seg_byte_size > curr_seg_byte_size) 1603 segment->SetByteSize(new_seg_byte_size); 1604 } else { 1605 // We need to change the base address of the segment and 1606 // adjust the child section offsets for all existing 1607 // children. 1608 const lldb::addr_t slide_amount = 1609 sect64_min_addr - curr_seg_min_addr; 1610 segment->Slide(slide_amount, false); 1611 segment->GetChildren().Slide(-slide_amount, false); 1612 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1613 } 1614 1615 // Grow the section size as needed. 1616 if (sect64.offset) { 1617 const lldb::addr_t segment_min_file_offset = 1618 segment->GetFileOffset(); 1619 const lldb::addr_t segment_max_file_offset = 1620 segment_min_file_offset + segment->GetFileSize(); 1621 1622 const lldb::addr_t section_min_file_offset = 1623 sect64.offset; 1624 const lldb::addr_t section_max_file_offset = 1625 section_min_file_offset + sect64.size; 1626 const lldb::addr_t new_file_offset = std::min( 1627 section_min_file_offset, segment_min_file_offset); 1628 const lldb::addr_t new_file_size = 1629 std::max(section_max_file_offset, 1630 segment_max_file_offset) - 1631 new_file_offset; 1632 segment->SetFileOffset(new_file_offset); 1633 segment->SetFileSize(new_file_size); 1634 } 1635 } else { 1636 // Create a fake section for the section's named segment 1637 segment_sp.reset(new Section( 1638 segment_sp, // Parent section 1639 module_sp, // Module to which this section belongs 1640 this, // Object file to which this section belongs 1641 ++segID << 8, // Section ID is the 1 based segment index 1642 // shifted right by 8 bits as not to 1643 // collide with any of the 256 section IDs 1644 // that are possible 1645 const_segname, // Name of this section 1646 eSectionTypeContainer, // This section is a container of 1647 // other sections. 1648 sect64.addr, // File VM address == addresses as they are 1649 // found in the object file 1650 sect64.size, // VM size in bytes of this section 1651 sect64.offset, // Offset to the data for this section in 1652 // the file 1653 sect64.offset ? sect64.size : 0, // Size in bytes of 1654 // this section as 1655 // found in the file 1656 sect64.align, 1657 load_cmd.flags)); // Flags for this section 1658 segment_sp->SetIsFake(true); 1659 segment_sp->SetPermissions(segment_permissions); 1660 m_sections_ap->AddSection(segment_sp); 1661 if (add_to_unified) 1662 unified_section_list.AddSection(segment_sp); 1663 segment_sp->SetIsEncrypted(segment_is_encrypted); 1664 } 1665 } 1666 assert(segment_sp.get()); 1667 1668 lldb::SectionType sect_type = eSectionTypeOther; 1669 1670 if (sect64.flags & 1671 (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1672 sect_type = eSectionTypeCode; 1673 else { 1674 uint32_t mach_sect_type = sect64.flags & SECTION_TYPE; 1675 static ConstString g_sect_name_objc_data("__objc_data"); 1676 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1677 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1678 static ConstString g_sect_name_objc_classrefs( 1679 "__objc_classrefs"); 1680 static ConstString g_sect_name_objc_superrefs( 1681 "__objc_superrefs"); 1682 static ConstString g_sect_name_objc_const("__objc_const"); 1683 static ConstString g_sect_name_objc_classlist( 1684 "__objc_classlist"); 1685 static ConstString g_sect_name_cfstring("__cfstring"); 1686 1687 static ConstString g_sect_name_dwarf_debug_abbrev( 1688 "__debug_abbrev"); 1689 static ConstString g_sect_name_dwarf_debug_aranges( 1690 "__debug_aranges"); 1691 static ConstString g_sect_name_dwarf_debug_frame( 1692 "__debug_frame"); 1693 static ConstString g_sect_name_dwarf_debug_info( 1694 "__debug_info"); 1695 static ConstString g_sect_name_dwarf_debug_line( 1696 "__debug_line"); 1697 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1698 static ConstString g_sect_name_dwarf_debug_macinfo( 1699 "__debug_macinfo"); 1700 static ConstString g_sect_name_dwarf_debug_pubnames( 1701 "__debug_pubnames"); 1702 static ConstString g_sect_name_dwarf_debug_pubtypes( 1703 "__debug_pubtypes"); 1704 static ConstString g_sect_name_dwarf_debug_ranges( 1705 "__debug_ranges"); 1706 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1707 static ConstString g_sect_name_dwarf_apple_names( 1708 "__apple_names"); 1709 static ConstString g_sect_name_dwarf_apple_types( 1710 "__apple_types"); 1711 static ConstString g_sect_name_dwarf_apple_namespaces( 1712 "__apple_namespac"); 1713 static ConstString g_sect_name_dwarf_apple_objc( 1714 "__apple_objc"); 1715 static ConstString g_sect_name_eh_frame("__eh_frame"); 1716 static ConstString g_sect_name_compact_unwind( 1717 "__unwind_info"); 1718 static ConstString g_sect_name_text("__text"); 1719 static ConstString g_sect_name_data("__data"); 1720 static ConstString g_sect_name_go_symtab("__gosymtab"); 1721 1722 if (section_name == g_sect_name_dwarf_debug_abbrev) 1723 sect_type = eSectionTypeDWARFDebugAbbrev; 1724 else if (section_name == g_sect_name_dwarf_debug_aranges) 1725 sect_type = eSectionTypeDWARFDebugAranges; 1726 else if (section_name == g_sect_name_dwarf_debug_frame) 1727 sect_type = eSectionTypeDWARFDebugFrame; 1728 else if (section_name == g_sect_name_dwarf_debug_info) 1729 sect_type = eSectionTypeDWARFDebugInfo; 1730 else if (section_name == g_sect_name_dwarf_debug_line) 1731 sect_type = eSectionTypeDWARFDebugLine; 1732 else if (section_name == g_sect_name_dwarf_debug_loc) 1733 sect_type = eSectionTypeDWARFDebugLoc; 1734 else if (section_name == g_sect_name_dwarf_debug_macinfo) 1735 sect_type = eSectionTypeDWARFDebugMacInfo; 1736 else if (section_name == g_sect_name_dwarf_debug_pubnames) 1737 sect_type = eSectionTypeDWARFDebugPubNames; 1738 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 1739 sect_type = eSectionTypeDWARFDebugPubTypes; 1740 else if (section_name == g_sect_name_dwarf_debug_ranges) 1741 sect_type = eSectionTypeDWARFDebugRanges; 1742 else if (section_name == g_sect_name_dwarf_debug_str) 1743 sect_type = eSectionTypeDWARFDebugStr; 1744 else if (section_name == g_sect_name_dwarf_apple_names) 1745 sect_type = eSectionTypeDWARFAppleNames; 1746 else if (section_name == g_sect_name_dwarf_apple_types) 1747 sect_type = eSectionTypeDWARFAppleTypes; 1748 else if (section_name == g_sect_name_dwarf_apple_namespaces) 1749 sect_type = eSectionTypeDWARFAppleNamespaces; 1750 else if (section_name == g_sect_name_dwarf_apple_objc) 1751 sect_type = eSectionTypeDWARFAppleObjC; 1752 else if (section_name == g_sect_name_objc_selrefs) 1753 sect_type = eSectionTypeDataCStringPointers; 1754 else if (section_name == g_sect_name_objc_msgrefs) 1755 sect_type = eSectionTypeDataObjCMessageRefs; 1756 else if (section_name == g_sect_name_eh_frame) 1757 sect_type = eSectionTypeEHFrame; 1758 else if (section_name == g_sect_name_compact_unwind) 1759 sect_type = eSectionTypeCompactUnwind; 1760 else if (section_name == g_sect_name_cfstring) 1761 sect_type = eSectionTypeDataObjCCFStrings; 1762 else if (section_name == g_sect_name_go_symtab) 1763 sect_type = eSectionTypeGoSymtab; 1764 else if (section_name == g_sect_name_objc_data || 1765 section_name == g_sect_name_objc_classrefs || 1766 section_name == g_sect_name_objc_superrefs || 1767 section_name == g_sect_name_objc_const || 1768 section_name == g_sect_name_objc_classlist) { 1769 sect_type = eSectionTypeDataPointers; 1770 } 1771 1772 if (sect_type == eSectionTypeOther) { 1773 switch (mach_sect_type) { 1774 // TODO: categorize sections by other flags for regular 1775 // sections 1776 case S_REGULAR: 1777 if (section_name == g_sect_name_text) 1778 sect_type = eSectionTypeCode; 1779 else if (section_name == g_sect_name_data) 1780 sect_type = eSectionTypeData; 1781 else 1782 sect_type = eSectionTypeOther; 1783 break; 1784 case S_ZEROFILL: 1785 sect_type = eSectionTypeZeroFill; 1786 break; 1787 case S_CSTRING_LITERALS: 1788 sect_type = eSectionTypeDataCString; 1789 break; // section with only literal C strings 1790 case S_4BYTE_LITERALS: 1791 sect_type = eSectionTypeData4; 1792 break; // section with only 4 byte literals 1793 case S_8BYTE_LITERALS: 1794 sect_type = eSectionTypeData8; 1795 break; // section with only 8 byte literals 1796 case S_LITERAL_POINTERS: 1797 sect_type = eSectionTypeDataPointers; 1798 break; // section with only pointers to literals 1799 case S_NON_LAZY_SYMBOL_POINTERS: 1800 sect_type = eSectionTypeDataPointers; 1801 break; // section with only non-lazy symbol pointers 1802 case S_LAZY_SYMBOL_POINTERS: 1803 sect_type = eSectionTypeDataPointers; 1804 break; // section with only lazy symbol pointers 1805 case S_SYMBOL_STUBS: 1806 sect_type = eSectionTypeCode; 1807 break; // section with only symbol stubs, byte size of 1808 // stub in the reserved2 field 1809 case S_MOD_INIT_FUNC_POINTERS: 1810 sect_type = eSectionTypeDataPointers; 1811 break; // section with only function pointers for 1812 // initialization 1813 case S_MOD_TERM_FUNC_POINTERS: 1814 sect_type = eSectionTypeDataPointers; 1815 break; // section with only function pointers for 1816 // termination 1817 case S_COALESCED: 1818 sect_type = eSectionTypeOther; 1819 break; 1820 case S_GB_ZEROFILL: 1821 sect_type = eSectionTypeZeroFill; 1822 break; 1823 case S_INTERPOSING: 1824 sect_type = eSectionTypeCode; 1825 break; // section with only pairs of function pointers for 1826 // interposing 1827 case S_16BYTE_LITERALS: 1828 sect_type = eSectionTypeData16; 1829 break; // section with only 16 byte literals 1830 case S_DTRACE_DOF: 1831 sect_type = eSectionTypeDebug; 1832 break; 1833 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1834 sect_type = eSectionTypeDataPointers; 1835 break; 1836 default: 1837 break; 1838 } 1839 } 1840 } 1841 1842 SectionSP section_sp(new Section( 1843 segment_sp, module_sp, this, ++sectID, section_name, 1844 sect_type, sect64.addr - segment_sp->GetFileAddress(), 1845 sect64.size, sect64.offset, 1846 sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1847 sect64.flags)); 1848 // Set the section to be encrypted to match the segment 1849 1850 bool section_is_encrypted = false; 1851 if (!segment_is_encrypted && load_cmd.filesize != 0) 1852 section_is_encrypted = 1853 encrypted_file_ranges.FindEntryThatContains( 1854 sect64.offset) != NULL; 1855 1856 section_sp->SetIsEncrypted(segment_is_encrypted || 1857 section_is_encrypted); 1858 section_sp->SetPermissions(segment_permissions); 1859 segment_sp->GetChildren().AddSection(section_sp); 1860 1861 if (segment_sp->IsFake()) { 1862 segment_sp.reset(); 1863 const_segname.Clear(); 1864 } 1865 } 1866 } 1867 if (segment_sp && is_dsym) { 1868 if (first_segment_sectID <= sectID) { 1869 lldb::user_id_t sect_uid; 1870 for (sect_uid = first_segment_sectID; sect_uid <= sectID; 1871 ++sect_uid) { 1872 SectionSP curr_section_sp( 1873 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1874 SectionSP next_section_sp; 1875 if (sect_uid + 1 <= sectID) 1876 next_section_sp = 1877 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1878 1879 if (curr_section_sp.get()) { 1880 if (curr_section_sp->GetByteSize() == 0) { 1881 if (next_section_sp.get() != NULL) 1882 curr_section_sp->SetByteSize( 1883 next_section_sp->GetFileAddress() - 1884 curr_section_sp->GetFileAddress()); 1885 else 1886 curr_section_sp->SetByteSize(load_cmd.vmsize); 1887 } 1888 } 1889 } 1890 } 1891 } 1892 } 1893 } 1894 } else if (load_cmd.cmd == LC_DYSYMTAB) { 1895 m_dysymtab.cmd = load_cmd.cmd; 1896 m_dysymtab.cmdsize = load_cmd.cmdsize; 1897 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1898 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1899 } 1900 1901 offset = load_cmd_offset + load_cmd.cmdsize; 1902 } 1903 1904 if (section_file_addresses_changed && module_sp.get()) { 1905 module_sp->SectionFileAddressesChanged(); 1906 } 1907 } 1908 } 1909 1910 class MachSymtabSectionInfo { 1911 public: 1912 MachSymtabSectionInfo(SectionList *section_list) 1913 : m_section_list(section_list), m_section_infos() { 1914 // Get the number of sections down to a depth of 1 to include 1915 // all segments and their sections, but no other sections that 1916 // may be added for debug map or 1917 m_section_infos.resize(section_list->GetNumSections(1)); 1918 } 1919 1920 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1921 if (n_sect == 0) 1922 return SectionSP(); 1923 if (n_sect < m_section_infos.size()) { 1924 if (!m_section_infos[n_sect].section_sp) { 1925 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1926 m_section_infos[n_sect].section_sp = section_sp; 1927 if (section_sp) { 1928 m_section_infos[n_sect].vm_range.SetBaseAddress( 1929 section_sp->GetFileAddress()); 1930 m_section_infos[n_sect].vm_range.SetByteSize( 1931 section_sp->GetByteSize()); 1932 } else { 1933 Host::SystemLog(Host::eSystemLogError, 1934 "error: unable to find section for section %u\n", 1935 n_sect); 1936 } 1937 } 1938 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1939 // Symbol is in section. 1940 return m_section_infos[n_sect].section_sp; 1941 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1942 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1943 file_addr) { 1944 // Symbol is in section with zero size, but has the same start 1945 // address as the section. This can happen with linker symbols 1946 // (symbols that start with the letter 'l' or 'L'. 1947 return m_section_infos[n_sect].section_sp; 1948 } 1949 } 1950 return m_section_list->FindSectionContainingFileAddress(file_addr); 1951 } 1952 1953 protected: 1954 struct SectionInfo { 1955 SectionInfo() : vm_range(), section_sp() {} 1956 1957 VMRange vm_range; 1958 SectionSP section_sp; 1959 }; 1960 SectionList *m_section_list; 1961 std::vector<SectionInfo> m_section_infos; 1962 }; 1963 1964 struct TrieEntry { 1965 TrieEntry() 1966 : name(), address(LLDB_INVALID_ADDRESS), flags(0), other(0), 1967 import_name() {} 1968 1969 void Clear() { 1970 name.Clear(); 1971 address = LLDB_INVALID_ADDRESS; 1972 flags = 0; 1973 other = 0; 1974 import_name.Clear(); 1975 } 1976 1977 void Dump() const { 1978 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1979 static_cast<unsigned long long>(address), 1980 static_cast<unsigned long long>(flags), 1981 static_cast<unsigned long long>(other), name.GetCString()); 1982 if (import_name) 1983 printf(" -> \"%s\"\n", import_name.GetCString()); 1984 else 1985 printf("\n"); 1986 } 1987 ConstString name; 1988 uint64_t address; 1989 uint64_t flags; 1990 uint64_t other; 1991 ConstString import_name; 1992 }; 1993 1994 struct TrieEntryWithOffset { 1995 lldb::offset_t nodeOffset; 1996 TrieEntry entry; 1997 1998 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 1999 2000 void Dump(uint32_t idx) const { 2001 printf("[%3u] 0x%16.16llx: ", idx, 2002 static_cast<unsigned long long>(nodeOffset)); 2003 entry.Dump(); 2004 } 2005 2006 bool operator<(const TrieEntryWithOffset &other) const { 2007 return (nodeOffset < other.nodeOffset); 2008 } 2009 }; 2010 2011 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 2012 const bool is_arm, 2013 std::vector<llvm::StringRef> &nameSlices, 2014 std::set<lldb::addr_t> &resolver_addresses, 2015 std::vector<TrieEntryWithOffset> &output) { 2016 if (!data.ValidOffset(offset)) 2017 return true; 2018 2019 const uint64_t terminalSize = data.GetULEB128(&offset); 2020 lldb::offset_t children_offset = offset + terminalSize; 2021 if (terminalSize != 0) { 2022 TrieEntryWithOffset e(offset); 2023 e.entry.flags = data.GetULEB128(&offset); 2024 const char *import_name = NULL; 2025 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2026 e.entry.address = 0; 2027 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2028 import_name = data.GetCStr(&offset); 2029 } else { 2030 e.entry.address = data.GetULEB128(&offset); 2031 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2032 e.entry.other = data.GetULEB128(&offset); 2033 uint64_t resolver_addr = e.entry.other; 2034 if (is_arm) 2035 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2036 resolver_addresses.insert(resolver_addr); 2037 } else 2038 e.entry.other = 0; 2039 } 2040 // Only add symbols that are reexport symbols with a valid import name 2041 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && 2042 import_name[0]) { 2043 std::string name; 2044 if (!nameSlices.empty()) { 2045 for (auto name_slice : nameSlices) 2046 name.append(name_slice.data(), name_slice.size()); 2047 } 2048 if (name.size() > 1) { 2049 // Skip the leading '_' 2050 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2051 } 2052 if (import_name) { 2053 // Skip the leading '_' 2054 e.entry.import_name.SetCString(import_name + 1); 2055 } 2056 output.push_back(e); 2057 } 2058 } 2059 2060 const uint8_t childrenCount = data.GetU8(&children_offset); 2061 for (uint8_t i = 0; i < childrenCount; ++i) { 2062 const char *cstr = data.GetCStr(&children_offset); 2063 if (cstr) 2064 nameSlices.push_back(llvm::StringRef(cstr)); 2065 else 2066 return false; // Corrupt data 2067 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2068 if (childNodeOffset) { 2069 if (!ParseTrieEntries(data, childNodeOffset, is_arm, nameSlices, 2070 resolver_addresses, output)) { 2071 return false; 2072 } 2073 } 2074 nameSlices.pop_back(); 2075 } 2076 return true; 2077 } 2078 2079 // Read the UUID out of a dyld_shared_cache file on-disk. 2080 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2081 const ByteOrder byte_order, 2082 const uint32_t addr_byte_size) { 2083 UUID dsc_uuid; 2084 DataBufferSP dsc_data_sp = dyld_shared_cache.MemoryMapFileContentsIfLocal( 2085 0, sizeof(struct lldb_copy_dyld_cache_header_v1)); 2086 if (dsc_data_sp) { 2087 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2088 2089 char version_str[7]; 2090 lldb::offset_t offset = 0; 2091 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2092 version_str[6] = '\0'; 2093 if (strcmp(version_str, "dyld_v") == 0) { 2094 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2095 uint8_t uuid_bytes[sizeof(uuid_t)]; 2096 memcpy(uuid_bytes, dsc_header_data.GetData(&offset, sizeof(uuid_t)), 2097 sizeof(uuid_t)); 2098 dsc_uuid.SetBytes(uuid_bytes); 2099 } 2100 } 2101 return dsc_uuid; 2102 } 2103 2104 size_t ObjectFileMachO::ParseSymtab() { 2105 Timer scoped_timer(LLVM_PRETTY_FUNCTION, 2106 "ObjectFileMachO::ParseSymtab () module = %s", 2107 m_file.GetFilename().AsCString("")); 2108 ModuleSP module_sp(GetModule()); 2109 if (!module_sp) 2110 return 0; 2111 2112 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2113 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2114 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2115 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2116 FunctionStarts function_starts; 2117 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2118 uint32_t i; 2119 FileSpecList dylib_files; 2120 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2121 static const llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2122 static const llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2123 static const llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2124 2125 for (i = 0; i < m_header.ncmds; ++i) { 2126 const lldb::offset_t cmd_offset = offset; 2127 // Read in the load command and load command size 2128 struct load_command lc; 2129 if (m_data.GetU32(&offset, &lc, 2) == NULL) 2130 break; 2131 // Watch for the symbol table load command 2132 switch (lc.cmd) { 2133 case LC_SYMTAB: 2134 symtab_load_command.cmd = lc.cmd; 2135 symtab_load_command.cmdsize = lc.cmdsize; 2136 // Read in the rest of the symtab load command 2137 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2138 0) // fill in symoff, nsyms, stroff, strsize fields 2139 return 0; 2140 if (symtab_load_command.symoff == 0) { 2141 if (log) 2142 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2143 return 0; 2144 } 2145 2146 if (symtab_load_command.stroff == 0) { 2147 if (log) 2148 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2149 return 0; 2150 } 2151 2152 if (symtab_load_command.nsyms == 0) { 2153 if (log) 2154 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2155 return 0; 2156 } 2157 2158 if (symtab_load_command.strsize == 0) { 2159 if (log) 2160 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2161 return 0; 2162 } 2163 break; 2164 2165 case LC_DYLD_INFO: 2166 case LC_DYLD_INFO_ONLY: 2167 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2168 dyld_info.cmd = lc.cmd; 2169 dyld_info.cmdsize = lc.cmdsize; 2170 } else { 2171 memset(&dyld_info, 0, sizeof(dyld_info)); 2172 } 2173 break; 2174 2175 case LC_LOAD_DYLIB: 2176 case LC_LOAD_WEAK_DYLIB: 2177 case LC_REEXPORT_DYLIB: 2178 case LC_LOADFVMLIB: 2179 case LC_LOAD_UPWARD_DYLIB: { 2180 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2181 const char *path = m_data.PeekCStr(name_offset); 2182 if (path) { 2183 FileSpec file_spec(path, false); 2184 // Strip the path if there is @rpath, @executable, etc so we just use 2185 // the basename 2186 if (path[0] == '@') 2187 file_spec.GetDirectory().Clear(); 2188 2189 if (lc.cmd == LC_REEXPORT_DYLIB) { 2190 m_reexported_dylibs.AppendIfUnique(file_spec); 2191 } 2192 2193 dylib_files.Append(file_spec); 2194 } 2195 } break; 2196 2197 case LC_FUNCTION_STARTS: 2198 function_starts_load_command.cmd = lc.cmd; 2199 function_starts_load_command.cmdsize = lc.cmdsize; 2200 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2201 NULL) // fill in symoff, nsyms, stroff, strsize fields 2202 memset(&function_starts_load_command, 0, 2203 sizeof(function_starts_load_command)); 2204 break; 2205 2206 default: 2207 break; 2208 } 2209 offset = cmd_offset + lc.cmdsize; 2210 } 2211 2212 if (symtab_load_command.cmd) { 2213 Symtab *symtab = m_symtab_ap.get(); 2214 SectionList *section_list = GetSectionList(); 2215 if (section_list == NULL) 2216 return 0; 2217 2218 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2219 const ByteOrder byte_order = m_data.GetByteOrder(); 2220 bool bit_width_32 = addr_byte_size == 4; 2221 const size_t nlist_byte_size = 2222 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2223 2224 DataExtractor nlist_data(NULL, 0, byte_order, addr_byte_size); 2225 DataExtractor strtab_data(NULL, 0, byte_order, addr_byte_size); 2226 DataExtractor function_starts_data(NULL, 0, byte_order, addr_byte_size); 2227 DataExtractor indirect_symbol_index_data(NULL, 0, byte_order, 2228 addr_byte_size); 2229 DataExtractor dyld_trie_data(NULL, 0, byte_order, addr_byte_size); 2230 2231 const addr_t nlist_data_byte_size = 2232 symtab_load_command.nsyms * nlist_byte_size; 2233 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2234 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2235 2236 ProcessSP process_sp(m_process_wp.lock()); 2237 Process *process = process_sp.get(); 2238 2239 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2240 2241 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2242 Target &target = process->GetTarget(); 2243 2244 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2245 2246 SectionSP linkedit_section_sp( 2247 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2248 // Reading mach file from memory in a process or core file... 2249 2250 if (linkedit_section_sp) { 2251 addr_t linkedit_load_addr = 2252 linkedit_section_sp->GetLoadBaseAddress(&target); 2253 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2254 // We might be trying to access the symbol table before the 2255 // __LINKEDIT's load 2256 // address has been set in the target. We can't fail to read the 2257 // symbol table, 2258 // so calculate the right address manually 2259 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2260 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2261 } 2262 2263 const addr_t linkedit_file_offset = 2264 linkedit_section_sp->GetFileOffset(); 2265 const addr_t symoff_addr = linkedit_load_addr + 2266 symtab_load_command.symoff - 2267 linkedit_file_offset; 2268 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2269 linkedit_file_offset; 2270 2271 bool data_was_read = false; 2272 2273 #if defined(__APPLE__) && \ 2274 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2275 if (m_header.flags & 0x80000000u && 2276 process->GetAddressByteSize() == sizeof(void *)) { 2277 // This mach-o memory file is in the dyld shared cache. If this 2278 // program is not remote and this is iOS, then this process will 2279 // share the same shared cache as the process we are debugging and 2280 // we can read the entire __LINKEDIT from the address space in this 2281 // process. This is a needed optimization that is used for local iOS 2282 // debugging only since all shared libraries in the shared cache do 2283 // not have corresponding files that exist in the file system of the 2284 // device. They have been combined into a single file. This means we 2285 // always have to load these files from memory. All of the symbol and 2286 // string tables from all of the __LINKEDIT sections from the shared 2287 // libraries in the shared cache have been merged into a single large 2288 // symbol and string table. Reading all of this symbol and string 2289 // table 2290 // data across can slow down debug launch times, so we optimize this 2291 // by 2292 // reading the memory for the __LINKEDIT section from this process. 2293 2294 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 2295 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 2296 bool use_lldb_cache = true; 2297 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2298 lldb_shared_cache != process_shared_cache) { 2299 use_lldb_cache = false; 2300 ModuleSP module_sp(GetModule()); 2301 if (module_sp) 2302 module_sp->ReportWarning("shared cache in process does not match " 2303 "lldb's own shared cache, startup will " 2304 "be slow."); 2305 } 2306 2307 PlatformSP platform_sp(target.GetPlatform()); 2308 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2309 data_was_read = true; 2310 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2311 eByteOrderLittle); 2312 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2313 eByteOrderLittle); 2314 if (function_starts_load_command.cmd) { 2315 const addr_t func_start_addr = 2316 linkedit_load_addr + function_starts_load_command.dataoff - 2317 linkedit_file_offset; 2318 function_starts_data.SetData( 2319 (void *)func_start_addr, 2320 function_starts_load_command.datasize, eByteOrderLittle); 2321 } 2322 } 2323 } 2324 #endif 2325 2326 if (!data_was_read) { 2327 // Always load dyld - the dynamic linker - from memory if we didn't 2328 // find a binary anywhere else. 2329 // lldb will not register dylib/framework/bundle loads/unloads if we 2330 // don't have the dyld symbols, 2331 // we force dyld to load from memory despite the user's 2332 // target.memory-module-load-level setting. 2333 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2334 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2335 DataBufferSP nlist_data_sp( 2336 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2337 if (nlist_data_sp) 2338 nlist_data.SetData(nlist_data_sp, 0, 2339 nlist_data_sp->GetByteSize()); 2340 // Load strings individually from memory when loading from memory 2341 // since shared cache 2342 // string tables contain strings for all symbols from all shared 2343 // cached libraries 2344 // DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, 2345 // strtab_data_byte_size)); 2346 // if (strtab_data_sp) 2347 // strtab_data.SetData (strtab_data_sp, 0, 2348 // strtab_data_sp->GetByteSize()); 2349 if (m_dysymtab.nindirectsyms != 0) { 2350 const addr_t indirect_syms_addr = linkedit_load_addr + 2351 m_dysymtab.indirectsymoff - 2352 linkedit_file_offset; 2353 DataBufferSP indirect_syms_data_sp( 2354 ReadMemory(process_sp, indirect_syms_addr, 2355 m_dysymtab.nindirectsyms * 4)); 2356 if (indirect_syms_data_sp) 2357 indirect_symbol_index_data.SetData( 2358 indirect_syms_data_sp, 0, 2359 indirect_syms_data_sp->GetByteSize()); 2360 } 2361 } else if (memory_module_load_level >= 2362 eMemoryModuleLoadLevelPartial) { 2363 if (function_starts_load_command.cmd) { 2364 const addr_t func_start_addr = 2365 linkedit_load_addr + function_starts_load_command.dataoff - 2366 linkedit_file_offset; 2367 DataBufferSP func_start_data_sp( 2368 ReadMemory(process_sp, func_start_addr, 2369 function_starts_load_command.datasize)); 2370 if (func_start_data_sp) 2371 function_starts_data.SetData(func_start_data_sp, 0, 2372 func_start_data_sp->GetByteSize()); 2373 } 2374 } 2375 } 2376 } 2377 } else { 2378 nlist_data.SetData(m_data, symtab_load_command.symoff, 2379 nlist_data_byte_size); 2380 strtab_data.SetData(m_data, symtab_load_command.stroff, 2381 strtab_data_byte_size); 2382 2383 if (dyld_info.export_size > 0) { 2384 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2385 dyld_info.export_size); 2386 } 2387 2388 if (m_dysymtab.nindirectsyms != 0) { 2389 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2390 m_dysymtab.nindirectsyms * 4); 2391 } 2392 if (function_starts_load_command.cmd) { 2393 function_starts_data.SetData(m_data, 2394 function_starts_load_command.dataoff, 2395 function_starts_load_command.datasize); 2396 } 2397 } 2398 2399 if (nlist_data.GetByteSize() == 0 && 2400 memory_module_load_level == eMemoryModuleLoadLevelComplete) { 2401 if (log) 2402 module_sp->LogMessage(log, "failed to read nlist data"); 2403 return 0; 2404 } 2405 2406 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2407 if (!have_strtab_data) { 2408 if (process) { 2409 if (strtab_addr == LLDB_INVALID_ADDRESS) { 2410 if (log) 2411 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2412 return 0; 2413 } 2414 } else { 2415 if (log) 2416 module_sp->LogMessage(log, "failed to read strtab data"); 2417 return 0; 2418 } 2419 } 2420 2421 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2422 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2423 const ConstString &g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2424 const ConstString &g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2425 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2426 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2427 SectionSP text_section_sp( 2428 section_list->FindSectionByName(g_segment_name_TEXT)); 2429 SectionSP data_section_sp( 2430 section_list->FindSectionByName(g_segment_name_DATA)); 2431 SectionSP data_dirty_section_sp( 2432 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2433 SectionSP data_const_section_sp( 2434 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2435 SectionSP objc_section_sp( 2436 section_list->FindSectionByName(g_segment_name_OBJC)); 2437 SectionSP eh_frame_section_sp; 2438 if (text_section_sp.get()) 2439 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2440 g_section_name_eh_frame); 2441 else 2442 eh_frame_section_sp = 2443 section_list->FindSectionByName(g_section_name_eh_frame); 2444 2445 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2446 2447 // lldb works best if it knows the start address of all functions in a 2448 // module. 2449 // Linker symbols or debug info are normally the best source of information 2450 // for start addr / size but 2451 // they may be stripped in a released binary. 2452 // Two additional sources of information exist in Mach-O binaries: 2453 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2454 // function's start address in the 2455 // binary, relative to the text section. 2456 // eh_frame - the eh_frame FDEs have the start addr & size of 2457 // each function 2458 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2459 // all modern binaries. 2460 // Binaries built to run on older releases may need to use eh_frame 2461 // information. 2462 2463 if (text_section_sp && function_starts_data.GetByteSize()) { 2464 FunctionStarts::Entry function_start_entry; 2465 function_start_entry.data = false; 2466 lldb::offset_t function_start_offset = 0; 2467 function_start_entry.addr = text_section_sp->GetFileAddress(); 2468 uint64_t delta; 2469 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2470 0) { 2471 // Now append the current entry 2472 function_start_entry.addr += delta; 2473 function_starts.Append(function_start_entry); 2474 } 2475 } else { 2476 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2477 // load command claiming an eh_frame 2478 // but it doesn't actually have the eh_frame content. And if we have a 2479 // dSYM, we don't need to do any 2480 // of this fill-in-the-missing-symbols works anyway - the debug info 2481 // should give us all the functions in 2482 // the module. 2483 if (text_section_sp.get() && eh_frame_section_sp.get() && 2484 m_type != eTypeDebugInfo) { 2485 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2486 eRegisterKindEHFrame, true); 2487 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2488 eh_frame.GetFunctionAddressAndSizeVector(functions); 2489 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2490 size_t count = functions.GetSize(); 2491 for (size_t i = 0; i < count; ++i) { 2492 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2493 functions.GetEntryAtIndex(i); 2494 if (func) { 2495 FunctionStarts::Entry function_start_entry; 2496 function_start_entry.addr = func->base - text_base_addr; 2497 function_starts.Append(function_start_entry); 2498 } 2499 } 2500 } 2501 } 2502 2503 const size_t function_starts_count = function_starts.GetSize(); 2504 2505 // For user process binaries (executables, dylibs, frameworks, bundles), if 2506 // we don't have 2507 // LC_FUNCTION_STARTS/eh_frame section in this binary, we're going to assume 2508 // the binary 2509 // has been stripped. Don't allow assembly language instruction emulation 2510 // because we don't 2511 // know proper function start boundaries. 2512 // 2513 // For all other types of binaries (kernels, stand-alone bare board 2514 // binaries, kexts), they 2515 // may not have LC_FUNCTION_STARTS / eh_frame sections - we should not make 2516 // any assumptions 2517 // about them based on that. 2518 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2519 m_allow_assembly_emulation_unwind_plans = false; 2520 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2521 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2522 2523 if (unwind_or_symbol_log) 2524 module_sp->LogMessage( 2525 unwind_or_symbol_log, 2526 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2527 } 2528 2529 const user_id_t TEXT_eh_frame_sectID = 2530 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2531 : static_cast<user_id_t>(NO_SECT); 2532 2533 lldb::offset_t nlist_data_offset = 0; 2534 2535 uint32_t N_SO_index = UINT32_MAX; 2536 2537 MachSymtabSectionInfo section_info(section_list); 2538 std::vector<uint32_t> N_FUN_indexes; 2539 std::vector<uint32_t> N_NSYM_indexes; 2540 std::vector<uint32_t> N_INCL_indexes; 2541 std::vector<uint32_t> N_BRAC_indexes; 2542 std::vector<uint32_t> N_COMM_indexes; 2543 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2544 typedef std::map<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2545 typedef std::map<const char *, uint32_t> ConstNameToSymbolIndexMap; 2546 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2547 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2548 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2549 // Any symbols that get merged into another will get an entry 2550 // in this map so we know 2551 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2552 uint32_t nlist_idx = 0; 2553 Symbol *symbol_ptr = NULL; 2554 2555 uint32_t sym_idx = 0; 2556 Symbol *sym = NULL; 2557 size_t num_syms = 0; 2558 std::string memory_symbol_name; 2559 uint32_t unmapped_local_symbols_found = 0; 2560 2561 std::vector<TrieEntryWithOffset> trie_entries; 2562 std::set<lldb::addr_t> resolver_addresses; 2563 2564 if (dyld_trie_data.GetByteSize() > 0) { 2565 std::vector<llvm::StringRef> nameSlices; 2566 ParseTrieEntries(dyld_trie_data, 0, is_arm, nameSlices, 2567 resolver_addresses, trie_entries); 2568 2569 ConstString text_segment_name("__TEXT"); 2570 SectionSP text_segment_sp = 2571 GetSectionList()->FindSectionByName(text_segment_name); 2572 if (text_segment_sp) { 2573 const lldb::addr_t text_segment_file_addr = 2574 text_segment_sp->GetFileAddress(); 2575 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) { 2576 for (auto &e : trie_entries) 2577 e.entry.address += text_segment_file_addr; 2578 } 2579 } 2580 } 2581 2582 typedef std::set<ConstString> IndirectSymbols; 2583 IndirectSymbols indirect_symbol_names; 2584 2585 #if defined(__APPLE__) && \ 2586 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2587 2588 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2589 // optimized by moving LOCAL 2590 // symbols out of the memory mapped portion of the DSC. The symbol 2591 // information has all been retained, 2592 // but it isn't available in the normal nlist data. However, there *are* 2593 // duplicate entries of *some* 2594 // LOCAL symbols in the normal nlist data. To handle this situation 2595 // correctly, we must first attempt 2596 // to parse any DSC unmapped symbol information. If we find any, we set a 2597 // flag that tells the normal 2598 // nlist parser to ignore all LOCAL symbols. 2599 2600 if (m_header.flags & 0x80000000u) { 2601 // Before we can start mapping the DSC, we need to make certain the target 2602 // process is actually 2603 // using the cache we can find. 2604 2605 // Next we need to determine the correct path for the dyld shared cache. 2606 2607 ArchSpec header_arch; 2608 GetArchitecture(header_arch); 2609 char dsc_path[PATH_MAX]; 2610 char dsc_path_development[PATH_MAX]; 2611 2612 snprintf( 2613 dsc_path, sizeof(dsc_path), "%s%s%s", 2614 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2615 */ 2616 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2617 header_arch.GetArchitectureName()); 2618 2619 snprintf( 2620 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2621 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2622 */ 2623 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2624 header_arch.GetArchitectureName(), ".development"); 2625 2626 FileSpec dsc_nondevelopment_filespec(dsc_path, false); 2627 FileSpec dsc_development_filespec(dsc_path_development, false); 2628 FileSpec dsc_filespec; 2629 2630 UUID dsc_uuid; 2631 UUID process_shared_cache_uuid; 2632 2633 if (process) { 2634 process_shared_cache_uuid = GetProcessSharedCacheUUID(process); 2635 } 2636 2637 // First see if we can find an exact match for the inferior process shared 2638 // cache UUID in 2639 // the development or non-development shared caches on disk. 2640 if (process_shared_cache_uuid.IsValid()) { 2641 if (dsc_development_filespec.Exists()) { 2642 UUID dsc_development_uuid = GetSharedCacheUUID( 2643 dsc_development_filespec, byte_order, addr_byte_size); 2644 if (dsc_development_uuid.IsValid() && 2645 dsc_development_uuid == process_shared_cache_uuid) { 2646 dsc_filespec = dsc_development_filespec; 2647 dsc_uuid = dsc_development_uuid; 2648 } 2649 } 2650 if (!dsc_uuid.IsValid() && dsc_nondevelopment_filespec.Exists()) { 2651 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2652 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2653 if (dsc_nondevelopment_uuid.IsValid() && 2654 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2655 dsc_filespec = dsc_nondevelopment_filespec; 2656 dsc_uuid = dsc_nondevelopment_uuid; 2657 } 2658 } 2659 } 2660 2661 // Failing a UUID match, prefer the development dyld_shared cache if both 2662 // are present. 