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/Debugger.h" 22 #include "lldb/Core/FileSpecList.h" 23 #include "lldb/Core/Module.h" 24 #include "lldb/Core/ModuleSpec.h" 25 #include "lldb/Core/PluginManager.h" 26 #include "lldb/Core/RangeMap.h" 27 #include "lldb/Core/RegisterValue.h" 28 #include "lldb/Core/Section.h" 29 #include "lldb/Core/StreamFile.h" 30 #include "lldb/Host/Host.h" 31 #include "lldb/Symbol/DWARFCallFrameInfo.h" 32 #include "lldb/Symbol/ObjectFile.h" 33 #include "lldb/Target/DynamicLoader.h" 34 #include "lldb/Target/MemoryRegionInfo.h" 35 #include "lldb/Target/Platform.h" 36 #include "lldb/Target/Process.h" 37 #include "lldb/Target/SectionLoadList.h" 38 #include "lldb/Target/Target.h" 39 #include "lldb/Target/Thread.h" 40 #include "lldb/Target/ThreadList.h" 41 #include "lldb/Utility/DataBufferLLVM.h" 42 #include "lldb/Utility/FileSpec.h" 43 #include "lldb/Utility/Log.h" 44 #include "lldb/Utility/Status.h" 45 #include "lldb/Utility/StreamString.h" 46 #include "lldb/Utility/Timer.h" 47 #include "lldb/Utility/UUID.h" 48 49 #include "lldb/Utility/SafeMachO.h" 50 51 #include "llvm/Support/MemoryBuffer.h" 52 53 #include "ObjectFileMachO.h" 54 55 #if defined(__APPLE__) && \ 56 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 57 // GetLLDBSharedCacheUUID() needs to call dlsym() 58 #include <dlfcn.h> 59 #endif 60 61 #ifndef __APPLE__ 62 #include "Utility/UuidCompatibility.h" 63 #else 64 #include <uuid/uuid.h> 65 #endif 66 67 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 68 using namespace lldb; 69 using namespace lldb_private; 70 using namespace llvm::MachO; 71 72 // Some structure definitions needed for parsing the dyld shared cache files 73 // found on iOS devices. 74 75 struct lldb_copy_dyld_cache_header_v1 { 76 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 77 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 78 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 79 uint32_t imagesOffset; 80 uint32_t imagesCount; 81 uint64_t dyldBaseAddress; 82 uint64_t codeSignatureOffset; 83 uint64_t codeSignatureSize; 84 uint64_t slideInfoOffset; 85 uint64_t slideInfoSize; 86 uint64_t localSymbolsOffset; 87 uint64_t localSymbolsSize; 88 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 89 // and later 90 }; 91 92 struct lldb_copy_dyld_cache_mapping_info { 93 uint64_t address; 94 uint64_t size; 95 uint64_t fileOffset; 96 uint32_t maxProt; 97 uint32_t initProt; 98 }; 99 100 struct lldb_copy_dyld_cache_local_symbols_info { 101 uint32_t nlistOffset; 102 uint32_t nlistCount; 103 uint32_t stringsOffset; 104 uint32_t stringsSize; 105 uint32_t entriesOffset; 106 uint32_t entriesCount; 107 }; 108 struct lldb_copy_dyld_cache_local_symbols_entry { 109 uint32_t dylibOffset; 110 uint32_t nlistStartIndex; 111 uint32_t nlistCount; 112 }; 113 114 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 115 public: 116 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 117 const DataExtractor &data) 118 : RegisterContextDarwin_x86_64(thread, 0) { 119 SetRegisterDataFrom_LC_THREAD(data); 120 } 121 122 void InvalidateAllRegisters() override { 123 // Do nothing... registers are always valid... 124 } 125 126 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 127 lldb::offset_t offset = 0; 128 SetError(GPRRegSet, Read, -1); 129 SetError(FPURegSet, Read, -1); 130 SetError(EXCRegSet, Read, -1); 131 bool done = false; 132 133 while (!done) { 134 int flavor = data.GetU32(&offset); 135 if (flavor == 0) 136 done = true; 137 else { 138 uint32_t i; 139 uint32_t count = data.GetU32(&offset); 140 switch (flavor) { 141 case GPRRegSet: 142 for (i = 0; i < count; ++i) 143 (&gpr.rax)[i] = data.GetU64(&offset); 144 SetError(GPRRegSet, Read, 0); 145 done = true; 146 147 break; 148 case FPURegSet: 149 // TODO: fill in FPU regs.... 150 // SetError (FPURegSet, Read, -1); 151 done = true; 152 153 break; 154 case EXCRegSet: 155 exc.trapno = data.GetU32(&offset); 156 exc.err = data.GetU32(&offset); 157 exc.faultvaddr = data.GetU64(&offset); 158 SetError(EXCRegSet, Read, 0); 159 done = true; 160 break; 161 case 7: 162 case 8: 163 case 9: 164 // fancy flavors that encapsulate of the above 165 // flavors... 