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 < 29; ++i) 671 gpr.x[i] = data.GetU64(&offset); 672 gpr.fp = data.GetU64(&offset); 673 gpr.lr = data.GetU64(&offset); 674 gpr.sp = data.GetU64(&offset); 675 gpr.pc = data.GetU64(&offset); 676 gpr.cpsr = data.GetU32(&offset); 677 SetError(GPRRegSet, Read, 0); 678 } 679 offset = next_thread_state; 680 break; 681 case FPURegSet: { 682 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 683 const int fpu_reg_buf_size = sizeof(fpu); 684 if (fpu_reg_buf_size == count && 685 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 686 fpu_reg_buf) == fpu_reg_buf_size) { 687 SetError(FPURegSet, Read, 0); 688 } else { 689 done = true; 690 } 691 } 692 offset = next_thread_state; 693 break; 694 case EXCRegSet: 695 if (count == 4) { 696 exc.far = data.GetU64(&offset); 697 exc.esr = data.GetU32(&offset); 698 exc.exception = data.GetU32(&offset); 699 SetError(EXCRegSet, Read, 0); 700 } 701 offset = next_thread_state; 702 break; 703 default: 704 done = true; 705 break; 706 } 707 } 708 } 709 710 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 711 const char *alt_name, size_t reg_byte_size, 712 Stream &data) { 713 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 714 if (reg_info == NULL) 715 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 716 if (reg_info) { 717 lldb_private::RegisterValue reg_value; 718 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 719 if (reg_info->byte_size >= reg_byte_size) 720 data.Write(reg_value.GetBytes(), reg_byte_size); 721 else { 722 data.Write(reg_value.GetBytes(), reg_info->byte_size); 723 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 724 ++i) 725 data.PutChar(0); 726 } 727 return reg_byte_size; 728 } 729 } 730 // Just write zeros if all else fails 731 for (size_t i = 0; i < reg_byte_size; ++i) 732 data.PutChar(0); 733 return reg_byte_size; 734 } 735 736 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 737 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 738 if (reg_ctx_sp) { 739 RegisterContext *reg_ctx = reg_ctx_sp.get(); 740 741 data.PutHex32(GPRRegSet); // Flavor 742 data.PutHex32(GPRWordCount); 743 WriteRegister(reg_ctx, "x0", NULL, 8, data); 744 WriteRegister(reg_ctx, "x1", NULL, 8, data); 745 WriteRegister(reg_ctx, "x2", NULL, 8, data); 746 WriteRegister(reg_ctx, "x3", NULL, 8, data); 747 WriteRegister(reg_ctx, "x4", NULL, 8, data); 748 WriteRegister(reg_ctx, "x5", NULL, 8, data); 749 WriteRegister(reg_ctx, "x6", NULL, 8, data); 750 WriteRegister(reg_ctx, "x7", NULL, 8, data); 751 WriteRegister(reg_ctx, "x8", NULL, 8, data); 752 WriteRegister(reg_ctx, "x9", NULL, 8, data); 753 WriteRegister(reg_ctx, "x10", NULL, 8, data); 754 WriteRegister(reg_ctx, "x11", NULL, 8, data); 755 WriteRegister(reg_ctx, "x12", NULL, 8, data); 756 WriteRegister(reg_ctx, "x13", NULL, 8, data); 757 WriteRegister(reg_ctx, "x14", NULL, 8, data); 758 WriteRegister(reg_ctx, "x15", NULL, 8, data); 759 WriteRegister(reg_ctx, "x16", NULL, 8, data); 760 WriteRegister(reg_ctx, "x17", NULL, 8, data); 761 WriteRegister(reg_ctx, "x18", NULL, 8, data); 762 WriteRegister(reg_ctx, "x19", NULL, 8, data); 763 WriteRegister(reg_ctx, "x20", NULL, 8, data); 764 WriteRegister(reg_ctx, "x21", NULL, 8, data); 765 WriteRegister(reg_ctx, "x22", NULL, 8, data); 766 WriteRegister(reg_ctx, "x23", NULL, 8, data); 767 WriteRegister(reg_ctx, "x24", NULL, 8, data); 768 WriteRegister(reg_ctx, "x25", NULL, 8, data); 769 WriteRegister(reg_ctx, "x26", NULL, 8, data); 770 WriteRegister(reg_ctx, "x27", NULL, 8, data); 771 WriteRegister(reg_ctx, "x28", NULL, 8, data); 772 WriteRegister(reg_ctx, "fp", NULL, 8, data); 773 WriteRegister(reg_ctx, "lr", NULL, 8, data); 774 WriteRegister(reg_ctx, "sp", NULL, 8, data); 775 WriteRegister(reg_ctx, "pc", NULL, 8, data); 776 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 777 778 // Write out the EXC registers 779 // data.PutHex32 (EXCRegSet); 780 // data.PutHex32 (EXCWordCount); 781 // WriteRegister (reg_ctx, "far", NULL, 8, data); 782 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 783 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 784 return true; 785 } 786 return false; 787 } 788 789 protected: 790 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 791 792 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 793 794 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 795 796 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 797 798 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 799 return 0; 800 } 801 802 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 803 return 0; 804 } 805 806 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 807 return 0; 808 } 809 810 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 811 return -1; 812 } 813 }; 814 815 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 816 switch (magic) { 817 case MH_MAGIC: 818 case MH_CIGAM: 819 return sizeof(struct mach_header); 820 821 case MH_MAGIC_64: 822 case MH_CIGAM_64: 823 return sizeof(struct mach_header_64); 824 break; 825 826 default: 827 break; 828 } 829 return 0; 830 } 831 832 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 833 834 void ObjectFileMachO::Initialize() { 835 PluginManager::RegisterPlugin( 836 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 837 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 838 } 839 840 void ObjectFileMachO::Terminate() { 841 PluginManager::UnregisterPlugin(CreateInstance); 842 } 843 844 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 845 static ConstString g_name("mach-o"); 846 return g_name; 847 } 848 849 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 850 return "Mach-o object file reader (32 and 64 bit)"; 851 } 852 853 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 854 DataBufferSP &data_sp, 855 lldb::offset_t data_offset, 856 const FileSpec *file, 857 lldb::offset_t file_offset, 858 lldb::offset_t length) { 859 if (!data_sp) { 860 data_sp = file->MemoryMapFileContentsIfLocal(file_offset, length); 861 data_offset = 0; 862 } 863 864 if (ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) { 865 // Update the data to contain the entire file if it doesn't already 866 if (data_sp->GetByteSize() < length) { 867 data_sp = file->MemoryMapFileContentsIfLocal(file_offset, length); 868 data_offset = 0; 869 } 870 std::unique_ptr<ObjectFile> objfile_ap(new ObjectFileMachO( 871 module_sp, data_sp, data_offset, file, file_offset, length)); 872 if (objfile_ap.get() && objfile_ap->ParseHeader()) 873 return objfile_ap.release(); 874 } 875 return NULL; 876 } 877 878 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 879 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 880 const ProcessSP &process_sp, lldb::addr_t header_addr) { 881 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 882 std::unique_ptr<ObjectFile> objfile_ap( 883 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 884 if (objfile_ap.get() && objfile_ap->ParseHeader()) 885 return objfile_ap.release(); 886 } 887 return NULL; 888 } 889 890 size_t ObjectFileMachO::GetModuleSpecifications( 891 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 892 lldb::offset_t data_offset, lldb::offset_t file_offset, 893 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 894 const size_t initial_count = specs.GetSize(); 895 896 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 897 DataExtractor data; 898 data.SetData(data_sp); 899 llvm::MachO::mach_header header; 900 if (ParseHeader(data, &data_offset, header)) { 901 size_t header_and_load_cmds = 902 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 903 if (header_and_load_cmds >= data_sp->GetByteSize()) { 904 data_sp = file.ReadFileContents(file_offset, header_and_load_cmds); 905 data.SetData(data_sp); 906 data_offset = MachHeaderSizeFromMagic(header.magic); 907 } 908 if (data_sp) { 909 ModuleSpec spec; 910 spec.GetFileSpec() = file; 911 spec.SetObjectOffset(file_offset); 912 spec.SetObjectSize(length); 913 914 if (GetArchitecture(header, data, data_offset, 915 spec.GetArchitecture())) { 916 if (spec.GetArchitecture().IsValid()) { 917 GetUUID(header, data, data_offset, spec.GetUUID()); 918 specs.Append(spec); 919 } 920 } 921 } 922 } 923 } 924 return specs.GetSize() - initial_count; 925 } 926 927 const ConstString &ObjectFileMachO::GetSegmentNameTEXT() { 928 static ConstString g_segment_name_TEXT("__TEXT"); 929 return g_segment_name_TEXT; 930 } 931 932 const ConstString &ObjectFileMachO::GetSegmentNameDATA() { 933 static ConstString g_segment_name_DATA("__DATA"); 934 return g_segment_name_DATA; 935 } 936 937 const ConstString &ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 938 static ConstString g_segment_name("__DATA_DIRTY"); 939 return g_segment_name; 940 } 941 942 const ConstString &ObjectFileMachO::GetSegmentNameDATA_CONST() { 943 static ConstString g_segment_name("__DATA_CONST"); 944 return g_segment_name; 945 } 946 947 const ConstString &ObjectFileMachO::GetSegmentNameOBJC() { 948 static ConstString g_segment_name_OBJC("__OBJC"); 949 return g_segment_name_OBJC; 950 } 951 952 const ConstString &ObjectFileMachO::GetSegmentNameLINKEDIT() { 953 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 954 return g_section_name_LINKEDIT; 955 } 956 957 const ConstString &ObjectFileMachO::GetSectionNameEHFrame() { 958 static ConstString g_section_name_eh_frame("__eh_frame"); 959 return g_section_name_eh_frame; 960 } 961 962 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 963 lldb::addr_t data_offset, 964 lldb::addr_t data_length) { 965 DataExtractor data; 966 data.SetData(data_sp, data_offset, data_length); 967 lldb::offset_t offset = 0; 968 uint32_t magic = data.GetU32(&offset); 969 return MachHeaderSizeFromMagic(magic) != 0; 970 } 971 972 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 973 DataBufferSP &data_sp, 974 lldb::offset_t data_offset, 975 const FileSpec *file, 976 lldb::offset_t file_offset, 977 lldb::offset_t length) 978 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 979 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 980 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 981 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 982 ::memset(&m_header, 0, sizeof(m_header)); 983 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 984 } 985 986 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 987 lldb::DataBufferSP &header_data_sp, 988 const lldb::ProcessSP &process_sp, 989 lldb::addr_t header_addr) 990 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 991 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 992 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 993 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 994 ::memset(&m_header, 0, sizeof(m_header)); 995 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 996 } 997 998 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 999 lldb::offset_t *data_offset_ptr, 1000 llvm::MachO::mach_header &header) { 1001 data.SetByteOrder(endian::InlHostByteOrder()); 1002 // Leave magic in the original byte order 1003 header.magic = data.GetU32(data_offset_ptr); 1004 bool can_parse = false; 1005 bool is_64_bit = false; 1006 switch (header.magic) { 1007 case MH_MAGIC: 1008 data.SetByteOrder(endian::InlHostByteOrder()); 1009 data.SetAddressByteSize(4); 1010 can_parse = true; 1011 break; 1012 1013 case MH_MAGIC_64: 1014 data.SetByteOrder(endian::InlHostByteOrder()); 1015 data.SetAddressByteSize(8); 1016 can_parse = true; 1017 is_64_bit = true; 1018 break; 1019 1020 case MH_CIGAM: 1021 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1022 ? eByteOrderLittle 1023 : eByteOrderBig); 1024 data.SetAddressByteSize(4); 1025 can_parse = true; 1026 break; 1027 1028 case MH_CIGAM_64: 1029 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1030 ? eByteOrderLittle 1031 : eByteOrderBig); 1032 data.SetAddressByteSize(8); 1033 is_64_bit = true; 1034 can_parse = true; 1035 break; 1036 1037 default: 1038 break; 1039 } 1040 1041 if (can_parse) { 1042 data.GetU32(data_offset_ptr, &header.cputype, 6); 1043 if (is_64_bit) 1044 *data_offset_ptr += 4; 1045 return true; 1046 } else { 1047 memset(&header, 0, sizeof(header)); 1048 } 1049 return false; 1050 } 1051 1052 bool ObjectFileMachO::ParseHeader() { 1053 ModuleSP module_sp(GetModule()); 1054 if (module_sp) { 1055 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1056 bool can_parse = false; 1057 lldb::offset_t offset = 0; 1058 m_data.SetByteOrder(endian::InlHostByteOrder()); 1059 // Leave magic in the original byte order 1060 m_header.magic = m_data.GetU32(&offset); 1061 switch (m_header.magic) { 1062 case MH_MAGIC: 1063 m_data.SetByteOrder(endian::InlHostByteOrder()); 1064 m_data.SetAddressByteSize(4); 1065 can_parse = true; 1066 break; 1067 1068 case MH_MAGIC_64: 1069 m_data.SetByteOrder(endian::InlHostByteOrder()); 1070 m_data.SetAddressByteSize(8); 1071 can_parse = true; 1072 break; 1073 1074 case MH_CIGAM: 1075 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1076 ? eByteOrderLittle 1077 : eByteOrderBig); 1078 m_data.SetAddressByteSize(4); 1079 can_parse = true; 1080 break; 1081 1082 case MH_CIGAM_64: 1083 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1084 ? eByteOrderLittle 1085 : eByteOrderBig); 1086 m_data.SetAddressByteSize(8); 1087 can_parse = true; 1088 break; 1089 1090 default: 1091 break; 1092 } 1093 1094 if (can_parse) { 1095 m_data.GetU32(&offset, &m_header.cputype, 6); 1096 1097 ArchSpec mach_arch; 1098 1099 if (GetArchitecture(mach_arch)) { 1100 // Check if the module has a required architecture 1101 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1102 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1103 return false; 1104 1105 if (SetModulesArchitecture(mach_arch)) { 1106 const size_t header_and_lc_size = 1107 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1108 if (m_data.GetByteSize() < header_and_lc_size) { 1109 DataBufferSP data_sp; 1110 ProcessSP process_sp(m_process_wp.lock()); 1111 if (process_sp) { 1112 data_sp = 1113 ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1114 } else { 1115 // Read in all only the load command data from the file on disk 1116 data_sp = 1117 m_file.ReadFileContents(m_file_offset, header_and_lc_size); 1118 if (data_sp->GetByteSize() != header_and_lc_size) 1119 return false; 1120 } 1121 if (data_sp) 1122 m_data.SetData(data_sp); 1123 } 1124 } 1125 return true; 1126 } 1127 } else { 1128 memset(&m_header, 0, sizeof(struct mach_header)); 1129 } 1130 } 1131 return false; 1132 } 1133 1134 ByteOrder ObjectFileMachO::GetByteOrder() const { 1135 return m_data.GetByteOrder(); 1136 } 1137 1138 bool ObjectFileMachO::IsExecutable() const { 1139 return m_header.filetype == MH_EXECUTE; 1140 } 1141 1142 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1143 return m_data.GetAddressByteSize(); 1144 } 1145 1146 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1147 Symtab *symtab = GetSymtab(); 1148 if (symtab) { 1149 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1150 if (symbol) { 1151 if (symbol->ValueIsAddress()) { 1152 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1153 if (section_sp) { 1154 const lldb::SectionType section_type = section_sp->GetType(); 1155 switch (section_type) { 1156 case eSectionTypeInvalid: 1157 return eAddressClassUnknown; 1158 1159 case eSectionTypeCode: 1160 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1161 // For ARM we have a bit in the n_desc field of the symbol 1162 // that tells us ARM/Thumb which is bit 0x0008. 1163 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1164 return eAddressClassCodeAlternateISA; 1165 } 1166 return eAddressClassCode; 1167 1168 case eSectionTypeContainer: 1169 return eAddressClassUnknown; 1170 1171 case eSectionTypeData: 1172 case eSectionTypeDataCString: 1173 case eSectionTypeDataCStringPointers: 1174 case eSectionTypeDataSymbolAddress: 1175 case eSectionTypeData4: 1176 case eSectionTypeData8: 1177 case eSectionTypeData16: 1178 case eSectionTypeDataPointers: 1179 case eSectionTypeZeroFill: 1180 case eSectionTypeDataObjCMessageRefs: 1181 case eSectionTypeDataObjCCFStrings: 1182 case eSectionTypeGoSymtab: 1183 return eAddressClassData; 1184 1185 case eSectionTypeDebug: 1186 case eSectionTypeDWARFDebugAbbrev: 1187 case eSectionTypeDWARFDebugAddr: 1188 case eSectionTypeDWARFDebugAranges: 1189 case eSectionTypeDWARFDebugFrame: 1190 case eSectionTypeDWARFDebugInfo: 1191 case eSectionTypeDWARFDebugLine: 1192 case eSectionTypeDWARFDebugLoc: 1193 case eSectionTypeDWARFDebugMacInfo: 1194 case eSectionTypeDWARFDebugMacro: 1195 case eSectionTypeDWARFDebugPubNames: 1196 case eSectionTypeDWARFDebugPubTypes: 1197 case eSectionTypeDWARFDebugRanges: 1198 case eSectionTypeDWARFDebugStr: 1199 case eSectionTypeDWARFDebugStrOffsets: 1200 case eSectionTypeDWARFAppleNames: 1201 case eSectionTypeDWARFAppleTypes: 1202 case eSectionTypeDWARFAppleNamespaces: 1203 case eSectionTypeDWARFAppleObjC: 1204 return eAddressClassDebug; 1205 1206 case eSectionTypeEHFrame: 1207 case eSectionTypeARMexidx: 1208 case eSectionTypeARMextab: 1209 case eSectionTypeCompactUnwind: 1210 return eAddressClassRuntime; 1211 1212 case eSectionTypeAbsoluteAddress: 1213 case eSectionTypeELFSymbolTable: 1214 case eSectionTypeELFDynamicSymbols: 1215 case eSectionTypeELFRelocationEntries: 1216 case eSectionTypeELFDynamicLinkInfo: 1217 case eSectionTypeOther: 1218 return eAddressClassUnknown; 1219 } 1220 } 1221 } 1222 1223 const SymbolType symbol_type = symbol->GetType(); 1224 switch (symbol_type) { 1225 case eSymbolTypeAny: 1226 return eAddressClassUnknown; 1227 case eSymbolTypeAbsolute: 1228 return eAddressClassUnknown; 1229 1230 case eSymbolTypeCode: 1231 case eSymbolTypeTrampoline: 1232 case eSymbolTypeResolver: 1233 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1234 // For ARM we have a bit in the n_desc field of the symbol 1235 // that tells us ARM/Thumb which is bit 0x0008. 1236 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1237 return eAddressClassCodeAlternateISA; 1238 } 1239 return eAddressClassCode; 1240 1241 case eSymbolTypeData: 1242 return eAddressClassData; 1243 case eSymbolTypeRuntime: 1244 return eAddressClassRuntime; 1245 case eSymbolTypeException: 1246 return eAddressClassRuntime; 1247 case eSymbolTypeSourceFile: 1248 return eAddressClassDebug; 1249 case eSymbolTypeHeaderFile: 1250 return eAddressClassDebug; 1251 case eSymbolTypeObjectFile: 1252 return eAddressClassDebug; 1253 case eSymbolTypeCommonBlock: 1254 return eAddressClassDebug; 1255 case eSymbolTypeBlock: 1256 return eAddressClassDebug; 1257 case eSymbolTypeLocal: 1258 return eAddressClassData; 1259 case eSymbolTypeParam: 1260 return eAddressClassData; 1261 case eSymbolTypeVariable: 1262 return eAddressClassData; 1263 case eSymbolTypeVariableType: 1264 return eAddressClassDebug; 1265 case eSymbolTypeLineEntry: 1266 return eAddressClassDebug; 1267 case eSymbolTypeLineHeader: 1268 return eAddressClassDebug; 1269 case eSymbolTypeScopeBegin: 1270 return eAddressClassDebug; 1271 case eSymbolTypeScopeEnd: 1272 return eAddressClassDebug; 1273 case eSymbolTypeAdditional: 1274 return eAddressClassUnknown; 1275 case eSymbolTypeCompiler: 1276 return eAddressClassDebug; 1277 case eSymbolTypeInstrumentation: 1278 return eAddressClassDebug; 1279 case eSymbolTypeUndefined: 1280 return eAddressClassUnknown; 1281 case eSymbolTypeObjCClass: 1282 return eAddressClassRuntime; 1283 case eSymbolTypeObjCMetaClass: 1284 return eAddressClassRuntime; 1285 case eSymbolTypeObjCIVar: 1286 return eAddressClassRuntime; 1287 case eSymbolTypeReExported: 1288 return eAddressClassRuntime; 1289 } 1290 } 1291 } 1292 return eAddressClassUnknown; 1293 } 1294 1295 Symtab *ObjectFileMachO::GetSymtab() { 1296 ModuleSP module_sp(GetModule()); 1297 if (module_sp) { 1298 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1299 if (m_symtab_ap.get() == NULL) { 1300 m_symtab_ap.reset(new Symtab(this)); 1301 std::lock_guard<std::recursive_mutex> symtab_guard( 1302 m_symtab_ap->GetMutex()); 1303 ParseSymtab(); 1304 m_symtab_ap->Finalize(); 1305 } 1306 } 1307 return m_symtab_ap.get(); 1308 } 1309 1310 bool ObjectFileMachO::IsStripped() { 1311 if (m_dysymtab.cmd == 0) { 1312 ModuleSP module_sp(GetModule()); 1313 if (module_sp) { 1314 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1315 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1316 const lldb::offset_t load_cmd_offset = offset; 1317 1318 load_command lc; 1319 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 1320 break; 1321 if (lc.cmd == LC_DYSYMTAB) { 1322 m_dysymtab.cmd = lc.cmd; 1323 m_dysymtab.cmdsize = lc.cmdsize; 1324 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1325 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1326 NULL) { 1327 // Clear m_dysymtab if we were unable to read all items from the 1328 // load command 1329 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1330 } 1331 } 1332 offset = load_cmd_offset + lc.cmdsize; 1333 } 1334 } 1335 } 1336 if (m_dysymtab.cmd) 1337 return m_dysymtab.nlocalsym <= 1; 1338 return false; 1339 } 1340 1341 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1342 if (!m_sections_ap.get()) { 1343 m_sections_ap.reset(new SectionList()); 1344 1345 const bool is_dsym = (m_header.filetype == MH_DSYM); 1346 lldb::user_id_t segID = 0; 1347 lldb::user_id_t sectID = 0; 1348 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1349 uint32_t i; 1350 const bool is_core = GetType() == eTypeCoreFile; 1351 // bool dump_sections = false; 1352 ModuleSP module_sp(GetModule()); 1353 // First look up any LC_ENCRYPTION_INFO load commands 1354 typedef RangeArray<uint32_t, uint32_t, 8> EncryptedFileRanges; 1355 EncryptedFileRanges encrypted_file_ranges; 1356 encryption_info_command encryption_cmd; 1357 for (i = 0; i < m_header.ncmds; ++i) { 1358 const lldb::offset_t load_cmd_offset = offset; 1359 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1360 break; 1361 1362 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for 1363 // the 3 fields we care about, so treat them the same. 1364 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1365 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1366 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1367 if (encryption_cmd.cryptid != 0) { 1368 EncryptedFileRanges::Entry entry; 1369 entry.SetRangeBase(encryption_cmd.cryptoff); 1370 entry.SetByteSize(encryption_cmd.cryptsize); 1371 encrypted_file_ranges.Append(entry); 1372 } 1373 } 1374 } 1375 offset = load_cmd_offset + encryption_cmd.cmdsize; 1376 } 1377 1378 bool section_file_addresses_changed = false; 1379 1380 offset = MachHeaderSizeFromMagic(m_header.magic); 1381 1382 struct segment_command_64 load_cmd; 1383 for (i = 0; i < m_header.ncmds; ++i) { 1384 const lldb::offset_t load_cmd_offset = offset; 1385 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1386 break; 1387 1388 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) { 1389 if (m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) { 1390 bool add_section = true; 1391 bool add_to_unified = true; 1392 ConstString const_segname(load_cmd.segname, 1393 std::min<size_t>(strlen(load_cmd.segname), 1394 sizeof(load_cmd.segname))); 1395 1396 SectionSP unified_section_sp( 1397 unified_section_list.FindSectionByName(const_segname)); 1398 if (is_dsym && unified_section_sp) { 1399 if (const_segname == GetSegmentNameLINKEDIT()) { 1400 // We need to keep the __LINKEDIT segment private to this object 1401 // file only 1402 add_to_unified = false; 1403 } else { 1404 // This is the dSYM file and this section has already been created 1405 // by 1406 // the object file, no need to create it. 1407 add_section = false; 1408 } 1409 } 1410 load_cmd.vmaddr = m_data.GetAddress(&offset); 1411 load_cmd.vmsize = m_data.GetAddress(&offset); 1412 load_cmd.fileoff = m_data.GetAddress(&offset); 1413 load_cmd.filesize = m_data.GetAddress(&offset); 1414 if (m_length != 0 && load_cmd.filesize != 0) { 1415 if (load_cmd.fileoff > m_length) { 1416 // We have a load command that says it extends past the end of the 1417 // file. This is likely 1418 // a corrupt file. We don't have any way to return an error 1419 // condition here (this method 1420 // was likely invoked from something like 1421 // ObjectFile::GetSectionList()) -- all we can do 1422 // is null out the SectionList vector and if a process has been 1423 // set up, dump a message 1424 // to stdout. The most common case here is core file debugging 1425 // with a truncated file. 1426 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 1427 ? "LC_SEGMENT_64" 1428 : "LC_SEGMENT"; 1429 module_sp->ReportWarning( 1430 "load command %u %s has a fileoff (0x%" PRIx64 1431 ") that extends beyond the end of the file (0x%" PRIx64 1432 "), ignoring this section", 1433 i, lc_segment_name, load_cmd.fileoff, m_length); 1434 1435 load_cmd.fileoff = 0; 1436 load_cmd.filesize = 0; 1437 } 1438 1439 if (load_cmd.fileoff + load_cmd.filesize > m_length) { 1440 // We have a load command that says it extends past the end of the 1441 // file. This is likely 1442 // a corrupt file. We don't have any way to return an error 1443 // condition here (this method 1444 // was likely invoked from something like 1445 // ObjectFile::GetSectionList()) -- all we can do 1446 // is null out the SectionList vector and if a process has been 1447 // set up, dump a message 1448 // to stdout. The most common case here is core file debugging 1449 // with a truncated file. 1450 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 1451 ? "LC_SEGMENT_64" 1452 : "LC_SEGMENT"; 1453 GetModule()->ReportWarning( 1454 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1455 ") that extends beyond the end of the file (0x%" PRIx64 1456 "), the segment will be truncated to match", 1457 i, lc_segment_name, load_cmd.fileoff + load_cmd.filesize, 1458 m_length); 1459 1460 // Tuncase the length 1461 load_cmd.filesize = m_length - load_cmd.fileoff; 1462 } 1463 } 1464 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) { 1465 uint32_t segment_permissions = 0; 1466 if (load_cmd.initprot & VM_PROT_READ) 1467 segment_permissions |= ePermissionsReadable; 1468 if (load_cmd.initprot & VM_PROT_WRITE) 1469 segment_permissions |= ePermissionsWritable; 1470 if (load_cmd.initprot & VM_PROT_EXECUTE) 1471 segment_permissions |= ePermissionsExecutable; 1472 1473 const bool segment_is_encrypted = 1474 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1475 1476 // Keep a list of mach segments around in case we need to 1477 // get at data that isn't stored in the abstracted Sections. 