2663 if (!dsc_filespec.Exists()) { 2664 if (dsc_development_filespec.Exists()) { 2665 dsc_filespec = dsc_development_filespec; 2666 } else { 2667 dsc_filespec = dsc_nondevelopment_filespec; 2668 } 2669 } 2670 2671 /* The dyld_cache_header has a pointer to the 2672 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2673 The dyld_cache_local_symbols_info structure gives us three things: 2674 1. The start and count of the nlist records in the dyld_shared_cache 2675 file 2676 2. The start and size of the strings for these nlist records 2677 3. The start and count of dyld_cache_local_symbols_entry entries 2678 2679 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2680 dyld shared cache. 2681 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2682 the dyld shared cache. 2683 The dyld_cache_local_symbols_entry also lists the start of this 2684 dylib/framework's nlist records 2685 and the count of how many nlist records there are for this 2686 dylib/framework. 2687 */ 2688 2689 // Process the dyld shared cache header to find the unmapped symbols 2690 2691 DataBufferSP dsc_data_sp = dsc_filespec.MemoryMapFileContentsIfLocal( 2692 0, sizeof(struct lldb_copy_dyld_cache_header_v1)); 2693 if (!dsc_uuid.IsValid()) { 2694 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2695 } 2696 if (dsc_data_sp) { 2697 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2698 2699 bool uuid_match = true; 2700 if (dsc_uuid.IsValid() && process) { 2701 if (process_shared_cache_uuid.IsValid() && 2702 dsc_uuid != process_shared_cache_uuid) { 2703 // The on-disk dyld_shared_cache file is not the same as the one in 2704 // this 2705 // process' memory, don't use it. 2706 uuid_match = false; 2707 ModuleSP module_sp(GetModule()); 2708 if (module_sp) 2709 module_sp->ReportWarning("process shared cache does not match " 2710 "on-disk dyld_shared_cache file, some " 2711 "symbol names will be missing."); 2712 } 2713 } 2714 2715 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2716 2717 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2718 2719 // If the mappingOffset points to a location inside the header, we've 2720 // opened an old dyld shared cache, and should not proceed further. 2721 if (uuid_match && 2722 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2723 2724 DataBufferSP dsc_mapping_info_data_sp = 2725 dsc_filespec.MemoryMapFileContentsIfLocal( 2726 mappingOffset, 2727 sizeof(struct lldb_copy_dyld_cache_mapping_info)); 2728 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2729 byte_order, addr_byte_size); 2730 offset = 0; 2731 2732 // The File addresses (from the in-memory Mach-O load commands) for 2733 // the shared libraries 2734 // in the shared library cache need to be adjusted by an offset to 2735 // match up with the 2736 // dylibOffset identifying field in the 2737 // dyld_cache_local_symbol_entry's. This offset is 2738 // recorded in mapping_offset_value. 2739 const uint64_t mapping_offset_value = 2740 dsc_mapping_info_data.GetU64(&offset); 2741 2742 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, 2743 localSymbolsOffset); 2744 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2745 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2746 2747 if (localSymbolsOffset && localSymbolsSize) { 2748 // Map the local symbols 2749 if (DataBufferSP dsc_local_symbols_data_sp = 2750 dsc_filespec.MemoryMapFileContentsIfLocal( 2751 localSymbolsOffset, localSymbolsSize)) { 2752 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2753 byte_order, addr_byte_size); 2754 2755 offset = 0; 2756 2757 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2758 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2759 UndefinedNameToDescMap undefined_name_to_desc; 2760 SymbolIndexToName reexport_shlib_needs_fixup; 2761 2762 // Read the local_symbols_infos struct in one shot 2763 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2764 dsc_local_symbols_data.GetU32(&offset, 2765 &local_symbols_info.nlistOffset, 6); 2766 2767 SectionSP text_section_sp( 2768 section_list->FindSectionByName(GetSegmentNameTEXT())); 2769 2770 uint32_t header_file_offset = 2771 (text_section_sp->GetFileAddress() - mapping_offset_value); 2772 2773 offset = local_symbols_info.entriesOffset; 2774 for (uint32_t entry_index = 0; 2775 entry_index < local_symbols_info.entriesCount; 2776 entry_index++) { 2777 struct lldb_copy_dyld_cache_local_symbols_entry 2778 local_symbols_entry; 2779 local_symbols_entry.dylibOffset = 2780 dsc_local_symbols_data.GetU32(&offset); 2781 local_symbols_entry.nlistStartIndex = 2782 dsc_local_symbols_data.GetU32(&offset); 2783 local_symbols_entry.nlistCount = 2784 dsc_local_symbols_data.GetU32(&offset); 2785 2786 if (header_file_offset == local_symbols_entry.dylibOffset) { 2787 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2788 2789 // The normal nlist code cannot correctly size the Symbols 2790 // array, we need to allocate it here. 2791 sym = symtab->Resize( 2792 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2793 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2794 num_syms = symtab->GetNumSymbols(); 2795 2796 nlist_data_offset = 2797 local_symbols_info.nlistOffset + 2798 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2799 uint32_t string_table_offset = 2800 local_symbols_info.stringsOffset; 2801 2802 for (uint32_t nlist_index = 0; 2803 nlist_index < local_symbols_entry.nlistCount; 2804 nlist_index++) { 2805 ///////////////////////////// 2806 { 2807 struct nlist_64 nlist; 2808 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize( 2809 nlist_data_offset, nlist_byte_size)) 2810 break; 2811 2812 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked( 2813 &nlist_data_offset); 2814 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked( 2815 &nlist_data_offset); 2816 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked( 2817 &nlist_data_offset); 2818 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked( 2819 &nlist_data_offset); 2820 nlist.n_value = 2821 dsc_local_symbols_data.GetAddress_unchecked( 2822 &nlist_data_offset); 2823 2824 SymbolType type = eSymbolTypeInvalid; 2825 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2826 string_table_offset + nlist.n_strx); 2827 2828 if (symbol_name == NULL) { 2829 // No symbol should be NULL, even the symbols with no 2830 // string values should have an offset zero which points 2831 // to an empty C-string 2832 Host::SystemLog( 2833 Host::eSystemLogError, 2834 "error: DSC unmapped local symbol[%u] has invalid " 2835 "string table offset 0x%x in %s, ignoring symbol\n", 2836 entry_index, nlist.n_strx, 2837 module_sp->GetFileSpec().GetPath().c_str()); 2838 continue; 2839 } 2840 if (symbol_name[0] == '\0') 2841 symbol_name = NULL; 2842 2843 const char *symbol_name_non_abi_mangled = NULL; 2844 2845 SectionSP symbol_section; 2846 uint32_t symbol_byte_size = 0; 2847 bool add_nlist = true; 2848 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2849 bool demangled_is_synthesized = false; 2850 bool is_gsym = false; 2851 bool set_value = true; 2852 2853 assert(sym_idx < num_syms); 2854 2855 sym[sym_idx].SetDebug(is_debug); 2856 2857 if (is_debug) { 2858 switch (nlist.n_type) { 2859 case N_GSYM: 2860 // global symbol: name,,NO_SECT,type,0 2861 // Sometimes the N_GSYM value contains the address. 2862 2863 // FIXME: In the .o files, we have a GSYM and a debug 2864 // symbol for all the ObjC data. They 2865 // have the same address, but we want to ensure that 2866 // we always find only the real symbol, 2867 // 'cause we don't currently correctly attribute the 2868 // GSYM one to the ObjCClass/Ivar/MetaClass 2869 // symbol type. This is a temporary hack to make sure 2870 // the ObjectiveC symbols get treated 2871 // correctly. To do this right, we should coalesce 2872 // all the GSYM & global symbols that have the 2873 // same address. 2874 2875 is_gsym = true; 2876 sym[sym_idx].SetExternal(true); 2877 2878 if (symbol_name && symbol_name[0] == '_' && 2879 symbol_name[1] == 'O') { 2880 llvm::StringRef symbol_name_ref(symbol_name); 2881 if (symbol_name_ref.startswith( 2882 g_objc_v2_prefix_class)) { 2883 symbol_name_non_abi_mangled = symbol_name + 1; 2884 symbol_name = 2885 symbol_name + g_objc_v2_prefix_class.size(); 2886 type = eSymbolTypeObjCClass; 2887 demangled_is_synthesized = true; 2888 2889 } else if (symbol_name_ref.startswith( 2890 g_objc_v2_prefix_metaclass)) { 2891 symbol_name_non_abi_mangled = symbol_name + 1; 2892 symbol_name = symbol_name + 2893 g_objc_v2_prefix_metaclass.size(); 2894 type = eSymbolTypeObjCMetaClass; 2895 demangled_is_synthesized = true; 2896 } else if (symbol_name_ref.startswith( 2897 g_objc_v2_prefix_ivar)) { 2898 symbol_name_non_abi_mangled = symbol_name + 1; 2899 symbol_name = 2900 symbol_name + g_objc_v2_prefix_ivar.size(); 2901 type = eSymbolTypeObjCIVar; 2902 demangled_is_synthesized = true; 2903 } 2904 } else { 2905 if (nlist.n_value != 0) 2906 symbol_section = section_info.GetSection( 2907 nlist.n_sect, nlist.n_value); 2908 type = eSymbolTypeData; 2909 } 2910 break; 2911 2912 case N_FNAME: 2913 // procedure name (f77 kludge): name,,NO_SECT,0,0 2914 type = eSymbolTypeCompiler; 2915 break; 2916 2917 case N_FUN: 2918 // procedure: name,,n_sect,linenumber,address 2919 if (symbol_name) { 2920 type = eSymbolTypeCode; 2921 symbol_section = section_info.GetSection( 2922 nlist.n_sect, nlist.n_value); 2923 2924 N_FUN_addr_to_sym_idx.insert( 2925 std::make_pair(nlist.n_value, sym_idx)); 2926 // We use the current number of symbols in the 2927 // symbol table in lieu of 2928 // using nlist_idx in case we ever start trimming 2929 // entries out 2930 N_FUN_indexes.push_back(sym_idx); 2931 } else { 2932 type = eSymbolTypeCompiler; 2933 2934 if (!N_FUN_indexes.empty()) { 2935 // Copy the size of the function into the original 2936 // STAB entry so we don't have 2937 // to hunt for it later 2938 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2939 ->SetByteSize(nlist.n_value); 2940 N_FUN_indexes.pop_back(); 2941 // We don't really need the end function STAB as 2942 // it contains the size which 2943 // we already placed with the original symbol, so 2944 // don't add it if we want a 2945 // minimal symbol table 2946 add_nlist = false; 2947 } 2948 } 2949 break; 2950 2951 case N_STSYM: 2952 // static symbol: name,,n_sect,type,address 2953 N_STSYM_addr_to_sym_idx.insert( 2954 std::make_pair(nlist.n_value, sym_idx)); 2955 symbol_section = section_info.GetSection( 2956 nlist.n_sect, nlist.n_value); 2957 if (symbol_name && symbol_name[0]) { 2958 type = ObjectFile::GetSymbolTypeFromName( 2959 symbol_name + 1, eSymbolTypeData); 2960 } 2961 break; 2962 2963 case N_LCSYM: 2964 // .lcomm symbol: name,,n_sect,type,address 2965 symbol_section = section_info.GetSection( 2966 nlist.n_sect, nlist.n_value); 2967 type = eSymbolTypeCommonBlock; 2968 break; 2969 2970 case N_BNSYM: 2971 // We use the current number of symbols in the symbol 2972 // table in lieu of 2973 // using nlist_idx in case we ever start trimming 2974 // entries out 2975 // Skip these if we want minimal symbol tables 2976 add_nlist = false; 2977 break; 2978 2979 case N_ENSYM: 2980 // Set the size of the N_BNSYM to the terminating 2981 // index of this N_ENSYM 2982 // so that we can always skip the entire symbol if we 2983 // need to navigate 2984 // more quickly at the source level when parsing STABS 2985 // Skip these if we want minimal symbol tables 2986 add_nlist = false; 2987 break; 2988 2989 case N_OPT: 2990 // emitted with gcc2_compiled and in gcc source 2991 type = eSymbolTypeCompiler; 2992 break; 2993 2994 case N_RSYM: 2995 // register sym: name,,NO_SECT,type,register 2996 type = eSymbolTypeVariable; 2997 break; 2998 2999 case N_SLINE: 3000 // src line: 0,,n_sect,linenumber,address 3001 symbol_section = section_info.GetSection( 3002 nlist.n_sect, nlist.n_value); 3003 type = eSymbolTypeLineEntry; 3004 break; 3005 3006 case N_SSYM: 3007 // structure elt: name,,NO_SECT,type,struct_offset 3008 type = eSymbolTypeVariableType; 3009 break; 3010 3011 case N_SO: 3012 // source file name 3013 type = eSymbolTypeSourceFile; 3014 if (symbol_name == NULL) { 3015 add_nlist = false; 3016 if (N_SO_index != UINT32_MAX) { 3017 // Set the size of the N_SO to the terminating 3018 // index of this N_SO 3019 // so that we can always skip the entire N_SO if 3020 // we need to navigate 3021 // more quickly at the source level when parsing 3022 // STABS 3023 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3024 symbol_ptr->SetByteSize(sym_idx); 3025 symbol_ptr->SetSizeIsSibling(true); 3026 } 3027 N_NSYM_indexes.clear(); 3028 N_INCL_indexes.clear(); 3029 N_BRAC_indexes.clear(); 3030 N_COMM_indexes.clear(); 3031 N_FUN_indexes.clear(); 3032 N_SO_index = UINT32_MAX; 3033 } else { 3034 // We use the current number of symbols in the 3035 // symbol table in lieu of 3036 // using nlist_idx in case we ever start trimming 3037 // entries out 3038 const bool N_SO_has_full_path = 3039 symbol_name[0] == '/'; 3040 if (N_SO_has_full_path) { 3041 if ((N_SO_index == sym_idx - 1) && 3042 ((sym_idx - 1) < num_syms)) { 3043 // We have two consecutive N_SO entries where 3044 // the first contains a directory 3045 // and the second contains a full path. 3046 sym[sym_idx - 1].GetMangled().SetValue( 3047 ConstString(symbol_name), false); 3048 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3049 add_nlist = false; 3050 } else { 3051 // This is the first entry in a N_SO that 3052 // contains a directory or 3053 // a full path to the source file 3054 N_SO_index = sym_idx; 3055 } 3056 } else if ((N_SO_index == sym_idx - 1) && 3057 ((sym_idx - 1) < num_syms)) { 3058 // This is usually the second N_SO entry that 3059 // contains just the filename, 3060 // so here we combine it with the first one if we 3061 // are minimizing the symbol table 3062 const char *so_path = 3063 sym[sym_idx - 1] 3064 .GetMangled() 3065 .GetDemangledName( 3066 lldb::eLanguageTypeUnknown) 3067 .AsCString(); 3068 if (so_path && so_path[0]) { 3069 std::string full_so_path(so_path); 3070 const size_t double_slash_pos = 3071 full_so_path.find("//"); 3072 if (double_slash_pos != std::string::npos) { 3073 // The linker has been generating bad N_SO 3074 // entries with doubled up paths 3075 // in the format "%s%s" where the first string 3076 // in the DW_AT_comp_dir, 3077 // and the second is the directory for the 3078 // source file so you end up with 3079 // a path that looks like "/tmp/src//tmp/src/" 3080 FileSpec so_dir(so_path, false); 3081 if (!so_dir.Exists()) { 3082 so_dir.SetFile( 3083 &full_so_path[double_slash_pos + 1], 3084 false); 3085 if (so_dir.Exists()) { 3086 // Trim off the incorrect path 3087 full_so_path.erase(0, 3088 double_slash_pos + 1); 3089 } 3090 } 3091 } 3092 if (*full_so_path.rbegin() != '/') 3093 full_so_path += '/'; 3094 full_so_path += symbol_name; 3095 sym[sym_idx - 1].GetMangled().SetValue( 3096 ConstString(full_so_path.c_str()), false); 3097 add_nlist = false; 3098 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3099 } 3100 } else { 3101 // This could be a relative path to a N_SO 3102 N_SO_index = sym_idx; 3103 } 3104 } 3105 break; 3106 3107 case N_OSO: 3108 // object file name: name,,0,0,st_mtime 3109 type = eSymbolTypeObjectFile; 3110 break; 3111 3112 case N_LSYM: 3113 // local sym: name,,NO_SECT,type,offset 3114 type = eSymbolTypeLocal; 3115 break; 3116 3117 //---------------------------------------------------------------------- 3118 // INCL scopes 3119 //---------------------------------------------------------------------- 3120 case N_BINCL: 3121 // include file beginning: name,,NO_SECT,0,sum 3122 // We use the current number of symbols in the symbol 3123 // table in lieu of 3124 // using nlist_idx in case we ever start trimming 3125 // entries out 3126 N_INCL_indexes.push_back(sym_idx); 3127 type = eSymbolTypeScopeBegin; 3128 break; 3129 3130 case N_EINCL: 3131 // include file end: name,,NO_SECT,0,0 3132 // Set the size of the N_BINCL to the terminating 3133 // index of this N_EINCL 3134 // so that we can always skip the entire symbol if we 3135 // need to navigate 3136 // more quickly at the source level when parsing STABS 3137 if (!N_INCL_indexes.empty()) { 3138 symbol_ptr = 3139 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3140 symbol_ptr->SetByteSize(sym_idx + 1); 3141 symbol_ptr->SetSizeIsSibling(true); 3142 N_INCL_indexes.pop_back(); 3143 } 3144 type = eSymbolTypeScopeEnd; 3145 break; 3146 3147 case N_SOL: 3148 // #included file name: name,,n_sect,0,address 3149 type = eSymbolTypeHeaderFile; 3150 3151 // We currently don't use the header files on darwin 3152 add_nlist = false; 3153 break; 3154 3155 case N_PARAMS: 3156 // compiler parameters: name,,NO_SECT,0,0 3157 type = eSymbolTypeCompiler; 3158 break; 3159 3160 case N_VERSION: 3161 // compiler version: name,,NO_SECT,0,0 3162 type = eSymbolTypeCompiler; 3163 break; 3164 3165 case N_OLEVEL: 3166 // compiler -O level: name,,NO_SECT,0,0 3167 type = eSymbolTypeCompiler; 3168 break; 3169 3170 case N_PSYM: 3171 // parameter: name,,NO_SECT,type,offset 3172 type = eSymbolTypeVariable; 3173 break; 3174 3175 case N_ENTRY: 3176 // alternate entry: name,,n_sect,linenumber,address 3177 symbol_section = section_info.GetSection( 3178 nlist.n_sect, nlist.n_value); 3179 type = eSymbolTypeLineEntry; 3180 break; 3181 3182 //---------------------------------------------------------------------- 3183 // Left and Right Braces 3184 //---------------------------------------------------------------------- 3185 case N_LBRAC: 3186 // left bracket: 0,,NO_SECT,nesting level,address 3187 // We use the current number of symbols in the symbol 3188 // table in lieu of 3189 // using nlist_idx in case we ever start trimming 3190 // entries out 3191 symbol_section = section_info.GetSection( 3192 nlist.n_sect, nlist.n_value); 3193 N_BRAC_indexes.push_back(sym_idx); 3194 type = eSymbolTypeScopeBegin; 3195 break; 3196 3197 case N_RBRAC: 3198 // right bracket: 0,,NO_SECT,nesting level,address 3199 // Set the size of the N_LBRAC to the terminating 3200 // index of this N_RBRAC 3201 // so that we can always skip the entire symbol if we 3202 // need to navigate 3203 // more quickly at the source level when parsing STABS 3204 symbol_section = section_info.GetSection( 3205 nlist.n_sect, nlist.n_value); 3206 if (!N_BRAC_indexes.empty()) { 3207 symbol_ptr = 3208 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3209 symbol_ptr->SetByteSize(sym_idx + 1); 3210 symbol_ptr->SetSizeIsSibling(true); 3211 N_BRAC_indexes.pop_back(); 3212 } 3213 type = eSymbolTypeScopeEnd; 3214 break; 3215 3216 case N_EXCL: 3217 // deleted include file: name,,NO_SECT,0,sum 3218 type = eSymbolTypeHeaderFile; 3219 break; 3220 3221 //---------------------------------------------------------------------- 3222 // COMM scopes 3223 //---------------------------------------------------------------------- 3224 case N_BCOMM: 3225 // begin common: name,,NO_SECT,0,0 3226 // We use the current number of symbols in the symbol 3227 // table in lieu of 3228 // using nlist_idx in case we ever start trimming 3229 // entries out 3230 type = eSymbolTypeScopeBegin; 3231 N_COMM_indexes.push_back(sym_idx); 3232 break; 3233 3234 case N_ECOML: 3235 // end common (local name): 0,,n_sect,0,address 3236 symbol_section = section_info.GetSection( 3237 nlist.n_sect, nlist.n_value); 3238 // Fall through 3239 3240 case N_ECOMM: 3241 // end common: name,,n_sect,0,0 3242 // Set the size of the N_BCOMM to the terminating 3243 // index of this N_ECOMM/N_ECOML 3244 // so that we can always skip the entire symbol if we 3245 // need to navigate 3246 // more quickly at the source level when parsing STABS 3247 if (!N_COMM_indexes.empty()) { 3248 symbol_ptr = 3249 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3250 symbol_ptr->SetByteSize(sym_idx + 1); 3251 symbol_ptr->SetSizeIsSibling(true); 3252 N_COMM_indexes.pop_back(); 3253 } 3254 type = eSymbolTypeScopeEnd; 3255 break; 3256 3257 case N_LENG: 3258 // second stab entry with length information 3259 type = eSymbolTypeAdditional; 3260 break; 3261 3262 default: 3263 break; 3264 } 3265 } else { 3266 // uint8_t n_pext = N_PEXT & nlist.n_type; 3267 uint8_t n_type = N_TYPE & nlist.n_type; 3268 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3269 3270 switch (n_type) { 3271 case N_INDR: { 3272 const char *reexport_name_cstr = 3273 strtab_data.PeekCStr(nlist.n_value); 3274 if (reexport_name_cstr && reexport_name_cstr[0]) { 3275 type = eSymbolTypeReExported; 3276 ConstString reexport_name( 3277 reexport_name_cstr + 3278 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3279 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3280 set_value = false; 3281 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3282 indirect_symbol_names.insert( 3283 ConstString(symbol_name + 3284 ((symbol_name[0] == '_') ? 