166 break; 167 168 default: 169 done = true; 170 break; 171 } 172 } 173 } 174 } 175 176 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 177 const char *alt_name, size_t reg_byte_size, 178 Stream &data) { 179 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 180 if (reg_info == NULL) 181 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 182 if (reg_info) { 183 lldb_private::RegisterValue reg_value; 184 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 185 if (reg_info->byte_size >= reg_byte_size) 186 data.Write(reg_value.GetBytes(), reg_byte_size); 187 else { 188 data.Write(reg_value.GetBytes(), reg_info->byte_size); 189 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 190 ++i) 191 data.PutChar(0); 192 } 193 return reg_byte_size; 194 } 195 } 196 // Just write zeros if all else fails 197 for (size_t i = 0; i < reg_byte_size; ++i) 198 data.PutChar(0); 199 return reg_byte_size; 200 } 201 202 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 203 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 204 if (reg_ctx_sp) { 205 RegisterContext *reg_ctx = reg_ctx_sp.get(); 206 207 data.PutHex32(GPRRegSet); // Flavor 208 data.PutHex32(GPRWordCount); 209 WriteRegister(reg_ctx, "rax", NULL, 8, data); 210 WriteRegister(reg_ctx, "rbx", NULL, 8, data); 211 WriteRegister(reg_ctx, "rcx", NULL, 8, data); 212 WriteRegister(reg_ctx, "rdx", NULL, 8, data); 213 WriteRegister(reg_ctx, "rdi", NULL, 8, data); 214 WriteRegister(reg_ctx, "rsi", NULL, 8, data); 215 WriteRegister(reg_ctx, "rbp", NULL, 8, data); 216 WriteRegister(reg_ctx, "rsp", NULL, 8, data); 217 WriteRegister(reg_ctx, "r8", NULL, 8, data); 218 WriteRegister(reg_ctx, "r9", NULL, 8, data); 219 WriteRegister(reg_ctx, "r10", NULL, 8, data); 220 WriteRegister(reg_ctx, "r11", NULL, 8, data); 221 WriteRegister(reg_ctx, "r12", NULL, 8, data); 222 WriteRegister(reg_ctx, "r13", NULL, 8, data); 223 WriteRegister(reg_ctx, "r14", NULL, 8, data); 224 WriteRegister(reg_ctx, "r15", NULL, 8, data); 225 WriteRegister(reg_ctx, "rip", NULL, 8, data); 226 WriteRegister(reg_ctx, "rflags", NULL, 8, data); 227 WriteRegister(reg_ctx, "cs", NULL, 8, data); 228 WriteRegister(reg_ctx, "fs", NULL, 8, data); 229 WriteRegister(reg_ctx, "gs", NULL, 8, data); 230 231 // // Write out the FPU registers 232 // const size_t fpu_byte_size = sizeof(FPU); 233 // size_t bytes_written = 0; 234 // data.PutHex32 (FPURegSet); 235 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 236 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 237 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 238 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 239 // data); // uint16_t fcw; // "fctrl" 240 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 241 // data); // uint16_t fsw; // "fstat" 242 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 243 // data); // uint8_t ftw; // "ftag" 244 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 245 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 246 // data); // uint16_t fop; // "fop" 247 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 248 // data); // uint32_t ip; // "fioff" 249 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 250 // data); // uint16_t cs; // "fiseg" 251 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 252 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 253 // data); // uint32_t dp; // "fooff" 254 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 255 // data); // uint16_t ds; // "foseg" 256 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 257 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 258 // data); // uint32_t mxcsr; 259 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 260 // 4, data);// uint32_t mxcsrmask; 261 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 