1478 m_mach_segments.push_back(load_cmd); 1479 1480 // Use a segment ID of the segment index shifted left by 8 so they 1481 // never conflict with any of the sections. 1482 SectionSP segment_sp; 1483 if (add_section && (const_segname || is_core)) { 1484 segment_sp.reset(new Section( 1485 module_sp, // Module to which this section belongs 1486 this, // Object file to which this sections belongs 1487 ++segID << 8, // Section ID is the 1 based segment index 1488 // shifted right by 8 bits as not to collide 1489 // with any of the 256 section IDs that are 1490 // possible 1491 const_segname, // Name of this section 1492 eSectionTypeContainer, // This section is a container of other 1493 // sections. 1494 load_cmd.vmaddr, // File VM address == addresses as they are 1495 // found in the object file 1496 load_cmd.vmsize, // VM size in bytes of this section 1497 load_cmd.fileoff, // Offset to the data for this section in 1498 // the file 1499 load_cmd.filesize, // Size in bytes of this section as found 1500 // in the file 1501 0, // Segments have no alignment information 1502 load_cmd.flags)); // Flags for this section 1503 1504 segment_sp->SetIsEncrypted(segment_is_encrypted); 1505 m_sections_ap->AddSection(segment_sp); 1506 segment_sp->SetPermissions(segment_permissions); 1507 if (add_to_unified) 1508 unified_section_list.AddSection(segment_sp); 1509 } else if (unified_section_sp) { 1510 if (is_dsym && 1511 unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1512 // Check to see if the module was read from memory? 1513 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) { 1514 // We have a module that is in memory and needs to have its 1515 // file address adjusted. We need to do this because when we 1516 // load a file from memory, its addresses will be slid 1517 // already, 1518 // yet the addresses in the new symbol file will still be 1519 // unslid. 1520 // Since everything is stored as section offset, this 1521 // shouldn't 1522 // cause any problems. 1523 1524 // Make sure we've parsed the symbol table from the 1525 // ObjectFile before we go around changing its Sections. 1526 module_sp->GetObjectFile()->GetSymtab(); 1527 // eh_frame would present the same problems but we parse that 1528 // on 1529 // a per-function basis as-needed so it's more difficult to 1530 // remove its use of the Sections. Realistically, the 1531 // environments 1532 // where this code path will be taken will not have eh_frame 1533 // sections. 1534 1535 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1536 1537 // Notify the module that the section addresses have been 1538 // changed once 1539 // we're done so any file-address caches can be updated. 1540 section_file_addresses_changed = true; 1541 } 1542 } 1543 m_sections_ap->AddSection(unified_section_sp); 1544 } 1545 1546 struct section_64 sect64; 1547 ::memset(§64, 0, sizeof(sect64)); 1548 // Push a section into our mach sections for the section at 1549 // index zero (NO_SECT) if we don't have any mach sections yet... 1550 if (m_mach_sections.empty()) 1551 m_mach_sections.push_back(sect64); 1552 uint32_t segment_sect_idx; 1553 const lldb::user_id_t first_segment_sectID = sectID + 1; 1554 1555 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1556 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1557 ++segment_sect_idx) { 1558 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1559 sizeof(sect64.sectname)) == NULL) 1560 break; 1561 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1562 sizeof(sect64.segname)) == NULL) 1563 break; 1564 sect64.addr = m_data.GetAddress(&offset); 1565 sect64.size = m_data.GetAddress(&offset); 1566 1567 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1568 break; 1569 1570 // Keep a list of mach sections around in case we need to 1571 // get at data that isn't stored in the abstracted Sections. 1572 m_mach_sections.push_back(sect64); 1573 1574 if (add_section) { 1575 ConstString section_name( 1576 sect64.sectname, std::min<size_t>(strlen(sect64.sectname), 1577 sizeof(sect64.sectname))); 1578 if (!const_segname) { 1579 // We have a segment with no name so we need to conjure up 1580 // segments that correspond to the section's segname if there 1581 // isn't already such a section. If there is such a section, 1582 // we resize the section so that it spans all sections. 1583 // We also mark these sections as fake so address matches 1584 // don't 1585 // hit if they land in the gaps between the child sections. 1586 const_segname.SetTrimmedCStringWithLength( 1587 sect64.segname, sizeof(sect64.segname)); 1588 segment_sp = 1589 unified_section_list.FindSectionByName(const_segname); 1590 if (segment_sp.get()) { 1591 Section *segment = segment_sp.get(); 1592 // Grow the section size as needed. 1593 const lldb::addr_t sect64_min_addr = sect64.addr; 1594 const lldb::addr_t sect64_max_addr = 1595 sect64_min_addr + sect64.size; 1596 const lldb::addr_t curr_seg_byte_size = 1597 segment->GetByteSize(); 1598 const lldb::addr_t curr_seg_min_addr = 1599 segment->GetFileAddress(); 1600 const lldb::addr_t curr_seg_max_addr = 1601 curr_seg_min_addr + curr_seg_byte_size; 1602 if (sect64_min_addr >= curr_seg_min_addr) { 1603 const lldb::addr_t new_seg_byte_size = 1604 sect64_max_addr - curr_seg_min_addr; 1605 // Only grow the section size if needed 1606 if (new_seg_byte_size > curr_seg_byte_size) 1607 segment->SetByteSize(new_seg_byte_size); 1608 } else { 1609 // We need to change the base address of the segment and 1610 // adjust the child section offsets for all existing 1611 // children. 1612 const lldb::addr_t slide_amount = 1613 sect64_min_addr - curr_seg_min_addr; 1614 segment->Slide(slide_amount, false); 1615 segment->GetChildren().Slide(-slide_amount, false); 1616 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1617 } 1618 1619 // Grow the section size as needed. 1620 if (sect64.offset) { 1621 const lldb::addr_t segment_min_file_offset = 1622 segment->GetFileOffset(); 1623 const lldb::addr_t segment_max_file_offset = 1624 segment_min_file_offset + segment->GetFileSize(); 1625 1626 const lldb::addr_t section_min_file_offset = 1627 sect64.offset; 1628 const lldb::addr_t section_max_file_offset = 1629 section_min_file_offset + sect64.size; 1630 const lldb::addr_t new_file_offset = std::min( 1631 section_min_file_offset, segment_min_file_offset); 1632 const lldb::addr_t new_file_size = 1633 std::max(section_max_file_offset, 1634 segment_max_file_offset) - 1635 new_file_offset; 1636 segment->SetFileOffset(new_file_offset); 1637 segment->SetFileSize(new_file_size); 1638 } 1639 } else { 1640 // Create a fake section for the section's named segment 1641 segment_sp.reset(new Section( 1642 segment_sp, // Parent section 1643 module_sp, // Module to which this section belongs 1644 this, // Object file to which this section belongs 1645 ++segID << 8, // Section ID is the 1 based segment index 1646 // shifted right by 8 bits as not to 1647 // collide with any of the 256 section IDs 1648 // that are possible 1649 const_segname, // Name of this section 1650 eSectionTypeContainer, // This section is a container of 1651 // other sections. 1652 sect64.addr, // File VM address == addresses as they are 1653 // found in the object file 1654 sect64.size, // VM size in bytes of this section 1655 sect64.offset, // Offset to the data for this section in 1656 // the file 1657 sect64.offset ? sect64.size : 0, // Size in bytes of 1658 // this section as 1659 // found in the file 1660 sect64.align, 1661 load_cmd.flags)); // Flags for this section 1662 segment_sp->SetIsFake(true); 1663 segment_sp->SetPermissions(segment_permissions); 1664 m_sections_ap->AddSection(segment_sp); 1665 if (add_to_unified) 1666 unified_section_list.AddSection(segment_sp); 1667 segment_sp->SetIsEncrypted(segment_is_encrypted); 1668 } 1669 } 1670 assert(segment_sp.get()); 1671 1672 lldb::SectionType sect_type = eSectionTypeOther; 1673 1674 if (sect64.flags & 1675 (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1676 sect_type = eSectionTypeCode; 1677 else { 1678 uint32_t mach_sect_type = sect64.flags & SECTION_TYPE; 1679 static ConstString g_sect_name_objc_data("__objc_data"); 1680 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1681 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1682 static ConstString g_sect_name_objc_classrefs( 1683 "__objc_classrefs"); 1684 static ConstString g_sect_name_objc_superrefs( 1685 "__objc_superrefs"); 1686 static ConstString g_sect_name_objc_const("__objc_const"); 1687 static ConstString g_sect_name_objc_classlist( 1688 "__objc_classlist"); 1689 static ConstString g_sect_name_cfstring("__cfstring"); 1690 1691 static ConstString g_sect_name_dwarf_debug_abbrev( 1692 "__debug_abbrev"); 1693 static ConstString g_sect_name_dwarf_debug_aranges( 1694 "__debug_aranges"); 1695 static ConstString g_sect_name_dwarf_debug_frame( 1696 "__debug_frame"); 1697 static ConstString g_sect_name_dwarf_debug_info( 1698 "__debug_info"); 1699 static ConstString g_sect_name_dwarf_debug_line( 1700 "__debug_line"); 1701 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1702 static ConstString g_sect_name_dwarf_debug_macinfo( 1703 "__debug_macinfo"); 1704 static ConstString g_sect_name_dwarf_debug_pubnames( 1705 "__debug_pubnames"); 1706 static ConstString g_sect_name_dwarf_debug_pubtypes( 1707 "__debug_pubtypes"); 1708 static ConstString g_sect_name_dwarf_debug_ranges( 1709 "__debug_ranges"); 1710 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1711 static ConstString g_sect_name_dwarf_apple_names( 1712 "__apple_names"); 1713 static ConstString g_sect_name_dwarf_apple_types( 1714 "__apple_types"); 1715 static ConstString g_sect_name_dwarf_apple_namespaces( 1716 "__apple_namespac"); 1717 static ConstString g_sect_name_dwarf_apple_objc( 1718 "__apple_objc"); 1719 static ConstString g_sect_name_eh_frame("__eh_frame"); 1720 static ConstString g_sect_name_compact_unwind( 1721 "__unwind_info"); 1722 static ConstString g_sect_name_text("__text"); 1723 static ConstString g_sect_name_data("__data"); 1724 static ConstString g_sect_name_go_symtab("__gosymtab"); 1725 1726 if (section_name == g_sect_name_dwarf_debug_abbrev) 1727 sect_type = eSectionTypeDWARFDebugAbbrev; 1728 else if (section_name == g_sect_name_dwarf_debug_aranges) 1729 sect_type = eSectionTypeDWARFDebugAranges; 1730 else if (section_name == g_sect_name_dwarf_debug_frame) 1731 sect_type = eSectionTypeDWARFDebugFrame; 1732 else if (section_name == g_sect_name_dwarf_debug_info) 1733 sect_type = eSectionTypeDWARFDebugInfo; 1734 else if (section_name == g_sect_name_dwarf_debug_line) 1735 sect_type = eSectionTypeDWARFDebugLine; 1736 else if (section_name == g_sect_name_dwarf_debug_loc) 1737 sect_type = eSectionTypeDWARFDebugLoc; 1738 else if (section_name == g_sect_name_dwarf_debug_macinfo) 1739 sect_type = eSectionTypeDWARFDebugMacInfo; 1740 else if (section_name == g_sect_name_dwarf_debug_pubnames) 1741 sect_type = eSectionTypeDWARFDebugPubNames; 1742 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 1743 sect_type = eSectionTypeDWARFDebugPubTypes; 1744 else if (section_name == g_sect_name_dwarf_debug_ranges) 1745 sect_type = eSectionTypeDWARFDebugRanges; 1746 else if (section_name == g_sect_name_dwarf_debug_str) 1747 sect_type = eSectionTypeDWARFDebugStr; 1748 else if (section_name == g_sect_name_dwarf_apple_names) 1749 sect_type = eSectionTypeDWARFAppleNames; 1750 else if (section_name == g_sect_name_dwarf_apple_types) 1751 sect_type = eSectionTypeDWARFAppleTypes; 1752 else if (section_name == g_sect_name_dwarf_apple_namespaces) 1753 sect_type = eSectionTypeDWARFAppleNamespaces; 1754 else if (section_name == g_sect_name_dwarf_apple_objc) 1755 sect_type = eSectionTypeDWARFAppleObjC; 1756 else if (section_name == g_sect_name_objc_selrefs) 1757 sect_type = eSectionTypeDataCStringPointers; 1758 else if (section_name == g_sect_name_objc_msgrefs) 1759 sect_type = eSectionTypeDataObjCMessageRefs; 1760 else if (section_name == g_sect_name_eh_frame) 1761 sect_type = eSectionTypeEHFrame; 1762 else if (section_name == g_sect_name_compact_unwind) 1763 sect_type = eSectionTypeCompactUnwind; 1764 else if (section_name == g_sect_name_cfstring) 1765 sect_type = eSectionTypeDataObjCCFStrings; 1766 else if (section_name == g_sect_name_go_symtab) 1767 sect_type = eSectionTypeGoSymtab; 1768 else if (section_name == g_sect_name_objc_data || 1769 section_name == g_sect_name_objc_classrefs || 1770 section_name == g_sect_name_objc_superrefs || 1771 section_name == g_sect_name_objc_const || 1772 section_name == g_sect_name_objc_classlist) { 1773 sect_type = eSectionTypeDataPointers; 1774 } 1775 1776 if (sect_type == eSectionTypeOther) { 1777 switch (mach_sect_type) { 1778 // TODO: categorize sections by other flags for regular 1779 // sections 1780 case S_REGULAR: 1781 if (section_name == g_sect_name_text) 1782 sect_type = eSectionTypeCode; 1783 else if (section_name == g_sect_name_data) 1784 sect_type = eSectionTypeData; 1785 else 1786 sect_type = eSectionTypeOther; 1787 break; 1788 case S_ZEROFILL: 1789 sect_type = eSectionTypeZeroFill; 1790 break; 1791 case S_CSTRING_LITERALS: 1792 sect_type = eSectionTypeDataCString; 1793 break; // section with only literal C strings 1794 case S_4BYTE_LITERALS: 1795 sect_type = eSectionTypeData4; 1796 break; // section with only 4 byte literals 1797 case S_8BYTE_LITERALS: 1798 sect_type = eSectionTypeData8; 1799 break; // section with only 8 byte literals 1800 case S_LITERAL_POINTERS: 1801 sect_type = eSectionTypeDataPointers; 1802 break; // section with only pointers to literals 1803 case S_NON_LAZY_SYMBOL_POINTERS: 1804 sect_type = eSectionTypeDataPointers; 1805 break; // section with only non-lazy symbol pointers 1806 case S_LAZY_SYMBOL_POINTERS: 1807 sect_type = eSectionTypeDataPointers; 1808 break; // section with only lazy symbol pointers 1809 case S_SYMBOL_STUBS: 1810 sect_type = eSectionTypeCode; 1811 break; // section with only symbol stubs, byte size of 1812 // stub in the reserved2 field 1813 case S_MOD_INIT_FUNC_POINTERS: 1814 sect_type = eSectionTypeDataPointers; 1815 break; // section with only function pointers for 1816 // initialization 1817 case S_MOD_TERM_FUNC_POINTERS: 1818 sect_type = eSectionTypeDataPointers; 1819 break; // section with only function pointers for 1820 // termination 1821 case S_COALESCED: 1822 sect_type = eSectionTypeOther; 1823 break; 1824 case S_GB_ZEROFILL: 1825 sect_type = eSectionTypeZeroFill; 1826 break; 1827 case S_INTERPOSING: 1828 sect_type = eSectionTypeCode; 1829 break; // section with only pairs of function pointers for 1830 // interposing 1831 case S_16BYTE_LITERALS: 1832 sect_type = eSectionTypeData16; 1833 break; // section with only 16 byte literals 1834 case S_DTRACE_DOF: 1835 sect_type = eSectionTypeDebug; 1836 break; 1837 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1838 sect_type = eSectionTypeDataPointers; 1839 break; 1840 default: 1841 break; 1842 } 1843 } 1844 } 1845 1846 SectionSP section_sp(new Section( 1847 segment_sp, module_sp, this, ++sectID, section_name, 1848 sect_type, sect64.addr - segment_sp->GetFileAddress(), 1849 sect64.size, sect64.offset, 1850 sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1851 sect64.flags)); 1852 // Set the section to be encrypted to match the segment 1853 1854 bool section_is_encrypted = false; 1855 if (!segment_is_encrypted && load_cmd.filesize != 0) 1856 section_is_encrypted = 1857 encrypted_file_ranges.FindEntryThatContains( 1858 sect64.offset) != NULL; 1859 1860 section_sp->SetIsEncrypted(segment_is_encrypted || 1861 section_is_encrypted); 1862 section_sp->SetPermissions(segment_permissions); 1863 segment_sp->GetChildren().AddSection(section_sp); 1864 1865 if (segment_sp->IsFake()) { 1866 segment_sp.reset(); 1867 const_segname.Clear(); 1868 } 1869 } 1870 } 1871 if (segment_sp && is_dsym) { 1872 if (first_segment_sectID <= sectID) { 1873 lldb::user_id_t sect_uid; 1874 for (sect_uid = first_segment_sectID; sect_uid <= sectID; 1875 ++sect_uid) { 1876 SectionSP curr_section_sp( 1877 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1878 SectionSP next_section_sp; 1879 if (sect_uid + 1 <= sectID) 1880 next_section_sp = 1881 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1882 1883 if (curr_section_sp.get()) { 1884 if (curr_section_sp->GetByteSize() == 0) { 1885 if (next_section_sp.get() != NULL) 1886 curr_section_sp->SetByteSize( 1887 next_section_sp->GetFileAddress() - 1888 curr_section_sp->GetFileAddress()); 1889 else 1890 curr_section_sp->SetByteSize(load_cmd.vmsize); 1891 } 1892 } 1893 } 1894 } 1895 } 1896 } 1897 } 1898 } else if (load_cmd.cmd == LC_DYSYMTAB) { 1899 m_dysymtab.cmd = load_cmd.cmd; 1900 m_dysymtab.cmdsize = load_cmd.cmdsize; 1901 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1902 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1903 } 1904 1905 offset = load_cmd_offset + load_cmd.cmdsize; 1906 } 1907 1908 if (section_file_addresses_changed && module_sp.get()) { 1909 module_sp->SectionFileAddressesChanged(); 1910 } 1911 } 1912 } 1913 1914 class MachSymtabSectionInfo { 1915 public: 1916 MachSymtabSectionInfo(SectionList *section_list) 1917 : m_section_list(section_list), m_section_infos() { 1918 // Get the number of sections down to a depth of 1 to include 1919 // all segments and their sections, but no other sections that 1920 // may be added for debug map or 1921 m_section_infos.resize(section_list->GetNumSections(1)); 1922 } 1923 1924 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1925 if (n_sect == 0) 1926 return SectionSP(); 1927 if (n_sect < m_section_infos.size()) { 1928 if (!m_section_infos[n_sect].section_sp) { 1929 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1930 m_section_infos[n_sect].section_sp = section_sp; 1931 if (section_sp) { 1932 m_section_infos[n_sect].vm_range.SetBaseAddress( 1933 section_sp->GetFileAddress()); 1934 m_section_infos[n_sect].vm_range.SetByteSize( 1935 section_sp->GetByteSize()); 1936 } else { 1937 Host::SystemLog(Host::eSystemLogError, 1938 "error: unable to find section for section %u\n", 1939 n_sect); 1940 } 1941 } 1942 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1943 // Symbol is in section. 1944 return m_section_infos[n_sect].section_sp; 1945 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1946 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1947 file_addr) { 1948 // Symbol is in section with zero size, but has the same start 1949 // address as the section. This can happen with linker symbols 1950 // (symbols that start with the letter 'l' or 'L'. 