1 : 0))); 3285 } else 3286 type = eSymbolTypeUndefined; 3287 } break; 3288 3289 case N_UNDF: 3290 if (symbol_name && symbol_name[0]) { 3291 ConstString undefined_name( 3292 symbol_name + 3293 ((symbol_name[0] == '_') ? 1 : 0)); 3294 undefined_name_to_desc[undefined_name] = 3295 nlist.n_desc; 3296 } 3297 // Fall through 3298 case N_PBUD: 3299 type = eSymbolTypeUndefined; 3300 break; 3301 3302 case N_ABS: 3303 type = eSymbolTypeAbsolute; 3304 break; 3305 3306 case N_SECT: { 3307 symbol_section = section_info.GetSection( 3308 nlist.n_sect, nlist.n_value); 3309 3310 if (symbol_section == NULL) { 3311 // TODO: warn about this? 3312 add_nlist = false; 3313 break; 3314 } 3315 3316 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3317 type = eSymbolTypeException; 3318 } else { 3319 uint32_t section_type = 3320 symbol_section->Get() & SECTION_TYPE; 3321 3322 switch (section_type) { 3323 case S_CSTRING_LITERALS: 3324 type = eSymbolTypeData; 3325 break; // section with only literal C strings 3326 case S_4BYTE_LITERALS: 3327 type = eSymbolTypeData; 3328 break; // section with only 4 byte literals 3329 case S_8BYTE_LITERALS: 3330 type = eSymbolTypeData; 3331 break; // section with only 8 byte literals 3332 case S_LITERAL_POINTERS: 3333 type = eSymbolTypeTrampoline; 3334 break; // section with only pointers to literals 3335 case S_NON_LAZY_SYMBOL_POINTERS: 3336 type = eSymbolTypeTrampoline; 3337 break; // section with only non-lazy symbol 3338 // pointers 3339 case S_LAZY_SYMBOL_POINTERS: 3340 type = eSymbolTypeTrampoline; 3341 break; // section with only lazy symbol pointers 3342 case S_SYMBOL_STUBS: 3343 type = eSymbolTypeTrampoline; 3344 break; // section with only symbol stubs, byte 3345 // size of stub in the reserved2 field 3346 case S_MOD_INIT_FUNC_POINTERS: 3347 type = eSymbolTypeCode; 3348 break; // section with only function pointers for 3349 // initialization 3350 case S_MOD_TERM_FUNC_POINTERS: 3351 type = eSymbolTypeCode; 3352 break; // section with only function pointers for 3353 // termination 3354 case S_INTERPOSING: 3355 type = eSymbolTypeTrampoline; 3356 break; // section with only pairs of function 3357 // pointers for interposing 3358 case S_16BYTE_LITERALS: 3359 type = eSymbolTypeData; 3360 break; // section with only 16 byte literals 3361 case S_DTRACE_DOF: 3362 type = eSymbolTypeInstrumentation; 3363 break; 3364 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3365 type = eSymbolTypeTrampoline; 3366 break; 3367 default: 3368 switch (symbol_section->GetType()) { 3369 case lldb::eSectionTypeCode: 3370 type = eSymbolTypeCode; 3371 break; 3372 case eSectionTypeData: 3373 case eSectionTypeDataCString: // Inlined C string 3374 // data 3375 case eSectionTypeDataCStringPointers: // Pointers 3376 // to C 3377 // string 3378 // data 3379 case eSectionTypeDataSymbolAddress: // Address of 3380 // a symbol in 3381 // the symbol 3382 // table 3383 case eSectionTypeData4: 3384 case eSectionTypeData8: 3385 case eSectionTypeData16: 3386 type = eSymbolTypeData; 3387 break; 3388 default: 3389 break; 3390 } 3391 break; 3392 } 3393 3394 if (type == eSymbolTypeInvalid) { 3395 const char *symbol_sect_name = 3396 symbol_section->GetName().AsCString(); 3397 if (symbol_section->IsDescendant( 3398 text_section_sp.get())) { 3399 if (symbol_section->IsClear( 3400 S_ATTR_PURE_INSTRUCTIONS | 3401 S_ATTR_SELF_MODIFYING_CODE | 3402 S_ATTR_SOME_INSTRUCTIONS)) 3403 type = eSymbolTypeData; 3404 else 3405 type = eSymbolTypeCode; 3406 } else if (symbol_section->IsDescendant( 3407 data_section_sp.get()) || 3408 symbol_section->IsDescendant( 3409 data_dirty_section_sp.get()) || 3410 symbol_section->IsDescendant( 3411 data_const_section_sp.get())) { 3412 if (symbol_sect_name && 3413 ::strstr(symbol_sect_name, "__objc") == 3414 symbol_sect_name) { 3415 type = eSymbolTypeRuntime; 3416 3417 if (symbol_name) { 3418 llvm::StringRef symbol_name_ref( 3419 symbol_name); 3420 if (symbol_name_ref.startswith("_OBJC_")) { 3421 static const llvm::StringRef 3422 g_objc_v2_prefix_class( 3423 "_OBJC_CLASS_$_"); 3424 static const llvm::StringRef 3425 g_objc_v2_prefix_metaclass( 3426 "_OBJC_METACLASS_$_"); 3427 static const llvm::StringRef 3428 g_objc_v2_prefix_ivar( 3429 "_OBJC_IVAR_$_"); 3430 if (symbol_name_ref.startswith( 3431 g_objc_v2_prefix_class)) { 3432 symbol_name_non_abi_mangled = 3433 symbol_name + 1; 3434 symbol_name = 3435 symbol_name + 3436 g_objc_v2_prefix_class.size(); 3437 type = eSymbolTypeObjCClass; 3438 demangled_is_synthesized = true; 3439 } else if ( 3440 symbol_name_ref.startswith( 3441 g_objc_v2_prefix_metaclass)) { 3442 symbol_name_non_abi_mangled = 3443 symbol_name + 1; 3444 symbol_name = 3445 symbol_name + 3446 g_objc_v2_prefix_metaclass.size(); 3447 type = eSymbolTypeObjCMetaClass; 3448 demangled_is_synthesized = true; 3449 } else if (symbol_name_ref.startswith( 3450 g_objc_v2_prefix_ivar)) { 3451 symbol_name_non_abi_mangled = 3452 symbol_name + 1; 3453 symbol_name = 3454 symbol_name + 3455 g_objc_v2_prefix_ivar.size(); 3456 type = eSymbolTypeObjCIVar; 3457 demangled_is_synthesized = true; 3458 } 3459 } 3460 } 3461 } else if (symbol_sect_name && 3462 ::strstr(symbol_sect_name, 3463 "__gcc_except_tab") == 3464 symbol_sect_name) { 3465 type = eSymbolTypeException; 3466 } else { 3467 type = eSymbolTypeData; 3468 } 3469 } else if (symbol_sect_name && 3470 ::strstr(symbol_sect_name, 3471 "__IMPORT") == 3472 symbol_sect_name) { 3473 type = eSymbolTypeTrampoline; 3474 } else if (symbol_section->IsDescendant( 3475 objc_section_sp.get())) { 3476 type = eSymbolTypeRuntime; 3477 if (symbol_name && symbol_name[0] == '.') { 3478 llvm::StringRef symbol_name_ref(symbol_name); 3479 static const llvm::StringRef 3480 g_objc_v1_prefix_class( 3481 ".objc_class_name_"); 3482 if (symbol_name_ref.startswith( 3483 g_objc_v1_prefix_class)) { 3484 symbol_name_non_abi_mangled = symbol_name; 3485 symbol_name = symbol_name + 3486 g_objc_v1_prefix_class.size(); 3487 type = eSymbolTypeObjCClass; 3488 demangled_is_synthesized = true; 3489 } 3490 } 3491 } 3492 } 3493 } 3494 } break; 3495 } 3496 } 3497 3498 if (add_nlist) { 3499 uint64_t symbol_value = nlist.n_value; 3500 if (symbol_name_non_abi_mangled) { 3501 sym[sym_idx].GetMangled().SetMangledName( 3502 ConstString(symbol_name_non_abi_mangled)); 3503 sym[sym_idx].GetMangled().SetDemangledName( 3504 ConstString(symbol_name)); 3505 } else { 3506 bool symbol_name_is_mangled = false; 3507 3508 if (symbol_name && symbol_name[0] == '_') { 3509 symbol_name_is_mangled = symbol_name[1] == '_'; 3510 symbol_name++; // Skip the leading underscore 3511 } 3512 3513 if (symbol_name) { 3514 ConstString const_symbol_name(symbol_name); 3515 sym[sym_idx].GetMangled().SetValue( 3516 const_symbol_name, symbol_name_is_mangled); 3517 if (is_gsym && is_debug) { 3518 const char *gsym_name = 3519 sym[sym_idx] 3520 .GetMangled() 3521 .GetName(lldb::eLanguageTypeUnknown, 3522 Mangled::ePreferMangled) 3523 .GetCString(); 3524 if (gsym_name) 3525 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3526 } 3527 } 3528 } 3529 if (symbol_section) { 3530 const addr_t section_file_addr = 3531 symbol_section->GetFileAddress(); 3532 if (symbol_byte_size == 0 && 3533 function_starts_count > 0) { 3534 addr_t symbol_lookup_file_addr = nlist.n_value; 3535 // Do an exact address match for non-ARM addresses, 3536 // else get the closest since 3537 // the symbol might be a thumb symbol which has an 3538 // address with bit zero set 3539 FunctionStarts::Entry *func_start_entry = 3540 function_starts.FindEntry( 3541 symbol_lookup_file_addr, !is_arm); 3542 if (is_arm && func_start_entry) { 3543 // Verify that the function start address is the 3544 // symbol address (ARM) 3545 // or the symbol address + 1 (thumb) 3546 if (func_start_entry->addr != 3547 symbol_lookup_file_addr && 3548 func_start_entry->addr != 3549 (symbol_lookup_file_addr + 1)) { 3550 // Not the right entry, NULL it out... 3551 func_start_entry = NULL; 3552 } 3553 } 3554 if (func_start_entry) { 3555 func_start_entry->data = true; 3556 3557 addr_t symbol_file_addr = func_start_entry->addr; 3558 uint32_t symbol_flags = 0; 3559 if (is_arm) { 3560 if (symbol_file_addr & 1) 3561 symbol_flags = 3562 MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3563 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3564 } 3565 3566 const FunctionStarts::Entry 3567 *next_func_start_entry = 3568 function_starts.FindNextEntry( 3569 func_start_entry); 3570 const addr_t section_end_file_addr = 3571 section_file_addr + 3572 symbol_section->GetByteSize(); 3573 if (next_func_start_entry) { 3574 addr_t next_symbol_file_addr = 3575 next_func_start_entry->addr; 3576 // Be sure the clear the Thumb address bit when 3577 // we calculate the size 3578 // from the current and next address 3579 if (is_arm) 3580 next_symbol_file_addr &= 3581 THUMB_ADDRESS_BIT_MASK; 3582 symbol_byte_size = std::min<lldb::addr_t>( 3583 next_symbol_file_addr - symbol_file_addr, 3584 section_end_file_addr - symbol_file_addr); 3585 } else { 3586 symbol_byte_size = 3587 section_end_file_addr - symbol_file_addr; 3588 } 3589 } 3590 } 3591 symbol_value -= section_file_addr; 3592 } 3593 3594 if (is_debug == false) { 3595 if (type == eSymbolTypeCode) { 3596 // See if we can find a N_FUN entry for any code 3597 // symbols. 3598 // If we do find a match, and the name matches, then 3599 // we 3600 // can merge the two into just the function symbol 3601 // to avoid 3602 // duplicate entries in the symbol table 3603 std::pair<ValueToSymbolIndexMap::const_iterator, 3604 ValueToSymbolIndexMap::const_iterator> 3605 range; 3606 range = N_FUN_addr_to_sym_idx.equal_range( 3607 nlist.n_value); 3608 if (range.first != range.second) { 3609 bool found_it = false; 3610 for (ValueToSymbolIndexMap::const_iterator pos = 3611 range.first; 3612 pos != range.second; ++pos) { 3613 if (sym[sym_idx].GetMangled().GetName( 3614 lldb::eLanguageTypeUnknown, 3615 Mangled::ePreferMangled) == 3616 sym[pos->second].GetMangled().GetName( 3617 lldb::eLanguageTypeUnknown, 3618 Mangled::ePreferMangled)) { 3619 m_nlist_idx_to_sym_idx[nlist_idx] = 3620 pos->second; 3621 // We just need the flags from the linker 3622 // symbol, so put these flags 3623 // into the N_FUN flags to avoid duplicate 3624 // symbols in the symbol table 3625 sym[pos->second].SetExternal( 3626 sym[sym_idx].IsExternal()); 3627 sym[pos->second].SetFlags(nlist.n_type << 16 | 3628 nlist.n_desc); 3629 if (resolver_addresses.find(nlist.n_value) != 3630 resolver_addresses.end()) 3631 sym[pos->second].SetType( 3632 eSymbolTypeResolver); 3633 sym[sym_idx].Clear(); 3634 found_it = true; 3635 break; 3636 } 3637 } 3638 if (found_it) 3639 continue; 3640 } else { 3641 if (resolver_addresses.find(nlist.n_value) != 3642 resolver_addresses.end()) 3643 type = eSymbolTypeResolver; 3644 } 3645 } else if (type == eSymbolTypeData || 3646 type == eSymbolTypeObjCClass || 3647 type == eSymbolTypeObjCMetaClass || 3648 type == eSymbolTypeObjCIVar) { 3649 // See if we can find a N_STSYM entry for any data 3650 // symbols. 3651 // If we do find a match, and the name matches, then 3652 // we 3653 // can merge the two into just the Static symbol to 3654 // avoid 3655 // duplicate entries in the symbol table 3656 std::pair<ValueToSymbolIndexMap::const_iterator, 3657 ValueToSymbolIndexMap::const_iterator> 3658 range; 3659 range = N_STSYM_addr_to_sym_idx.equal_range( 3660 nlist.n_value); 3661 if (range.first != range.second) { 3662 bool found_it = false; 3663 for (ValueToSymbolIndexMap::const_iterator pos = 3664 range.first; 3665 pos != range.second; ++pos) { 3666 if (sym[sym_idx].GetMangled().GetName( 3667 lldb::eLanguageTypeUnknown, 3668 Mangled::ePreferMangled) == 3669 sym[pos->second].GetMangled().GetName( 3670 lldb::eLanguageTypeUnknown, 3671 Mangled::ePreferMangled)) { 3672 m_nlist_idx_to_sym_idx[nlist_idx] = 3673 pos->second; 3674 // We just need the flags from the linker 3675 // symbol, so put these flags 3676 // into the N_STSYM flags to avoid duplicate 3677 // symbols in the symbol table 3678 sym[pos->second].SetExternal( 3679 sym[sym_idx].IsExternal()); 3680 sym[pos->second].SetFlags(nlist.n_type << 16 | 3681 nlist.n_desc); 3682 sym[sym_idx].Clear(); 3683 found_it = true; 3684 break; 3685 } 3686 } 3687 if (found_it) 3688 continue; 3689 } else { 3690 const char *gsym_name = 3691 sym[sym_idx] 3692 .GetMangled() 3693 .GetName(lldb::eLanguageTypeUnknown, 3694 Mangled::ePreferMangled) 3695 .GetCString(); 3696 if (gsym_name) { 3697 // Combine N_GSYM stab entries with the non stab 3698 // symbol 3699 ConstNameToSymbolIndexMap::const_iterator pos = 3700 N_GSYM_name_to_sym_idx.find(gsym_name); 3701 if (pos != N_GSYM_name_to_sym_idx.end()) { 3702 const uint32_t GSYM_sym_idx = pos->second; 3703 m_nlist_idx_to_sym_idx[nlist_idx] = 3704 GSYM_sym_idx; 3705 // Copy the address, because often the N_GSYM 3706 // address has an invalid address of zero 3707 // when the global is a common symbol 3708 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3709 symbol_section); 3710 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3711 symbol_value); 3712 // We just need the flags from the linker 3713 // symbol, so put these flags 3714 // into the N_GSYM flags to avoid duplicate 3715 // symbols in the symbol table 3716 sym[GSYM_sym_idx].SetFlags( 3717 nlist.n_type << 16 | nlist.n_desc); 3718 sym[sym_idx].Clear(); 3719 continue; 3720 } 3721 } 3722 } 3723 } 3724 } 3725 3726 sym[sym_idx].SetID(nlist_idx); 3727 sym[sym_idx].SetType(type); 3728 if (set_value) { 3729 sym[sym_idx].GetAddressRef().SetSection( 3730 symbol_section); 3731 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3732 } 3733 sym[sym_idx].SetFlags(nlist.n_type << 16 | 3734 nlist.n_desc); 3735 3736 if (symbol_byte_size > 0) 3737 sym[sym_idx].SetByteSize(symbol_byte_size); 3738 3739 if (demangled_is_synthesized) 3740 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3741 ++sym_idx; 3742 } else { 3743 sym[sym_idx].Clear(); 3744 } 3745 } 3746 ///////////////////////////// 3747 } 3748 break; // No more entries to consider 3749 } 3750 } 3751 3752 for (const auto &pos : reexport_shlib_needs_fixup) { 3753 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3754 if (undef_pos != undefined_name_to_desc.end()) { 3755 const uint8_t dylib_ordinal = 3756 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3757 if (dylib_ordinal > 0 && 3758 dylib_ordinal < dylib_files.GetSize()) 3759 sym[pos.first].SetReExportedSymbolSharedLibrary( 3760 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3761 } 3762 } 3763 } 3764 } 3765 } 3766 } 3767 } 3768 3769 // Must reset this in case it was mutated above! 3770 nlist_data_offset = 0; 3771 #endif 3772 3773 if (nlist_data.GetByteSize() > 0) { 3774 3775 // If the sym array was not created while parsing the DSC unmapped 3776 // symbols, create it now. 3777 if (sym == NULL) { 3778 sym = symtab->Resize(symtab_load_command.nsyms + 3779 m_dysymtab.nindirectsyms); 3780 num_syms = symtab->GetNumSymbols(); 3781 } 3782 3783 if (unmapped_local_symbols_found) { 3784 assert(m_dysymtab.ilocalsym == 0); 3785 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3786 nlist_idx = m_dysymtab.nlocalsym; 3787 } else { 3788 nlist_idx = 0; 3789 } 3790 3791 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3792 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3793 UndefinedNameToDescMap undefined_name_to_desc; 3794 SymbolIndexToName reexport_shlib_needs_fixup; 3795 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 3796 struct nlist_64 nlist; 3797 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, 3798 nlist_byte_size)) 3799 break; 3800 3801 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3802 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 3803 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 3804 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 3805 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 3806 3807 SymbolType type = eSymbolTypeInvalid; 3808 const char *symbol_name = NULL; 3809 3810 if (have_strtab_data) { 3811 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3812 3813 if (symbol_name == NULL) { 3814 // No symbol should be NULL, even the symbols with no 3815 // string values should have an offset zero which points 3816 // to an empty C-string 3817 Host::SystemLog(Host::eSystemLogError, 3818 "error: symbol[%u] has invalid string table offset " 3819 "0x%x in %s, ignoring symbol\n", 3820 nlist_idx, nlist.n_strx, 3821 module_sp->GetFileSpec().GetPath().c_str()); 3822 continue; 3823 } 3824 if (symbol_name[0] == '\0') 3825 symbol_name = NULL; 3826 } else { 3827 const addr_t str_addr = strtab_addr + nlist.n_strx; 3828 Error str_error; 3829 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3830 str_error)) 3831 symbol_name = memory_symbol_name.c_str(); 3832 } 3833 const char *symbol_name_non_abi_mangled = NULL; 3834 3835 SectionSP symbol_section; 3836 lldb::addr_t symbol_byte_size = 0; 3837 bool add_nlist = true; 3838 bool is_gsym = false; 3839 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3840 bool demangled_is_synthesized = false; 3841 bool set_value = true; 3842 assert(sym_idx < num_syms); 3843 3844 sym[sym_idx].SetDebug(is_debug); 3845 3846 if (is_debug) { 3847 switch (nlist.n_type) { 3848 case N_GSYM: 3849 // global symbol: name,,NO_SECT,type,0 3850 // Sometimes the N_GSYM value contains the address. 