262 // sizeof(MMSReg), data); 263 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 264 // sizeof(MMSReg), data); 265 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 266 // sizeof(MMSReg), data); 267 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 268 // sizeof(MMSReg), data); 269 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 270 // sizeof(MMSReg), data); 271 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 272 // sizeof(MMSReg), data); 273 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 274 // sizeof(MMSReg), data); 275 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 276 // sizeof(MMSReg), data); 277 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 278 // sizeof(XMMReg), data); 279 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 280 // sizeof(XMMReg), data); 281 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 282 // sizeof(XMMReg), data); 283 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 284 // sizeof(XMMReg), data); 285 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 286 // sizeof(XMMReg), data); 287 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 288 // sizeof(XMMReg), data); 289 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 290 // sizeof(XMMReg), data); 291 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 292 // sizeof(XMMReg), data); 293 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 294 // sizeof(XMMReg), data); 295 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 296 // sizeof(XMMReg), data); 297 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 298 // sizeof(XMMReg), data); 299 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 300 // sizeof(XMMReg), data); 301 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 302 // sizeof(XMMReg), data); 303 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 304 // sizeof(XMMReg), data); 305 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 306 // sizeof(XMMReg), data); 307 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 308 // sizeof(XMMReg), data); 309 // 310 // // Fill rest with zeros 311 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 312 // i) 313 // data.PutChar(0); 314 315 // Write out the EXC registers 316 data.PutHex32(EXCRegSet); 317 data.PutHex32(EXCWordCount); 318 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 319 WriteRegister(reg_ctx, "err", NULL, 4, data); 320 WriteRegister(reg_ctx, "faultvaddr", NULL, 8, data); 321 return true; 322 } 323 return false; 324 } 325 326 protected: 327 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 328 329 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 330 331 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 332 333 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 334 return 0; 335 } 336 337 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 338 return 0; 339 } 340 341 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 342 return 0; 343 } 344 }; 345 346 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 347 public: 348 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 349 const DataExtractor &data) 350 : RegisterContextDarwin_i386(thread, 0) { 351 SetRegisterDataFrom_LC_THREAD(data); 352 } 353 354 void InvalidateAllRegisters() override { 355 // Do nothing... registers are always valid... 356 } 357 358 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 359 lldb::offset_t offset = 0; 360 SetError(GPRRegSet, Read, -1); 361 SetError(FPURegSet, Read, -1); 362 SetError(EXCRegSet, Read, -1); 363 bool done = false; 364 365 while (!done) { 366 int flavor = data.GetU32(&offset); 367 if (flavor == 0) 368 done = true; 369 else { 370 uint32_t i; 371 uint32_t count = data.GetU32(&offset); 372 switch (flavor) { 373 case GPRRegSet: 374 for (i = 0; i < count; ++i) 375 (&gpr.eax)[i] = data.GetU32(&offset); 376 SetError(GPRRegSet, Read, 0); 377 done = true; 378 379 break; 380 case FPURegSet: 381 // TODO: fill in FPU regs.... 382 // SetError (FPURegSet, Read, -1); 383 done = true; 384 385 break; 386 case EXCRegSet: 387 exc.trapno = data.GetU32(&offset); 388 exc.err = data.GetU32(&offset); 389 exc.faultvaddr = data.GetU32(&offset); 390 SetError(EXCRegSet, Read, 0); 391 done = true; 392 break; 393 case 7: 394 case 8: 395 case 9: 396 // fancy flavors that encapsulate of the above 397 // flavors... 