1951 return m_section_infos[n_sect].section_sp; 1952 } 1953 } 1954 return m_section_list->FindSectionContainingFileAddress(file_addr); 1955 } 1956 1957 protected: 1958 struct SectionInfo { 1959 SectionInfo() : vm_range(), section_sp() {} 1960 1961 VMRange vm_range; 1962 SectionSP section_sp; 1963 }; 1964 SectionList *m_section_list; 1965 std::vector<SectionInfo> m_section_infos; 1966 }; 1967 1968 struct TrieEntry { 1969 TrieEntry() 1970 : name(), address(LLDB_INVALID_ADDRESS), flags(0), other(0), 1971 import_name() {} 1972 1973 void Clear() { 1974 name.Clear(); 1975 address = LLDB_INVALID_ADDRESS; 1976 flags = 0; 1977 other = 0; 1978 import_name.Clear(); 1979 } 1980 1981 void Dump() const { 1982 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1983 static_cast<unsigned long long>(address), 1984 static_cast<unsigned long long>(flags), 1985 static_cast<unsigned long long>(other), name.GetCString()); 1986 if (import_name) 1987 printf(" -> \"%s\"\n", import_name.GetCString()); 1988 else 1989 printf("\n"); 1990 } 1991 ConstString name; 1992 uint64_t address; 1993 uint64_t flags; 1994 uint64_t other; 1995 ConstString import_name; 1996 }; 1997 1998 struct TrieEntryWithOffset { 1999 lldb::offset_t nodeOffset; 2000 TrieEntry entry; 2001 2002 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 2003 2004 void Dump(uint32_t idx) const { 2005 printf("[%3u] 0x%16.16llx: ", idx, 2006 static_cast<unsigned long long>(nodeOffset)); 2007 entry.Dump(); 2008 } 2009 2010 bool operator<(const TrieEntryWithOffset &other) const { 2011 return (nodeOffset < other.nodeOffset); 2012 } 2013 }; 2014 2015 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 2016 const bool is_arm, 2017 std::vector<llvm::StringRef> &nameSlices, 2018 std::set<lldb::addr_t> &resolver_addresses, 2019 std::vector<TrieEntryWithOffset> &output) { 2020 if (!data.ValidOffset(offset)) 2021 return true; 2022 2023 const uint64_t terminalSize = data.GetULEB128(&offset); 2024 lldb::offset_t children_offset = offset + terminalSize; 2025 if (terminalSize != 0) { 2026 TrieEntryWithOffset e(offset); 2027 e.entry.flags = data.GetULEB128(&offset); 2028 const char *import_name = NULL; 2029 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2030 e.entry.address = 0; 2031 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2032 import_name = data.GetCStr(&offset); 2033 } else { 2034 e.entry.address = data.GetULEB128(&offset); 2035 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2036 e.entry.other = data.GetULEB128(&offset); 2037 uint64_t resolver_addr = e.entry.other; 2038 if (is_arm) 2039 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2040 resolver_addresses.insert(resolver_addr); 2041 } else 2042 e.entry.other = 0; 2043 } 2044 // Only add symbols that are reexport symbols with a valid import name 2045 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && 2046 import_name[0]) { 2047 std::string name; 2048 if (!nameSlices.empty()) { 2049 for (auto name_slice : nameSlices) 2050 name.append(name_slice.data(), name_slice.size()); 2051 } 2052 if (name.size() > 1) { 2053 // Skip the leading '_' 2054 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2055 } 2056 if (import_name) { 2057 // Skip the leading '_' 2058 e.entry.import_name.SetCString(import_name + 1); 2059 } 2060 output.push_back(e); 2061 } 2062 } 2063 2064 const uint8_t childrenCount = data.GetU8(&children_offset); 2065 for (uint8_t i = 0; i < childrenCount; ++i) { 2066 const char *cstr = data.GetCStr(&children_offset); 2067 if (cstr) 2068 nameSlices.push_back(llvm::StringRef(cstr)); 2069 else 2070 return false; // Corrupt data 2071 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2072 if (childNodeOffset) { 2073 if (!ParseTrieEntries(data, childNodeOffset, is_arm, nameSlices, 2074 resolver_addresses, output)) { 2075 return false; 2076 } 2077 } 2078 nameSlices.pop_back(); 2079 } 2080 return true; 2081 } 2082 2083 // Read the UUID out of a dyld_shared_cache file on-disk. 2084 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2085 const ByteOrder byte_order, 2086 const uint32_t addr_byte_size) { 2087 UUID dsc_uuid; 2088 DataBufferSP dsc_data_sp = dyld_shared_cache.MemoryMapFileContentsIfLocal( 2089 0, sizeof(struct lldb_copy_dyld_cache_header_v1)); 2090 if (dsc_data_sp) { 2091 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2092 2093 char version_str[7]; 2094 lldb::offset_t offset = 0; 2095 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2096 version_str[6] = '\0'; 2097 if (strcmp(version_str, "dyld_v") == 0) { 2098 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2099 uint8_t uuid_bytes[sizeof(uuid_t)]; 2100 memcpy(uuid_bytes, dsc_header_data.GetData(&offset, sizeof(uuid_t)), 2101 sizeof(uuid_t)); 2102 dsc_uuid.SetBytes(uuid_bytes); 2103 } 2104 } 2105 return dsc_uuid; 2106 } 2107 2108 size_t ObjectFileMachO::ParseSymtab() { 2109 Timer scoped_timer(LLVM_PRETTY_FUNCTION, 2110 "ObjectFileMachO::ParseSymtab () module = %s", 2111 m_file.GetFilename().AsCString("")); 2112 ModuleSP module_sp(GetModule()); 2113 if (!module_sp) 2114 return 0; 2115 2116 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2117 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2118 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2119 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2120 FunctionStarts function_starts; 2121 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2122 uint32_t i; 2123 FileSpecList dylib_files; 2124 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2125 static const llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2126 static const llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2127 static const llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2128 2129 for (i = 0; i < m_header.ncmds; ++i) { 2130 const lldb::offset_t cmd_offset = offset; 2131 // Read in the load command and load command size 2132 struct load_command lc; 2133 if (m_data.GetU32(&offset, &lc, 2) == NULL) 2134 break; 2135 // Watch for the symbol table load command 2136 switch (lc.cmd) { 2137 case LC_SYMTAB: 2138 symtab_load_command.cmd = lc.cmd; 2139 symtab_load_command.cmdsize = lc.cmdsize; 2140 // Read in the rest of the symtab load command 2141 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2142 0) // fill in symoff, nsyms, stroff, strsize fields 2143 return 0; 2144 if (symtab_load_command.symoff == 0) { 2145 if (log) 2146 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2147 return 0; 2148 } 2149 2150 if (symtab_load_command.stroff == 0) { 2151 if (log) 2152 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2153 return 0; 2154 } 2155 2156 if (symtab_load_command.nsyms == 0) { 2157 if (log) 2158 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2159 return 0; 2160 } 2161 2162 if (symtab_load_command.strsize == 0) { 2163 if (log) 2164 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2165 return 0; 2166 } 2167 break; 2168 2169 case LC_DYLD_INFO: 2170 case LC_DYLD_INFO_ONLY: 2171 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2172 dyld_info.cmd = lc.cmd; 2173 dyld_info.cmdsize = lc.cmdsize; 2174 } else { 2175 memset(&dyld_info, 0, sizeof(dyld_info)); 2176 } 2177 break; 2178 2179 case LC_LOAD_DYLIB: 2180 case LC_LOAD_WEAK_DYLIB: 2181 case LC_REEXPORT_DYLIB: 2182 case LC_LOADFVMLIB: 2183 case LC_LOAD_UPWARD_DYLIB: { 2184 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2185 const char *path = m_data.PeekCStr(name_offset); 2186 if (path) { 2187 FileSpec file_spec(path, false); 2188 // Strip the path if there is @rpath, @executable, etc so we just use 2189 // the basename 2190 if (path[0] == '@') 2191 file_spec.GetDirectory().Clear(); 2192 2193 if (lc.cmd == LC_REEXPORT_DYLIB) { 2194 m_reexported_dylibs.AppendIfUnique(file_spec); 2195 } 2196 2197 dylib_files.Append(file_spec); 2198 } 2199 } break; 2200 2201 case LC_FUNCTION_STARTS: 2202 function_starts_load_command.cmd = lc.cmd; 2203 function_starts_load_command.cmdsize = lc.cmdsize; 2204 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2205 NULL) // fill in symoff, nsyms, stroff, strsize fields 2206 memset(&function_starts_load_command, 0, 2207 sizeof(function_starts_load_command)); 2208 break; 2209 2210 default: 2211 break; 2212 } 2213 offset = cmd_offset + lc.cmdsize; 2214 } 2215 2216 if (symtab_load_command.cmd) { 2217 Symtab *symtab = m_symtab_ap.get(); 2218 SectionList *section_list = GetSectionList(); 2219 if (section_list == NULL) 2220 return 0; 2221 2222 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2223 const ByteOrder byte_order = m_data.GetByteOrder(); 2224 bool bit_width_32 = addr_byte_size == 4; 2225 const size_t nlist_byte_size = 2226 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2227 2228 DataExtractor nlist_data(NULL, 0, byte_order, addr_byte_size); 2229 DataExtractor strtab_data(NULL, 0, byte_order, addr_byte_size); 2230 DataExtractor function_starts_data(NULL, 0, byte_order, addr_byte_size); 2231 DataExtractor indirect_symbol_index_data(NULL, 0, byte_order, 2232 addr_byte_size); 2233 DataExtractor dyld_trie_data(NULL, 0, byte_order, addr_byte_size); 2234 2235 const addr_t nlist_data_byte_size = 2236 symtab_load_command.nsyms * nlist_byte_size; 2237 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2238 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2239 2240 ProcessSP process_sp(m_process_wp.lock()); 2241 Process *process = process_sp.get(); 2242 2243 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2244 2245 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2246 Target &target = process->GetTarget(); 2247 2248 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2249 2250 SectionSP linkedit_section_sp( 2251 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2252 // Reading mach file from memory in a process or core file... 2253 2254 if (linkedit_section_sp) { 2255 addr_t linkedit_load_addr = 2256 linkedit_section_sp->GetLoadBaseAddress(&target); 2257 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2258 // We might be trying to access the symbol table before the 2259 // __LINKEDIT's load 2260 // address has been set in the target. We can't fail to read the 2261 // symbol table, 2262 // so calculate the right address manually 2263 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2264 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2265 } 2266 2267 const addr_t linkedit_file_offset = 2268 linkedit_section_sp->GetFileOffset(); 2269 const addr_t symoff_addr = linkedit_load_addr + 2270 symtab_load_command.symoff - 2271 linkedit_file_offset; 2272 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2273 linkedit_file_offset; 2274 2275 bool data_was_read = false; 2276 2277 #if defined(__APPLE__) && \ 2278 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2279 if (m_header.flags & 0x80000000u && 2280 process->GetAddressByteSize() == sizeof(void *)) { 2281 // This mach-o memory file is in the dyld shared cache. If this 2282 // program is not remote and this is iOS, then this process will 2283 // share the same shared cache as the process we are debugging and 2284 // we can read the entire __LINKEDIT from the address space in this 2285 // process. This is a needed optimization that is used for local iOS 2286 // debugging only since all shared libraries in the shared cache do 2287 // not have corresponding files that exist in the file system of the 2288 // device. They have been combined into a single file. This means we 2289 // always have to load these files from memory. All of the symbol and 2290 // string tables from all of the __LINKEDIT sections from the shared 2291 // libraries in the shared cache have been merged into a single large 2292 // symbol and string table. Reading all of this symbol and string 2293 // table 2294 // data across can slow down debug launch times, so we optimize this 2295 // by 2296 // reading the memory for the __LINKEDIT section from this process. 2297 2298 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 2299 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 2300 bool use_lldb_cache = true; 2301 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2302 lldb_shared_cache != process_shared_cache) { 2303 use_lldb_cache = false; 2304 ModuleSP module_sp(GetModule()); 2305 if (module_sp) 2306 module_sp->ReportWarning("shared cache in process does not match " 2307 "lldb's own shared cache, startup will " 2308 "be slow."); 2309 } 2310 2311 PlatformSP platform_sp(target.GetPlatform()); 2312 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2313 data_was_read = true; 2314 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2315 eByteOrderLittle); 2316 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2317 eByteOrderLittle); 2318 if (function_starts_load_command.cmd) { 2319 const addr_t func_start_addr = 2320 linkedit_load_addr + function_starts_load_command.dataoff - 2321 linkedit_file_offset; 2322 function_starts_data.SetData( 2323 (void *)func_start_addr, 2324 function_starts_load_command.datasize, eByteOrderLittle); 2325 } 2326 } 2327 } 2328 #endif 2329 2330 if (!data_was_read) { 2331 // Always load dyld - the dynamic linker - from memory if we didn't 2332 // find a binary anywhere else. 2333 // lldb will not register dylib/framework/bundle loads/unloads if we 2334 // don't have the dyld symbols, 2335 // we force dyld to load from memory despite the user's 2336 // target.memory-module-load-level setting. 2337 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2338 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2339 DataBufferSP nlist_data_sp( 2340 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2341 if (nlist_data_sp) 2342 nlist_data.SetData(nlist_data_sp, 0, 2343 nlist_data_sp->GetByteSize()); 2344 // Load strings individually from memory when loading from memory 2345 // since shared cache 2346 // string tables contain strings for all symbols from all shared 2347 // cached libraries 2348 // DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, 2349 // strtab_data_byte_size)); 2350 // if (strtab_data_sp) 2351 // strtab_data.SetData (strtab_data_sp, 0, 2352 // strtab_data_sp->GetByteSize()); 2353 if (m_dysymtab.nindirectsyms != 0) { 2354 const addr_t indirect_syms_addr = linkedit_load_addr + 2355 m_dysymtab.indirectsymoff - 2356 linkedit_file_offset; 2357 DataBufferSP indirect_syms_data_sp( 2358 ReadMemory(process_sp, indirect_syms_addr, 2359 m_dysymtab.nindirectsyms * 4)); 2360 if (indirect_syms_data_sp) 2361 indirect_symbol_index_data.SetData( 2362 indirect_syms_data_sp, 0, 2363 indirect_syms_data_sp->GetByteSize()); 2364 } 2365 } else if (memory_module_load_level >= 2366 eMemoryModuleLoadLevelPartial) { 2367 if (function_starts_load_command.cmd) { 2368 const addr_t func_start_addr = 2369 linkedit_load_addr + function_starts_load_command.dataoff - 2370 linkedit_file_offset; 2371 DataBufferSP func_start_data_sp( 2372 ReadMemory(process_sp, func_start_addr, 2373 function_starts_load_command.datasize)); 2374 if (func_start_data_sp) 2375 function_starts_data.SetData(func_start_data_sp, 0, 2376 func_start_data_sp->GetByteSize()); 2377 } 2378 } 2379 } 2380 } 2381 } else { 2382 nlist_data.SetData(m_data, symtab_load_command.symoff, 2383 nlist_data_byte_size); 2384 strtab_data.SetData(m_data, symtab_load_command.stroff, 2385 strtab_data_byte_size); 2386 2387 if (dyld_info.export_size > 0) { 2388 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2389 dyld_info.export_size); 2390 } 2391 2392 if (m_dysymtab.nindirectsyms != 0) { 2393 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2394 m_dysymtab.nindirectsyms * 4); 2395 } 2396 if (function_starts_load_command.cmd) { 2397 function_starts_data.SetData(m_data, 2398 function_starts_load_command.dataoff, 2399 function_starts_load_command.datasize); 2400 } 2401 } 2402 2403 if (nlist_data.GetByteSize() == 0 && 2404 memory_module_load_level == eMemoryModuleLoadLevelComplete) { 2405 if (log) 2406 module_sp->LogMessage(log, "failed to read nlist data"); 2407 return 0; 2408 } 2409 2410 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2411 if (!have_strtab_data) { 2412 if (process) { 2413 if (strtab_addr == LLDB_INVALID_ADDRESS) { 2414 if (log) 2415 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2416 return 0; 2417 } 2418 } else { 2419 if (log) 2420 module_sp->LogMessage(log, "failed to read strtab data"); 2421 return 0; 2422 } 2423 } 2424 2425 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2426 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2427 const ConstString &g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2428 const ConstString &g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2429 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2430 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2431 SectionSP text_section_sp( 2432 section_list->FindSectionByName(g_segment_name_TEXT)); 2433 SectionSP data_section_sp( 2434 section_list->FindSectionByName(g_segment_name_DATA)); 2435 SectionSP data_dirty_section_sp( 2436 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2437 SectionSP data_const_section_sp( 2438 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2439 SectionSP objc_section_sp( 2440 section_list->FindSectionByName(g_segment_name_OBJC)); 2441 SectionSP eh_frame_section_sp; 2442 if (text_section_sp.get()) 2443 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2444 g_section_name_eh_frame); 2445 else 2446 eh_frame_section_sp = 2447 section_list->FindSectionByName(g_section_name_eh_frame); 2448 2449 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2450 2451 // lldb works best if it knows the start address of all functions in a 2452 // module. 2453 // Linker symbols or debug info are normally the best source of information 2454 // for start addr / size but 2455 // they may be stripped in a released binary. 2456 // Two additional sources of information exist in Mach-O binaries: 2457 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2458 // function's start address in the 2459 // binary, relative to the text section. 2460 // eh_frame - the eh_frame FDEs have the start addr & size of 2461 // each function 2462 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2463 // all modern binaries. 2464 // Binaries built to run on older releases may need to use eh_frame 2465 // information. 2466 2467 if (text_section_sp && function_starts_data.GetByteSize()) { 2468 FunctionStarts::Entry function_start_entry; 2469 function_start_entry.data = false; 2470 lldb::offset_t function_start_offset = 0; 2471 function_start_entry.addr = text_section_sp->GetFileAddress(); 2472 uint64_t delta; 2473 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2474 0) { 2475 // Now append the current entry 2476 function_start_entry.addr += delta; 2477 function_starts.Append(function_start_entry); 2478 } 2479 } else { 2480 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2481 // load command claiming an eh_frame 2482 // but it doesn't actually have the eh_frame content. And if we have a 2483 // dSYM, we don't need to do any 2484 // of this fill-in-the-missing-symbols works anyway - the debug info 2485 // should give us all the functions in 2486 // the module. 2487 if (text_section_sp.get() && eh_frame_section_sp.get() && 2488 m_type != eTypeDebugInfo) { 2489 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2490 eRegisterKindEHFrame, true); 2491 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2492 eh_frame.GetFunctionAddressAndSizeVector(functions); 2493 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2494 size_t count = functions.GetSize(); 2495 for (size_t i = 0; i < count; ++i) { 2496 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2497 functions.GetEntryAtIndex(i); 2498 if (func) { 2499 FunctionStarts::Entry function_start_entry; 2500 function_start_entry.addr = func->base - text_base_addr; 2501 function_starts.Append(function_start_entry); 2502 } 2503 } 2504 } 2505 } 2506 2507 const size_t function_starts_count = function_starts.GetSize(); 2508 2509 // For user process binaries (executables, dylibs, frameworks, bundles), if 2510 // we don't have 2511 // LC_FUNCTION_STARTS/eh_frame section in this binary, we're going to assume 2512 // the binary 2513 // has been stripped. Don't allow assembly language instruction emulation 2514 // because we don't 2515 // know proper function start boundaries. 2516 // 2517 // For all other types of binaries (kernels, stand-alone bare board 2518 // binaries, kexts), they 2519 // may not have LC_FUNCTION_STARTS / eh_frame sections - we should not make 2520 // any assumptions 2521 // about them based on that. 2522 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2523 m_allow_assembly_emulation_unwind_plans = false; 2524 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2525 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2526 2527 if (unwind_or_symbol_log) 2528 module_sp->LogMessage( 2529 unwind_or_symbol_log, 2530 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2531 } 2532 2533 const user_id_t TEXT_eh_frame_sectID = 2534 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2535 : static_cast<user_id_t>(NO_SECT); 2536 2537 lldb::offset_t nlist_data_offset = 0; 2538 2539 uint32_t N_SO_index = UINT32_MAX; 2540 2541 MachSymtabSectionInfo section_info(section_list); 2542 std::vector<uint32_t> N_FUN_indexes; 2543 std::vector<uint32_t> N_NSYM_indexes; 2544 std::vector<uint32_t> N_INCL_indexes; 2545 std::vector<uint32_t> N_BRAC_indexes; 2546 std::vector<uint32_t> N_COMM_indexes; 2547 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2548 typedef std::map<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2549 typedef std::map<const char *, uint32_t> ConstNameToSymbolIndexMap; 2550 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2551 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2552 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2553 // Any symbols that get merged into another will get an entry 2554 // in this map so we know 2555 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2556 uint32_t nlist_idx = 0; 2557 Symbol *symbol_ptr = NULL; 2558 2559 uint32_t sym_idx = 0; 2560 Symbol *sym = NULL; 2561 size_t num_syms = 0; 2562 std::string memory_symbol_name; 2563 uint32_t unmapped_local_symbols_found = 0; 2564 2565 std::vector<TrieEntryWithOffset> trie_entries; 2566 std::set<lldb::addr_t> resolver_addresses; 2567 2568 if (dyld_trie_data.GetByteSize() > 0) { 2569 std::vector<llvm::StringRef> nameSlices; 2570 ParseTrieEntries(dyld_trie_data, 0, is_arm, nameSlices, 2571 resolver_addresses, trie_entries); 2572 2573 ConstString text_segment_name("__TEXT"); 2574 SectionSP text_segment_sp = 2575 GetSectionList()->FindSectionByName(text_segment_name); 2576 if (text_segment_sp) { 2577 const lldb::addr_t text_segment_file_addr = 2578 text_segment_sp->GetFileAddress(); 2579 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) { 2580 for (auto &e : trie_entries) 2581 e.entry.address += text_segment_file_addr; 2582 } 2583 } 2584 } 2585 2586 typedef std::set<ConstString> IndirectSymbols; 2587 IndirectSymbols indirect_symbol_names; 2588 2589 #if defined(__APPLE__) && \ 2590 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2591 2592 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2593 // optimized by moving LOCAL 2594 // symbols out of the memory mapped portion of the DSC. The symbol 2595 // information has all been retained, 2596 // but it isn't available in the normal nlist data. However, there *are* 2597 // duplicate entries of *some* 2598 // LOCAL symbols in the normal nlist data. To handle this situation 2599 // correctly, we must first attempt 2600 // to parse any DSC unmapped symbol information. If we find any, we set a 2601 // flag that tells the normal 2602 // nlist parser to ignore all LOCAL symbols. 2603 2604 if (m_header.flags & 0x80000000u) { 2605 // Before we can start mapping the DSC, we need to make certain the target 2606 // process is actually 2607 // using the cache we can find. 2608 2609 // Next we need to determine the correct path for the dyld shared cache. 