3851 3852 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3853 // the ObjC data. They 3854 // have the same address, but we want to ensure that we always find 3855 // only the real symbol, 3856 // 'cause we don't currently correctly attribute the GSYM one to the 3857 // ObjCClass/Ivar/MetaClass 3858 // symbol type. This is a temporary hack to make sure the 3859 // ObjectiveC symbols get treated 3860 // correctly. To do this right, we should coalesce all the GSYM & 3861 // global symbols that have the 3862 // same address. 3863 is_gsym = true; 3864 sym[sym_idx].SetExternal(true); 3865 3866 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3867 llvm::StringRef symbol_name_ref(symbol_name); 3868 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3869 symbol_name_non_abi_mangled = symbol_name + 1; 3870 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3871 type = eSymbolTypeObjCClass; 3872 demangled_is_synthesized = true; 3873 3874 } else if (symbol_name_ref.startswith( 3875 g_objc_v2_prefix_metaclass)) { 3876 symbol_name_non_abi_mangled = symbol_name + 1; 3877 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3878 type = eSymbolTypeObjCMetaClass; 3879 demangled_is_synthesized = true; 3880 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3881 symbol_name_non_abi_mangled = symbol_name + 1; 3882 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3883 type = eSymbolTypeObjCIVar; 3884 demangled_is_synthesized = true; 3885 } 3886 } else { 3887 if (nlist.n_value != 0) 3888 symbol_section = 3889 section_info.GetSection(nlist.n_sect, nlist.n_value); 3890 type = eSymbolTypeData; 3891 } 3892 break; 3893 3894 case N_FNAME: 3895 // procedure name (f77 kludge): name,,NO_SECT,0,0 3896 type = eSymbolTypeCompiler; 3897 break; 3898 3899 case N_FUN: 3900 // procedure: name,,n_sect,linenumber,address 3901 if (symbol_name) { 3902 type = eSymbolTypeCode; 3903 symbol_section = 3904 section_info.GetSection(nlist.n_sect, nlist.n_value); 3905 3906 N_FUN_addr_to_sym_idx.insert( 3907 std::make_pair(nlist.n_value, sym_idx)); 3908 // We use the current number of symbols in the symbol table in 3909 // lieu of 3910 // using nlist_idx in case we ever start trimming entries out 3911 N_FUN_indexes.push_back(sym_idx); 3912 } else { 3913 type = eSymbolTypeCompiler; 3914 3915 if (!N_FUN_indexes.empty()) { 3916 // Copy the size of the function into the original STAB entry so 3917 // we don't have 3918 // to hunt for it later 3919 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3920 ->SetByteSize(nlist.n_value); 3921 N_FUN_indexes.pop_back(); 3922 // We don't really need the end function STAB as it contains the 3923 // size which 3924 // we already placed with the original symbol, so don't add it 3925 // if we want a 3926 // minimal symbol table 3927 add_nlist = false; 3928 } 3929 } 3930 break; 3931 3932 case N_STSYM: 3933 // static symbol: name,,n_sect,type,address 3934 N_STSYM_addr_to_sym_idx.insert( 3935 std::make_pair(nlist.n_value, sym_idx)); 3936 symbol_section = 3937 section_info.GetSection(nlist.n_sect, nlist.n_value); 3938 if (symbol_name && symbol_name[0]) { 3939 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3940 eSymbolTypeData); 3941 } 3942 break; 3943 3944 case N_LCSYM: 3945 // .lcomm symbol: name,,n_sect,type,address 3946 symbol_section = 3947 section_info.GetSection(nlist.n_sect, nlist.n_value); 3948 type = eSymbolTypeCommonBlock; 3949 break; 3950 3951 case N_BNSYM: 3952 // We use the current number of symbols in the symbol table in lieu 3953 // of 3954 // using nlist_idx in case we ever start trimming entries out 3955 // Skip these if we want minimal symbol tables 3956 add_nlist = false; 3957 break; 3958 3959 case N_ENSYM: 3960 // Set the size of the N_BNSYM to the terminating index of this 3961 // N_ENSYM 3962 // so that we can always skip the entire symbol if we need to 3963 // navigate 3964 // more quickly at the source level when parsing STABS 3965 // Skip these if we want minimal symbol tables 3966 add_nlist = false; 3967 break; 3968 3969 case N_OPT: 3970 // emitted with gcc2_compiled and in gcc source 3971 type = eSymbolTypeCompiler; 3972 break; 3973 3974 case N_RSYM: 3975 // register sym: name,,NO_SECT,type,register 3976 type = eSymbolTypeVariable; 3977 break; 3978 3979 case N_SLINE: 3980 // src line: 0,,n_sect,linenumber,address 3981 symbol_section = 3982 section_info.GetSection(nlist.n_sect, nlist.n_value); 3983 type = eSymbolTypeLineEntry; 3984 break; 3985 3986 case N_SSYM: 3987 // structure elt: name,,NO_SECT,type,struct_offset 3988 type = eSymbolTypeVariableType; 3989 break; 3990 3991 case N_SO: 3992 // source file name 3993 type = eSymbolTypeSourceFile; 3994 if (symbol_name == NULL) { 3995 add_nlist = false; 3996 if (N_SO_index != UINT32_MAX) { 3997 // Set the size of the N_SO to the terminating index of this 3998 // N_SO 3999 // so that we can always skip the entire N_SO if we need to 4000 // navigate 4001 // more quickly at the source level when parsing STABS 4002 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 4003 symbol_ptr->SetByteSize(sym_idx); 4004 symbol_ptr->SetSizeIsSibling(true); 4005 } 4006 N_NSYM_indexes.clear(); 4007 N_INCL_indexes.clear(); 4008 N_BRAC_indexes.clear(); 4009 N_COMM_indexes.clear(); 4010 N_FUN_indexes.clear(); 4011 N_SO_index = UINT32_MAX; 4012 } else { 4013 // We use the current number of symbols in the symbol table in 4014 // lieu of 4015 // using nlist_idx in case we ever start trimming entries out 4016 const bool N_SO_has_full_path = symbol_name[0] == '/'; 4017 if (N_SO_has_full_path) { 4018 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 4019 // We have two consecutive N_SO entries where the first 4020 // contains a directory 4021 // and the second contains a full path. 4022 sym[sym_idx - 1].GetMangled().SetValue( 4023 ConstString(symbol_name), false); 4024 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4025 add_nlist = false; 4026 } else { 4027 // This is the first entry in a N_SO that contains a directory 4028 // or 4029 // a full path to the source file 4030 N_SO_index = sym_idx; 4031 } 4032 } else if ((N_SO_index == sym_idx - 1) && 4033 ((sym_idx - 1) < num_syms)) { 4034 // This is usually the second N_SO entry that contains just the 4035 // filename, 4036 // so here we combine it with the first one if we are minimizing 4037 // the symbol table 4038 const char *so_path = 4039 sym[sym_idx - 1] 4040 .GetMangled() 4041 .GetDemangledName(lldb::eLanguageTypeUnknown) 4042 .AsCString(); 4043 if (so_path && so_path[0]) { 4044 std::string full_so_path(so_path); 4045 const size_t double_slash_pos = full_so_path.find("//"); 4046 if (double_slash_pos != std::string::npos) { 4047 // The linker has been generating bad N_SO entries with 4048 // doubled up paths 4049 // in the format "%s%s" where the first string in the 4050 // DW_AT_comp_dir, 4051 // and the second is the directory for the source file so 4052 // you end up with 4053 // a path that looks like "/tmp/src//tmp/src/" 4054 FileSpec so_dir(so_path, false); 4055 if (!so_dir.Exists()) { 4056 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 4057 false); 4058 if (so_dir.Exists()) { 4059 // Trim off the incorrect path 4060 full_so_path.erase(0, double_slash_pos + 1); 4061 } 4062 } 4063 } 4064 if (*full_so_path.rbegin() != '/') 4065 full_so_path += '/'; 4066 full_so_path += symbol_name; 4067 sym[sym_idx - 1].GetMangled().SetValue( 4068 ConstString(full_so_path.c_str()), false); 4069 add_nlist = false; 4070 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4071 } 4072 } else { 4073 // This could be a relative path to a N_SO 4074 N_SO_index = sym_idx; 4075 } 4076 } 4077 break; 4078 4079 case N_OSO: 4080 // object file name: name,,0,0,st_mtime 4081 type = eSymbolTypeObjectFile; 4082 break; 4083 4084 case N_LSYM: 4085 // local sym: name,,NO_SECT,type,offset 4086 type = eSymbolTypeLocal; 4087 break; 4088 4089 //---------------------------------------------------------------------- 4090 // INCL scopes 4091 //---------------------------------------------------------------------- 4092 case N_BINCL: 4093 // include file beginning: name,,NO_SECT,0,sum 4094 // We use the current number of symbols in the symbol table in lieu 4095 // of 4096 // using nlist_idx in case we ever start trimming entries out 4097 N_INCL_indexes.push_back(sym_idx); 4098 type = eSymbolTypeScopeBegin; 4099 break; 4100 4101 case N_EINCL: 4102 // include file end: name,,NO_SECT,0,0 4103 // Set the size of the N_BINCL to the terminating index of this 4104 // N_EINCL 4105 // so that we can always skip the entire symbol if we need to 4106 // navigate 4107 // more quickly at the source level when parsing STABS 4108 if (!N_INCL_indexes.empty()) { 4109 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4110 symbol_ptr->SetByteSize(sym_idx + 1); 4111 symbol_ptr->SetSizeIsSibling(true); 4112 N_INCL_indexes.pop_back(); 4113 } 4114 type = eSymbolTypeScopeEnd; 4115 break; 4116 4117 case N_SOL: 4118 // #included file name: name,,n_sect,0,address 4119 type = eSymbolTypeHeaderFile; 4120 4121 // We currently don't use the header files on darwin 4122 add_nlist = false; 4123 break; 4124 4125 case N_PARAMS: 4126 // compiler parameters: name,,NO_SECT,0,0 4127 type = eSymbolTypeCompiler; 4128 break; 4129 4130 case N_VERSION: 4131 // compiler version: name,,NO_SECT,0,0 4132 type = eSymbolTypeCompiler; 4133 break; 4134 4135 case N_OLEVEL: 4136 // compiler -O level: name,,NO_SECT,0,0 4137 type = eSymbolTypeCompiler; 4138 break; 4139 4140 case N_PSYM: 4141 // parameter: name,,NO_SECT,type,offset 4142 type = eSymbolTypeVariable; 4143 break; 4144 4145 case N_ENTRY: 4146 // alternate entry: name,,n_sect,linenumber,address 4147 symbol_section = 4148 section_info.GetSection(nlist.n_sect, nlist.n_value); 4149 type = eSymbolTypeLineEntry; 4150 break; 4151 4152 //---------------------------------------------------------------------- 4153 // Left and Right Braces 4154 //---------------------------------------------------------------------- 4155 case N_LBRAC: 4156 // left bracket: 0,,NO_SECT,nesting level,address 4157 // We use the current number of symbols in the symbol table in lieu 4158 // of 4159 // using nlist_idx in case we ever start trimming entries out 4160 symbol_section = 4161 section_info.GetSection(nlist.n_sect, nlist.n_value); 4162 N_BRAC_indexes.push_back(sym_idx); 4163 type = eSymbolTypeScopeBegin; 4164 break; 4165 4166 case N_RBRAC: 4167 // right bracket: 0,,NO_SECT,nesting level,address 4168 // Set the size of the N_LBRAC to the terminating index of this 4169 // N_RBRAC 4170 // so that we can always skip the entire symbol if we need to 4171 // navigate 4172 // more quickly at the source level when parsing STABS 4173 symbol_section = 4174 section_info.GetSection(nlist.n_sect, nlist.n_value); 4175 if (!N_BRAC_indexes.empty()) { 4176 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4177 symbol_ptr->SetByteSize(sym_idx + 1); 4178 symbol_ptr->SetSizeIsSibling(true); 4179 N_BRAC_indexes.pop_back(); 4180 } 4181 type = eSymbolTypeScopeEnd; 4182 break; 4183 4184 case N_EXCL: 4185 // deleted include file: name,,NO_SECT,0,sum 4186 type = eSymbolTypeHeaderFile; 4187 break; 4188 4189 //---------------------------------------------------------------------- 4190 // COMM scopes 4191 //---------------------------------------------------------------------- 4192 case N_BCOMM: 4193 // begin common: name,,NO_SECT,0,0 4194 // We use the current number of symbols in the symbol table in lieu 4195 // of 4196 // using nlist_idx in case we ever start trimming entries out 4197 type = eSymbolTypeScopeBegin; 4198 N_COMM_indexes.push_back(sym_idx); 4199 break; 4200 4201 case N_ECOML: 4202 // end common (local name): 0,,n_sect,0,address 4203 symbol_section = 4204 section_info.GetSection(nlist.n_sect, nlist.n_value); 4205 LLVM_FALLTHROUGH; 4206 4207 case N_ECOMM: 4208 // end common: name,,n_sect,0,0 4209 // Set the size of the N_BCOMM to the terminating index of this 4210 // N_ECOMM/N_ECOML 4211 // so that we can always skip the entire symbol if we need to 4212 // navigate 4213 // more quickly at the source level when parsing STABS 4214 if (!N_COMM_indexes.empty()) { 4215 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4216 symbol_ptr->SetByteSize(sym_idx + 1); 4217 symbol_ptr->SetSizeIsSibling(true); 4218 N_COMM_indexes.pop_back(); 4219 } 4220 type = eSymbolTypeScopeEnd; 4221 break; 4222 4223 case N_LENG: 4224 // second stab entry with length information 4225 type = eSymbolTypeAdditional; 4226 break; 4227 4228 default: 4229 break; 4230 } 4231 } else { 4232 // uint8_t n_pext = N_PEXT & nlist.n_type; 4233 uint8_t n_type = N_TYPE & nlist.n_type; 4234 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4235 4236 switch (n_type) { 4237 case N_INDR: { 4238 const char *reexport_name_cstr = 4239 strtab_data.PeekCStr(nlist.n_value); 4240 if (reexport_name_cstr && reexport_name_cstr[0]) { 4241 type = eSymbolTypeReExported; 4242 ConstString reexport_name( 4243 reexport_name_cstr + 4244 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4245 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4246 set_value = false; 4247 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4248 indirect_symbol_names.insert( 4249 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4250 } else 4251 type = eSymbolTypeUndefined; 4252 } break; 4253 4254 case N_UNDF: 4255 if (symbol_name && symbol_name[0]) { 4256 ConstString undefined_name(symbol_name + 4257 ((symbol_name[0] == '_') ? 1 : 0)); 4258 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4259 } 4260 LLVM_FALLTHROUGH; 4261 4262 case N_PBUD: 4263 type = eSymbolTypeUndefined; 4264 break; 4265 4266 case N_ABS: 4267 type = eSymbolTypeAbsolute; 4268 break; 4269 4270 case N_SECT: { 4271 symbol_section = 4272 section_info.GetSection(nlist.n_sect, nlist.n_value); 4273 4274 if (!symbol_section) { 4275 // TODO: warn about this? 4276 add_nlist = false; 4277 break; 4278 } 4279 4280 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4281 type = eSymbolTypeException; 4282 } else { 4283 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4284 4285 switch (section_type) { 4286 case S_CSTRING_LITERALS: 4287 type = eSymbolTypeData; 4288 break; // section with only literal C strings 4289 case S_4BYTE_LITERALS: 4290 type = eSymbolTypeData; 4291 break; // section with only 4 byte literals 4292 case S_8BYTE_LITERALS: 4293 type = eSymbolTypeData; 4294 break; // section with only 8 byte literals 4295 case S_LITERAL_POINTERS: 4296 type = eSymbolTypeTrampoline; 4297 break; // section with only pointers to literals 4298 case S_NON_LAZY_SYMBOL_POINTERS: 4299 type = eSymbolTypeTrampoline; 4300 break; // section with only non-lazy symbol pointers 4301 case S_LAZY_SYMBOL_POINTERS: 4302 type = eSymbolTypeTrampoline; 4303 break; // section with only lazy symbol pointers 4304 case S_SYMBOL_STUBS: 4305 type = eSymbolTypeTrampoline; 4306 break; // section with only symbol stubs, byte size of stub in 4307 // the reserved2 field 4308 case S_MOD_INIT_FUNC_POINTERS: 4309 type = eSymbolTypeCode; 4310 break; // section with only function pointers for initialization 4311 case S_MOD_TERM_FUNC_POINTERS: 4312 type = eSymbolTypeCode; 4313 break; // section with only function pointers for termination 4314 case S_INTERPOSING: 4315 type = eSymbolTypeTrampoline; 4316 break; // section with only pairs of function pointers for 4317 // interposing 4318 case S_16BYTE_LITERALS: 4319 type = eSymbolTypeData; 4320 break; // section with only 16 byte literals 4321 case S_DTRACE_DOF: 4322 type = eSymbolTypeInstrumentation; 4323 break; 4324 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4325 type = eSymbolTypeTrampoline; 4326 break; 4327 default: 4328 switch (symbol_section->GetType()) { 4329 case lldb::eSectionTypeCode: 4330 type = eSymbolTypeCode; 4331 break; 4332 case eSectionTypeData: 4333 case eSectionTypeDataCString: // Inlined C string data 4334 case eSectionTypeDataCStringPointers: // Pointers to C string 4335 // data 4336 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4337 // the symbol table 4338 case eSectionTypeData4: 4339 case eSectionTypeData8: 4340 case eSectionTypeData16: 4341 type = eSymbolTypeData; 4342 break; 4343 default: 4344 break; 4345 } 4346 break; 4347 } 4348 4349 if (type == eSymbolTypeInvalid) { 4350 const char *symbol_sect_name = 4351 symbol_section->GetName().AsCString(); 4352 if (symbol_section->IsDescendant(text_section_sp.get())) { 4353 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4354 S_ATTR_SELF_MODIFYING_CODE | 4355 S_ATTR_SOME_INSTRUCTIONS)) 4356 type = eSymbolTypeData; 4357 else 4358 type = eSymbolTypeCode; 4359 } else if (symbol_section->IsDescendant( 4360 data_section_sp.get()) || 4361 symbol_section->IsDescendant( 4362 data_dirty_section_sp.get()) || 4363 symbol_section->IsDescendant( 4364 data_const_section_sp.get())) { 4365 if (symbol_sect_name && 4366 ::strstr(symbol_sect_name, "__objc") == 4367 symbol_sect_name) { 4368 type = eSymbolTypeRuntime; 4369 4370 if (symbol_name) { 4371 llvm::StringRef symbol_name_ref(symbol_name); 4372 if (symbol_name_ref.startswith("_OBJC_")) { 4373 static const llvm::StringRef g_objc_v2_prefix_class( 4374 "_OBJC_CLASS_$_"); 4375 static const llvm::StringRef g_objc_v2_prefix_metaclass( 4376 "_OBJC_METACLASS_$_"); 4377 static const llvm::StringRef g_objc_v2_prefix_ivar( 4378 "_OBJC_IVAR_$_"); 4379 if (symbol_name_ref.startswith( 4380 g_objc_v2_prefix_class)) { 4381 symbol_name_non_abi_mangled = symbol_name + 1; 4382 symbol_name = 4383 symbol_name + g_objc_v2_prefix_class.size(); 4384 type = eSymbolTypeObjCClass; 4385 demangled_is_synthesized = true; 4386 } else if (symbol_name_ref.startswith( 4387 g_objc_v2_prefix_metaclass)) { 4388 symbol_name_non_abi_mangled = symbol_name + 1; 4389 symbol_name = 4390 symbol_name + g_objc_v2_prefix_metaclass.size(); 4391 type = eSymbolTypeObjCMetaClass; 4392 demangled_is_synthesized = true; 4393 } else if (symbol_name_ref.startswith( 4394 g_objc_v2_prefix_ivar)) { 4395 symbol_name_non_abi_mangled = symbol_name + 1; 4396 symbol_name = 4397 symbol_name + g_objc_v2_prefix_ivar.size(); 4398 type = eSymbolTypeObjCIVar; 4399 demangled_is_synthesized = true; 4400 } 4401 } 4402 } 4403 } else if (symbol_sect_name && 4404 