398 break; 399 400 default: 401 done = true; 402 break; 403 } 404 } 405 } 406 } 407 408 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 409 const char *alt_name, size_t reg_byte_size, 410 Stream &data) { 411 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 412 if (reg_info == NULL) 413 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 414 if (reg_info) { 415 lldb_private::RegisterValue reg_value; 416 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 417 if (reg_info->byte_size >= reg_byte_size) 418 data.Write(reg_value.GetBytes(), reg_byte_size); 419 else { 420 data.Write(reg_value.GetBytes(), reg_info->byte_size); 421 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 422 ++i) 423 data.PutChar(0); 424 } 425 return reg_byte_size; 426 } 427 } 428 // Just write zeros if all else fails 429 for (size_t i = 0; i < reg_byte_size; ++i) 430 data.PutChar(0); 431 return reg_byte_size; 432 } 433 434 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 435 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 436 if (reg_ctx_sp) { 437 RegisterContext *reg_ctx = reg_ctx_sp.get(); 438 439 data.PutHex32(GPRRegSet); // Flavor 440 data.PutHex32(GPRWordCount); 441 WriteRegister(reg_ctx, "eax", NULL, 4, data); 442 WriteRegister(reg_ctx, "ebx", NULL, 4, data); 443 WriteRegister(reg_ctx, "ecx", NULL, 4, data); 444 WriteRegister(reg_ctx, "edx", NULL, 4, data); 445 WriteRegister(reg_ctx, "edi", NULL, 4, data); 446 WriteRegister(reg_ctx, "esi", NULL, 4, data); 447 WriteRegister(reg_ctx, "ebp", NULL, 4, data); 448 WriteRegister(reg_ctx, "esp", NULL, 4, data); 449 WriteRegister(reg_ctx, "ss", NULL, 4, data); 450 WriteRegister(reg_ctx, "eflags", NULL, 4, data); 451 WriteRegister(reg_ctx, "eip", NULL, 4, data); 452 WriteRegister(reg_ctx, "cs", NULL, 4, data); 453 WriteRegister(reg_ctx, "ds", NULL, 4, data); 454 WriteRegister(reg_ctx, "es", NULL, 4, data); 455 WriteRegister(reg_ctx, "fs", NULL, 4, data); 456 WriteRegister(reg_ctx, "gs", NULL, 4, data); 457 458 // Write out the EXC registers 459 data.PutHex32(EXCRegSet); 460 data.PutHex32(EXCWordCount); 461 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 462 WriteRegister(reg_ctx, "err", NULL, 4, data); 463 WriteRegister(reg_ctx, "faultvaddr", NULL, 4, data); 464 return true; 465 } 466 return false; 467 } 468 469 protected: 470 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 471 472 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 473 474 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 475 476 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 477 return 0; 478 } 479 480 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 481 return 0; 482 } 483 484 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 485 return 0; 486 } 487 }; 488 489 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 490 public: 491 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 492 const DataExtractor &data) 493 : RegisterContextDarwin_arm(thread, 0) { 494 SetRegisterDataFrom_LC_THREAD(data); 495 } 496 497 void InvalidateAllRegisters() override { 498 // Do nothing... registers are always valid... 499 } 500 501 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 502 lldb::offset_t offset = 0; 503 SetError(GPRRegSet, Read, -1); 504 SetError(FPURegSet, Read, -1); 505 SetError(EXCRegSet, Read, -1); 506 bool done = false; 507 508 while (!done) { 509 int flavor = data.GetU32(&offset); 510 uint32_t count = data.GetU32(&offset); 511 lldb::offset_t next_thread_state = offset + (count * 4); 512 switch (flavor) { 513 case GPRAltRegSet: 514 case GPRRegSet: 515 for (uint32_t i = 0; i < count; ++i) { 516 gpr.r[i] = data.GetU32(&offset); 517 } 518 519 // Note that gpr.cpsr is also copied by the above loop; this loop 520 // technically extends 521 // one element past the end of the gpr.r[] array. 