2610 2611 ArchSpec header_arch; 2612 GetArchitecture(header_arch); 2613 char dsc_path[PATH_MAX]; 2614 char dsc_path_development[PATH_MAX]; 2615 2616 snprintf( 2617 dsc_path, sizeof(dsc_path), "%s%s%s", 2618 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2619 */ 2620 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2621 header_arch.GetArchitectureName()); 2622 2623 snprintf( 2624 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2625 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2626 */ 2627 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2628 header_arch.GetArchitectureName(), ".development"); 2629 2630 FileSpec dsc_nondevelopment_filespec(dsc_path, false); 2631 FileSpec dsc_development_filespec(dsc_path_development, false); 2632 FileSpec dsc_filespec; 2633 2634 UUID dsc_uuid; 2635 UUID process_shared_cache_uuid; 2636 2637 if (process) { 2638 process_shared_cache_uuid = GetProcessSharedCacheUUID(process); 2639 } 2640 2641 // First see if we can find an exact match for the inferior process shared 2642 // cache UUID in 2643 // the development or non-development shared caches on disk. 2644 if (process_shared_cache_uuid.IsValid()) { 2645 if (dsc_development_filespec.Exists()) { 2646 UUID dsc_development_uuid = GetSharedCacheUUID( 2647 dsc_development_filespec, byte_order, addr_byte_size); 2648 if (dsc_development_uuid.IsValid() && 2649 dsc_development_uuid == process_shared_cache_uuid) { 2650 dsc_filespec = dsc_development_filespec; 2651 dsc_uuid = dsc_development_uuid; 2652 } 2653 } 2654 if (!dsc_uuid.IsValid() && dsc_nondevelopment_filespec.Exists()) { 2655 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2656 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2657 if (dsc_nondevelopment_uuid.IsValid() && 2658 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2659 dsc_filespec = dsc_nondevelopment_filespec; 2660 dsc_uuid = dsc_nondevelopment_uuid; 2661 } 2662 } 2663 } 2664 2665 // Failing a UUID match, prefer the development dyld_shared cache if both 2666 // are present. 2667 if (!dsc_filespec.Exists()) { 2668 if (dsc_development_filespec.Exists()) { 2669 dsc_filespec = dsc_development_filespec; 2670 } else { 2671 dsc_filespec = dsc_nondevelopment_filespec; 2672 } 2673 } 2674 2675 /* The dyld_cache_header has a pointer to the 2676 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2677 The dyld_cache_local_symbols_info structure gives us three things: 2678 1. The start and count of the nlist records in the dyld_shared_cache 2679 file 2680 2. The start and size of the strings for these nlist records 2681 3. The start and count of dyld_cache_local_symbols_entry entries 2682 2683 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2684 dyld shared cache. 2685 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2686 the dyld shared cache. 2687 The dyld_cache_local_symbols_entry also lists the start of this 2688 dylib/framework's nlist records 2689 and the count of how many nlist records there are for this 2690 dylib/framework. 2691 */ 2692 2693 // Process the dyld shared cache header to find the unmapped symbols 2694 2695 DataBufferSP dsc_data_sp = dsc_filespec.MemoryMapFileContentsIfLocal( 2696 0, sizeof(struct lldb_copy_dyld_cache_header_v1)); 2697 if (!dsc_uuid.IsValid()) { 2698 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2699 } 2700 if (dsc_data_sp) { 2701 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2702 2703 bool uuid_match = true; 2704 if (dsc_uuid.IsValid() && process) { 2705 if (process_shared_cache_uuid.IsValid() && 2706 dsc_uuid != process_shared_cache_uuid) { 2707 // The on-disk dyld_shared_cache file is not the same as the one in 2708 // this 2709 // process' memory, don't use it. 2710 uuid_match = false; 2711 ModuleSP module_sp(GetModule()); 2712 if (module_sp) 2713 module_sp->ReportWarning("process shared cache does not match " 2714 "on-disk dyld_shared_cache file, some " 2715 "symbol names will be missing."); 2716 } 2717 } 2718 2719 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2720 2721 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2722 2723 // If the mappingOffset points to a location inside the header, we've 2724 // opened an old dyld shared cache, and should not proceed further. 2725 if (uuid_match && 2726 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2727 2728 DataBufferSP dsc_mapping_info_data_sp = 2729 dsc_filespec.MemoryMapFileContentsIfLocal( 2730 mappingOffset, 2731 sizeof(struct lldb_copy_dyld_cache_mapping_info)); 2732 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2733 byte_order, addr_byte_size); 2734 offset = 0; 2735 2736 // The File addresses (from the in-memory Mach-O load commands) for 2737 // the shared libraries 2738 // in the shared library cache need to be adjusted by an offset to 2739 // match up with the 2740 // dylibOffset identifying field in the 2741 // dyld_cache_local_symbol_entry's. This offset is 2742 // recorded in mapping_offset_value. 2743 const uint64_t mapping_offset_value = 2744 dsc_mapping_info_data.GetU64(&offset); 2745 2746 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, 2747 localSymbolsOffset); 2748 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2749 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2750 2751 if (localSymbolsOffset && localSymbolsSize) { 2752 // Map the local symbols 2753 if (DataBufferSP dsc_local_symbols_data_sp = 2754 dsc_filespec.MemoryMapFileContentsIfLocal( 2755 localSymbolsOffset, localSymbolsSize)) { 2756 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2757 byte_order, addr_byte_size); 2758 2759 offset = 0; 2760 2761 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2762 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2763 UndefinedNameToDescMap undefined_name_to_desc; 2764 SymbolIndexToName reexport_shlib_needs_fixup; 2765 2766 // Read the local_symbols_infos struct in one shot 2767 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2768 dsc_local_symbols_data.GetU32(&offset, 2769 &local_symbols_info.nlistOffset, 6); 2770 2771 SectionSP text_section_sp( 2772 section_list->FindSectionByName(GetSegmentNameTEXT())); 2773 2774 uint32_t header_file_offset = 2775 (text_section_sp->GetFileAddress() - mapping_offset_value); 2776 2777 offset = local_symbols_info.entriesOffset; 2778 for (uint32_t entry_index = 0; 2779 entry_index < local_symbols_info.entriesCount; 2780 entry_index++) { 2781 struct lldb_copy_dyld_cache_local_symbols_entry 2782 local_symbols_entry; 2783 local_symbols_entry.dylibOffset = 2784 dsc_local_symbols_data.GetU32(&offset); 2785 local_symbols_entry.nlistStartIndex = 2786 dsc_local_symbols_data.GetU32(&offset); 2787 local_symbols_entry.nlistCount = 2788 dsc_local_symbols_data.GetU32(&offset); 2789 2790 if (header_file_offset == local_symbols_entry.dylibOffset) { 2791 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2792 2793 // The normal nlist code cannot correctly size the Symbols 2794 // array, we need to allocate it here. 2795 sym = symtab->Resize( 2796 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2797 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2798 num_syms = symtab->GetNumSymbols(); 2799 2800 nlist_data_offset = 2801 local_symbols_info.nlistOffset + 2802 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2803 uint32_t string_table_offset = 2804 local_symbols_info.stringsOffset; 2805 2806 for (uint32_t nlist_index = 0; 2807 nlist_index < local_symbols_entry.nlistCount; 2808 nlist_index++) { 2809 ///////////////////////////// 2810 { 2811 struct nlist_64 nlist; 2812 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize( 2813 nlist_data_offset, nlist_byte_size)) 2814 break; 2815 2816 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked( 2817 &nlist_data_offset); 2818 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked( 2819 &nlist_data_offset); 2820 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked( 2821 &nlist_data_offset); 2822 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked( 2823 &nlist_data_offset); 2824 nlist.n_value = 2825 dsc_local_symbols_data.GetAddress_unchecked( 2826 &nlist_data_offset); 2827 2828 SymbolType type = eSymbolTypeInvalid; 2829 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2830 string_table_offset + nlist.n_strx); 2831 2832 if (symbol_name == NULL) { 2833 // No symbol should be NULL, even the symbols with no 2834 // string values should have an offset zero which points 2835 // to an empty C-string 2836 Host::SystemLog( 2837 Host::eSystemLogError, 2838 "error: DSC unmapped local symbol[%u] has invalid " 2839 "string table offset 0x%x in %s, ignoring symbol\n", 2840 entry_index, nlist.n_strx, 2841 module_sp->GetFileSpec().GetPath().c_str()); 2842 continue; 2843 } 2844 if (symbol_name[0] == '\0') 2845 symbol_name = NULL; 2846 2847 const char *symbol_name_non_abi_mangled = NULL; 2848 2849 SectionSP symbol_section; 2850 uint32_t symbol_byte_size = 0; 2851 bool add_nlist = true; 2852 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2853 bool demangled_is_synthesized = false; 2854 bool is_gsym = false; 2855 bool set_value = true; 2856 2857 assert(sym_idx < num_syms); 2858 2859 sym[sym_idx].SetDebug(is_debug); 2860 2861 if (is_debug) { 2862 switch (nlist.n_type) { 2863 case N_GSYM: 2864 // global symbol: name,,NO_SECT,type,0 2865 // Sometimes the N_GSYM value contains the address. 2866 2867 // FIXME: In the .o files, we have a GSYM and a debug 2868 // symbol for all the ObjC data. They 2869 // have the same address, but we want to ensure that 2870 // we always find only the real symbol, 2871 // 'cause we don't currently correctly attribute the 2872 // GSYM one to the ObjCClass/Ivar/MetaClass 2873 // symbol type. This is a temporary hack to make sure 2874 // the ObjectiveC symbols get treated 2875 // correctly. To do this right, we should coalesce 2876 // all the GSYM & global symbols that have the 2877 // same address. 2878 2879 is_gsym = true; 2880 sym[sym_idx].SetExternal(true); 2881 2882 if (symbol_name && symbol_name[0] == '_' && 2883 symbol_name[1] == 'O') { 2884 llvm::StringRef symbol_name_ref(symbol_name); 2885 if (symbol_name_ref.startswith( 2886 g_objc_v2_prefix_class)) { 2887 symbol_name_non_abi_mangled = symbol_name + 1; 2888 symbol_name = 2889 symbol_name + g_objc_v2_prefix_class.size(); 2890 type = eSymbolTypeObjCClass; 2891 demangled_is_synthesized = true; 2892 2893 } else if (symbol_name_ref.startswith( 2894 g_objc_v2_prefix_metaclass)) { 2895 symbol_name_non_abi_mangled = symbol_name + 1; 2896 symbol_name = symbol_name + 2897 g_objc_v2_prefix_metaclass.size(); 2898 type = eSymbolTypeObjCMetaClass; 2899 demangled_is_synthesized = true; 2900 } else if (symbol_name_ref.startswith( 2901 g_objc_v2_prefix_ivar)) { 2902 symbol_name_non_abi_mangled = symbol_name + 1; 2903 symbol_name = 2904 symbol_name + g_objc_v2_prefix_ivar.size(); 2905 type = eSymbolTypeObjCIVar; 2906 demangled_is_synthesized = true; 2907 } 2908 } else { 2909 if (nlist.n_value != 0) 2910 symbol_section = section_info.GetSection( 2911 nlist.n_sect, nlist.n_value); 2912 type = eSymbolTypeData; 2913 } 2914 break; 2915 2916 case N_FNAME: 2917 // procedure name (f77 kludge): name,,NO_SECT,0,0 2918 type = eSymbolTypeCompiler; 2919 break; 2920 2921 case N_FUN: 2922 // procedure: name,,n_sect,linenumber,address 2923 if (symbol_name) { 2924 type = eSymbolTypeCode; 2925 symbol_section = section_info.GetSection( 2926 nlist.n_sect, nlist.n_value); 2927 2928 N_FUN_addr_to_sym_idx.insert( 2929 std::make_pair(nlist.n_value, sym_idx)); 2930 // We use the current number of symbols in the 2931 // symbol table in lieu of 2932 // using nlist_idx in case we ever start trimming 2933 // entries out 2934 N_FUN_indexes.push_back(sym_idx); 2935 } else { 2936 type = eSymbolTypeCompiler; 2937 2938 if (!N_FUN_indexes.empty()) { 2939 // Copy the size of the function into the original 2940 // STAB entry so we don't have 2941 // to hunt for it later 2942 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2943 ->SetByteSize(nlist.n_value); 2944 N_FUN_indexes.pop_back(); 2945 // We don't really need the end function STAB as 2946 // it contains the size which 2947 // we already placed with the original symbol, so 2948 // don't add it if we want a 2949 // minimal symbol table 2950 add_nlist = false; 2951 } 2952 } 2953 break; 2954 2955 case N_STSYM: 2956 // static symbol: name,,n_sect,type,address 2957 N_STSYM_addr_to_sym_idx.insert( 2958 std::make_pair(nlist.n_value, sym_idx)); 2959 symbol_section = section_info.GetSection( 2960 nlist.n_sect, nlist.n_value); 2961 if (symbol_name && symbol_name[0]) { 2962 type = ObjectFile::GetSymbolTypeFromName( 2963 symbol_name + 1, eSymbolTypeData); 2964 } 2965 break; 2966 2967 case N_LCSYM: 2968 // .lcomm symbol: name,,n_sect,type,address 2969 symbol_section = section_info.GetSection( 2970 nlist.n_sect, nlist.n_value); 2971 type = eSymbolTypeCommonBlock; 2972 break; 2973 2974 case N_BNSYM: 2975 // We use the current number of symbols in the symbol 2976 // table in lieu of 2977 // using nlist_idx in case we ever start trimming 2978 // entries out 2979 // Skip these if we want minimal symbol tables 2980 add_nlist = false; 2981 break; 2982 2983 case N_ENSYM: 2984 // Set the size of the N_BNSYM to the terminating 2985 // index of this N_ENSYM 2986 // so that we can always skip the entire symbol if we 2987 // need to navigate 2988 // more quickly at the source level when parsing STABS 2989 // Skip these if we want minimal symbol tables 2990 add_nlist = false; 2991 break; 2992 2993 case N_OPT: 2994 // emitted with gcc2_compiled and in gcc source 2995 type = eSymbolTypeCompiler; 2996 break; 2997 2998 case N_RSYM: 2999 // register sym: name,,NO_SECT,type,register 3000 type = eSymbolTypeVariable; 3001 break; 3002 3003 case N_SLINE: 3004 // src line: 0,,n_sect,linenumber,address 3005 symbol_section = section_info.GetSection( 3006 nlist.n_sect, nlist.n_value); 3007 type = eSymbolTypeLineEntry; 3008 break; 3009 3010 case N_SSYM: 3011 // structure elt: name,,NO_SECT,type,struct_offset 3012 type = eSymbolTypeVariableType; 3013 break; 3014 3015 case N_SO: 3016 // source file name 3017 type = eSymbolTypeSourceFile; 3018 if (symbol_name == NULL) { 3019 add_nlist = false; 3020 if (N_SO_index != UINT32_MAX) { 3021 // Set the size of the N_SO to the terminating 3022 // index of this N_SO 3023 // so that we can always skip the entire N_SO if 3024 // we need to navigate 3025 // more quickly at the source level when parsing 3026 // STABS 3027 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3028 symbol_ptr->SetByteSize(sym_idx); 3029 symbol_ptr->SetSizeIsSibling(true); 3030 } 3031 N_NSYM_indexes.clear(); 3032 N_INCL_indexes.clear(); 3033 N_BRAC_indexes.clear(); 3034 N_COMM_indexes.clear(); 3035 N_FUN_indexes.clear(); 3036 N_SO_index = UINT32_MAX; 3037 } else { 3038 // We use the current number of symbols in the 3039 // symbol table in lieu of 3040 // using nlist_idx in case we ever start trimming 3041 // entries out 3042 const bool N_SO_has_full_path = 3043 symbol_name[0] == '/'; 3044 if (N_SO_has_full_path) { 3045 if ((N_SO_index == sym_idx - 1) && 3046 ((sym_idx - 1) < num_syms)) { 3047 // We have two consecutive N_SO entries where 3048 // the first contains a directory 3049 // and the second contains a full path. 3050 sym[sym_idx - 1].GetMangled().SetValue( 3051 ConstString(symbol_name), false); 3052 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3053 add_nlist = false; 3054 } else { 3055 // This is the first entry in a N_SO that 3056 // contains a directory or 3057 // a full path to the source file 3058 N_SO_index = sym_idx; 3059 } 3060 } else if ((N_SO_index == sym_idx - 1) && 3061 ((sym_idx - 1) < num_syms)) { 3062 // This is usually the second N_SO entry that 3063 // contains just the filename, 3064 // so here we combine it with the first one if we 3065 // are minimizing the symbol table 3066 const char *so_path = 3067 sym[sym_idx - 1] 3068 .GetMangled() 3069 .GetDemangledName( 3070 lldb::eLanguageTypeUnknown) 3071 .AsCString(); 3072 if (so_path && so_path[0]) { 3073 std::string full_so_path(so_path); 3074 const size_t double_slash_pos = 3075 full_so_path.find("//"); 3076 if (double_slash_pos != std::string::npos) { 3077 // The linker has been generating bad N_SO 3078 // entries with doubled up paths 3079 // in the format "%s%s" where the first string 3080 // in the DW_AT_comp_dir, 3081 // and the second is the directory for the 3082 // source file so you end up with 3083 // a path that looks like "/tmp/src//tmp/src/" 3084 FileSpec so_dir(so_path, false); 3085 if (!so_dir.Exists()) { 3086 so_dir.SetFile( 3087 &full_so_path[double_slash_pos + 1], 3088 false); 3089 if (so_dir.Exists()) { 3090 // Trim off the incorrect path 3091 full_so_path.erase(0, 3092 double_slash_pos + 1); 3093 } 3094 } 3095 } 3096 if (*full_so_path.rbegin() != '/') 3097 full_so_path += '/'; 3098 full_so_path += symbol_name; 3099 sym[sym_idx - 1].GetMangled().SetValue( 3100 ConstString(full_so_path.c_str()), false); 3101 add_nlist = false; 3102 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3103 } 3104 } else { 3105 // This could be a relative path to a N_SO 3106 N_SO_index = sym_idx; 3107 } 3108 } 3109 break; 3110 3111 case N_OSO: 3112 // object file name: name,,0,0,st_mtime 3113 type = eSymbolTypeObjectFile; 3114 break; 3115 3116 case N_LSYM: 3117 // local sym: name,,NO_SECT,type,offset 3118 type = eSymbolTypeLocal; 3119 break; 3120 3121 //---------------------------------------------------------------------- 3122 // INCL scopes 3123 //---------------------------------------------------------------------- 3124 case N_BINCL: 3125 // include file beginning: name,,NO_SECT,0,sum 3126 // We use the current number of symbols in the symbol 3127 // table in lieu of 3128 // using nlist_idx in case we ever start trimming 3129 // entries out 3130 N_INCL_indexes.push_back(sym_idx); 3131 type = eSymbolTypeScopeBegin; 3132 break; 3133 3134 case N_EINCL: 3135 // include file end: name,,NO_SECT,0,0 3136 // Set the size of the N_BINCL to the terminating 3137 // index of this N_EINCL 3138 // so that we can always skip the entire symbol if we 3139 // need to navigate 3140 // more quickly at the source level when parsing STABS 3141 if (!N_INCL_indexes.empty()) { 3142 symbol_ptr = 3143 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3144 symbol_ptr->SetByteSize(sym_idx + 1); 3145 symbol_ptr->SetSizeIsSibling(true); 3146 N_INCL_indexes.pop_back(); 3147 } 3148 type = eSymbolTypeScopeEnd; 3149 break; 3150 3151 case N_SOL: 3152 // #included file name: name,,n_sect,0,address 3153 type = eSymbolTypeHeaderFile; 3154 3155 // We currently don't use the header files on darwin 3156 add_nlist = false; 3157 break; 3158 3159 case N_PARAMS: 3160 // compiler parameters: name,,NO_SECT,0,0 3161 type = eSymbolTypeCompiler; 3162 break; 3163 3164 case N_VERSION: 3165 // compiler version: name,,NO_SECT,0,0 3166 type = eSymbolTypeCompiler; 3167 break; 3168 3169 case N_OLEVEL: 3170 // compiler -O level: name,,NO_SECT,0,0 3171 type = eSymbolTypeCompiler; 3172 break; 3173 3174 case N_PSYM: 3175 // parameter: name,,NO_SECT,type,offset 3176 type = eSymbolTypeVariable; 3177 break; 3178 3179 case N_ENTRY: 3180 // alternate entry: name,,n_sect,linenumber,address 3181 symbol_section = section_info.GetSection( 3182 nlist.n_sect, nlist.n_value); 3183 type = eSymbolTypeLineEntry; 3184 break; 3185 3186 //---------------------------------------------------------------------- 3187 // Left and Right Braces 3188 //---------------------------------------------------------------------- 3189 case N_LBRAC: 3190 // left bracket: 0,,NO_SECT,nesting level,address 3191 // We use the current number of symbols in the symbol 3192 // table in lieu of 3193 // using nlist_idx in case we ever start trimming 3194 // entries out 3195 symbol_section = section_info.GetSection( 3196 nlist.n_sect, nlist.n_value); 3197 N_BRAC_indexes.push_back(sym_idx); 3198 type = eSymbolTypeScopeBegin; 3199 break; 3200 3201 case N_RBRAC: 3202 // right bracket: 0,,NO_SECT,nesting level,address 3203 // Set the size of the N_LBRAC to the terminating 3204 // index of this N_RBRAC 3205 // so that we can always skip the entire symbol if we 3206 // need to navigate 3207 // more quickly at the source level when parsing STABS 3208 symbol_section = section_info.GetSection( 3209 nlist.n_sect, nlist.n_value); 3210 if (!N_BRAC_indexes.empty()) { 3211 symbol_ptr = 3212 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3213 symbol_ptr->SetByteSize(sym_idx + 1); 3214 symbol_ptr->SetSizeIsSibling(true); 3215 N_BRAC_indexes.pop_back(); 3216 } 3217 type = eSymbolTypeScopeEnd; 3218 break; 3219 3220 case N_EXCL: 3221 // deleted include file: name,,NO_SECT,0,sum 3222 type = eSymbolTypeHeaderFile; 3223 break; 3224 3225 //---------------------------------------------------------------------- 3226 // COMM scopes 3227 //---------------------------------------------------------------------- 3228 case N_BCOMM: 3229 // begin common: name,,NO_SECT,0,0 3230 // We use the current number of symbols in the symbol 3231 // table in lieu of 3232 // using nlist_idx in case we ever start trimming 3233 // entries out 3234 type = eSymbolTypeScopeBegin; 3235 N_COMM_indexes.push_back(sym_idx); 3236 break; 3237 3238 case N_ECOML: 3239 // end common (local name): 0,,n_sect,0,address 3240 symbol_section = section_info.GetSection( 3241 nlist.n_sect, nlist.n_value); 3242 // Fall through 3243 3244 case N_ECOMM: 3245 // end common: name,,n_sect,0,0 3246 // Set the size of the N_BCOMM to the terminating 3247 // index of this N_ECOMM/N_ECOML 3248 // so that we can always skip the entire symbol if we 3249 // need to navigate 3250 // more quickly at the source level when parsing STABS 3251 if (!N_COMM_indexes.empty()) { 3252 symbol_ptr = 3253 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3254 symbol_ptr->SetByteSize(sym_idx + 1); 3255 symbol_ptr->SetSizeIsSibling(true); 3256 N_COMM_indexes.pop_back(); 3257 } 3258 type = eSymbolTypeScopeEnd; 3259 break; 3260 3261 case N_LENG: 3262 // second stab entry with length information 3263 type = eSymbolTypeAdditional; 3264 break; 3265 3266 default: 3267 break; 3268 } 3269 } else { 3270 // uint8_t n_pext = N_PEXT & nlist.n_type; 3271 uint8_t n_type = N_TYPE & nlist.n_type; 3272 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3273 3274 switch (n_type) { 3275 case N_INDR: { 3276 const char *reexport_name_cstr = 3277 strtab_data.PeekCStr(nlist.n_value); 3278 if (reexport_name_cstr && reexport_name_cstr[0]) { 3279 type = eSymbolTypeReExported; 3280 ConstString reexport_name( 3281 reexport_name_cstr + 3282 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3283 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3284 set_value = false; 3285 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3286 indirect_symbol_names.insert( 3287 ConstString(symbol_name + 3288 ((symbol_name[0] == '_') ? 1 : 0))); 3289 } else 3290 type = eSymbolTypeUndefined; 3291 } break; 3292 3293 case N_UNDF: 3294 if (symbol_name && symbol_name[0]) { 3295 ConstString undefined_name( 3296 symbol_name + 3297 ((symbol_name[0] == '_') ? 