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4405 symbol_sect_name) { 4406 type = eSymbolTypeException; 4407 } else { 4408 type = eSymbolTypeData; 4409 } 4410 } else if (symbol_sect_name && 4411 ::strstr(symbol_sect_name, "__IMPORT") == 4412 symbol_sect_name) { 4413 type = eSymbolTypeTrampoline; 4414 } else if (symbol_section->IsDescendant( 4415 objc_section_sp.get())) { 4416 type = eSymbolTypeRuntime; 4417 if (symbol_name && symbol_name[0] == '.') { 4418 llvm::StringRef symbol_name_ref(symbol_name); 4419 static const llvm::StringRef g_objc_v1_prefix_class( 4420 ".objc_class_name_"); 4421 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4422 symbol_name_non_abi_mangled = symbol_name; 4423 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4424 type = eSymbolTypeObjCClass; 4425 demangled_is_synthesized = true; 4426 } 4427 } 4428 } 4429 } 4430 } 4431 } break; 4432 } 4433 } 4434 4435 if (add_nlist) { 4436 uint64_t symbol_value = nlist.n_value; 4437 4438 if (symbol_name_non_abi_mangled) { 4439 sym[sym_idx].GetMangled().SetMangledName( 4440 ConstString(symbol_name_non_abi_mangled)); 4441 sym[sym_idx].GetMangled().SetDemangledName( 4442 ConstString(symbol_name)); 4443 } else { 4444 bool symbol_name_is_mangled = false; 4445 4446 if (symbol_name && symbol_name[0] == '_') { 4447 symbol_name_is_mangled = symbol_name[1] == '_'; 4448 symbol_name++; // Skip the leading underscore 4449 } 4450 4451 if (symbol_name) { 4452 ConstString const_symbol_name(symbol_name); 4453 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4454 symbol_name_is_mangled); 4455 } 4456 } 4457 4458 if (is_gsym) { 4459 const char *gsym_name = sym[sym_idx] 4460 .GetMangled() 4461 .GetName(lldb::eLanguageTypeUnknown, 4462 Mangled::ePreferMangled) 4463 .GetCString(); 4464 if (gsym_name) 4465 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4466 } 4467 4468 if (symbol_section) { 4469 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4470 if (symbol_byte_size == 0 && function_starts_count > 0) { 4471 addr_t symbol_lookup_file_addr = nlist.n_value; 4472 // Do an exact address match for non-ARM addresses, else get the 4473 // closest since 4474 // the symbol might be a thumb symbol which has an address with 4475 // bit zero set 4476 FunctionStarts::Entry *func_start_entry = 4477 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4478 if (is_arm && func_start_entry) { 4479 // Verify that the function start address is the symbol address 4480 // (ARM) 4481 // or the symbol address + 1 (thumb) 4482 if (func_start_entry->addr != symbol_lookup_file_addr && 4483 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4484 // Not the right entry, NULL it out... 4485 func_start_entry = NULL; 4486 } 4487 } 4488 if (func_start_entry) { 4489 func_start_entry->data = true; 4490 4491 addr_t symbol_file_addr = func_start_entry->addr; 4492 if (is_arm) 4493 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4494 4495 const FunctionStarts::Entry *next_func_start_entry = 4496 function_starts.FindNextEntry(func_start_entry); 4497 const addr_t section_end_file_addr = 4498 section_file_addr + symbol_section->GetByteSize(); 4499 if (next_func_start_entry) { 4500 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4501 // Be sure the clear the Thumb address bit when we calculate 4502 // the size 4503 // from the current and next address 4504 if (is_arm) 4505 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4506 symbol_byte_size = std::min<lldb::addr_t>( 4507 next_symbol_file_addr - symbol_file_addr, 4508 section_end_file_addr - symbol_file_addr); 4509 } else { 4510 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4511 } 4512 } 4513 } 4514 symbol_value -= section_file_addr; 4515 } 4516 4517 if (is_debug == false) { 4518 if (type == eSymbolTypeCode) { 4519 // See if we can find a N_FUN entry for any code symbols. 4520 // If we do find a match, and the name matches, then we 4521 // can merge the two into just the function symbol to avoid 4522 // duplicate entries in the symbol table 4523 std::pair<ValueToSymbolIndexMap::const_iterator, 4524 ValueToSymbolIndexMap::const_iterator> 4525 range; 4526 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4527 if (range.first != range.second) { 4528 bool found_it = false; 4529 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4530 pos != range.second; ++pos) { 4531 if (sym[sym_idx].GetMangled().GetName( 4532 lldb::eLanguageTypeUnknown, 4533 Mangled::ePreferMangled) == 4534 sym[pos->second].GetMangled().GetName( 4535 lldb::eLanguageTypeUnknown, 4536 Mangled::ePreferMangled)) { 4537 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4538 // We just need the flags from the linker symbol, so put 4539 // these flags 4540 // into the N_FUN flags to avoid duplicate symbols in the 4541 // symbol table 4542 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4543 sym[pos->second].SetFlags(nlist.n_type << 16 | 4544 nlist.n_desc); 4545 if (resolver_addresses.find(nlist.n_value) != 4546 resolver_addresses.end()) 4547 sym[pos->second].SetType(eSymbolTypeResolver); 4548 sym[sym_idx].Clear(); 4549 found_it = true; 4550 break; 4551 } 4552 } 4553 if (found_it) 4554 continue; 4555 } else { 4556 if (resolver_addresses.find(nlist.n_value) != 4557 resolver_addresses.end()) 4558 type = eSymbolTypeResolver; 4559 } 4560 } else if (type == eSymbolTypeData || 4561 type == eSymbolTypeObjCClass || 4562 type == eSymbolTypeObjCMetaClass || 4563 type == eSymbolTypeObjCIVar) { 4564 // See if we can find a N_STSYM entry for any data symbols. 4565 // If we do find a match, and the name matches, then we 4566 // can merge the two into just the Static symbol to avoid 4567 // duplicate entries in the symbol table 4568 std::pair<ValueToSymbolIndexMap::const_iterator, 4569 ValueToSymbolIndexMap::const_iterator> 4570 range; 4571 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4572 if (range.first != range.second) { 4573 bool found_it = false; 4574 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4575 pos != range.second; ++pos) { 4576 if (sym[sym_idx].GetMangled().GetName( 4577 lldb::eLanguageTypeUnknown, 4578 Mangled::ePreferMangled) == 4579 sym[pos->second].GetMangled().GetName( 4580 lldb::eLanguageTypeUnknown, 4581 Mangled::ePreferMangled)) { 4582 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4583 // We just need the flags from the linker symbol, so put 4584 // these flags 4585 // into the N_STSYM flags to avoid duplicate symbols in the 4586 // symbol table 4587 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4588 sym[pos->second].SetFlags(nlist.n_type << 16 | 4589 nlist.n_desc); 4590 sym[sym_idx].Clear(); 4591 found_it = true; 4592 break; 4593 } 4594 } 4595 if (found_it) 4596 continue; 4597 } else { 4598 // Combine N_GSYM stab entries with the non stab symbol 4599 const char *gsym_name = sym[sym_idx] 4600 .GetMangled() 4601 .GetName(lldb::eLanguageTypeUnknown, 4602 Mangled::ePreferMangled) 4603 .GetCString(); 4604 if (gsym_name) { 4605 ConstNameToSymbolIndexMap::const_iterator pos = 4606 N_GSYM_name_to_sym_idx.find(gsym_name); 4607 if (pos != N_GSYM_name_to_sym_idx.end()) { 4608 const uint32_t GSYM_sym_idx = pos->second; 4609 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4610 // Copy the address, because often the N_GSYM address has an 4611 // invalid address of zero 4612 // when the global is a common symbol 4613 sym[GSYM_sym_idx].GetAddressRef().SetSection( 4614 symbol_section); 4615 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4616 // We just need the flags from the linker symbol, so put 4617 // these flags 4618 // into the N_GSYM flags to avoid duplicate symbols in the 4619 // symbol table 4620 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 4621 nlist.n_desc); 4622 sym[sym_idx].Clear(); 4623 continue; 4624 } 4625 } 4626 } 4627 } 4628 } 4629 4630 sym[sym_idx].SetID(nlist_idx); 4631 sym[sym_idx].SetType(type); 4632 if (set_value) { 4633 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4634 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4635 } 4636 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4637 4638 if (symbol_byte_size > 0) 4639 sym[sym_idx].SetByteSize(symbol_byte_size); 4640 4641 if (demangled_is_synthesized) 4642 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4643 4644 ++sym_idx; 4645 } else { 4646 sym[sym_idx].Clear(); 4647 } 4648 } 4649 4650 for (const auto &pos : reexport_shlib_needs_fixup) { 4651 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4652 if (undef_pos != undefined_name_to_desc.end()) { 4653 const uint8_t dylib_ordinal = 4654 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4655 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4656 sym[pos.first].SetReExportedSymbolSharedLibrary( 4657 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4658 } 4659 } 4660 } 4661 4662 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4663 4664 if (function_starts_count > 0) { 4665 uint32_t num_synthetic_function_symbols = 0; 4666 for (i = 0; i < function_starts_count; ++i) { 4667 if (function_starts.GetEntryRef(i).data == false) 4668 ++num_synthetic_function_symbols; 4669 } 4670 4671 if (num_synthetic_function_symbols > 0) { 4672 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4673 num_syms = sym_idx + num_synthetic_function_symbols; 4674 sym = symtab->Resize(num_syms); 4675 } 4676 for (i = 0; i < function_starts_count; ++i) { 4677 const FunctionStarts::Entry *func_start_entry = 4678 function_starts.GetEntryAtIndex(i); 4679 if (func_start_entry->data == false) { 4680 addr_t symbol_file_addr = func_start_entry->addr; 4681 uint32_t symbol_flags = 0; 4682 if (is_arm) { 4683 if (symbol_file_addr & 1) 4684 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4685 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4686 } 4687 Address symbol_addr; 4688 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4689 SectionSP symbol_section(symbol_addr.GetSection()); 4690 uint32_t symbol_byte_size = 0; 4691 if (symbol_section) { 4692 const addr_t section_file_addr = 4693 symbol_section->GetFileAddress(); 4694 const FunctionStarts::Entry *next_func_start_entry = 4695 function_starts.FindNextEntry(func_start_entry); 4696 const addr_t section_end_file_addr = 4697 section_file_addr + symbol_section->GetByteSize(); 4698 if (next_func_start_entry) { 4699 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4700 if (is_arm) 4701 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4702 symbol_byte_size = std::min<lldb::addr_t>( 4703 next_symbol_file_addr - symbol_file_addr, 4704 section_end_file_addr - symbol_file_addr); 4705 } else { 4706 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4707 } 4708 sym[sym_idx].SetID(synthetic_sym_id++); 4709 sym[sym_idx].GetMangled().SetDemangledName( 4710 GetNextSyntheticSymbolName()); 4711 sym[sym_idx].SetType(eSymbolTypeCode); 4712 sym[sym_idx].SetIsSynthetic(true); 4713 sym[sym_idx].GetAddressRef() = symbol_addr; 4714 if (symbol_flags) 4715 sym[sym_idx].SetFlags(symbol_flags); 4716 if (symbol_byte_size) 4717 sym[sym_idx].SetByteSize(symbol_byte_size); 4718 ++sym_idx; 4719 } 4720 } 4721 } 4722 } 4723 } 4724 } 4725 4726 // Trim our symbols down to just what we ended up with after 4727 // removing any symbols. 4728 if (sym_idx < num_syms) { 4729 num_syms = sym_idx; 4730 sym = symtab->Resize(num_syms); 4731 } 4732 4733 // Now synthesize indirect symbols 4734 if (m_dysymtab.nindirectsyms != 0) { 4735 if (indirect_symbol_index_data.GetByteSize()) { 4736 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4737 m_nlist_idx_to_sym_idx.end(); 4738 4739 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4740 ++sect_idx) { 4741 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4742 S_SYMBOL_STUBS) { 4743 uint32_t symbol_stub_byte_size = 4744 m_mach_sections[sect_idx].reserved2; 4745 if (symbol_stub_byte_size == 0) 4746 continue; 4747 4748 const uint32_t num_symbol_stubs = 4749 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4750 4751 if (num_symbol_stubs == 0) 4752 continue; 4753 4754 const uint32_t symbol_stub_index_offset = 4755 m_mach_sections[sect_idx].reserved1; 4756 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; 4757 ++stub_idx) { 4758 const uint32_t symbol_stub_index = 4759 symbol_stub_index_offset + stub_idx; 4760 const lldb::addr_t symbol_stub_addr = 4761 m_mach_sections[sect_idx].addr + 4762 (stub_idx * symbol_stub_byte_size); 4763 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4764 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4765 symbol_stub_offset, 4)) { 4766 const uint32_t stub_sym_id = 4767 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4768 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4769 continue; 4770 4771 NListIndexToSymbolIndexMap::const_iterator index_pos = 4772 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4773 Symbol *stub_symbol = NULL; 4774 if (index_pos != end_index_pos) { 4775 // We have a remapping from the original nlist index to 4776 // a current symbol index, so just look this up by index 4777 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4778 } else { 4779 // We need to lookup a symbol using the original nlist 4780 // symbol index since this index is coming from the 4781 // S_SYMBOL_STUBS 4782 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4783 } 4784 4785 if (stub_symbol) { 4786 Address so_addr(symbol_stub_addr, section_list); 4787 4788 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4789 // Change the external symbol into a trampoline that makes 4790 // sense 4791 // These symbols were N_UNDF N_EXT, and are useless to us, 4792 // so we 4793 // can re-use them so we don't have to make up a synthetic 4794 // symbol 4795 // for no good reason. 4796 if (resolver_addresses.find(symbol_stub_addr) == 4797 resolver_addresses.end()) 4798 stub_symbol->SetType(eSymbolTypeTrampoline); 4799 else 4800 stub_symbol->SetType(eSymbolTypeResolver); 4801 stub_symbol->SetExternal(false); 4802 stub_symbol->GetAddressRef() = so_addr; 4803 stub_symbol->SetByteSize(symbol_stub_byte_size); 4804 } else { 4805 // Make a synthetic symbol to describe the trampoline stub 4806 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4807 if (sym_idx >= num_syms) { 4808 sym = symtab->Resize(++num_syms); 4809 stub_symbol = NULL; // this pointer no longer valid 4810 } 4811 sym[sym_idx].SetID(synthetic_sym_id++); 4812 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4813 if (resolver_addresses.find(symbol_stub_addr) == 4814 resolver_addresses.end()) 4815 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4816 else 4817 sym[sym_idx].SetType(eSymbolTypeResolver); 4818 sym[sym_idx].SetIsSynthetic(true); 4819 sym[sym_idx].GetAddressRef() = so_addr; 4820 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4821 ++sym_idx; 4822 } 4823 } else { 4824 if (log) 4825 log->Warning("symbol stub referencing symbol table symbol " 4826 "%u that isn't in our minimal symbol table, " 4827 "fix this!!!", 4828 stub_sym_id); 4829 } 4830 } 4831 } 4832 } 4833 } 4834 } 4835 } 4836 4837 if (!trie_entries.empty()) { 4838 for (const auto &e : trie_entries) { 4839 if (e.entry.import_name) { 4840 // Only add indirect symbols from the Trie entries if we 4841 // didn't have a N_INDR nlist entry for this already 4842 if (indirect_symbol_names.find(e.entry.name) == 4843 indirect_symbol_names.end()) { 4844 // Make a synthetic symbol to describe re-exported symbol. 4845 if (sym_idx >= num_syms) 4846 sym = symtab->Resize(++num_syms); 4847 sym[sym_idx].SetID(synthetic_sym_id++); 4848 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4849 sym[sym_idx].SetType(eSymbolTypeReExported); 4850 sym[sym_idx].SetIsSynthetic(true); 4851 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4852 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4853 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4854 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4855 } 4856 ++sym_idx; 4857 } 4858 } 4859 } 4860 } 4861 4862 // StreamFile s(stdout, false); 4863 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4864 // symtab->Dump(&s, NULL, eSortOrderNone); 4865 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4866 // sizes 4867 symtab->CalculateSymbolSizes(); 4868 4869 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4870 // symtab->Dump(&s, NULL, eSortOrderNone); 4871 4872 return symtab->GetNumSymbols(); 4873 } 4874 return 0; 4875 } 4876 4877 void ObjectFileMachO::Dump(Stream *s) { 4878 ModuleSP module_sp(GetModule()); 4879 if (module_sp) { 4880 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4881 s->Printf("%p: ", static_cast<void *>(this)); 4882 s->Indent(); 4883 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4884 s->PutCString("ObjectFileMachO64"); 4885 else 4886 s->PutCString("ObjectFileMachO32"); 4887 4888 ArchSpec header_arch; 4889 GetArchitecture(header_arch); 4890 4891 *s << ", file = '" << m_file 4892 << "', arch = " << header_arch.GetArchitectureName() << "\n"; 4893 4894 SectionList *sections = GetSectionList(); 4895 if (sections) 4896 sections->Dump(s, NULL, true, UINT32_MAX); 4897 4898 if (m_symtab_ap.get()) 4899 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4900 } 4901 } 4902 4903 bool ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4904 const lldb_private::DataExtractor &data, 4905 lldb::offset_t lc_offset, 4906 lldb_private::UUID &uuid) { 4907 uint32_t i; 4908 struct uuid_command load_cmd; 4909 4910 lldb::offset_t offset = lc_offset; 4911 for (i = 0; i < header.ncmds; ++i) { 4912 const lldb::offset_t cmd_offset = offset; 4913 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4914 break; 4915 4916 if (load_cmd.cmd == LC_UUID) { 4917 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4918 4919 if (uuid_bytes) { 4920 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4921 // We pretend these object files have no UUID to prevent crashing. 4922 4923 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4924 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4925 0xbb, 0x14, 0xf0, 0x0d}; 4926 4927 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4928 return false; 4929 4930 uuid.SetBytes(uuid_bytes); 4931 return true; 4932 } 4933 return false; 4934 } 4935 offset = cmd_offset + load_cmd.cmdsize; 4936 } 4937 return false; 4938 } 4939 4940 bool ObjectFileMachO::GetArchitecture(const llvm::MachO::mach_header &header, 4941 const lldb_private::DataExtractor &data, 4942 lldb::offset_t lc_offset, 4943 ArchSpec &arch) { 4944 arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4945 4946 if (arch.IsValid()) { 4947 llvm::Triple &triple = arch.GetTriple(); 4948 4949 // Set OS to an unspecified unknown or a "*" so it can match any OS 4950 triple.setOS(llvm::Triple::UnknownOS); 4951 triple.setOSName(llvm::StringRef()); 4952 4953 if (header.filetype == MH_PRELOAD) { 4954 if (header.cputype == CPU_TYPE_ARM) { 4955 // If this is a 32-bit arm binary, and it's a standalone binary, 4956 // force the Vendor to Apple so we don't accidentally pick up 4957 // the generic armv7 ABI at runtime. Apple's armv7 ABI always uses 4958 // r7 for the frame pointer register; most other armv7 ABIs use a 4959 // combination of r7 and r11. 