522 523 SetError(GPRRegSet, Read, 0); 524 offset = next_thread_state; 525 break; 526 527 case FPURegSet: { 528 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 529 const int fpu_reg_buf_size = sizeof(fpu.floats); 530 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 531 fpu_reg_buf) == fpu_reg_buf_size) { 532 offset += fpu_reg_buf_size; 533 fpu.fpscr = data.GetU32(&offset); 534 SetError(FPURegSet, Read, 0); 535 } else { 536 done = true; 537 } 538 } 539 offset = next_thread_state; 540 break; 541 542 case EXCRegSet: 543 if (count == 3) { 544 exc.exception = data.GetU32(&offset); 545 exc.fsr = data.GetU32(&offset); 546 exc.far = data.GetU32(&offset); 547 SetError(EXCRegSet, Read, 0); 548 } 549 done = true; 550 offset = next_thread_state; 551 break; 552 553 // Unknown register set flavor, stop trying to parse. 554 default: 555 done = true; 556 } 557 } 558 } 559 560 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 561 const char *alt_name, size_t reg_byte_size, 562 Stream &data) { 563 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 564 if (reg_info == NULL) 565 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 566 if (reg_info) { 567 lldb_private::RegisterValue reg_value; 568 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 569 if (reg_info->byte_size >= reg_byte_size) 570 data.Write(reg_value.GetBytes(), reg_byte_size); 571 else { 572 data.Write(reg_value.GetBytes(), reg_info->byte_size); 573 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 574 ++i) 575 data.PutChar(0); 576 } 577 return reg_byte_size; 578 } 579 } 580 // Just write zeros if all else fails 581 for (size_t i = 0; i < reg_byte_size; ++i) 582 data.PutChar(0); 583 return reg_byte_size; 584 } 585 586 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 587 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 588 if (reg_ctx_sp) { 589 RegisterContext *reg_ctx = reg_ctx_sp.get(); 590 591 data.PutHex32(GPRRegSet); // Flavor 592 data.PutHex32(GPRWordCount); 593 WriteRegister(reg_ctx, "r0", NULL, 4, data); 594 WriteRegister(reg_ctx, "r1", NULL, 4, data); 595 WriteRegister(reg_ctx, "r2", NULL, 4, data); 596 WriteRegister(reg_ctx, "r3", NULL, 4, data); 597 WriteRegister(reg_ctx, "r4", NULL, 4, data); 598 WriteRegister(reg_ctx, "r5", NULL, 4, data); 599 WriteRegister(reg_ctx, "r6", NULL, 4, data); 600 WriteRegister(reg_ctx, "r7", NULL, 4, data); 601 WriteRegister(reg_ctx, "r8", NULL, 4, data); 602 WriteRegister(reg_ctx, "r9", NULL, 4, data); 603 WriteRegister(reg_ctx, "r10", NULL, 4, data); 604 WriteRegister(reg_ctx, "r11", NULL, 4, data); 605 WriteRegister(reg_ctx, "r12", NULL, 4, data); 606 WriteRegister(reg_ctx, "sp", NULL, 4, data); 607 WriteRegister(reg_ctx, "lr", NULL, 4, data); 608 WriteRegister(reg_ctx, "pc", NULL, 4, data); 609 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 610 611 // Write out the EXC registers 612 // data.PutHex32 (EXCRegSet); 613 // data.PutHex32 (EXCWordCount); 614 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 615 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 616 // WriteRegister (reg_ctx, "far", NULL, 4, data); 617 return true; 618 } 619 return false; 620 } 621 622 protected: 623 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 624 625 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 626 627 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 628 629 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 630 631 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 632 return 0; 633 } 634 635 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 636 return 0; 637 } 638 639 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 640 return 0; 641 } 642 643 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 644 return -1; 645 } 646 }; 647 648 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 649 public: 650 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 651 const DataExtractor &data) 652 : RegisterContextDarwin_arm64(thread, 0) { 653 SetRegisterDataFrom_LC_THREAD(data); 654 } 655 656 void InvalidateAllRegisters() override { 657 // Do nothing... registers are always valid... 