1 : 0)); 3298 undefined_name_to_desc[undefined_name] = 3299 nlist.n_desc; 3300 } 3301 // Fall through 3302 case N_PBUD: 3303 type = eSymbolTypeUndefined; 3304 break; 3305 3306 case N_ABS: 3307 type = eSymbolTypeAbsolute; 3308 break; 3309 3310 case N_SECT: { 3311 symbol_section = section_info.GetSection( 3312 nlist.n_sect, nlist.n_value); 3313 3314 if (symbol_section == NULL) { 3315 // TODO: warn about this? 3316 add_nlist = false; 3317 break; 3318 } 3319 3320 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3321 type = eSymbolTypeException; 3322 } else { 3323 uint32_t section_type = 3324 symbol_section->Get() & SECTION_TYPE; 3325 3326 switch (section_type) { 3327 case S_CSTRING_LITERALS: 3328 type = eSymbolTypeData; 3329 break; // section with only literal C strings 3330 case S_4BYTE_LITERALS: 3331 type = eSymbolTypeData; 3332 break; // section with only 4 byte literals 3333 case S_8BYTE_LITERALS: 3334 type = eSymbolTypeData; 3335 break; // section with only 8 byte literals 3336 case S_LITERAL_POINTERS: 3337 type = eSymbolTypeTrampoline; 3338 break; // section with only pointers to literals 3339 case S_NON_LAZY_SYMBOL_POINTERS: 3340 type = eSymbolTypeTrampoline; 3341 break; // section with only non-lazy symbol 3342 // pointers 3343 case S_LAZY_SYMBOL_POINTERS: 3344 type = eSymbolTypeTrampoline; 3345 break; // section with only lazy symbol pointers 3346 case S_SYMBOL_STUBS: 3347 type = eSymbolTypeTrampoline; 3348 break; // section with only symbol stubs, byte 3349 // size of stub in the reserved2 field 3350 case S_MOD_INIT_FUNC_POINTERS: 3351 type = eSymbolTypeCode; 3352 break; // section with only function pointers for 3353 // initialization 3354 case S_MOD_TERM_FUNC_POINTERS: 3355 type = eSymbolTypeCode; 3356 break; // section with only function pointers for 3357 // termination 3358 case S_INTERPOSING: 3359 type = eSymbolTypeTrampoline; 3360 break; // section with only pairs of function 3361 // pointers for interposing 3362 case S_16BYTE_LITERALS: 3363 type = eSymbolTypeData; 3364 break; // section with only 16 byte literals 3365 case S_DTRACE_DOF: 3366 type = eSymbolTypeInstrumentation; 3367 break; 3368 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3369 type = eSymbolTypeTrampoline; 3370 break; 3371 default: 3372 switch (symbol_section->GetType()) { 3373 case lldb::eSectionTypeCode: 3374 type = eSymbolTypeCode; 3375 break; 3376 case eSectionTypeData: 3377 case eSectionTypeDataCString: // Inlined C string 3378 // data 3379 case eSectionTypeDataCStringPointers: // Pointers 3380 // to C 3381 // string 3382 // data 3383 case eSectionTypeDataSymbolAddress: // Address of 3384 // a symbol in 3385 // the symbol 3386 // table 3387 case eSectionTypeData4: 3388 case eSectionTypeData8: 3389 case eSectionTypeData16: 3390 type = eSymbolTypeData; 3391 break; 3392 default: 3393 break; 3394 } 3395 break; 3396 } 3397 3398 if (type == eSymbolTypeInvalid) { 3399 const char *symbol_sect_name = 3400 symbol_section->GetName().AsCString(); 3401 if (symbol_section->IsDescendant( 3402 text_section_sp.get())) { 3403 if (symbol_section->IsClear( 3404 S_ATTR_PURE_INSTRUCTIONS | 3405 S_ATTR_SELF_MODIFYING_CODE | 3406 S_ATTR_SOME_INSTRUCTIONS)) 3407 type = eSymbolTypeData; 3408 else 3409 type = eSymbolTypeCode; 3410 } else if (symbol_section->IsDescendant( 3411 data_section_sp.get()) || 3412 symbol_section->IsDescendant( 3413 data_dirty_section_sp.get()) || 3414 symbol_section->IsDescendant( 3415 data_const_section_sp.get())) { 3416 if (symbol_sect_name && 3417 ::strstr(symbol_sect_name, "__objc") == 3418 symbol_sect_name) { 3419 type = eSymbolTypeRuntime; 3420 3421 if (symbol_name) { 3422 llvm::StringRef symbol_name_ref( 3423 symbol_name); 3424 if (symbol_name_ref.startswith("_OBJC_")) { 3425 static const llvm::StringRef 3426 g_objc_v2_prefix_class( 3427 "_OBJC_CLASS_$_"); 3428 static const llvm::StringRef 3429 g_objc_v2_prefix_metaclass( 3430 "_OBJC_METACLASS_$_"); 3431 static const llvm::StringRef 3432 g_objc_v2_prefix_ivar( 3433 "_OBJC_IVAR_$_"); 3434 if (symbol_name_ref.startswith( 3435 g_objc_v2_prefix_class)) { 3436 symbol_name_non_abi_mangled = 3437 symbol_name + 1; 3438 symbol_name = 3439 symbol_name + 3440 g_objc_v2_prefix_class.size(); 3441 type = eSymbolTypeObjCClass; 3442 demangled_is_synthesized = true; 3443 } else if ( 3444 symbol_name_ref.startswith( 3445 g_objc_v2_prefix_metaclass)) { 3446 symbol_name_non_abi_mangled = 3447 symbol_name + 1; 3448 symbol_name = 3449 symbol_name + 3450 g_objc_v2_prefix_metaclass.size(); 3451 type = eSymbolTypeObjCMetaClass; 3452 demangled_is_synthesized = true; 3453 } else if (symbol_name_ref.startswith( 3454 g_objc_v2_prefix_ivar)) { 3455 symbol_name_non_abi_mangled = 3456 symbol_name + 1; 3457 symbol_name = 3458 symbol_name + 3459 g_objc_v2_prefix_ivar.size(); 3460 type = eSymbolTypeObjCIVar; 3461 demangled_is_synthesized = true; 3462 } 3463 } 3464 } 3465 } else if (symbol_sect_name && 3466 ::strstr(symbol_sect_name, 3467 "__gcc_except_tab") == 3468 symbol_sect_name) { 3469 type = eSymbolTypeException; 3470 } else { 3471 type = eSymbolTypeData; 3472 } 3473 } else if (symbol_sect_name && 3474 ::strstr(symbol_sect_name, 3475 "__IMPORT") == 3476 symbol_sect_name) { 3477 type = eSymbolTypeTrampoline; 3478 } else if (symbol_section->IsDescendant( 3479 objc_section_sp.get())) { 3480 type = eSymbolTypeRuntime; 3481 if (symbol_name && symbol_name[0] == '.') { 3482 llvm::StringRef symbol_name_ref(symbol_name); 3483 static const llvm::StringRef 3484 g_objc_v1_prefix_class( 3485 ".objc_class_name_"); 3486 if (symbol_name_ref.startswith( 3487 g_objc_v1_prefix_class)) { 3488 symbol_name_non_abi_mangled = symbol_name; 3489 symbol_name = symbol_name + 3490 g_objc_v1_prefix_class.size(); 3491 type = eSymbolTypeObjCClass; 3492 demangled_is_synthesized = true; 3493 } 3494 } 3495 } 3496 } 3497 } 3498 } break; 3499 } 3500 } 3501 3502 if (add_nlist) { 3503 uint64_t symbol_value = nlist.n_value; 3504 if (symbol_name_non_abi_mangled) { 3505 sym[sym_idx].GetMangled().SetMangledName( 3506 ConstString(symbol_name_non_abi_mangled)); 3507 sym[sym_idx].GetMangled().SetDemangledName( 3508 ConstString(symbol_name)); 3509 } else { 3510 bool symbol_name_is_mangled = false; 3511 3512 if (symbol_name && symbol_name[0] == '_') { 3513 symbol_name_is_mangled = symbol_name[1] == '_'; 3514 symbol_name++; // Skip the leading underscore 3515 } 3516 3517 if (symbol_name) { 3518 ConstString const_symbol_name(symbol_name); 3519 sym[sym_idx].GetMangled().SetValue( 3520 const_symbol_name, symbol_name_is_mangled); 3521 if (is_gsym && is_debug) { 3522 const char *gsym_name = 3523 sym[sym_idx] 3524 .GetMangled() 3525 .GetName(lldb::eLanguageTypeUnknown, 3526 Mangled::ePreferMangled) 3527 .GetCString(); 3528 if (gsym_name) 3529 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3530 } 3531 } 3532 } 3533 if (symbol_section) { 3534 const addr_t section_file_addr = 3535 symbol_section->GetFileAddress(); 3536 if (symbol_byte_size == 0 && 3537 function_starts_count > 0) { 3538 addr_t symbol_lookup_file_addr = nlist.n_value; 3539 // Do an exact address match for non-ARM addresses, 3540 // else get the closest since 3541 // the symbol might be a thumb symbol which has an 3542 // address with bit zero set 3543 FunctionStarts::Entry *func_start_entry = 3544 function_starts.FindEntry( 3545 symbol_lookup_file_addr, !is_arm); 3546 if (is_arm && func_start_entry) { 3547 // Verify that the function start address is the 3548 // symbol address (ARM) 3549 // or the symbol address + 1 (thumb) 3550 if (func_start_entry->addr != 3551 symbol_lookup_file_addr && 3552 func_start_entry->addr != 3553 (symbol_lookup_file_addr + 1)) { 3554 // Not the right entry, NULL it out... 3555 func_start_entry = NULL; 3556 } 3557 } 3558 if (func_start_entry) { 3559 func_start_entry->data = true; 3560 3561 addr_t symbol_file_addr = func_start_entry->addr; 3562 uint32_t symbol_flags = 0; 3563 if (is_arm) { 3564 if (symbol_file_addr & 1) 3565 symbol_flags = 3566 MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3567 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3568 } 3569 3570 const FunctionStarts::Entry 3571 *next_func_start_entry = 3572 function_starts.FindNextEntry( 3573 func_start_entry); 3574 const addr_t section_end_file_addr = 3575 section_file_addr + 3576 symbol_section->GetByteSize(); 3577 if (next_func_start_entry) { 3578 addr_t next_symbol_file_addr = 3579 next_func_start_entry->addr; 3580 // Be sure the clear the Thumb address bit when 3581 // we calculate the size 3582 // from the current and next address 3583 if (is_arm) 3584 next_symbol_file_addr &= 3585 THUMB_ADDRESS_BIT_MASK; 3586 symbol_byte_size = std::min<lldb::addr_t>( 3587 next_symbol_file_addr - symbol_file_addr, 3588 section_end_file_addr - symbol_file_addr); 3589 } else { 3590 symbol_byte_size = 3591 section_end_file_addr - symbol_file_addr; 3592 } 3593 } 3594 } 3595 symbol_value -= section_file_addr; 3596 } 3597 3598 if (is_debug == false) { 3599 if (type == eSymbolTypeCode) { 3600 // See if we can find a N_FUN entry for any code 3601 // symbols. 3602 // If we do find a match, and the name matches, then 3603 // we 3604 // can merge the two into just the function symbol 3605 // to avoid 3606 // duplicate entries in the symbol table 3607 std::pair<ValueToSymbolIndexMap::const_iterator, 3608 ValueToSymbolIndexMap::const_iterator> 3609 range; 3610 range = N_FUN_addr_to_sym_idx.equal_range( 3611 nlist.n_value); 3612 if (range.first != range.second) { 3613 bool found_it = false; 3614 for (ValueToSymbolIndexMap::const_iterator pos = 3615 range.first; 3616 pos != range.second; ++pos) { 3617 if (sym[sym_idx].GetMangled().GetName( 3618 lldb::eLanguageTypeUnknown, 3619 Mangled::ePreferMangled) == 3620 sym[pos->second].GetMangled().GetName( 3621 lldb::eLanguageTypeUnknown, 3622 Mangled::ePreferMangled)) { 3623 m_nlist_idx_to_sym_idx[nlist_idx] = 3624 pos->second; 3625 // We just need the flags from the linker 3626 // symbol, so put these flags 3627 // into the N_FUN flags to avoid duplicate 3628 // symbols in the symbol table 3629 sym[pos->second].SetExternal( 3630 sym[sym_idx].IsExternal()); 3631 sym[pos->second].SetFlags(nlist.n_type << 16 | 3632 nlist.n_desc); 3633 if (resolver_addresses.find(nlist.n_value) != 3634 resolver_addresses.end()) 3635 sym[pos->second].SetType( 3636 eSymbolTypeResolver); 3637 sym[sym_idx].Clear(); 3638 found_it = true; 3639 break; 3640 } 3641 } 3642 if (found_it) 3643 continue; 3644 } else { 3645 if (resolver_addresses.find(nlist.n_value) != 3646 resolver_addresses.end()) 3647 type = eSymbolTypeResolver; 3648 } 3649 } else if (type == eSymbolTypeData || 3650 type == eSymbolTypeObjCClass || 3651 type == eSymbolTypeObjCMetaClass || 3652 type == eSymbolTypeObjCIVar) { 3653 // See if we can find a N_STSYM entry for any data 3654 // symbols. 3655 // If we do find a match, and the name matches, then 3656 // we 3657 // can merge the two into just the Static symbol to 3658 // avoid 3659 // duplicate entries in the symbol table 3660 std::pair<ValueToSymbolIndexMap::const_iterator, 3661 ValueToSymbolIndexMap::const_iterator> 3662 range; 3663 range = N_STSYM_addr_to_sym_idx.equal_range( 3664 nlist.n_value); 3665 if (range.first != range.second) { 3666 bool found_it = false; 3667 for (ValueToSymbolIndexMap::const_iterator pos = 3668 range.first; 3669 pos != range.second; ++pos) { 3670 if (sym[sym_idx].GetMangled().GetName( 3671 lldb::eLanguageTypeUnknown, 3672 Mangled::ePreferMangled) == 3673 sym[pos->second].GetMangled().GetName( 3674 lldb::eLanguageTypeUnknown, 3675 Mangled::ePreferMangled)) { 3676 m_nlist_idx_to_sym_idx[nlist_idx] = 3677 pos->second; 3678 // We just need the flags from the linker 3679 // symbol, so put these flags 3680 // into the N_STSYM flags to avoid duplicate 3681 // symbols in the symbol table 3682 sym[pos->second].SetExternal( 3683 sym[sym_idx].IsExternal()); 3684 sym[pos->second].SetFlags(nlist.n_type << 16 | 3685 nlist.n_desc); 3686 sym[sym_idx].Clear(); 3687 found_it = true; 3688 break; 3689 } 3690 } 3691 if (found_it) 3692 continue; 3693 } else { 3694 const char *gsym_name = 3695 sym[sym_idx] 3696 .GetMangled() 3697 .GetName(lldb::eLanguageTypeUnknown, 3698 Mangled::ePreferMangled) 3699 .GetCString(); 3700 if (gsym_name) { 3701 // Combine N_GSYM stab entries with the non stab 3702 // symbol 3703 ConstNameToSymbolIndexMap::const_iterator pos = 3704 N_GSYM_name_to_sym_idx.find(gsym_name); 3705 if (pos != N_GSYM_name_to_sym_idx.end()) { 3706 const uint32_t GSYM_sym_idx = pos->second; 3707 m_nlist_idx_to_sym_idx[nlist_idx] = 3708 GSYM_sym_idx; 3709 // Copy the address, because often the N_GSYM 3710 // address has an invalid address of zero 3711 // when the global is a common symbol 3712 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3713 symbol_section); 3714 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3715 symbol_value); 3716 // We just need the flags from the linker 3717 // symbol, so put these flags 3718 // into the N_GSYM flags to avoid duplicate 3719 // symbols in the symbol table 3720 sym[GSYM_sym_idx].SetFlags( 3721 nlist.n_type << 16 | nlist.n_desc); 3722 sym[sym_idx].Clear(); 3723 continue; 3724 } 3725 } 3726 } 3727 } 3728 } 3729 3730 sym[sym_idx].SetID(nlist_idx); 3731 sym[sym_idx].SetType(type); 3732 if (set_value) { 3733 sym[sym_idx].GetAddressRef().SetSection( 3734 symbol_section); 3735 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3736 } 3737 sym[sym_idx].SetFlags(nlist.n_type << 16 | 3738 nlist.n_desc); 3739 3740 if (symbol_byte_size > 0) 3741 sym[sym_idx].SetByteSize(symbol_byte_size); 3742 3743 if (demangled_is_synthesized) 3744 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3745 ++sym_idx; 3746 } else { 3747 sym[sym_idx].Clear(); 3748 } 3749 } 3750 ///////////////////////////// 3751 } 3752 break; // No more entries to consider 3753 } 3754 } 3755 3756 for (const auto &pos : reexport_shlib_needs_fixup) { 3757 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3758 if (undef_pos != undefined_name_to_desc.end()) { 3759 const uint8_t dylib_ordinal = 3760 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3761 if (dylib_ordinal > 0 && 3762 dylib_ordinal < dylib_files.GetSize()) 3763 sym[pos.first].SetReExportedSymbolSharedLibrary( 3764 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3765 } 3766 } 3767 } 3768 } 3769 } 3770 } 3771 } 3772 3773 // Must reset this in case it was mutated above! 3774 nlist_data_offset = 0; 3775 #endif 3776 3777 if (nlist_data.GetByteSize() > 0) { 3778 3779 // If the sym array was not created while parsing the DSC unmapped 3780 // symbols, create it now. 3781 if (sym == NULL) { 3782 sym = symtab->Resize(symtab_load_command.nsyms + 3783 m_dysymtab.nindirectsyms); 3784 num_syms = symtab->GetNumSymbols(); 3785 } 3786 3787 if (unmapped_local_symbols_found) { 3788 assert(m_dysymtab.ilocalsym == 0); 3789 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3790 nlist_idx = m_dysymtab.nlocalsym; 3791 } else { 3792 nlist_idx = 0; 3793 } 3794 3795 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3796 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3797 UndefinedNameToDescMap undefined_name_to_desc; 3798 SymbolIndexToName reexport_shlib_needs_fixup; 3799 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 3800 struct nlist_64 nlist; 3801 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, 3802 nlist_byte_size)) 3803 break; 3804 3805 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3806 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 3807 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 3808 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 3809 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 3810 3811 SymbolType type = eSymbolTypeInvalid; 3812 const char *symbol_name = NULL; 3813 3814 if (have_strtab_data) { 3815 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3816 3817 if (symbol_name == NULL) { 3818 // No symbol should be NULL, even the symbols with no 3819 // string values should have an offset zero which points 3820 // to an empty C-string 3821 Host::SystemLog(Host::eSystemLogError, 3822 "error: symbol[%u] has invalid string table offset " 3823 "0x%x in %s, ignoring symbol\n", 3824 nlist_idx, nlist.n_strx, 3825 module_sp->GetFileSpec().GetPath().c_str()); 3826 continue; 3827 } 3828 if (symbol_name[0] == '\0') 3829 symbol_name = NULL; 3830 } else { 3831 const addr_t str_addr = strtab_addr + nlist.n_strx; 3832 Error str_error; 3833 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3834 str_error)) 3835 symbol_name = memory_symbol_name.c_str(); 3836 } 3837 const char *symbol_name_non_abi_mangled = NULL; 3838 3839 SectionSP symbol_section; 3840 lldb::addr_t symbol_byte_size = 0; 3841 bool add_nlist = true; 3842 bool is_gsym = false; 3843 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3844 bool demangled_is_synthesized = false; 3845 bool set_value = true; 3846 assert(sym_idx < num_syms); 3847 3848 sym[sym_idx].SetDebug(is_debug); 3849 3850 if (is_debug) { 3851 switch (nlist.n_type) { 3852 case N_GSYM: 3853 // global symbol: name,,NO_SECT,type,0 3854 // Sometimes the N_GSYM value contains the address. 3855 3856 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3857 // the ObjC data. They 3858 // have the same address, but we want to ensure that we always find 3859 // only the real symbol, 3860 // 'cause we don't currently correctly attribute the GSYM one to the 3861 // ObjCClass/Ivar/MetaClass 3862 // symbol type. This is a temporary hack to make sure the 3863 // ObjectiveC symbols get treated 3864 // correctly. To do this right, we should coalesce all the GSYM & 3865 // global symbols that have the 3866 // same address. 3867 is_gsym = true; 3868 sym[sym_idx].SetExternal(true); 3869 3870 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3871 llvm::StringRef symbol_name_ref(symbol_name); 3872 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3873 symbol_name_non_abi_mangled = symbol_name + 1; 3874 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3875 type = eSymbolTypeObjCClass; 3876 demangled_is_synthesized = true; 3877 3878 } else if (symbol_name_ref.startswith( 3879 g_objc_v2_prefix_metaclass)) { 3880 symbol_name_non_abi_mangled = symbol_name + 1; 3881 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3882 type = eSymbolTypeObjCMetaClass; 3883 demangled_is_synthesized = true; 3884 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3885 symbol_name_non_abi_mangled = symbol_name + 1; 3886 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3887 type = eSymbolTypeObjCIVar; 3888 demangled_is_synthesized = true; 3889 } 3890 } else { 3891 if (nlist.n_value != 0) 3892 symbol_section = 3893 section_info.GetSection(nlist.n_sect, nlist.n_value); 3894 type = eSymbolTypeData; 3895 } 3896 break; 3897 3898 case N_FNAME: 3899 // procedure name (f77 kludge): name,,NO_SECT,0,0 3900 type = eSymbolTypeCompiler; 3901 break; 3902 3903 case N_FUN: 3904 // procedure: name,,n_sect,linenumber,address 3905 if (symbol_name) { 3906 type = eSymbolTypeCode; 3907 symbol_section = 3908 section_info.GetSection(nlist.n_sect, nlist.n_value); 3909 3910 N_FUN_addr_to_sym_idx.insert( 3911 std::make_pair(nlist.n_value, sym_idx)); 3912 // We use the current number of symbols in the symbol table in 3913 // lieu of 3914 // using nlist_idx in case we ever start trimming entries out 3915 N_FUN_indexes.push_back(sym_idx); 3916 } else { 3917 type = eSymbolTypeCompiler; 3918 3919 if (!N_FUN_indexes.empty()) { 3920 // Copy the size of the function into the original STAB entry so 3921 // we don't have 3922 // to hunt for it later 3923 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3924 ->SetByteSize(nlist.n_value); 3925 N_FUN_indexes.pop_back(); 3926 // We don't really need the end function STAB as it contains the 3927 // size which 3928 // we already placed with the original symbol, so don't add it 3929 // if we want a 3930 // minimal symbol table 3931 add_nlist = false; 3932 } 3933 } 3934 break; 3935 3936 case N_STSYM: 3937 // static symbol: name,,n_sect,type,address 3938 N_STSYM_addr_to_sym_idx.insert( 3939 std::make_pair(nlist.n_value, sym_idx)); 3940 symbol_section = 3941 section_info.GetSection(nlist.n_sect, nlist.n_value); 3942 if (symbol_name && symbol_name[0]) { 3943 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3944 eSymbolTypeData); 3945 } 3946 break; 3947 3948 case N_LCSYM: 3949 // .lcomm symbol: name,,n_sect,type,address 3950 symbol_section = 3951 section_info.GetSection(nlist.n_sect, nlist.n_value); 3952 type = eSymbolTypeCommonBlock; 3953 break; 3954 3955 case N_BNSYM: 3956 // We use the current number of symbols in the symbol table in lieu 3957 // of 3958 // using nlist_idx in case we ever start trimming entries out 3959 // Skip these if we want minimal symbol tables 3960 add_nlist = false; 3961 break; 3962 3963 case N_ENSYM: 3964 // Set the size of the N_BNSYM to the terminating index of this 3965 // N_ENSYM 3966 // so that we can always skip the entire symbol if we need to 3967 // navigate 3968 // more quickly at the source level when parsing STABS 3969 // Skip these if we want minimal symbol tables 3970 add_nlist = false; 3971 break; 3972 3973 case N_OPT: 3974 // emitted with gcc2_compiled and in gcc source 3975 type = eSymbolTypeCompiler; 3976 break; 3977 3978 case N_RSYM: 3979 // register sym: name,,NO_SECT,type,register 3980 type = eSymbolTypeVariable; 3981 break; 3982 3983 case N_SLINE: 3984 // src line: 0,,n_sect,linenumber,address 3985 symbol_section = 3986 section_info.GetSection(nlist.n_sect, nlist.n_value); 3987 type = eSymbolTypeLineEntry; 3988 break; 3989 3990 case N_SSYM: 3991 // structure elt: name,,NO_SECT,type,struct_offset 3992 type = eSymbolTypeVariableType; 3993 break; 3994 3995 case N_SO: 3996 // source file name 3997 type = eSymbolTypeSourceFile; 3998 if (symbol_name == NULL) { 3999 add_nlist = false; 4000 if (N_SO_index != UINT32_MAX) { 4001 // Set the size of the N_SO to the terminating index of this 4002 // N_SO 4003 // so that we can always skip the entire N_SO if we need to 4004 // navigate 4005 // more quickly at the source level when parsing STABS 4006 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 4007 symbol_ptr->SetByteSize(sym_idx); 4008 symbol_ptr->SetSizeIsSibling(true); 4009 } 4010 N_NSYM_indexes.clear(); 4011 N_INCL_indexes.clear(); 4012 N_BRAC_indexes.clear(); 4013 N_COMM_indexes.clear(); 4014 N_FUN_indexes.clear(); 4015 N_SO_index = UINT32_MAX; 4016 } else { 4017 // We use the current number of symbols in the symbol table in 4018 // lieu of 4019 // using nlist_idx in case we ever start trimming entries out 4020 const bool N_SO_has_full_path = symbol_name[0] == '/'; 4021 if (N_SO_has_full_path) { 4022 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 4023 // We have two consecutive N_SO entries where the first 4024 // contains a directory 4025 // and the second contains a full path. 4026 sym[sym_idx - 1].GetMangled().SetValue( 4027 ConstString(symbol_name), false); 4028 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4029 add_nlist = false; 4030 } else { 4031 // This is the first entry in a N_SO that contains a directory 4032 // or 4033 // a full path to the source file 4034 N_SO_index = sym_idx; 4035 } 4036 } else if ((N_SO_index == sym_idx - 1) && 4037 ((sym_idx - 1) < num_syms)) { 4038 // This is usually the second N_SO entry that contains just the 4039 // filename, 4040 // so here we combine it with the first one if we are