4960 triple.setVendor(llvm::Triple::Apple); 4961 } else { 4962 // Set vendor to an unspecified unknown or a "*" so it can match any 4963 // vendor 4964 // This is required for correct behavior of EFI debugging on x86_64 4965 triple.setVendor(llvm::Triple::UnknownVendor); 4966 triple.setVendorName(llvm::StringRef()); 4967 } 4968 return true; 4969 } else { 4970 struct load_command load_cmd; 4971 4972 lldb::offset_t offset = lc_offset; 4973 for (uint32_t i = 0; i < header.ncmds; ++i) { 4974 const lldb::offset_t cmd_offset = offset; 4975 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4976 break; 4977 4978 switch (load_cmd.cmd) { 4979 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4980 triple.setOS(llvm::Triple::IOS); 4981 return true; 4982 4983 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4984 triple.setOS(llvm::Triple::MacOSX); 4985 return true; 4986 4987 case llvm::MachO::LC_VERSION_MIN_TVOS: 4988 triple.setOS(llvm::Triple::TvOS); 4989 return true; 4990 4991 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4992 triple.setOS(llvm::Triple::WatchOS); 4993 return true; 4994 4995 default: 4996 break; 4997 } 4998 4999 offset = cmd_offset + load_cmd.cmdsize; 5000 } 5001 5002 if (header.filetype != MH_KEXT_BUNDLE) { 5003 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 5004 // so lets not say our Vendor is Apple, leave it as an unspecified 5005 // unknown 5006 triple.setVendor(llvm::Triple::UnknownVendor); 5007 triple.setVendorName(llvm::StringRef()); 5008 } 5009 } 5010 } 5011 return arch.IsValid(); 5012 } 5013 5014 bool ObjectFileMachO::GetUUID(lldb_private::UUID *uuid) { 5015 ModuleSP module_sp(GetModule()); 5016 if (module_sp) { 5017 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5018 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5019 return GetUUID(m_header, m_data, offset, *uuid); 5020 } 5021 return false; 5022 } 5023 5024 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5025 uint32_t count = 0; 5026 ModuleSP module_sp(GetModule()); 5027 if (module_sp) { 5028 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5029 struct load_command load_cmd; 5030 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5031 std::vector<std::string> rpath_paths; 5032 std::vector<std::string> rpath_relative_paths; 5033 const bool resolve_path = false; // Don't resolve the dependent file paths 5034 // since they may not reside on this system 5035 uint32_t i; 5036 for (i = 0; i < m_header.ncmds; ++i) { 5037 const uint32_t cmd_offset = offset; 5038 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5039 break; 5040 5041 switch (load_cmd.cmd) { 5042 case LC_RPATH: 5043 case LC_LOAD_DYLIB: 5044 case LC_LOAD_WEAK_DYLIB: 5045 case LC_REEXPORT_DYLIB: 5046 case LC_LOAD_DYLINKER: 5047 case LC_LOADFVMLIB: 5048 case LC_LOAD_UPWARD_DYLIB: { 5049 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5050 const char *path = m_data.PeekCStr(name_offset); 5051 if (path) { 5052 if (load_cmd.cmd == LC_RPATH) 5053 rpath_paths.push_back(path); 5054 else { 5055 if (path[0] == '@') { 5056 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5057 rpath_relative_paths.push_back(path + strlen("@rpath")); 5058 } else { 5059 FileSpec file_spec(path, resolve_path); 5060 if (files.AppendIfUnique(file_spec)) 5061 count++; 5062 } 5063 } 5064 } 5065 } break; 5066 5067 default: 5068 break; 5069 } 5070 offset = cmd_offset + load_cmd.cmdsize; 5071 } 5072 5073 if (!rpath_paths.empty()) { 5074 // Fixup all LC_RPATH values to be absolute paths 5075 FileSpec this_file_spec(m_file); 5076 this_file_spec.ResolvePath(); 5077 std::string loader_path("@loader_path"); 5078 std::string executable_path("@executable_path"); 5079 for (auto &rpath : rpath_paths) { 5080 if (rpath.find(loader_path) == 0) { 5081 rpath.erase(0, loader_path.size()); 5082 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5083 } else if (rpath.find(executable_path) == 0) { 5084 rpath.erase(0, executable_path.size()); 5085 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5086 } 5087 } 5088 5089 for (const auto &rpath_relative_path : rpath_relative_paths) { 5090 for (const auto &rpath : rpath_paths) { 5091 std::string path = rpath; 5092 path += rpath_relative_path; 5093 // It is OK to resolve this path because we must find a file on 5094 // disk for us to accept it anyway if it is rpath relative. 5095 FileSpec file_spec(path, true); 5096 // Remove any redundant parts of the path (like "../foo") since 5097 // LC_RPATH values often contain "..". 5098 file_spec.NormalizePath(); 5099 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) { 5100 count++; 5101 break; 5102 } 5103 } 5104 } 5105 } 5106 } 5107 return count; 5108 } 5109 5110 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5111 // If the object file is not an executable it can't hold the entry point. 5112 // m_entry_point_address 5113 // is initialized to an invalid address, so we can just return that. 5114 // If m_entry_point_address is valid it means we've found it already, so 5115 // return the cached value. 5116 5117 if (!IsExecutable() || m_entry_point_address.IsValid()) 5118 return m_entry_point_address; 5119 5120 // Otherwise, look for the UnixThread or Thread command. The data for the 5121 // Thread command is given in 5122 // /usr/include/mach-o.h, but it is basically: 5123 // 5124 // uint32_t flavor - this is the flavor argument you would pass to 5125 // thread_get_state 5126 // uint32_t count - this is the count of longs in the thread state data 5127 // struct XXX_thread_state state - this is the structure from 5128 // <machine/thread_status.h> corresponding to the flavor. 5129 // <repeat this trio> 5130 // 5131 // So we just keep reading the various register flavors till we find the GPR 5132 // one, then read the PC out of there. 5133 // FIXME: We will need to have a "RegisterContext data provider" class at some 5134 // point that can get all the registers 5135 // out of data in this form & attach them to a given thread. That should 5136 // underlie the MacOS X User process plugin, 5137 // and we'll also need it for the MacOS X Core File process plugin. When we 5138 // have that we can also use it here. 5139 // 5140 // For now we hard-code the offsets and flavors we need: 5141 // 5142 // 5143 5144 ModuleSP module_sp(GetModule()); 5145 if (module_sp) { 5146 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5147 struct load_command load_cmd; 5148 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5149 uint32_t i; 5150 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5151 bool done = false; 5152 5153 for (i = 0; i < m_header.ncmds; ++i) { 5154 const lldb::offset_t cmd_offset = offset; 5155 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5156 break; 5157 5158 switch (load_cmd.cmd) { 5159 case LC_UNIXTHREAD: 5160 case LC_THREAD: { 5161 while (offset < cmd_offset + load_cmd.cmdsize) { 5162 uint32_t flavor = m_data.GetU32(&offset); 5163 uint32_t count = m_data.GetU32(&offset); 5164 if (count == 0) { 5165 // We've gotten off somehow, log and exit; 5166 return m_entry_point_address; 5167 } 5168 5169 switch (m_header.cputype) { 5170 case llvm::MachO::CPU_TYPE_ARM: 5171 if (flavor == 1 || 5172 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 from 5173 // mach/arm/thread_status.h 5174 { 5175 offset += 60; // This is the offset of pc in the GPR thread state 5176 // data structure. 5177 start_address = m_data.GetU32(&offset); 5178 done = true; 5179 } 5180 break; 5181 case llvm::MachO::CPU_TYPE_ARM64: 5182 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5183 { 5184 offset += 256; // This is the offset of pc in the GPR thread state 5185 // data structure. 5186 start_address = m_data.GetU64(&offset); 5187 done = true; 5188 } 5189 break; 5190 case llvm::MachO::CPU_TYPE_I386: 5191 if (flavor == 5192 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5193 { 5194 offset += 40; // This is the offset of eip in the GPR thread state 5195 // data structure. 5196 start_address = m_data.GetU32(&offset); 5197 done = true; 5198 } 5199 break; 5200 case llvm::MachO::CPU_TYPE_X86_64: 5201 if (flavor == 5202 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5203 { 5204 offset += 16 * 8; // This is the offset of rip in the GPR thread 5205 // state data structure. 5206 start_address = m_data.GetU64(&offset); 5207 done = true; 5208 } 5209 break; 5210 default: 5211 return m_entry_point_address; 5212 } 5213 // Haven't found the GPR flavor yet, skip over the data for this 5214 // flavor: 5215 if (done) 5216 break; 5217 offset += count * 4; 5218 } 5219 } break; 5220 case LC_MAIN: { 5221 ConstString text_segment_name("__TEXT"); 5222 uint64_t entryoffset = m_data.GetU64(&offset); 5223 SectionSP text_segment_sp = 5224 GetSectionList()->FindSectionByName(text_segment_name); 5225 if (text_segment_sp) { 5226 done = true; 5227 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5228 } 5229 } break; 5230 5231 default: 5232 break; 5233 } 5234 if (done) 5235 break; 5236 5237 // Go to the next load command: 5238 offset = cmd_offset + load_cmd.cmdsize; 5239 } 5240 5241 if (start_address != LLDB_INVALID_ADDRESS) { 5242 // We got the start address from the load commands, so now resolve that 5243 // address in the sections 5244 // of this ObjectFile: 5245 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5246 start_address, GetSectionList())) { 5247 m_entry_point_address.Clear(); 5248 } 5249 } else { 5250 // We couldn't read the UnixThread load command - maybe it wasn't there. 5251 // As a fallback look for the 5252 // "start" symbol in the main executable. 5253 5254 ModuleSP module_sp(GetModule()); 5255 5256 if (module_sp) { 5257 SymbolContextList contexts; 5258 SymbolContext context; 5259 if (module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5260 eSymbolTypeCode, contexts)) { 5261 if (contexts.GetContextAtIndex(0, context)) 5262 m_entry_point_address = context.symbol->GetAddress(); 5263 } 5264 } 5265 } 5266 } 5267 5268 return m_entry_point_address; 5269 } 5270 5271 lldb_private::Address ObjectFileMachO::GetHeaderAddress() { 5272 lldb_private::Address header_addr; 5273 SectionList *section_list = GetSectionList(); 5274 if (section_list) { 5275 SectionSP text_segment_sp( 5276 section_list->FindSectionByName(GetSegmentNameTEXT())); 5277 if (text_segment_sp) { 5278 header_addr.SetSection(text_segment_sp); 5279 header_addr.SetOffset(0); 5280 } 5281 } 5282 return header_addr; 5283 } 5284 5285 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5286 ModuleSP module_sp(GetModule()); 5287 if (module_sp) { 5288 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5289 if (!m_thread_context_offsets_valid) { 5290 m_thread_context_offsets_valid = true; 5291 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5292 FileRangeArray::Entry file_range; 5293 thread_command thread_cmd; 5294 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5295 const uint32_t cmd_offset = offset; 5296 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 5297 break; 5298 5299 if (thread_cmd.cmd == LC_THREAD) { 5300 file_range.SetRangeBase(offset); 5301 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5302 m_thread_context_offsets.Append(file_range); 5303 } 5304 offset = cmd_offset + thread_cmd.cmdsize; 5305 } 5306 } 5307 } 5308 return m_thread_context_offsets.GetSize(); 5309 } 5310 5311 lldb::RegisterContextSP 5312 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5313 lldb_private::Thread &thread) { 5314 lldb::RegisterContextSP reg_ctx_sp; 5315 5316 ModuleSP module_sp(GetModule()); 5317 if (module_sp) { 5318 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5319 if (!m_thread_context_offsets_valid) 5320 GetNumThreadContexts(); 5321 5322 const FileRangeArray::Entry *thread_context_file_range = 5323 m_thread_context_offsets.GetEntryAtIndex(idx); 5324 if (thread_context_file_range) { 5325 5326 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5327 thread_context_file_range->GetByteSize()); 5328 5329 switch (m_header.cputype) { 5330 case llvm::MachO::CPU_TYPE_ARM64: 5331 reg_ctx_sp.reset(new RegisterContextDarwin_arm64_Mach(thread, data)); 5332 break; 5333 5334 case llvm::MachO::CPU_TYPE_ARM: 5335 reg_ctx_sp.reset(new RegisterContextDarwin_arm_Mach(thread, data)); 5336 break; 5337 5338 case llvm::MachO::CPU_TYPE_I386: 5339 reg_ctx_sp.reset(new RegisterContextDarwin_i386_Mach(thread, data)); 5340 break; 5341 5342 case llvm::MachO::CPU_TYPE_X86_64: 5343 reg_ctx_sp.reset(new RegisterContextDarwin_x86_64_Mach(thread, data)); 5344 break; 5345 } 5346 } 5347 } 5348 return reg_ctx_sp; 5349 } 5350 5351 ObjectFile::Type ObjectFileMachO::CalculateType() { 5352 switch (m_header.filetype) { 5353 case MH_OBJECT: // 0x1u 5354 if (GetAddressByteSize() == 4) { 5355 // 32 bit kexts are just object files, but they do have a valid 5356 // UUID load command. 5357 UUID uuid; 5358 if (GetUUID(&uuid)) { 5359 // this checking for the UUID load command is not enough 5360 // we could eventually look for the symbol named 5361 // "OSKextGetCurrentIdentifier" as this is required of kexts 5362 if (m_strata == eStrataInvalid) 5363 m_strata = eStrataKernel; 5364 return eTypeSharedLibrary; 5365 } 5366 } 5367 return eTypeObjectFile; 5368 5369 case MH_EXECUTE: 5370 return eTypeExecutable; // 0x2u 5371 case MH_FVMLIB: 5372 return eTypeSharedLibrary; // 0x3u 5373 case MH_CORE: 5374 return eTypeCoreFile; // 0x4u 5375 case MH_PRELOAD: 5376 return eTypeSharedLibrary; // 0x5u 5377 case MH_DYLIB: 5378 return eTypeSharedLibrary; // 0x6u 5379 case MH_DYLINKER: 5380 return eTypeDynamicLinker; // 0x7u 5381 case MH_BUNDLE: 5382 return eTypeSharedLibrary; // 0x8u 5383 case MH_DYLIB_STUB: 5384 return eTypeStubLibrary; // 0x9u 5385 case MH_DSYM: 5386 return eTypeDebugInfo; // 0xAu 5387 case MH_KEXT_BUNDLE: 5388 return eTypeSharedLibrary; // 0xBu 5389 default: 5390 break; 5391 } 5392 return eTypeUnknown; 5393 } 5394 5395 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5396 switch (m_header.filetype) { 5397 case MH_OBJECT: // 0x1u 5398 { 5399 // 32 bit kexts are just object files, but they do have a valid 5400 // UUID load command. 5401 UUID uuid; 5402 if (GetUUID(&uuid)) { 5403 // this checking for the UUID load command is not enough 5404 // we could eventually look for the symbol named 5405 // "OSKextGetCurrentIdentifier" as this is required of kexts 5406 if (m_type == eTypeInvalid) 5407 m_type = eTypeSharedLibrary; 5408 5409 return eStrataKernel; 5410 } 5411 } 5412 return eStrataUnknown; 5413 5414 case MH_EXECUTE: // 0x2u 5415 // Check for the MH_DYLDLINK bit in the flags 5416 if (m_header.flags & MH_DYLDLINK) { 5417 return eStrataUser; 5418 } else { 5419 SectionList *section_list = GetSectionList(); 5420 if (section_list) { 5421 static ConstString g_kld_section_name("__KLD"); 5422 if (section_list->FindSectionByName(g_kld_section_name)) 5423 return eStrataKernel; 5424 } 5425 } 5426 return eStrataRawImage; 5427 5428 case MH_FVMLIB: 5429 return eStrataUser; // 0x3u 5430 case MH_CORE: 5431 return eStrataUnknown; // 0x4u 5432 case MH_PRELOAD: 5433 return eStrataRawImage; // 0x5u 5434 case MH_DYLIB: 5435 return eStrataUser; // 0x6u 5436 case MH_DYLINKER: 5437 return eStrataUser; // 0x7u 5438 case MH_BUNDLE: 5439 return eStrataUser; // 0x8u 5440 case MH_DYLIB_STUB: 5441 return eStrataUser; // 0x9u 5442 case MH_DSYM: 5443 return eStrataUnknown; // 0xAu 5444 case MH_KEXT_BUNDLE: 5445 return eStrataKernel; // 0xBu 5446 default: 5447 break; 5448 } 5449 return eStrataUnknown; 5450 } 5451 5452 uint32_t ObjectFileMachO::GetVersion(uint32_t *versions, 5453 uint32_t num_versions) { 5454 ModuleSP module_sp(GetModule()); 5455 if (module_sp) { 5456 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5457 struct dylib_command load_cmd; 5458 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5459 uint32_t version_cmd = 0; 5460 uint64_t version = 0; 5461 uint32_t i; 5462 for (i = 0; i < m_header.ncmds; ++i) { 5463 const lldb::offset_t cmd_offset = offset; 5464 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5465 break; 5466 5467 if (load_cmd.cmd == LC_ID_DYLIB) { 5468 if (version_cmd == 0) { 5469 version_cmd = load_cmd.cmd; 5470 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 5471 break; 5472 version = load_cmd.dylib.current_version; 5473 } 5474 break; // Break for now unless there is another more complete version 5475 // number load command in the future. 