658 } 659 660 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 661 lldb::offset_t offset = 0; 662 SetError(GPRRegSet, Read, -1); 663 SetError(FPURegSet, Read, -1); 664 SetError(EXCRegSet, Read, -1); 665 bool done = false; 666 while (!done) { 667 int flavor = data.GetU32(&offset); 668 uint32_t count = data.GetU32(&offset); 669 lldb::offset_t next_thread_state = offset + (count * 4); 670 switch (flavor) { 671 case GPRRegSet: 672 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 673 // 32-bit register) 674 if (count >= (33 * 2) + 1) { 675 for (uint32_t i = 0; i < 29; ++i) 676 gpr.x[i] = data.GetU64(&offset); 677 gpr.fp = data.GetU64(&offset); 678 gpr.lr = data.GetU64(&offset); 679 gpr.sp = data.GetU64(&offset); 680 gpr.pc = data.GetU64(&offset); 681 gpr.cpsr = data.GetU32(&offset); 682 SetError(GPRRegSet, Read, 0); 683 } 684 offset = next_thread_state; 685 break; 686 case FPURegSet: { 687 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 688 const int fpu_reg_buf_size = sizeof(fpu); 689 if (fpu_reg_buf_size == count && 690 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 691 fpu_reg_buf) == fpu_reg_buf_size) { 692 SetError(FPURegSet, Read, 0); 693 } else { 694 done = true; 695 } 696 } 697 offset = next_thread_state; 698 break; 699 case EXCRegSet: 700 if (count == 4) { 701 exc.far = data.GetU64(&offset); 702 exc.esr = data.GetU32(&offset); 703 exc.exception = data.GetU32(&offset); 704 SetError(EXCRegSet, Read, 0); 705 } 706 offset = next_thread_state; 707 break; 708 default: 709 done = true; 710 break; 711 } 712 } 713 } 714 715 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 716 const char *alt_name, size_t reg_byte_size, 717 Stream &data) { 718 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 719 if (reg_info == NULL) 720 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 721 if (reg_info) { 722 lldb_private::RegisterValue reg_value; 723 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 724 if (reg_info->byte_size >= reg_byte_size) 725 data.Write(reg_value.GetBytes(), reg_byte_size); 726 else { 727 data.Write(reg_value.GetBytes(), reg_info->byte_size); 728 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 729 ++i) 730 data.PutChar(0); 731 } 732 return reg_byte_size; 733 } 734 } 735 // Just write zeros if all else fails 736 for (size_t i = 0; i < reg_byte_size; ++i) 737 data.PutChar(0); 738 return reg_byte_size; 739 } 740 741 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 742 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 743 if (reg_ctx_sp) { 744 RegisterContext *reg_ctx = reg_ctx_sp.get(); 745 746 data.PutHex32(GPRRegSet); // Flavor 747 data.PutHex32(GPRWordCount); 748 WriteRegister(reg_ctx, "x0", NULL, 8, data); 749 WriteRegister(reg_ctx, "x1", NULL, 8, data); 750 WriteRegister(reg_ctx, "x2", NULL, 8, data); 751 WriteRegister(reg_ctx, "x3", NULL, 8, data); 752 WriteRegister(reg_ctx, "x4", NULL, 8, data); 753 WriteRegister(reg_ctx, "x5", NULL, 8, data); 754 WriteRegister(reg_ctx, "x6", NULL, 8, data); 755 WriteRegister(reg_ctx, "x7", NULL, 8, data); 756 WriteRegister(reg_ctx, "x8", NULL, 8, data); 757 WriteRegister(reg_ctx, "x9", NULL, 8, data); 758 WriteRegister(reg_ctx, "x10", NULL, 8, data); 759 WriteRegister(reg_ctx, "x11", NULL, 8, data); 760 WriteRegister(reg_ctx, "x12", NULL, 8, data); 761 WriteRegister(reg_ctx, "x13", NULL, 8, data); 762 WriteRegister(reg_ctx, "x14", NULL, 8, data); 763 WriteRegister(reg_ctx, "x15", NULL, 8, data); 764 WriteRegister(reg_ctx, "x16", NULL, 8, data); 765 WriteRegister(reg_ctx, "x17", NULL, 8, data); 766 WriteRegister(reg_ctx, "x18", NULL, 8, data); 767 WriteRegister(reg_ctx, "x19", NULL, 8, data); 