minimizing 4041 // the symbol table 4042 const char *so_path = 4043 sym[sym_idx - 1] 4044 .GetMangled() 4045 .GetDemangledName(lldb::eLanguageTypeUnknown) 4046 .AsCString(); 4047 if (so_path && so_path[0]) { 4048 std::string full_so_path(so_path); 4049 const size_t double_slash_pos = full_so_path.find("//"); 4050 if (double_slash_pos != std::string::npos) { 4051 // The linker has been generating bad N_SO entries with 4052 // doubled up paths 4053 // in the format "%s%s" where the first string in the 4054 // DW_AT_comp_dir, 4055 // and the second is the directory for the source file so 4056 // you end up with 4057 // a path that looks like "/tmp/src//tmp/src/" 4058 FileSpec so_dir(so_path, false); 4059 if (!so_dir.Exists()) { 4060 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 4061 false); 4062 if (so_dir.Exists()) { 4063 // Trim off the incorrect path 4064 full_so_path.erase(0, double_slash_pos + 1); 4065 } 4066 } 4067 } 4068 if (*full_so_path.rbegin() != '/') 4069 full_so_path += '/'; 4070 full_so_path += symbol_name; 4071 sym[sym_idx - 1].GetMangled().SetValue( 4072 ConstString(full_so_path.c_str()), false); 4073 add_nlist = false; 4074 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4075 } 4076 } else { 4077 // This could be a relative path to a N_SO 4078 N_SO_index = sym_idx; 4079 } 4080 } 4081 break; 4082 4083 case N_OSO: 4084 // object file name: name,,0,0,st_mtime 4085 type = eSymbolTypeObjectFile; 4086 break; 4087 4088 case N_LSYM: 4089 // local sym: name,,NO_SECT,type,offset 4090 type = eSymbolTypeLocal; 4091 break; 4092 4093 //---------------------------------------------------------------------- 4094 // INCL scopes 4095 //---------------------------------------------------------------------- 4096 case N_BINCL: 4097 // include file beginning: name,,NO_SECT,0,sum 4098 // We use the current number of symbols in the symbol table in lieu 4099 // of 4100 // using nlist_idx in case we ever start trimming entries out 4101 N_INCL_indexes.push_back(sym_idx); 4102 type = eSymbolTypeScopeBegin; 4103 break; 4104 4105 case N_EINCL: 4106 // include file end: name,,NO_SECT,0,0 4107 // Set the size of the N_BINCL to the terminating index of this 4108 // N_EINCL 4109 // so that we can always skip the entire symbol if we need to 4110 // navigate 4111 // more quickly at the source level when parsing STABS 4112 if (!N_INCL_indexes.empty()) { 4113 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4114 symbol_ptr->SetByteSize(sym_idx + 1); 4115 symbol_ptr->SetSizeIsSibling(true); 4116 N_INCL_indexes.pop_back(); 4117 } 4118 type = eSymbolTypeScopeEnd; 4119 break; 4120 4121 case N_SOL: 4122 // #included file name: name,,n_sect,0,address 4123 type = eSymbolTypeHeaderFile; 4124 4125 // We currently don't use the header files on darwin 4126 add_nlist = false; 4127 break; 4128 4129 case N_PARAMS: 4130 // compiler parameters: name,,NO_SECT,0,0 4131 type = eSymbolTypeCompiler; 4132 break; 4133 4134 case N_VERSION: 4135 // compiler version: name,,NO_SECT,0,0 4136 type = eSymbolTypeCompiler; 4137 break; 4138 4139 case N_OLEVEL: 4140 // compiler -O level: name,,NO_SECT,0,0 4141 type = eSymbolTypeCompiler; 4142 break; 4143 4144 case N_PSYM: 4145 // parameter: name,,NO_SECT,type,offset 4146 type = eSymbolTypeVariable; 4147 break; 4148 4149 case N_ENTRY: 4150 // alternate entry: name,,n_sect,linenumber,address 4151 symbol_section = 4152 section_info.GetSection(nlist.n_sect, nlist.n_value); 4153 type = eSymbolTypeLineEntry; 4154 break; 4155 4156 //---------------------------------------------------------------------- 4157 // Left and Right Braces 4158 //---------------------------------------------------------------------- 4159 case N_LBRAC: 4160 // left bracket: 0,,NO_SECT,nesting level,address 4161 // We use the current number of symbols in the symbol table in lieu 4162 // of 4163 // using nlist_idx in case we ever start trimming entries out 4164 symbol_section = 4165 section_info.GetSection(nlist.n_sect, nlist.n_value); 4166 N_BRAC_indexes.push_back(sym_idx); 4167 type = eSymbolTypeScopeBegin; 4168 break; 4169 4170 case N_RBRAC: 4171 // right bracket: 0,,NO_SECT,nesting level,address 4172 // Set the size of the N_LBRAC to the terminating index of this 4173 // N_RBRAC 4174 // so that we can always skip the entire symbol if we need to 4175 // navigate 4176 // more quickly at the source level when parsing STABS 4177 symbol_section = 4178 section_info.GetSection(nlist.n_sect, nlist.n_value); 4179 if (!N_BRAC_indexes.empty()) { 4180 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4181 symbol_ptr->SetByteSize(sym_idx + 1); 4182 symbol_ptr->SetSizeIsSibling(true); 4183 N_BRAC_indexes.pop_back(); 4184 } 4185 type = eSymbolTypeScopeEnd; 4186 break; 4187 4188 case N_EXCL: 4189 // deleted include file: name,,NO_SECT,0,sum 4190 type = eSymbolTypeHeaderFile; 4191 break; 4192 4193 //---------------------------------------------------------------------- 4194 // COMM scopes 4195 //---------------------------------------------------------------------- 4196 case N_BCOMM: 4197 // begin common: name,,NO_SECT,0,0 4198 // We use the current number of symbols in the symbol table in lieu 4199 // of 4200 // using nlist_idx in case we ever start trimming entries out 4201 type = eSymbolTypeScopeBegin; 4202 N_COMM_indexes.push_back(sym_idx); 4203 break; 4204 4205 case N_ECOML: 4206 // end common (local name): 0,,n_sect,0,address 4207 symbol_section = 4208 section_info.GetSection(nlist.n_sect, nlist.n_value); 4209 LLVM_FALLTHROUGH; 4210 4211 case N_ECOMM: 4212 // end common: name,,n_sect,0,0 4213 // Set the size of the N_BCOMM to the terminating index of this 4214 // N_ECOMM/N_ECOML 4215 // so that we can always skip the entire symbol if we need to 4216 // navigate 4217 // more quickly at the source level when parsing STABS 4218 if (!N_COMM_indexes.empty()) { 4219 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4220 symbol_ptr->SetByteSize(sym_idx + 1); 4221 symbol_ptr->SetSizeIsSibling(true); 4222 N_COMM_indexes.pop_back(); 4223 } 4224 type = eSymbolTypeScopeEnd; 4225 break; 4226 4227 case N_LENG: 4228 // second stab entry with length information 4229 type = eSymbolTypeAdditional; 4230 break; 4231 4232 default: 4233 break; 4234 } 4235 } else { 4236 // uint8_t n_pext = N_PEXT & nlist.n_type; 4237 uint8_t n_type = N_TYPE & nlist.n_type; 4238 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4239 4240 switch (n_type) { 4241 case N_INDR: { 4242 const char *reexport_name_cstr = 4243 strtab_data.PeekCStr(nlist.n_value); 4244 if (reexport_name_cstr && reexport_name_cstr[0]) { 4245 type = eSymbolTypeReExported; 4246 ConstString reexport_name( 4247 reexport_name_cstr + 4248 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4249 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4250 set_value = false; 4251 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4252 indirect_symbol_names.insert( 4253 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4254 } else 4255 type = eSymbolTypeUndefined; 4256 } break; 4257 4258 case N_UNDF: 4259 if (symbol_name && symbol_name[0]) { 4260 ConstString undefined_name(symbol_name + 4261 ((symbol_name[0] == '_') ? 1 : 0)); 4262 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4263 } 4264 LLVM_FALLTHROUGH; 4265 4266 case N_PBUD: 4267 type = eSymbolTypeUndefined; 4268 break; 4269 4270 case N_ABS: 4271 type = eSymbolTypeAbsolute; 4272 break; 4273 4274 case N_SECT: { 4275 symbol_section = 4276 section_info.GetSection(nlist.n_sect, nlist.n_value); 4277 4278 if (!symbol_section) { 4279 // TODO: warn about this? 4280 add_nlist = false; 4281 break; 4282 } 4283 4284 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4285 type = eSymbolTypeException; 4286 } else { 4287 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4288 4289 switch (section_type) { 4290 case S_CSTRING_LITERALS: 4291 type = eSymbolTypeData; 4292 break; // section with only literal C strings 4293 case S_4BYTE_LITERALS: 4294 type = eSymbolTypeData; 4295 break; // section with only 4 byte literals 4296 case S_8BYTE_LITERALS: 4297 type = eSymbolTypeData; 4298 break; // section with only 8 byte literals 4299 case S_LITERAL_POINTERS: 4300 type = eSymbolTypeTrampoline; 4301 break; // section with only pointers to literals 4302 case S_NON_LAZY_SYMBOL_POINTERS: 4303 type = eSymbolTypeTrampoline; 4304 break; // section with only non-lazy symbol pointers 4305 case S_LAZY_SYMBOL_POINTERS: 4306 type = eSymbolTypeTrampoline; 4307 break; // section with only lazy symbol pointers 4308 case S_SYMBOL_STUBS: 4309 type = eSymbolTypeTrampoline; 4310 break; // section with only symbol stubs, byte size of stub in 4311 // the reserved2 field 4312 case S_MOD_INIT_FUNC_POINTERS: 4313 type = eSymbolTypeCode; 4314 break; // section with only function pointers for initialization 4315 case S_MOD_TERM_FUNC_POINTERS: 4316 type = eSymbolTypeCode; 4317 break; // section with only function pointers for termination 4318 case S_INTERPOSING: 4319 type = eSymbolTypeTrampoline; 4320 break; // section with only pairs of function pointers for 4321 // interposing 4322 case S_16BYTE_LITERALS: 4323 type = eSymbolTypeData; 4324 break; // section with only 16 byte literals 4325 case S_DTRACE_DOF: 4326 type = eSymbolTypeInstrumentation; 4327 break; 4328 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4329 type = eSymbolTypeTrampoline; 4330 break; 4331 default: 4332 switch (symbol_section->GetType()) { 4333 case lldb::eSectionTypeCode: 4334 type = eSymbolTypeCode; 4335 break; 4336 case eSectionTypeData: 4337 case eSectionTypeDataCString: // Inlined C string data 4338 case eSectionTypeDataCStringPointers: // Pointers to C string 4339 // data 4340 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4341 // the symbol table 4342 case eSectionTypeData4: 4343 case eSectionTypeData8: 4344 case eSectionTypeData16: 4345 type = eSymbolTypeData; 4346 break; 4347 default: 4348 break; 4349 } 4350 break; 4351 } 4352 4353 if (type == eSymbolTypeInvalid) { 4354 const char *symbol_sect_name = 4355 symbol_section->GetName().AsCString(); 4356 if (symbol_section->IsDescendant(text_section_sp.get())) { 4357 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4358 S_ATTR_SELF_MODIFYING_CODE | 4359 S_ATTR_SOME_INSTRUCTIONS)) 4360 type = eSymbolTypeData; 4361 else 4362 type = eSymbolTypeCode; 4363 } else if (symbol_section->IsDescendant( 4364 data_section_sp.get()) || 4365 symbol_section->IsDescendant( 4366 data_dirty_section_sp.get()) || 4367 symbol_section->IsDescendant( 4368 data_const_section_sp.get())) { 4369 if (symbol_sect_name && 4370 ::strstr(symbol_sect_name, "__objc") == 4371 symbol_sect_name) { 4372 type = eSymbolTypeRuntime; 4373 4374 if (symbol_name) { 4375 llvm::StringRef symbol_name_ref(symbol_name); 4376 if (symbol_name_ref.startswith("_OBJC_")) { 4377 static const llvm::StringRef g_objc_v2_prefix_class( 4378 "_OBJC_CLASS_$_"); 4379 static const llvm::StringRef g_objc_v2_prefix_metaclass( 4380 "_OBJC_METACLASS_$_"); 4381 static const llvm::StringRef g_objc_v2_prefix_ivar( 4382 "_OBJC_IVAR_$_"); 4383 if (symbol_name_ref.startswith( 4384 g_objc_v2_prefix_class)) { 4385 symbol_name_non_abi_mangled = symbol_name + 1; 4386 symbol_name = 4387 symbol_name + g_objc_v2_prefix_class.size(); 4388 type = eSymbolTypeObjCClass; 4389 demangled_is_synthesized = true; 4390 } else if (symbol_name_ref.startswith( 4391 g_objc_v2_prefix_metaclass)) { 4392 symbol_name_non_abi_mangled = symbol_name + 1; 4393 symbol_name = 4394 symbol_name + g_objc_v2_prefix_metaclass.size(); 4395 type = eSymbolTypeObjCMetaClass; 4396 demangled_is_synthesized = true; 4397 } else if (symbol_name_ref.startswith( 4398 g_objc_v2_prefix_ivar)) { 4399 symbol_name_non_abi_mangled = symbol_name + 1; 4400 symbol_name = 4401 symbol_name + g_objc_v2_prefix_ivar.size(); 4402 type = eSymbolTypeObjCIVar; 4403 demangled_is_synthesized = true; 4404 } 4405 } 4406 } 4407 } else if (symbol_sect_name && 4408 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4409 symbol_sect_name) { 4410 type = eSymbolTypeException; 4411 } else { 4412 type = eSymbolTypeData; 4413 } 4414 } else if (symbol_sect_name && 4415 ::strstr(symbol_sect_name, "__IMPORT") == 4416 symbol_sect_name) { 4417 type = eSymbolTypeTrampoline; 4418 } else if (symbol_section->IsDescendant( 4419 objc_section_sp.get())) { 4420 type = eSymbolTypeRuntime; 4421 if (symbol_name && symbol_name[0] == '.') { 4422 llvm::StringRef symbol_name_ref(symbol_name); 4423 static const llvm::StringRef g_objc_v1_prefix_class( 4424 ".objc_class_name_"); 4425 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4426 symbol_name_non_abi_mangled = symbol_name; 4427 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4428 type = eSymbolTypeObjCClass; 4429 demangled_is_synthesized = true; 4430 } 4431 } 4432 } 4433 } 4434 } 4435 } break; 4436 } 4437 } 4438 4439 if (add_nlist) { 4440 uint64_t symbol_value = nlist.n_value; 4441 4442 if (symbol_name_non_abi_mangled) { 4443 sym[sym_idx].GetMangled().SetMangledName( 4444 ConstString(symbol_name_non_abi_mangled)); 4445 sym[sym_idx].GetMangled().SetDemangledName( 4446 ConstString(symbol_name)); 4447 } else { 4448 bool symbol_name_is_mangled = false; 4449 4450 if (symbol_name && symbol_name[0] == '_') { 4451 symbol_name_is_mangled = symbol_name[1] == '_'; 4452 symbol_name++; // Skip the leading underscore 4453 } 4454 4455 if (symbol_name) { 4456 ConstString const_symbol_name(symbol_name); 4457 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4458 symbol_name_is_mangled); 4459 } 4460 } 4461 4462 if (is_gsym) { 4463 const char *gsym_name = sym[sym_idx] 4464 .GetMangled() 4465 .GetName(lldb::eLanguageTypeUnknown, 4466 Mangled::ePreferMangled) 4467 .GetCString(); 4468 if (gsym_name) 4469 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4470 } 4471 4472 if (symbol_section) { 4473 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4474 if (symbol_byte_size == 0 && function_starts_count > 0) { 4475 addr_t symbol_lookup_file_addr = nlist.n_value; 4476 // Do an exact address match for non-ARM addresses, else get the 4477 // closest since 4478 // the symbol might be a thumb symbol which has an address with 4479 // bit zero set 4480 FunctionStarts::Entry *func_start_entry = 4481 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4482 if (is_arm && func_start_entry) { 4483 // Verify that the function start address is the symbol address 4484 // (ARM) 4485 // or the symbol address + 1 (thumb) 4486 if (func_start_entry->addr != symbol_lookup_file_addr && 4487 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4488 // Not the right entry, NULL it out... 4489 func_start_entry = NULL; 4490 } 4491 } 4492 if (func_start_entry) { 4493 func_start_entry->data = true; 4494 4495 addr_t symbol_file_addr = func_start_entry->addr; 4496 if (is_arm) 4497 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4498 4499 const FunctionStarts::Entry *next_func_start_entry = 4500 function_starts.FindNextEntry(func_start_entry); 4501 const addr_t section_end_file_addr = 4502 section_file_addr + symbol_section->GetByteSize(); 4503 if (next_func_start_entry) { 4504 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4505 // Be sure the clear the Thumb address bit when we calculate 4506 // the size 4507 // from the current and next address 4508 if (is_arm) 4509 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4510 symbol_byte_size = std::min<lldb::addr_t>( 4511 next_symbol_file_addr - symbol_file_addr, 4512 section_end_file_addr - symbol_file_addr); 4513 } else { 4514 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4515 } 4516 } 4517 } 4518 symbol_value -= section_file_addr; 4519 } 4520 4521 if (is_debug == false) { 4522 if (type == eSymbolTypeCode) { 4523 // See if we can find a N_FUN entry for any code symbols. 4524 // If we do find a match, and the name matches, then we 4525 // can merge the two into just the function symbol to avoid 4526 // duplicate entries in the symbol table 4527 std::pair<ValueToSymbolIndexMap::const_iterator, 4528 ValueToSymbolIndexMap::const_iterator> 4529 range; 4530 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4531 if (range.first != range.second) { 4532 bool found_it = false; 4533 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4534 pos != range.second; ++pos) { 4535 if (sym[sym_idx].GetMangled().GetName( 4536 lldb::eLanguageTypeUnknown, 4537 Mangled::ePreferMangled) == 4538 sym[pos->second].GetMangled().GetName( 4539 lldb::eLanguageTypeUnknown, 4540 Mangled::ePreferMangled)) { 4541 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4542 // We just need the flags from the linker symbol, so put 4543 // these flags 4544 // into the N_FUN flags to avoid duplicate symbols in the 4545 // symbol table 4546 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4547 sym[pos->second].SetFlags(nlist.n_type << 16 | 4548 nlist.n_desc); 4549 if (resolver_addresses.find(nlist.n_value) != 4550 resolver_addresses.end()) 4551 sym[pos->second].SetType(eSymbolTypeResolver); 4552 sym[sym_idx].Clear(); 4553 found_it = true; 4554 break; 4555 } 4556 } 4557 if (found_it) 4558 continue; 4559 } else { 4560 if (resolver_addresses.find(nlist.n_value) != 4561 resolver_addresses.end()) 4562 type = eSymbolTypeResolver; 4563 } 4564 } else if (type == eSymbolTypeData || 4565 type == eSymbolTypeObjCClass || 4566 type == eSymbolTypeObjCMetaClass || 4567 type == eSymbolTypeObjCIVar) { 4568 // See if we can find a N_STSYM entry for any data symbols. 4569 // If we do find a match, and the name matches, then we 4570 // can merge the two into just the Static symbol to avoid 4571 // duplicate entries in the symbol table 4572 std::pair<ValueToSymbolIndexMap::const_iterator, 4573 ValueToSymbolIndexMap::const_iterator> 4574 range; 4575 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4576 if (range.first != range.second) { 4577 bool found_it = false; 4578 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4579 pos != range.second; ++pos) { 4580 if (sym[sym_idx].GetMangled().GetName( 4581 lldb::eLanguageTypeUnknown, 4582 Mangled::ePreferMangled) == 4583 sym[pos->second].GetMangled().GetName( 4584 lldb::eLanguageTypeUnknown, 4585 Mangled::ePreferMangled)) { 4586 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4587 // We just need the flags from the linker symbol, so put 4588 // these flags 4589 // into the N_STSYM flags to avoid duplicate symbols in the 4590 // symbol table 4591 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4592 sym[pos->second].SetFlags(nlist.n_type << 16 | 4593 nlist.n_desc); 4594 sym[sym_idx].Clear(); 4595 found_it = true; 4596 break; 4597 } 4598 } 4599 if (found_it) 4600 continue; 4601 } else { 4602 // Combine N_GSYM stab entries with the non stab symbol 4603 const char *gsym_name = sym[sym_idx] 4604 .GetMangled() 4605 .GetName(lldb::eLanguageTypeUnknown, 4606 Mangled::ePreferMangled) 4607 .GetCString(); 4608 if (gsym_name) { 4609 ConstNameToSymbolIndexMap::const_iterator pos = 4610 N_GSYM_name_to_sym_idx.find(gsym_name); 4611 if (pos != N_GSYM_name_to_sym_idx.end()) { 4612 const uint32_t GSYM_sym_idx = pos->second; 4613 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4614 // Copy the address, because often the N_GSYM address has an 4615 // invalid address of zero 4616 // when the global is a common symbol 4617 sym[GSYM_sym_idx].GetAddressRef().SetSection( 4618 symbol_section); 4619 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4620 // We just need the flags from the linker symbol, so put 4621 // these flags 4622 // into the N_GSYM flags to avoid duplicate symbols in the 4623 // symbol table 4624 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 4625 nlist.n_desc); 4626 sym[sym_idx].Clear(); 4627 continue; 4628 } 4629 } 4630 } 4631 } 4632 } 4633 4634 sym[sym_idx].SetID(nlist_idx); 4635 sym[sym_idx].SetType(type); 4636 if (set_value) { 4637 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4638 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4639 } 4640 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4641 4642 if (symbol_byte_size > 0) 4643 sym[sym_idx].SetByteSize(symbol_byte_size); 4644 4645 if (demangled_is_synthesized) 4646 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4647 4648 ++sym_idx; 4649 } else { 4650 sym[sym_idx].Clear(); 4651 } 4652 } 4653 4654 for (const auto &pos : reexport_shlib_needs_fixup) { 4655 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4656 if (undef_pos != undefined_name_to_desc.end()) { 4657 const uint8_t dylib_ordinal = 4658 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4659 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4660 sym[pos.first].SetReExportedSymbolSharedLibrary( 4661 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4662 } 4663 } 4664 } 4665 4666 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4667 4668 if (function_starts_count > 0) { 4669 uint32_t num_synthetic_function_symbols = 0; 4670 for (i = 0; i < function_starts_count; ++i) { 4671 if (function_starts.GetEntryRef(i).data == false) 4672 ++num_synthetic_function_symbols; 4673 } 4674 4675 if (num_synthetic_function_symbols > 0) { 4676 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4677 num_syms = sym_idx + num_synthetic_function_symbols; 4678 sym = symtab->Resize(num_syms); 4679 } 4680 for (i = 0; i < function_starts_count; ++i) { 4681 const FunctionStarts::Entry *func_start_entry = 4682 function_starts.GetEntryAtIndex(i); 4683 if (func_start_entry->data == false) { 4684 addr_t symbol_file_addr = func_start_entry->addr; 4685 uint32_t symbol_flags = 0; 4686 if (is_arm) { 4687 if (symbol_file_addr & 1) 4688 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4689 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4690 } 4691 Address symbol_addr; 4692 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4693 SectionSP symbol_section(symbol_addr.GetSection()); 4694 uint32_t symbol_byte_size = 0; 4695 if (symbol_section) { 4696 const addr_t section_file_addr = 4697 symbol_section->GetFileAddress(); 4698 const FunctionStarts::Entry *next_func_start_entry = 4699 function_starts.FindNextEntry(func_start_entry); 4700 const addr_t section_end_file_addr = 4701 section_file_addr + symbol_section->GetByteSize(); 4702 if (next_func_start_entry) { 4703 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4704 if (is_arm) 4705 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4706 symbol_byte_size = std::min<lldb::addr_t>( 4707 next_symbol_file_addr - symbol_file_addr, 4708 section_end_file_addr - symbol_file_addr); 4709 } else { 4710 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4711 } 4712 sym[sym_idx].SetID(synthetic_sym_id++); 4713 sym[sym_idx].GetMangled().SetDemangledName( 4714 GetNextSyntheticSymbolName()); 4715 sym[sym_idx].SetType(eSymbolTypeCode); 4716 sym[sym_idx].SetIsSynthetic(true); 4717 sym[sym_idx].GetAddressRef() = symbol_addr; 4718 if (symbol_flags) 4719 sym[sym_idx].SetFlags(symbol_flags); 4720 if (symbol_byte_size) 4721 sym[sym_idx].SetByteSize(symbol_byte_size); 4722 ++sym_idx; 4723 } 4724 } 4725 } 4726 } 4727 } 4728 } 4729 4730 // Trim our symbols down to just what we ended up with after 4731 // removing any symbols. 