5476 } 5477 offset = cmd_offset + load_cmd.cmdsize; 5478 } 5479 5480 if (version_cmd == LC_ID_DYLIB) { 5481 if (versions != NULL && num_versions > 0) { 5482 if (num_versions > 0) 5483 versions[0] = (version & 0xFFFF0000ull) >> 16; 5484 if (num_versions > 1) 5485 versions[1] = (version & 0x0000FF00ull) >> 8; 5486 if (num_versions > 2) 5487 versions[2] = (version & 0x000000FFull); 5488 // Fill in an remaining version numbers with invalid values 5489 for (i = 3; i < num_versions; ++i) 5490 versions[i] = UINT32_MAX; 5491 } 5492 // The LC_ID_DYLIB load command has a version with 3 version numbers 5493 // in it, so always return 3 5494 return 3; 5495 } 5496 } 5497 return false; 5498 } 5499 5500 bool ObjectFileMachO::GetArchitecture(ArchSpec &arch) { 5501 ModuleSP module_sp(GetModule()); 5502 if (module_sp) { 5503 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5504 return GetArchitecture(m_header, m_data, 5505 MachHeaderSizeFromMagic(m_header.magic), arch); 5506 } 5507 return false; 5508 } 5509 5510 UUID ObjectFileMachO::GetProcessSharedCacheUUID(Process *process) { 5511 UUID uuid; 5512 if (process && process->GetDynamicLoader()) { 5513 DynamicLoader *dl = process->GetDynamicLoader(); 5514 addr_t load_address; 5515 LazyBool using_shared_cache; 5516 LazyBool private_shared_cache; 5517 dl->GetSharedCacheInformation(load_address, uuid, using_shared_cache, 5518 private_shared_cache); 5519 } 5520 return uuid; 5521 } 5522 5523 UUID ObjectFileMachO::GetLLDBSharedCacheUUID() { 5524 UUID uuid; 5525 #if defined(__APPLE__) && \ 5526 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5527 uint8_t *(*dyld_get_all_image_infos)(void); 5528 dyld_get_all_image_infos = 5529 (uint8_t * (*)())dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5530 if (dyld_get_all_image_infos) { 5531 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5532 if (dyld_all_image_infos_address) { 5533 uint32_t *version = (uint32_t *) 5534 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5535 if (*version >= 13) { 5536 uuid_t *sharedCacheUUID_address = 0; 5537 int wordsize = sizeof(uint8_t *); 5538 if (wordsize == 8) { 5539 sharedCacheUUID_address = 5540 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5541 160); // sharedCacheUUID <mach-o/dyld_images.h> 5542 } else { 5543 sharedCacheUUID_address = 5544 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5545 84); // sharedCacheUUID <mach-o/dyld_images.h> 5546 } 5547 uuid.SetBytes(sharedCacheUUID_address); 5548 } 5549 } 5550 } 5551 #endif 5552 return uuid; 5553 } 5554 5555 uint32_t ObjectFileMachO::GetMinimumOSVersion(uint32_t *versions, 5556 uint32_t num_versions) { 5557 if (m_min_os_versions.empty()) { 5558 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5559 bool success = false; 5560 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5561 const lldb::offset_t load_cmd_offset = offset; 5562 5563 version_min_command lc; 5564 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5565 break; 5566 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5567 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5568 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5569 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5570 if (m_data.GetU32(&offset, &lc.version, 5571 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5572 const uint32_t xxxx = lc.version >> 16; 5573 const uint32_t yy = (lc.version >> 8) & 0xffu; 5574 const uint32_t zz = lc.version & 0xffu; 5575 if (xxxx) { 5576 m_min_os_versions.push_back(xxxx); 5577 m_min_os_versions.push_back(yy); 5578 m_min_os_versions.push_back(zz); 5579 } 5580 success = true; 5581 } 5582 } 5583 offset = load_cmd_offset + lc.cmdsize; 5584 } 5585 5586 if (success == false) { 5587 // Push an invalid value so we don't keep trying to 5588 m_min_os_versions.push_back(UINT32_MAX); 5589 } 5590 } 5591 5592 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) { 5593 if (versions != NULL && num_versions > 0) { 5594 for (size_t i = 0; i < num_versions; ++i) { 5595 if (i < m_min_os_versions.size()) 5596 versions[i] = m_min_os_versions[i]; 5597 else 5598 versions[i] = 0; 5599 } 5600 } 5601 return m_min_os_versions.size(); 5602 } 5603 // Call the superclasses version that will empty out the data 5604 return ObjectFile::GetMinimumOSVersion(versions, num_versions); 5605 } 5606 5607 uint32_t ObjectFileMachO::GetSDKVersion(uint32_t *versions, 5608 uint32_t num_versions) { 5609 if (m_sdk_versions.empty()) { 5610 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5611 bool success = false; 5612 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5613 const lldb::offset_t load_cmd_offset = offset; 5614 5615 version_min_command lc; 5616 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5617 break; 5618 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5619 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5620 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5621 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5622 if (m_data.GetU32(&offset, &lc.version, 5623 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5624 const uint32_t xxxx = lc.sdk >> 16; 5625 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5626 const uint32_t zz = lc.sdk & 0xffu; 5627 if (xxxx) { 5628 m_sdk_versions.push_back(xxxx); 5629 m_sdk_versions.push_back(yy); 5630 m_sdk_versions.push_back(zz); 5631 } 5632 success = true; 5633 } 5634 } 5635 offset = load_cmd_offset + lc.cmdsize; 5636 } 5637 5638 if (success == false) { 5639 // Push an invalid value so we don't keep trying to 5640 m_sdk_versions.push_back(UINT32_MAX); 5641 } 5642 } 5643 5644 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) { 5645 if (versions != NULL && num_versions > 0) { 5646 for (size_t i = 0; i < num_versions; ++i) { 5647 if (i < m_sdk_versions.size()) 5648 versions[i] = m_sdk_versions[i]; 5649 else 5650 versions[i] = 0; 5651 } 5652 } 5653 return m_sdk_versions.size(); 5654 } 5655 // Call the superclasses version that will empty out the data 5656 return ObjectFile::GetSDKVersion(versions, num_versions); 5657 } 5658 5659 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5660 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5661 } 5662 5663 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5664 return m_allow_assembly_emulation_unwind_plans; 5665 } 5666 5667 //------------------------------------------------------------------ 5668 // PluginInterface protocol 5669 //------------------------------------------------------------------ 5670 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5671 return GetPluginNameStatic(); 5672 } 5673 5674 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5675 5676 Section *ObjectFileMachO::GetMachHeaderSection() { 5677 // Find the first address of the mach header which is the first non-zero 5678 // file sized section whose file offset is zero. This is the base file address 5679 // of the mach-o file which can be subtracted from the vmaddr of the other 5680 // segments found in memory and added to the load address 5681 ModuleSP module_sp = GetModule(); 5682 if (module_sp) { 5683 SectionList *section_list = GetSectionList(); 5684 if (section_list) { 5685 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5686 const size_t num_sections = section_list->GetSize(); 5687 5688 for (size_t sect_idx = 0; sect_idx < num_sections && 5689 mach_base_file_addr == LLDB_INVALID_ADDRESS; 5690 ++sect_idx) { 5691 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5692 if (section && section->GetFileSize() > 0 && 5693 section->GetFileOffset() == 0 && 5694 section->IsThreadSpecific() == false && 5695 module_sp.get() == section->GetModule().get()) { 5696 return section; 5697 } 5698 } 5699 } 5700 } 5701 return nullptr; 5702 } 5703 5704 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 5705 lldb::addr_t mach_header_load_address, const Section *mach_header_section, 5706 const Section *section) { 5707 ModuleSP module_sp = GetModule(); 5708 if (module_sp && mach_header_section && section && 5709 mach_header_load_address != LLDB_INVALID_ADDRESS) { 5710 lldb::addr_t mach_header_file_addr = mach_header_section->GetFileAddress(); 5711 if (mach_header_file_addr != LLDB_INVALID_ADDRESS) { 5712 if (section && section->GetFileSize() > 0 && 5713 section->IsThreadSpecific() == false && 5714 module_sp.get() == section->GetModule().get()) { 5715 // Ignore __LINKEDIT and __DWARF segments 5716 if (section->GetName() == GetSegmentNameLINKEDIT()) { 5717 // Only map __LINKEDIT if we have an in memory image and this isn't 5718 // a kernel binary like a kext or mach_kernel. 5719 const bool is_memory_image = (bool)m_process_wp.lock(); 5720 const Strata strata = GetStrata(); 5721 if (is_memory_image == false || strata == eStrataKernel) 5722 return LLDB_INVALID_ADDRESS; 5723 } 5724 return section->GetFileAddress() - mach_header_file_addr + 5725 mach_header_load_address; 5726 } 5727 } 5728 } 5729 return LLDB_INVALID_ADDRESS; 5730 } 5731 5732 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 5733 bool value_is_offset) { 5734 ModuleSP module_sp = GetModule(); 5735 if (module_sp) { 5736 size_t num_loaded_sections = 0; 5737 SectionList *section_list = GetSectionList(); 5738 if (section_list) { 5739 const size_t num_sections = section_list->GetSize(); 5740 5741 if (value_is_offset) { 5742 // "value" is an offset to apply to each top level segment 5743 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5744 // Iterate through the object file sections to find all 5745 // of the sections that size on disk (to avoid __PAGEZERO) 5746 // and load them 5747 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5748 if (section_sp && section_sp->GetFileSize() > 0 && 5749 section_sp->IsThreadSpecific() == false && 5750 module_sp.get() == section_sp->GetModule().get()) { 5751 // Ignore __LINKEDIT and __DWARF segments 5752 if (section_sp->GetName() == GetSegmentNameLINKEDIT()) { 5753 // Only map __LINKEDIT if we have an in memory image and this 5754 // isn't 5755 // a kernel binary like a kext or mach_kernel. 5756 const bool is_memory_image = (bool)m_process_wp.lock(); 5757 const Strata strata = GetStrata(); 5758 if (is_memory_image == false || strata == eStrataKernel) 5759 continue; 5760 } 5761 if (target.GetSectionLoadList().SetSectionLoadAddress( 5762 section_sp, section_sp->GetFileAddress() + value)) 5763 ++num_loaded_sections; 5764 } 5765 } 5766 } else { 5767 // "value" is the new base address of the mach_header, adjust each 5768 // section accordingly 5769 5770 Section *mach_header_section = GetMachHeaderSection(); 5771 if (mach_header_section) { 5772 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5773 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5774 5775 lldb::addr_t section_load_addr = 5776 CalculateSectionLoadAddressForMemoryImage( 5777 value, mach_header_section, section_sp.get()); 5778 if (section_load_addr != LLDB_INVALID_ADDRESS) { 5779 if (target.GetSectionLoadList().SetSectionLoadAddress( 5780 section_sp, section_load_addr)) 5781 ++num_loaded_sections; 5782 } 5783 } 5784 } 5785 } 5786 } 5787 return num_loaded_sections > 0; 5788 } 5789 return false; 5790 } 5791 5792 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 5793 const FileSpec &outfile, Error &error) { 5794 if (process_sp) { 5795 Target &target = process_sp->GetTarget(); 5796 const ArchSpec target_arch = target.GetArchitecture(); 5797 const llvm::Triple &target_triple = target_arch.GetTriple(); 5798 if (target_triple.getVendor() == llvm::Triple::Apple && 5799 (target_triple.getOS() == llvm::Triple::MacOSX || 5800 target_triple.getOS() == llvm::Triple::IOS || 5801 target_triple.getOS() == llvm::Triple::WatchOS || 5802 target_triple.getOS() == llvm::Triple::TvOS)) { 5803 bool make_core = false; 5804 switch (target_arch.GetMachine()) { 5805 case llvm::Triple::aarch64: 5806 case llvm::Triple::arm: 5807 case llvm::Triple::thumb: 5808 case llvm::Triple::x86: 5809 case llvm::Triple::x86_64: 5810 make_core = true; 5811 break; 5812 default: 5813 error.SetErrorStringWithFormat("unsupported core architecture: %s", 5814 target_triple.str().c_str()); 5815 break; 5816 } 5817 5818 if (make_core) { 5819 std::vector<segment_command_64> segment_load_commands; 5820 // uint32_t range_info_idx = 0; 5821 MemoryRegionInfo range_info; 5822 Error range_error = process_sp->GetMemoryRegionInfo(0, range_info); 5823 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 5824 const ByteOrder byte_order = target_arch.GetByteOrder(); 5825 if (range_error.Success()) { 5826 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 5827 const addr_t addr = range_info.GetRange().GetRangeBase(); 5828 const addr_t size = range_info.GetRange().GetByteSize(); 5829 5830 if (size == 0) 5831 break; 5832 5833 // Calculate correct protections 5834 uint32_t prot = 0; 5835 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 5836 prot |= VM_PROT_READ; 5837 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 5838 prot |= VM_PROT_WRITE; 5839 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 5840 prot |= VM_PROT_EXECUTE; 5841 5842 // printf ("[%3u] [0x%16.16" PRIx64 " - 5843 // 0x%16.16" PRIx64 ") %c%c%c\n", 5844 // range_info_idx, 5845 // addr, 5846 // size, 5847 // (prot & VM_PROT_READ ) ? 'r' : 5848 // '-', 5849 // (prot & VM_PROT_WRITE ) ? 'w' : 5850 // '-', 5851 // (prot & VM_PROT_EXECUTE) ? 'x' : 5852 // '-'); 5853 5854 if (prot != 0) { 5855 uint32_t cmd_type = LC_SEGMENT_64; 5856 uint32_t segment_size = sizeof(segment_command_64); 5857 if (addr_byte_size == 4) { 5858 cmd_type = LC_SEGMENT; 5859 segment_size = sizeof(segment_command); 5860 } 5861 segment_command_64 segment = { 5862 cmd_type, // uint32_t cmd; 5863 segment_size, // uint32_t cmdsize; 5864 {0}, // char segname[16]; 5865 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 5866 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 5867 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 5868 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 5869 prot, // uint32_t maxprot; 5870 prot, // uint32_t initprot; 5871 0, // uint32_t nsects; 5872 0}; // uint32_t flags; 5873 segment_load_commands.push_back(segment); 5874 } else { 5875 // No protections and a size of 1 used to be returned from old 5876 // debugservers when we asked about a region that was past the 5877 // last memory region and it indicates the end... 5878 if (size == 1) 5879 break; 5880 } 5881 5882 range_error = process_sp->GetMemoryRegionInfo( 5883 range_info.GetRange().GetRangeEnd(), range_info); 5884 if (range_error.Fail()) 5885 break; 5886 } 5887 5888 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 5889 5890 mach_header_64 mach_header; 5891 if (addr_byte_size == 8) { 5892 mach_header.magic = MH_MAGIC_64; 5893 } else { 5894 mach_header.magic = MH_MAGIC; 5895 } 5896 mach_header.cputype = target_arch.GetMachOCPUType(); 5897 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 5898 mach_header.filetype = MH_CORE; 5899 mach_header.ncmds = segment_load_commands.size(); 5900 mach_header.flags = 0; 5901 mach_header.reserved = 0; 5902 ThreadList &thread_list = process_sp->GetThreadList(); 5903 const uint32_t num_threads = thread_list.GetSize(); 5904 5905 // Make an array of LC_THREAD data items. Each one contains 5906 // the contents of the LC_THREAD load command. The data doesn't 5907 // contain the load command + load command size, we will 5908 // add the load command and load command size as we emit the data. 5909 std::vector<StreamString> LC_THREAD_datas(num_threads); 5910 for (auto &LC_THREAD_data : LC_THREAD_datas) { 5911 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 5912 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 5913 LC_THREAD_data.SetByteOrder(byte_order); 5914 } 5915 for (uint32_t thread_idx = 0; thread_idx < num_threads; 5916 ++thread_idx) { 5917 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 5918 if (thread_sp) { 5919 switch (mach_header.cputype) { 5920 case llvm::MachO::CPU_TYPE_ARM64: 5921 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 5922 thread_sp.get(), LC_THREAD_datas[thread_idx]); 5923 break; 5924 5925 case llvm::MachO::CPU_TYPE_ARM: 5926 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 5927 thread_sp.get(), LC_THREAD_datas[thread_idx]); 5928 break; 5929 5930 case llvm::MachO::CPU_TYPE_I386: 5931 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 5932 thread_sp.get(), LC_THREAD_datas[thread_idx]); 5933 break; 5934 5935 case llvm::MachO::CPU_TYPE_X86_64: 5936 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 5937 thread_sp.get(), LC_THREAD_datas[thread_idx]); 5938 break; 5939 } 5940 } 5941 } 5942 5943 // The size of the load command is the size of the segments... 5944 if (addr_byte_size == 8) { 5945 mach_header.sizeofcmds = segment_load_commands.size() * 5946 sizeof(struct segment_command_64); 5947 } else { 5948 mach_header.sizeofcmds = 5949 segment_load_commands.size() * sizeof(struct segment_command); 5950 } 5951 5952 // and the size of all LC_THREAD load command 5953 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 5954 ++mach_header.ncmds; 5955 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 5956 } 5957 5958 printf("mach_header: 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x " 5959 "0x%8.8x 0x%8.8x\n", 5960 mach_header.magic, mach_header.cputype, mach_header.cpusubtype, 5961 mach_header.filetype, mach_header.ncmds, 5962 mach_header.sizeofcmds, mach_header.flags, 5963 mach_header.reserved); 5964 5965 // Write the mach header 5966 buffer.PutHex32(mach_header.magic); 5967 buffer.PutHex32(mach_header.cputype); 5968 buffer.PutHex32(mach_header.cpusubtype); 5969 buffer.PutHex32(mach_header.filetype); 5970 buffer.PutHex32(mach_header.ncmds); 5971 buffer.PutHex32(mach_header.sizeofcmds); 5972 buffer.PutHex32(mach_header.flags); 5973 if (addr_byte_size == 8) { 5974 buffer.PutHex32(mach_header.reserved); 5975 } 5976 5977 // Skip the mach header and all load commands and align to the next 5978 // 0x1000 byte boundary 5979 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 5980 if (file_offset & 0x00000fff) { 5981 file_offset += 0x00001000ull; 5982 file_offset &= (~0x00001000ull + 1); 5983 } 5984 5985 for (auto &segment : segment_load_commands) { 5986 segment.fileoff = file_offset; 5987 file_offset += segment.filesize; 5988 } 5989 5990 // Write out all of the LC_THREAD load commands 5991 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 5992 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 5993 buffer.PutHex32(LC_THREAD); 5994 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 5995 buffer.Write(LC_THREAD_data.GetData(), LC_THREAD_data_size); 5996 } 5997 5998 // Write out all of the segment load commands 5999 for (const auto &segment : segment_load_commands) { 6000 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6001 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6002 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6003 segment.cmd, segment.cmdsize, segment.vmaddr, 6004 segment.vmaddr + segment.vmsize, segment.fileoff, 6005 segment.filesize, segment.maxprot, segment.initprot, 6006 segment.nsects, segment.flags); 6007 6008 buffer.PutHex32(segment.cmd); 6009 buffer.PutHex32(segment.cmdsize); 6010 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6011 if (addr_byte_size == 8) { 6012 buffer.PutHex64(segment.vmaddr); 6013 buffer.PutHex64(segment.vmsize); 6014 buffer.PutHex64(segment.fileoff); 6015 buffer.PutHex64(segment.filesize); 6016 } else { 6017 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6018 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6019 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6020 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6021 } 6022 buffer.PutHex32(segment.maxprot); 6023 buffer.PutHex32(segment.initprot); 6024 buffer.PutHex32(segment.nsects); 6025 buffer.PutHex32(segment.flags); 6026 } 6027 6028 File core_file; 6029 std::string core_file_path(outfile.GetPath()); 6030 error = core_file.Open(core_file_path.c_str(), 6031 File::eOpenOptionWrite | 6032 File::eOpenOptionTruncate | 6033 File::eOpenOptionCanCreate); 6034 if (error.Success()) { 6035 // Read 1 page at a time 6036 uint8_t bytes[0x1000]; 6037 // Write the mach header and load commands out to the core file 6038 size_t bytes_written = buffer.GetString().size(); 6039 error = core_file.Write(buffer.GetString().data(), bytes_written); 6040 if (error.Success()) { 6041 // Now write the file data for all memory segments in the process 6042 for (const auto &segment : segment_load_commands) { 6043 if (core_file.SeekFromStart(segment.fileoff) == -1) { 6044 error.SetErrorStringWithFormat( 6045 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6046 segment.fileoff, core_file_path.c_str()); 6047 break; 6048 } 6049 6050 printf("Saving %" PRId64 6051 " bytes of data for memory region at 0x%" PRIx64 "\n", 6052 segment.vmsize, segment.vmaddr); 6053 addr_t bytes_left = segment.vmsize; 6054 addr_t addr = segment.vmaddr; 6055 Error memory_read_error; 6056 while (bytes_left > 0 && error.Success()) { 6057 const size_t bytes_to_read = 6058 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6059 const size_t bytes_read = process_sp->ReadMemory( 6060 addr, bytes, bytes_to_read, memory_read_error); 6061 if (bytes_read == bytes_to_read) { 6062 size_t bytes_written = bytes_read; 6063 error = core_file.Write(bytes, bytes_written); 6064 bytes_left -= bytes_read; 6065 addr += bytes_read; 6066 } else { 6067 // Some pages within regions are not readable, those 6068 // should be zero filled 6069 memset(bytes, 0, bytes_to_read); 6070 size_t bytes_written = bytes_to_read; 6071 error = core_file.Write(bytes, bytes_written); 6072 bytes_left -= bytes_to_read; 6073 addr += bytes_to_read; 6074 } 6075 } 6076 } 6077 } 6078 } 6079 } else { 6080 error.SetErrorString( 6081 "process doesn't support getting memory region info"); 6082 } 6083 } 6084 return true; // This is the right plug to handle saving core files for 6085 // this process 6086 } 6087 } 6088 return false; 6089 } 6090