768 WriteRegister(reg_ctx, "x20", NULL, 8, data); 769 WriteRegister(reg_ctx, "x21", NULL, 8, data); 770 WriteRegister(reg_ctx, "x22", NULL, 8, data); 771 WriteRegister(reg_ctx, "x23", NULL, 8, data); 772 WriteRegister(reg_ctx, "x24", NULL, 8, data); 773 WriteRegister(reg_ctx, "x25", NULL, 8, data); 774 WriteRegister(reg_ctx, "x26", NULL, 8, data); 775 WriteRegister(reg_ctx, "x27", NULL, 8, data); 776 WriteRegister(reg_ctx, "x28", NULL, 8, data); 777 WriteRegister(reg_ctx, "fp", NULL, 8, data); 778 WriteRegister(reg_ctx, "lr", NULL, 8, data); 779 WriteRegister(reg_ctx, "sp", NULL, 8, data); 780 WriteRegister(reg_ctx, "pc", NULL, 8, data); 781 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 782 783 // Write out the EXC registers 784 // data.PutHex32 (EXCRegSet); 785 // data.PutHex32 (EXCWordCount); 786 // WriteRegister (reg_ctx, "far", NULL, 8, data); 787 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 788 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 789 return true; 790 } 791 return false; 792 } 793 794 protected: 795 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 796 797 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 798 799 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 800 801 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 802 803 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 804 return 0; 805 } 806 807 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 808 return 0; 809 } 810 811 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 812 return 0; 813 } 814 815 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 816 return -1; 817 } 818 }; 819 820 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 821 switch (magic) { 822 case MH_MAGIC: 823 case MH_CIGAM: 824 return sizeof(struct mach_header); 825 826 case MH_MAGIC_64: 827 case MH_CIGAM_64: 828 return sizeof(struct mach_header_64); 829 break; 830 831 default: 832 break; 833 } 834 return 0; 835 } 836 837 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 838 839 void ObjectFileMachO::Initialize() { 840 PluginManager::RegisterPlugin( 841 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 842 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 843 } 844 845 void ObjectFileMachO::Terminate() { 846 PluginManager::UnregisterPlugin(CreateInstance); 847 } 848 849 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 850 static ConstString g_name("mach-o"); 851 return g_name; 852 } 853 854 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 855 return "Mach-o object file reader (32 and 64 bit)"; 856 } 857 858 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 859 DataBufferSP &data_sp, 860 lldb::offset_t data_offset, 861 const FileSpec *file, 862 lldb::offset_t file_offset, 863 lldb::offset_t length) { 864 if (!data_sp) { 865 data_sp = 866 DataBufferLLVM::CreateSliceFromPath(file->GetPath(), length, file_offset); 867 if (!data_sp) 868 return nullptr; 869 data_offset = 0; 870 } 871 872 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 873 return nullptr; 874 875 // Update the data to contain the entire file if it doesn't already 876 if (data_sp->GetByteSize() < length) { 877 data_sp = 878 DataBufferLLVM::CreateSliceFromPath(file->GetPath(), length, file_offset); 879 if (!data_sp) 880 return nullptr; 881 data_offset = 0; 882 } 883 auto objfile_ap = llvm::make_unique<ObjectFileMachO>( 884 module_sp, data_sp, data_offset, file, file_offset, length); 885 if (!objfile_ap || !objfile_ap->ParseHeader()) 886 return nullptr; 887 888 return objfile_ap.release(); 889 } 890 891 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 892 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 893 const ProcessSP &process_sp, lldb::addr_t header_addr) { 894 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 895 std::unique_ptr<ObjectFile> objfile_ap( 896 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 897 if (objfile_ap.get() && objfile_ap->ParseHeader()) 898 return objfile_ap.release(); 