4732 if (sym_idx < num_syms) { 4733 num_syms = sym_idx; 4734 sym = symtab->Resize(num_syms); 4735 } 4736 4737 // Now synthesize indirect symbols 4738 if (m_dysymtab.nindirectsyms != 0) { 4739 if (indirect_symbol_index_data.GetByteSize()) { 4740 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4741 m_nlist_idx_to_sym_idx.end(); 4742 4743 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4744 ++sect_idx) { 4745 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4746 S_SYMBOL_STUBS) { 4747 uint32_t symbol_stub_byte_size = 4748 m_mach_sections[sect_idx].reserved2; 4749 if (symbol_stub_byte_size == 0) 4750 continue; 4751 4752 const uint32_t num_symbol_stubs = 4753 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4754 4755 if (num_symbol_stubs == 0) 4756 continue; 4757 4758 const uint32_t symbol_stub_index_offset = 4759 m_mach_sections[sect_idx].reserved1; 4760 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; 4761 ++stub_idx) { 4762 const uint32_t symbol_stub_index = 4763 symbol_stub_index_offset + stub_idx; 4764 const lldb::addr_t symbol_stub_addr = 4765 m_mach_sections[sect_idx].addr + 4766 (stub_idx * symbol_stub_byte_size); 4767 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4768 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4769 symbol_stub_offset, 4)) { 4770 const uint32_t stub_sym_id = 4771 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4772 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4773 continue; 4774 4775 NListIndexToSymbolIndexMap::const_iterator index_pos = 4776 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4777 Symbol *stub_symbol = NULL; 4778 if (index_pos != end_index_pos) { 4779 // We have a remapping from the original nlist index to 4780 // a current symbol index, so just look this up by index 4781 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4782 } else { 4783 // We need to lookup a symbol using the original nlist 4784 // symbol index since this index is coming from the 4785 // S_SYMBOL_STUBS 4786 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4787 } 4788 4789 if (stub_symbol) { 4790 Address so_addr(symbol_stub_addr, section_list); 4791 4792 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4793 // Change the external symbol into a trampoline that makes 4794 // sense 4795 // These symbols were N_UNDF N_EXT, and are useless to us, 4796 // so we 4797 // can re-use them so we don't have to make up a synthetic 4798 // symbol 4799 // for no good reason. 4800 if (resolver_addresses.find(symbol_stub_addr) == 4801 resolver_addresses.end()) 4802 stub_symbol->SetType(eSymbolTypeTrampoline); 4803 else 4804 stub_symbol->SetType(eSymbolTypeResolver); 4805 stub_symbol->SetExternal(false); 4806 stub_symbol->GetAddressRef() = so_addr; 4807 stub_symbol->SetByteSize(symbol_stub_byte_size); 4808 } else { 4809 // Make a synthetic symbol to describe the trampoline stub 4810 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4811 if (sym_idx >= num_syms) { 4812 sym = symtab->Resize(++num_syms); 4813 stub_symbol = NULL; // this pointer no longer valid 4814 } 4815 sym[sym_idx].SetID(synthetic_sym_id++); 4816 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4817 if (resolver_addresses.find(symbol_stub_addr) == 4818 resolver_addresses.end()) 4819 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4820 else 4821 sym[sym_idx].SetType(eSymbolTypeResolver); 4822 sym[sym_idx].SetIsSynthetic(true); 4823 sym[sym_idx].GetAddressRef() = so_addr; 4824 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4825 ++sym_idx; 4826 } 4827 } else { 4828 if (log) 4829 log->Warning("symbol stub referencing symbol table symbol " 4830 "%u that isn't in our minimal symbol table, " 4831 "fix this!!!", 4832 stub_sym_id); 4833 } 4834 } 4835 } 4836 } 4837 } 4838 } 4839 } 4840 4841 if (!trie_entries.empty()) { 4842 for (const auto &e : trie_entries) { 4843 if (e.entry.import_name) { 4844 // Only add indirect symbols from the Trie entries if we 4845 // didn't have a N_INDR nlist entry for this already 4846 if (indirect_symbol_names.find(e.entry.name) == 4847 indirect_symbol_names.end()) { 4848 // Make a synthetic symbol to describe re-exported symbol. 4849 if (sym_idx >= num_syms) 4850 sym = symtab->Resize(++num_syms); 4851 sym[sym_idx].SetID(synthetic_sym_id++); 4852 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4853 sym[sym_idx].SetType(eSymbolTypeReExported); 4854 sym[sym_idx].SetIsSynthetic(true); 4855 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4856 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4857 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4858 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4859 } 4860 ++sym_idx; 4861 } 4862 } 4863 } 4864 } 4865 4866 // StreamFile s(stdout, false); 4867 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4868 // symtab->Dump(&s, NULL, eSortOrderNone); 4869 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4870 // sizes 4871 symtab->CalculateSymbolSizes(); 4872 4873 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4874 // symtab->Dump(&s, NULL, eSortOrderNone); 4875 4876 return symtab->GetNumSymbols(); 4877 } 4878 return 0; 4879 } 4880 4881 void ObjectFileMachO::Dump(Stream *s) { 4882 ModuleSP module_sp(GetModule()); 4883 if (module_sp) { 4884 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4885 s->Printf("%p: ", static_cast<void *>(this)); 4886 s->Indent(); 4887 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4888 s->PutCString("ObjectFileMachO64"); 4889 else 4890 s->PutCString("ObjectFileMachO32"); 4891 4892 ArchSpec header_arch; 4893 GetArchitecture(header_arch); 4894 4895 *s << ", file = '" << m_file 4896 << "', arch = " << header_arch.GetArchitectureName() << "\n"; 4897 4898 SectionList *sections = GetSectionList(); 4899 if (sections) 4900 sections->Dump(s, NULL, true, UINT32_MAX); 4901 4902 if (m_symtab_ap.get()) 4903 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4904 } 4905 } 4906 4907 bool ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4908 const lldb_private::DataExtractor &data, 4909 lldb::offset_t lc_offset, 4910 lldb_private::UUID &uuid) { 4911 uint32_t i; 4912 struct uuid_command load_cmd; 4913 4914 lldb::offset_t offset = lc_offset; 4915 for (i = 0; i < header.ncmds; ++i) { 4916 const lldb::offset_t cmd_offset = offset; 4917 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4918 break; 4919 4920 if (load_cmd.cmd == LC_UUID) { 4921 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4922 4923 if (uuid_bytes) { 4924 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4925 // We pretend these object files have no UUID to prevent crashing. 4926 4927 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4928 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4929 0xbb, 0x14, 0xf0, 0x0d}; 4930 4931 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4932 return false; 4933 4934 uuid.SetBytes(uuid_bytes); 4935 return true; 4936 } 4937 return false; 4938 } 4939 offset = cmd_offset + load_cmd.cmdsize; 4940 } 4941 return false; 4942 } 4943 4944 bool ObjectFileMachO::GetArchitecture(const llvm::MachO::mach_header &header, 4945 const lldb_private::DataExtractor &data, 4946 lldb::offset_t lc_offset, 4947 ArchSpec &arch) { 4948 arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4949 4950 if (arch.IsValid()) { 4951 llvm::Triple &triple = arch.GetTriple(); 4952 4953 // Set OS to an unspecified unknown or a "*" so it can match any OS 4954 triple.setOS(llvm::Triple::UnknownOS); 4955 triple.setOSName(llvm::StringRef()); 4956 4957 if (header.filetype == MH_PRELOAD) { 4958 if (header.cputype == CPU_TYPE_ARM) { 4959 // If this is a 32-bit arm binary, and it's a standalone binary, 4960 // force the Vendor to Apple so we don't accidentally pick up 4961 // the generic armv7 ABI at runtime. Apple's armv7 ABI always uses 4962 // r7 for the frame pointer register; most other armv7 ABIs use a 4963 // combination of r7 and r11. 4964 triple.setVendor(llvm::Triple::Apple); 4965 } else { 4966 // Set vendor to an unspecified unknown or a "*" so it can match any 4967 // vendor 4968 // This is required for correct behavior of EFI debugging on x86_64 4969 triple.setVendor(llvm::Triple::UnknownVendor); 4970 triple.setVendorName(llvm::StringRef()); 4971 } 4972 return true; 4973 } else { 4974 struct load_command load_cmd; 4975 4976 lldb::offset_t offset = lc_offset; 4977 for (uint32_t i = 0; i < header.ncmds; ++i) { 4978 const lldb::offset_t cmd_offset = offset; 4979 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4980 break; 4981 4982 switch (load_cmd.cmd) { 4983 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4984 triple.setOS(llvm::Triple::IOS); 4985 return true; 4986 4987 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4988 triple.setOS(llvm::Triple::MacOSX); 4989 return true; 4990 4991 case llvm::MachO::LC_VERSION_MIN_TVOS: 4992 triple.setOS(llvm::Triple::TvOS); 4993 return true; 4994 4995 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4996 triple.setOS(llvm::Triple::WatchOS); 4997 return true; 4998 4999 default: 5000 break; 5001 } 5002 5003 offset = cmd_offset + load_cmd.cmdsize; 5004 } 5005 5006 if (header.filetype != MH_KEXT_BUNDLE) { 5007 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 5008 // so lets not say our Vendor is Apple, leave it as an unspecified 5009 // unknown 5010 triple.setVendor(llvm::Triple::UnknownVendor); 5011 triple.setVendorName(llvm::StringRef()); 5012 } 5013 } 5014 } 5015 return arch.IsValid(); 5016 } 5017 5018 bool ObjectFileMachO::GetUUID(lldb_private::UUID *uuid) { 5019 ModuleSP module_sp(GetModule()); 5020 if (module_sp) { 5021 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5022 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5023 return GetUUID(m_header, m_data, offset, *uuid); 5024 } 5025 return false; 5026 } 5027 5028 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5029 uint32_t count = 0; 5030 ModuleSP module_sp(GetModule()); 5031 if (module_sp) { 5032 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5033 struct load_command load_cmd; 5034 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5035 std::vector<std::string> rpath_paths; 5036 std::vector<std::string> rpath_relative_paths; 5037 const bool resolve_path = false; // Don't resolve the dependent file paths 5038 // since they may not reside on this system 5039 uint32_t i; 5040 for (i = 0; i < m_header.ncmds; ++i) { 5041 const uint32_t cmd_offset = offset; 5042 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5043 break; 5044 5045 switch (load_cmd.cmd) { 5046 case LC_RPATH: 5047 case LC_LOAD_DYLIB: 5048 case LC_LOAD_WEAK_DYLIB: 5049 case LC_REEXPORT_DYLIB: 5050 case LC_LOAD_DYLINKER: 5051 case LC_LOADFVMLIB: 5052 case LC_LOAD_UPWARD_DYLIB: { 5053 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5054 const char *path = m_data.PeekCStr(name_offset); 5055 if (path) { 5056 if (load_cmd.cmd == LC_RPATH) 5057 rpath_paths.push_back(path); 5058 else { 5059 if (path[0] == '@') { 5060 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5061 rpath_relative_paths.push_back(path + strlen("@rpath")); 5062 } else { 5063 FileSpec file_spec(path, resolve_path); 5064 if (files.AppendIfUnique(file_spec)) 5065 count++; 5066 } 5067 } 5068 } 5069 } break; 5070 5071 default: 5072 break; 5073 } 5074 offset = cmd_offset + load_cmd.cmdsize; 5075 } 5076 5077 if (!rpath_paths.empty()) { 5078 // Fixup all LC_RPATH values to be absolute paths 5079 FileSpec this_file_spec(m_file); 5080 this_file_spec.ResolvePath(); 5081 std::string loader_path("@loader_path"); 5082 std::string executable_path("@executable_path"); 5083 for (auto &rpath : rpath_paths) { 5084 if (rpath.find(loader_path) == 0) { 5085 rpath.erase(0, loader_path.size()); 5086 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5087 } else if (rpath.find(executable_path) == 0) { 5088 rpath.erase(0, executable_path.size()); 5089 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5090 } 5091 } 5092 5093 for (const auto &rpath_relative_path : rpath_relative_paths) { 5094 for (const auto &rpath : rpath_paths) { 5095 std::string path = rpath; 5096 path += rpath_relative_path; 5097 // It is OK to resolve this path because we must find a file on 5098 // disk for us to accept it anyway if it is rpath relative. 5099 FileSpec file_spec(path, true); 5100 // Remove any redundant parts of the path (like "../foo") since 5101 // LC_RPATH values often contain "..". 5102 file_spec = file_spec.GetNormalizedPath(); 5103 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) { 5104 count++; 5105 break; 5106 } 5107 } 5108 } 5109 } 5110 } 5111 return count; 5112 } 5113 5114 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5115 // If the object file is not an executable it can't hold the entry point. 5116 // m_entry_point_address 5117 // is initialized to an invalid address, so we can just return that. 5118 // If m_entry_point_address is valid it means we've found it already, so 5119 // return the cached value. 5120 5121 if (!IsExecutable() || m_entry_point_address.IsValid()) 5122 return m_entry_point_address; 5123 5124 // Otherwise, look for the UnixThread or Thread command. The data for the 5125 // Thread command is given in 5126 // /usr/include/mach-o.h, but it is basically: 5127 // 5128 // uint32_t flavor - this is the flavor argument you would pass to 5129 // thread_get_state 5130 // uint32_t count - this is the count of longs in the thread state data 5131 // struct XXX_thread_state state - this is the structure from 5132 // <machine/thread_status.h> corresponding to the flavor. 5133 // <repeat this trio> 5134 // 5135 // So we just keep reading the various register flavors till we find the GPR 5136 // one, then read the PC out of there. 5137 // FIXME: We will need to have a "RegisterContext data provider" class at some 5138 // point that can get all the registers 5139 // out of data in this form & attach them to a given thread. That should 5140 // underlie the MacOS X User process plugin, 5141 // and we'll also need it for the MacOS X Core File process plugin. When we 5142 // have that we can also use it here. 5143 // 5144 // For now we hard-code the offsets and flavors we need: 5145 // 5146 // 5147 5148 ModuleSP module_sp(GetModule()); 5149 if (module_sp) { 5150 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5151 struct load_command load_cmd; 5152 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5153 uint32_t i; 5154 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5155 bool done = false; 5156 5157 for (i = 0; i < m_header.ncmds; ++i) { 5158 const lldb::offset_t cmd_offset = offset; 5159 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5160 break; 5161 5162 switch (load_cmd.cmd) { 5163 case LC_UNIXTHREAD: 5164 case LC_THREAD: { 5165 while (offset < cmd_offset + load_cmd.cmdsize) { 5166 uint32_t flavor = m_data.GetU32(&offset); 5167 uint32_t count = m_data.GetU32(&offset); 5168 if (count == 0) { 5169 // We've gotten off somehow, log and exit; 5170 return m_entry_point_address; 5171 } 5172 5173 switch (m_header.cputype) { 5174 case llvm::MachO::CPU_TYPE_ARM: 5175 if (flavor == 1 || 5176 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 from 5177 // mach/arm/thread_status.h 5178 { 5179 offset += 60; // This is the offset of pc in the GPR thread state 5180 // data structure. 5181 start_address = m_data.GetU32(&offset); 5182 done = true; 5183 } 5184 break; 5185 case llvm::MachO::CPU_TYPE_ARM64: 5186 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5187 { 5188 offset += 256; // This is the offset of pc in the GPR thread state 5189 // data structure. 5190 start_address = m_data.GetU64(&offset); 5191 done = true; 5192 } 5193 break; 5194 case llvm::MachO::CPU_TYPE_I386: 5195 if (flavor == 5196 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5197 { 5198 offset += 40; // This is the offset of eip in the GPR thread state 5199 // data structure. 5200 start_address = m_data.GetU32(&offset); 5201 done = true; 5202 } 5203 break; 5204 case llvm::MachO::CPU_TYPE_X86_64: 5205 if (flavor == 5206 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5207 { 5208 offset += 16 * 8; // This is the offset of rip in the GPR thread 5209 // state data structure. 5210 start_address = m_data.GetU64(&offset); 5211 done = true; 5212 } 5213 break; 5214 default: 5215 return m_entry_point_address; 5216 } 5217 // Haven't found the GPR flavor yet, skip over the data for this 5218 // flavor: 5219 if (done) 5220 break; 5221 offset += count * 4; 5222 } 5223 } break; 5224 case LC_MAIN: { 5225 ConstString text_segment_name("__TEXT"); 5226 uint64_t entryoffset = m_data.GetU64(&offset); 5227 SectionSP text_segment_sp = 5228 GetSectionList()->FindSectionByName(text_segment_name); 5229 if (text_segment_sp) { 5230 done = true; 5231 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5232 } 5233 } break; 5234 5235 default: 5236 break; 5237 } 5238 if (done) 5239 break; 5240 5241 // Go to the next load command: 5242 offset = cmd_offset + load_cmd.cmdsize; 5243 } 5244 5245 if (start_address != LLDB_INVALID_ADDRESS) { 5246 // We got the start address from the load commands, so now resolve that 5247 // address in the sections 5248 // of this ObjectFile: 5249 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5250 start_address, GetSectionList())) { 5251 m_entry_point_address.Clear(); 5252 } 5253 } else { 5254 // We couldn't read the UnixThread load command - maybe it wasn't there. 5255 // As a fallback look for the 5256 // "start" symbol in the main executable. 5257 5258 ModuleSP module_sp(GetModule()); 5259 5260 if (module_sp) { 5261 SymbolContextList contexts; 5262 SymbolContext context; 5263 if (module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5264 eSymbolTypeCode, contexts)) { 5265 if (contexts.GetContextAtIndex(0, context)) 5266 m_entry_point_address = context.symbol->GetAddress(); 5267 } 5268 } 5269 } 5270 } 5271 5272 return m_entry_point_address; 5273 } 5274 5275 lldb_private::Address ObjectFileMachO::GetHeaderAddress() { 5276 lldb_private::Address header_addr; 5277 SectionList *section_list = GetSectionList(); 5278 if (section_list) { 5279 SectionSP text_segment_sp( 5280 section_list->FindSectionByName(GetSegmentNameTEXT())); 5281 if (text_segment_sp) { 5282 header_addr.SetSection(text_segment_sp); 5283 header_addr.SetOffset(0); 5284 } 5285 } 5286 return header_addr; 5287 } 5288 5289 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5290 ModuleSP module_sp(GetModule()); 5291 if (module_sp) { 5292 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5293 if (!m_thread_context_offsets_valid) { 5294 m_thread_context_offsets_valid = true; 5295 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5296 FileRangeArray::Entry file_range; 5297 thread_command thread_cmd; 5298 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5299 const uint32_t cmd_offset = offset; 5300 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 5301 break; 5302 5303 if (thread_cmd.cmd == LC_THREAD) { 5304 file_range.SetRangeBase(offset); 5305 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5306 m_thread_context_offsets.Append(file_range); 5307 } 5308 offset = cmd_offset + thread_cmd.cmdsize; 5309 } 5310 } 5311 } 5312 return m_thread_context_offsets.GetSize(); 5313 } 5314 5315 lldb::RegisterContextSP 5316 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5317 lldb_private::Thread &thread) { 5318 lldb::RegisterContextSP reg_ctx_sp; 5319 5320 ModuleSP module_sp(GetModule()); 5321 if (module_sp) { 5322 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5323 if (!m_thread_context_offsets_valid) 5324 GetNumThreadContexts(); 5325 5326 const FileRangeArray::Entry *thread_context_file_range = 5327 m_thread_context_offsets.GetEntryAtIndex(idx); 5328 if (thread_context_file_range) { 5329 5330 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5331 thread_context_file_range->GetByteSize()); 5332 5333 switch (m_header.cputype) { 5334 case llvm::MachO::CPU_TYPE_ARM64: 5335 reg_ctx_sp.reset(new RegisterContextDarwin_arm64_Mach(thread, data)); 5336 break; 5337 5338 case llvm::MachO::CPU_TYPE_ARM: 5339 reg_ctx_sp.reset(new RegisterContextDarwin_arm_Mach(thread, data)); 5340 break; 5341 5342 case llvm::MachO::CPU_TYPE_I386: 5343 reg_ctx_sp.reset(new RegisterContextDarwin_i386_Mach(thread, data)); 5344 break; 5345 5346 case llvm::MachO::CPU_TYPE_X86_64: 5347 reg_ctx_sp.reset(new RegisterContextDarwin_x86_64_Mach(thread, data)); 5348 break; 5349 } 5350 } 5351 } 5352 return reg_ctx_sp; 5353 } 5354 5355 ObjectFile::Type ObjectFileMachO::CalculateType() { 5356 switch (m_header.filetype) { 5357 case MH_OBJECT: // 0x1u 5358 if (GetAddressByteSize() == 4) { 5359 // 32 bit kexts are just object files, but they do have a valid 5360 // UUID load command. 5361 UUID uuid; 5362 if (GetUUID(&uuid)) { 5363 // this checking for the UUID load command is not enough 5364 // we could eventually look for the symbol named 5365 // "OSKextGetCurrentIdentifier" as this is required of kexts 5366 if (m_strata == eStrataInvalid) 5367 m_strata = eStrataKernel; 5368 return eTypeSharedLibrary; 5369 } 5370 } 5371 return eTypeObjectFile; 5372 5373 case MH_EXECUTE: 5374 return eTypeExecutable; // 0x2u 5375 case MH_FVMLIB: 5376 return eTypeSharedLibrary; // 0x3u 5377 case MH_CORE: 5378 return eTypeCoreFile; // 0x4u 5379 case MH_PRELOAD: 5380 return eTypeSharedLibrary; // 0x5u 5381 case MH_DYLIB: 5382 return eTypeSharedLibrary; // 0x6u 5383 case MH_DYLINKER: 5384 return eTypeDynamicLinker; // 0x7u 5385 case MH_BUNDLE: 5386 return eTypeSharedLibrary; // 0x8u 5387 case MH_DYLIB_STUB: 5388 return eTypeStubLibrary; // 0x9u 5389 case MH_DSYM: 5390 return eTypeDebugInfo; // 0xAu 5391 case MH_KEXT_BUNDLE: 5392 return eTypeSharedLibrary; // 0xBu 5393 default: 5394 break; 5395 } 5396 return eTypeUnknown; 5397 } 5398 5399 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5400 switch (m_header.filetype) { 5401 case MH_OBJECT: // 0x1u 5402 { 5403 // 32 bit kexts are just object files, but they do have a valid 5404 // UUID load command. 5405 UUID uuid; 5406 if (GetUUID(&uuid)) { 5407 // this checking for the UUID load command is not enough 5408 // we could eventually look for the symbol named 5409 // "OSKextGetCurrentIdentifier" as this is required of kexts 5410 if (m_type == eTypeInvalid) 5411 m_type = eTypeSharedLibrary; 5412 5413 return eStrataKernel; 5414 } 5415 } 5416 return eStrataUnknown; 5417 5418 case MH_EXECUTE: // 0x2u 5419 // Check for the MH_DYLDLINK bit in the flags 5420 if (m_header.flags & MH_DYLDLINK) { 5421 return eStrataUser; 5422 } else { 5423 SectionList *section_list = GetSectionList(); 5424 if (section_list) { 5425 static ConstString g_kld_section_name("__KLD"); 5426 if (section_list->FindSectionByName(g_kld_section_name)) 5427 return eStrataKernel; 5428 } 5429 } 5430 return eStrataRawImage; 5431 5432 case MH_FVMLIB: 5433 return eStrataUser; // 0x3u 5434 case MH_CORE: 5435 return eStrataUnknown; // 0x4u 5436 case MH_PRELOAD: 5437 return eStrataRawImage; // 0x5u 5438 case MH_DYLIB: 5439 return eStrataUser; // 0x6u 5440 case MH_DYLINKER: 5441 return eStrataUser; // 0x7u 5442 case MH_BUNDLE: 5443 return eStrataUser; // 0x8u 5444 case MH_DYLIB_STUB: 5445 return eStrataUser; // 0x9u 5446 case MH_DSYM: 5447 return eStrataUnknown; // 0xAu 5448 case MH_KEXT_BUNDLE: 5449 return eStrataKernel; // 0xBu 5450 default: 5451 break; 5452 } 5453 return eStrataUnknown; 5454 } 5455 5456 uint32_t ObjectFileMachO::GetVersion(uint32_t *versions, 5457 uint32_t num_versions) { 5458 ModuleSP module_sp(GetModule()); 5459 if (module_sp) { 5460 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5461 struct dylib_command load_cmd; 5462 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5463 uint32_t version_cmd = 0; 5464 uint64_t version = 0; 5465 uint32_t i; 5466 for (i = 0; i < m_header.ncmds; ++i) { 5467 const lldb::offset_t cmd_offset = offset; 5468 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5469 break; 5470 5471 if (load_cmd.cmd == LC_ID_DYLIB) { 5472 if (version_cmd == 0) { 5473 version_cmd = load_cmd.cmd; 5474 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 5475 break; 5476 version = load_cmd.dylib.current_version; 5477 } 5478 break; // Break for now unless there is another more complete version 5479 // number load command in the future. 5480 } 5481 offset = cmd_offset + load_cmd.cmdsize; 5482 } 5483 5484 if (version_cmd == LC_ID_DYLIB) { 5485 if (versions != NULL && num_versions > 0) { 5486 if (num_versions > 0) 5487 versions[0] = (version & 0xFFFF0000ull) >> 16; 5488 if (num_versions > 1) 5489 versions[1] = (version & 0x0000FF00ull) >> 8; 5490 if (num_versions > 2) 5491 versions[2] = (version & 0x000000FFull); 5492 // Fill in an remaining version numbers with invalid values 5493 for (i = 3; i < num_versions; ++i) 5494 versions[i] = UINT32_MAX; 5495 } 5496 // The LC_ID_DYLIB load command has a version with 3 version numbers 5497 // in it, so always return 3 5498 return 3; 5499 } 5500 } 5501 return false; 5502 } 5503 5504 bool ObjectFileMachO::GetArchitecture(ArchSpec &arch) { 5505 ModuleSP module_sp(GetModule()); 5506 if (module_sp) { 5507 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5508 return GetArchitecture(m_header, m_data, 5509 MachHeaderSizeFromMagic(m_header.magic), arch); 5510 } 5511 return false; 5512 } 5513 5514 UUID ObjectFileMachO::GetProcessSharedCacheUUID(Process *process) { 5515 UUID uuid; 5516 if (process && process->GetDynamicLoader()) { 5517 DynamicLoader *dl = process->GetDynamicLoader(); 5518 addr_t load_address; 5519 LazyBool using_shared_cache; 5520 LazyBool private_shared_cache; 5521 dl->GetSharedCacheInformation(load_address, uuid, using_shared_cache, 5522 private_shared_cache); 5523 } 5524 return uuid; 5525 } 5526 5527 UUID ObjectFileMachO::GetLLDBSharedCacheUUID() { 5528 UUID uuid; 5529 #if defined(__APPLE__) && \ 5530 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5531 uint8_t *(*dyld_get_all_image_infos)(void); 5532 dyld_get_all_image_infos = 5533 (uint8_t * (*)())dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5534 if (dyld_get_all_image_infos) { 5535 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5536 if (dyld_all_image_infos_address) { 5537 uint32_t *version = (uint32_t *) 5538 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5539 if (*version >= 13) { 5540 uuid_t *sharedCacheUUID_address = 0; 5541 int wordsize = sizeof(uint8_t *); 5542 if (wordsize == 8) { 5543 sharedCacheUUID_address = 5544 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5545 160); // sharedCacheUUID <mach-o/dyld_images.h> 5546 } else { 5547 sharedCacheUUID_address = 5548 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5549 84); // sharedCacheUUID <mach-o/dyld_images.h> 5550 } 5551 uuid.SetBytes(sharedCacheUUID_address); 5552 } 5553 } 5554 } 5555 #endif 5556 return uuid; 5557 } 5558 5559 uint32_t ObjectFileMachO::GetMinimumOSVersion(uint32_t *versions, 5560 uint32_t num_versions) { 5561 if (m_min_os_versions.empty()) { 5562 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5563 bool success = false; 5564 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5565 const lldb::offset_t load_cmd_offset = offset; 5566 5567 version_min_command lc; 5568 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5569 break; 5570 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5571 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5572 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5573 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5574 if (m_data.GetU32(&offset, &lc.version, 5575 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5576 const uint32_t xxxx = lc.version >> 16; 5577 const uint32_t yy = (lc.version >> 8) & 0xffu; 5578 const uint32_t zz = lc.version & 0xffu; 5579 if (xxxx) { 5580 m_min_os_versions.push_back(xxxx); 5581 m_min_os_versions.push_back(yy); 5582 m_min_os_versions.push_back(zz); 5583 } 5584 success = true; 5585 } 5586 } 5587 offset = load_cmd_offset + lc.cmdsize; 5588 } 5589 5590 if (success == false) { 5591 // Push an invalid value so we don't keep trying to 5592 m_min_os_versions.push_back(UINT32_MAX); 5593 } 5594 } 5595 5596 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) { 5597 if (versions != NULL && num_versions > 0) { 5598 for (size_t i = 0; i < num_versions; ++i) { 5599 if (i < m_min_os_versions.size()) 5600 versions[i] = m_min_os_versions[i]; 5601 else 5602 versions[i] = 0; 5603 } 5604 } 5605 return m_min_os_versions.size(); 5606 } 5607 // Call the superclasses version that will empty out the data 5608 return ObjectFile::GetMinimumOSVersion(versions, num_versions); 5609 } 5610 5611 uint32_t ObjectFileMachO::GetSDKVersion(uint32_t *versions, 5612 uint32_t num_versions) { 5613 if (m_sdk_versions.empty()) { 5614 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5615 bool success = false; 5616 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5617 const lldb::offset_t load_cmd_offset = offset; 5618 5619 version_min_command lc; 5620 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5621 break; 5622 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5623 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5624 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5625 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5626 if (m_data.GetU32(&offset, &lc.version, 5627 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5628 const uint32_t xxxx = lc.sdk >> 16; 5629 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5630 const uint32_t zz = lc.sdk & 0xffu; 5631 if (xxxx) { 5632 m_sdk_versions.push_back(xxxx); 5633 m_sdk_versions.push_back(yy); 5634 m_sdk_versions.push_back(zz); 5635 } 5636 success = true; 5637 } 5638 } 5639 offset = load_cmd_offset + lc.cmdsize; 5640 } 5641 5642 if (success == false) { 5643 // Push an invalid value so we don't keep trying to 5644 m_sdk_versions.push_back(UINT32_MAX); 5645 } 5646 } 5647 5648 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) { 5649 if (versions != NULL && num_versions > 0) { 5650 for (size_t i = 0; i < num_versions; ++i) { 5651 if (i < m_sdk_versions.size()) 5652 versions[i] = m_sdk_versions[i]; 5653 else 5654 versions[i] = 0; 5655 } 5656 } 5657 return m_sdk_versions.size(); 5658 } 5659 // Call the superclasses version that will empty out the data 5660 return ObjectFile::GetSDKVersion(versions, num_versions); 5661 } 5662 5663 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5664 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5665 } 5666 5667 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5668 return m_allow_assembly_emulation_unwind_plans; 5669 } 5670 5671 //------------------------------------------------------------------ 5672 // PluginInterface protocol 5673 //------------------------------------------------------------------ 5674 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5675 return GetPluginNameStatic(); 5676 } 5677 5678 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5679 5680 Section *ObjectFileMachO::GetMachHeaderSection() { 5681 // Find the first address of the mach header which is the first non-zero 5682 // file sized section whose file offset is zero. This is the base file address 5683 // of the mach-o file which can be subtracted from the vmaddr of the other 5684 // segments found in memory and added to the load address 5685 ModuleSP module_sp = GetModule(); 5686 if (module_sp) { 5687 SectionList *section_list = GetSectionList(); 5688 if (section_list) { 5689 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5690 const size_t num_sections = section_list->GetSize(); 5691 5692 for (size_t sect_idx = 0; sect_idx < num_sections && 5693 mach_base_file_addr == LLDB_INVALID_ADDRESS; 5694 ++sect_idx) { 5695 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5696 if (section && section->GetFileSize() > 0 && 5697 section->GetFileOffset() == 0 && 5698 section->IsThreadSpecific() == false && 5699 module_sp.get() == section->GetModule().get()) { 5700 return section; 5701 } 5702 } 5703 } 5704 } 5705 return nullptr; 5706 } 5707 5708 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 5709 lldb::addr_t mach_header_load_address, const Section *mach_header_section, 5710 const Section *section) { 5711 ModuleSP module_sp = GetModule(); 5712 if (module_sp && mach_header_section && section && 5713 mach_header_load_address != LLDB_INVALID_ADDRESS) { 5714 lldb::addr_t mach_header_file_addr = mach_header_section->GetFileAddress(); 5715 if (mach_header_file_addr != LLDB_INVALID_ADDRESS) { 5716 if (section && section->GetFileSize() > 0 && 5717 section->IsThreadSpecific() == false && 5718 module_sp.get() == section->GetModule().get()) { 5719 // Ignore __LINKEDIT and __DWARF segments 5720 if (section->GetName() == GetSegmentNameLINKEDIT()) { 5721 // Only map __LINKEDIT if we have an in memory image and this isn't 5722 // a kernel binary like a kext or mach_kernel. 5723 const bool is_memory_image = (bool)m_process_wp.lock(); 5724 const Strata strata = GetStrata(); 5725 if (is_memory_image == false || strata == eStrataKernel) 5726 return LLDB_INVALID_ADDRESS; 5727 } 5728 return section->GetFileAddress() - mach_header_file_addr + 5729 mach_header_load_address; 5730 } 5731 } 5732 } 5733 return LLDB_INVALID_ADDRESS; 5734 } 5735 5736 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 5737 bool value_is_offset) { 5738 ModuleSP module_sp = GetModule(); 5739 if (module_sp) { 5740 size_t num_loaded_sections = 0; 5741 SectionList *section_list = GetSectionList(); 5742 if (section_list) { 5743 const size_t num_sections = section_list->GetSize(); 5744 5745 if (value_is_offset) { 5746 // "value" is an offset to apply to each top level segment 5747 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5748 // Iterate through the object file sections to find all 5749 // of the sections that size on disk (to avoid __PAGEZERO) 5750 // and load them 5751 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5752 if (section_sp && section_sp->GetFileSize() > 0 && 5753 section_sp->IsThreadSpecific() == false && 5754 module_sp.get() == section_sp->GetModule().get()) { 5755 // Ignore __LINKEDIT and __DWARF segments 5756 if (section_sp->GetName() == GetSegmentNameLINKEDIT()) { 5757 // Only map __LINKEDIT if we have an in memory image and this 5758 // isn't 5759 // a kernel binary like a kext or mach_kernel. 5760 const bool is_memory_image = (bool)m_process_wp.lock(); 5761 const Strata strata = GetStrata(); 5762 if (is_memory_image == false || strata == eStrataKernel) 5763 continue; 5764 } 5765 if (target.GetSectionLoadList().SetSectionLoadAddress( 5766 section_sp, section_sp->GetFileAddress() + value)) 5767 ++num_loaded_sections; 5768 } 5769 } 5770 } else { 5771 // "value" is the new base address of the mach_header, adjust each 5772 // section accordingly 5773 5774 Section *mach_header_section = GetMachHeaderSection(); 5775 if (mach_header_section) { 5776 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5777 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5778 5779 lldb::addr_t section_load_addr = 5780 CalculateSectionLoadAddressForMemoryImage( 5781 value, mach_header_section, section_sp.get()); 5782 if (section_load_addr != LLDB_INVALID_ADDRESS) { 5783 if (target.GetSectionLoadList().SetSectionLoadAddress( 5784 section_sp, section_load_addr)) 5785 ++num_loaded_sections; 5786 } 5787 } 5788 } 5789 } 5790 } 5791 return num_loaded_sections > 0; 5792 } 5793 return false; 5794 } 5795 5796 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 5797 const FileSpec &outfile, Error &error) { 5798 if (process_sp) { 5799 Target &target = process_sp->GetTarget(); 5800 const ArchSpec target_arch = target.GetArchitecture(); 5801 const llvm::Triple &target_triple = target_arch.GetTriple(); 5802 if (target_triple.getVendor() == llvm::Triple::Apple && 5803 (target_triple.getOS() == llvm::Triple::MacOSX || 5804 target_triple.getOS() == llvm::Triple::IOS || 5805 target_triple.getOS() == llvm::Triple::WatchOS || 5806 target_triple.getOS() == llvm::Triple::TvOS)) { 5807 bool make_core = false; 5808 switch (target_arch.GetMachine()) { 5809 case llvm::Triple::aarch64: 5810 case llvm::Triple::arm: 5811 case llvm::Triple::thumb: 5812 case llvm::Triple::x86: 5813 case llvm::Triple::x86_64: 5814 make_core = true; 5815 break; 5816 default: 5817 error.SetErrorStringWithFormat("unsupported core architecture: %s", 5818 target_triple.str().c_str()); 5819 break; 5820 } 5821 5822 if (make_core) { 5823 std::vector<segment_command_64> segment_load_commands; 5824 // uint32_t range_info_idx = 0; 5825 MemoryRegionInfo range_info; 5826 Error range_error = process_sp->GetMemoryRegionInfo(0, range_info); 5827 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 5828 const ByteOrder byte_order = target_arch.GetByteOrder(); 5829 if (range_error.Success()) { 5830 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 5831 const addr_t addr = range_info.GetRange().GetRangeBase(); 5832 const addr_t size = range_info.GetRange().GetByteSize(); 5833 5834 if (size == 0) 5835 break; 5836 5837 // Calculate correct protections 5838 uint32_t prot = 0; 5839 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 5840 prot |= VM_PROT_READ; 5841 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 5842 prot |= VM_PROT_WRITE; 5843 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 5844 prot |= VM_PROT_EXECUTE; 5845 5846 // printf ("[%3u] [0x%16.16" PRIx64 " - 5847 // 0x%16.16" PRIx64 ") %c%c%c\n", 5848 // range_info_idx, 5849 // addr, 5850 // size, 5851 // (prot & VM_PROT_READ ) ? 'r' : 5852 // '-', 5853 // (prot & VM_PROT_WRITE ) ? 'w' : 5854 // '-', 5855 // (prot & VM_PROT_EXECUTE) ? 'x' : 5856 // '-'); 5857 5858 if (prot != 0) { 5859 uint32_t cmd_type = LC_SEGMENT_64; 5860 uint32_t segment_size = sizeof(segment_command_64); 5861 if (addr_byte_size == 4) { 5862 cmd_type = LC_SEGMENT; 5863 segment_size = sizeof(segment_command); 5864 } 5865 segment_command_64 segment = { 5866 cmd_type, // uint32_t cmd; 5867 segment_size, // uint32_t cmdsize; 5868 {0}, // char segname[16]; 5869 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 5870 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 5871 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 5872 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 5873 prot, // uint32_t maxprot; 5874 prot, // uint32_t initprot; 5875 0, // uint32_t nsects; 5876 0}; // uint32_t flags; 5877 segment_load_commands.push_back(segment); 5878 } else { 5879 // No protections and a size of 1 used to be returned from old 5880 // debugservers when we asked about a region that was past the 5881 // last memory region and it indicates the end... 5882 if (size == 1) 5883 break; 5884 } 5885 5886 range_error = process_sp->GetMemoryRegionInfo( 5887 range_info.GetRange().GetRangeEnd(), range_info); 5888 if (range_error.Fail()) 5889 break; 5890 } 5891 5892 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 5893 5894 mach_header_64 mach_header; 5895 if (addr_byte_size == 8) { 5896 mach_header.magic = MH_MAGIC_64; 5897 } else { 5898 mach_header.magic = MH_MAGIC; 5899 } 5900 mach_header.cputype = target_arch.GetMachOCPUType(); 5901 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 5902 mach_header.filetype = MH_CORE; 5903 mach_header.ncmds = segment_load_commands.size(); 5904 mach_header.flags = 0; 5905 mach_header.reserved = 0; 5906 ThreadList &thread_list = process_sp->GetThreadList(); 5907 const uint32_t num_threads = thread_list.GetSize(); 5908 5909 // Make an array of LC_THREAD data items. Each one contains 5910 // the contents of the LC_THREAD load command. The data doesn't 5911 // contain the load command + load command size, we will 5912 // add the load command and load command size as we emit the data. 5913 std::vector<StreamString> LC_THREAD_datas(num_threads); 5914 for (auto &LC_THREAD_data : LC_THREAD_datas) { 5915 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 5916 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 5917 LC_THREAD_data.SetByteOrder(byte_order); 5918 } 5919 for (uint32_t thread_idx = 0; thread_idx < num_threads; 5920 ++thread_idx) { 5921 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 5922 if (thread_sp) { 5923 switch (mach_header.cputype) { 5924 case llvm::MachO::CPU_TYPE_ARM64: 5925 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 5926 thread_sp.get(), LC_THREAD_datas[thread_idx]); 5927 break; 5928 5929 case llvm::MachO::CPU_TYPE_ARM: 5930 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 5931 thread_sp.get(), LC_THREAD_datas[thread_idx]); 5932 break; 5933 5934 case llvm::MachO::CPU_TYPE_I386: 5935 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 5936 thread_sp.get(), LC_THREAD_datas[thread_idx]); 5937 break; 5938 5939 case llvm::MachO::CPU_TYPE_X86_64: 5940 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 5941 thread_sp.get(), LC_THREAD_datas[thread_idx]); 5942 break; 5943 } 5944 } 5945 } 5946 5947 // The size of the load command is the size of the segments... 5948 if (addr_byte_size == 8) { 5949 mach_header.sizeofcmds = segment_load_commands.size() * 5950 sizeof(struct segment_command_64); 5951 } else { 5952 mach_header.sizeofcmds = 5953 segment_load_commands.size() * sizeof(struct segment_command); 5954 } 5955 5956 // and the size of all LC_THREAD load command 5957 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 5958 ++mach_header.ncmds; 5959 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 5960 } 5961 5962 printf("mach_header: 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x " 5963 "0x%8.8x 0x%8.8x\n", 5964 mach_header.magic, mach_header.cputype, mach_header.cpusubtype, 5965 mach_header.filetype, mach_header.ncmds, 5966 mach_header.sizeofcmds, mach_header.flags, 5967 mach_header.reserved); 5968 5969 // Write the mach header 5970 buffer.PutHex32(mach_header.magic); 5971 buffer.PutHex32(mach_header.cputype); 5972 buffer.PutHex32(mach_header.cpusubtype); 5973 buffer.PutHex32(mach_header.filetype); 5974 buffer.PutHex32(mach_header.ncmds); 5975 buffer.PutHex32(mach_header.sizeofcmds); 5976 buffer.PutHex32(mach_header.flags); 5977 if (addr_byte_size == 8) { 5978 buffer.PutHex32(mach_header.reserved); 5979 } 5980 5981 // Skip the mach header and all load commands and align to the next 5982 // 0x1000 byte boundary 5983 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 5984 if (file_offset & 0x00000fff) { 5985 file_offset += 0x00001000ull; 5986 file_offset &= (~0x00001000ull + 1); 5987 } 5988 5989 for (auto &segment : segment_load_commands) { 5990 segment.fileoff = file_offset; 5991 file_offset += segment.filesize; 5992 } 5993 5994 // Write out all of the LC_THREAD load commands 5995 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 5996 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 5997 buffer.PutHex32(LC_THREAD); 5998 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 5999 buffer.Write(LC_THREAD_data.GetString().data(), 6000 LC_THREAD_data_size); 6001 } 6002 6003 // Write out all of the segment load commands 6004 for (const auto &segment : segment_load_commands) { 6005 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6006 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6007 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6008 segment.cmd, segment.cmdsize, segment.vmaddr, 6009 segment.vmaddr + segment.vmsize, segment.fileoff, 6010 segment.filesize, segment.maxprot, segment.initprot, 6011 segment.nsects, segment.flags); 6012 6013 buffer.PutHex32(segment.cmd); 6014 buffer.PutHex32(segment.cmdsize); 6015 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6016 if (addr_byte_size == 8) { 6017 buffer.PutHex64(segment.vmaddr); 6018 buffer.PutHex64(segment.vmsize); 6019 buffer.PutHex64(segment.fileoff); 6020 buffer.PutHex64(segment.filesize); 6021 } else { 6022 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6023 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6024 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6025 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6026 } 6027 buffer.PutHex32(segment.maxprot); 6028 buffer.PutHex32(segment.initprot); 6029 buffer.PutHex32(segment.nsects); 6030 buffer.PutHex32(segment.flags); 6031 } 6032 6033 File core_file; 6034 std::string core_file_path(outfile.GetPath()); 6035 error = core_file.Open(core_file_path.c_str(), 6036 File::eOpenOptionWrite | 6037 File::eOpenOptionTruncate | 6038 File::eOpenOptionCanCreate); 6039 if (error.Success()) { 6040 // Read 1 page at a time 6041 uint8_t bytes[0x1000]; 6042 // Write the mach header and load commands out to the core file 6043 size_t bytes_written = buffer.GetString().size(); 6044 error = core_file.Write(buffer.GetString().data(), bytes_written); 6045 if (error.Success()) { 6046 // Now write the file data for all memory segments in the process 6047 for (const auto &segment : segment_load_commands) { 6048 if (core_file.SeekFromStart(segment.fileoff) == -1) { 6049 error.SetErrorStringWithFormat( 6050 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6051 segment.fileoff, core_file_path.c_str()); 6052 break; 6053 } 6054 6055 printf("Saving %" PRId64 6056 " bytes of data for memory region at 0x%" PRIx64 "\n", 6057 segment.vmsize, segment.vmaddr); 6058 addr_t bytes_left = segment.vmsize; 6059 addr_t addr = segment.vmaddr; 6060 Error memory_read_error; 6061 while (bytes_left > 0 && error.Success()) { 6062 const size_t bytes_to_read = 6063 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6064 const size_t bytes_read = process_sp->ReadMemory( 6065 addr, bytes, bytes_to_read, memory_read_error); 6066 if (bytes_read == bytes_to_read) { 6067 size_t bytes_written = bytes_read; 6068 error = core_file.Write(bytes, bytes_written); 6069 bytes_left -= bytes_read; 6070 addr += bytes_read; 6071 } else { 6072 // Some pages within regions are not readable, those 6073 // should be zero filled 6074 memset(bytes, 0, bytes_to_read); 6075 size_t bytes_written = bytes_to_read; 6076 error = core_file.Write(bytes, bytes_written); 6077 bytes_left -= bytes_to_read; 6078 addr += bytes_to_read; 6079 } 6080 } 6081 } 6082 } 6083 } 6084 } else { 6085 error.SetErrorString( 6086 "process doesn't support getting memory region info"); 6087 } 6088 } 6089 return true; // This is the right plug to handle saving core files for 6090 // this process 6091 } 6092 } 6093 return false; 6094 } 6095