899 } 900 return NULL; 901 } 902 903 size_t ObjectFileMachO::GetModuleSpecifications( 904 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 905 lldb::offset_t data_offset, lldb::offset_t file_offset, 906 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 907 const size_t initial_count = specs.GetSize(); 908 909 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 910 DataExtractor data; 911 data.SetData(data_sp); 912 llvm::MachO::mach_header header; 913 if (ParseHeader(data, &data_offset, header)) { 914 size_t header_and_load_cmds = 915 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 916 if (header_and_load_cmds >= data_sp->GetByteSize()) { 917 data_sp = DataBufferLLVM::CreateSliceFromPath( 918 file.GetPath(), header_and_load_cmds, file_offset); 919 data.SetData(data_sp); 920 data_offset = MachHeaderSizeFromMagic(header.magic); 921 } 922 if (data_sp) { 923 ModuleSpec spec; 924 spec.GetFileSpec() = file; 925 spec.SetObjectOffset(file_offset); 926 spec.SetObjectSize(length); 927 928 if (GetArchitecture(header, data, data_offset, 929 spec.GetArchitecture())) { 930 if (spec.GetArchitecture().IsValid()) { 931 GetUUID(header, data, data_offset, spec.GetUUID()); 932 specs.Append(spec); 933 } 934 } 935 } 936 } 937 } 938 return specs.GetSize() - initial_count; 939 } 940 941 const ConstString &ObjectFileMachO::GetSegmentNameTEXT() { 942 static ConstString g_segment_name_TEXT("__TEXT"); 943 return g_segment_name_TEXT; 944 } 945 946 const ConstString &ObjectFileMachO::GetSegmentNameDATA() { 947 static ConstString g_segment_name_DATA("__DATA"); 948 return g_segment_name_DATA; 949 } 950 951 const ConstString &ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 952 static ConstString g_segment_name("__DATA_DIRTY"); 953 return g_segment_name; 954 } 955 956 const ConstString &ObjectFileMachO::GetSegmentNameDATA_CONST() { 957 static ConstString g_segment_name("__DATA_CONST"); 958 return g_segment_name; 959 } 960 961 const ConstString &ObjectFileMachO::GetSegmentNameOBJC() { 962 static ConstString g_segment_name_OBJC("__OBJC"); 963 return g_segment_name_OBJC; 964 } 965 966 const ConstString &ObjectFileMachO::GetSegmentNameLINKEDIT() { 967 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 968 return g_section_name_LINKEDIT; 969 } 970 971 const ConstString &ObjectFileMachO::GetSectionNameEHFrame() { 972 static ConstString g_section_name_eh_frame("__eh_frame"); 973 return g_section_name_eh_frame; 974 } 975 976 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 977 lldb::addr_t data_offset, 978 lldb::addr_t data_length) { 979 DataExtractor data; 980 data.SetData(data_sp, data_offset, data_length); 981 lldb::offset_t offset = 0; 982 uint32_t magic = data.GetU32(&offset); 983 return MachHeaderSizeFromMagic(magic) != 0; 984 } 985 986 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 987 DataBufferSP &data_sp, 988 lldb::offset_t data_offset, 989 const FileSpec *file, 990 lldb::offset_t file_offset, 991 lldb::offset_t length) 992 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 993 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 994 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 995 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 996 ::memset(&m_header, 0, sizeof(m_header)); 997 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 998 } 999 1000 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 1001 lldb::DataBufferSP &header_data_sp, 1002 const lldb::ProcessSP &process_sp, 1003 lldb::addr_t header_addr) 1004 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 1005 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 1006 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 1007 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 1008 ::memset(&m_header, 0, sizeof(m_header)); 1009 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1010 } 1011 1012 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 1013 lldb::offset_t *data_offset_ptr, 1014 llvm::MachO::mach_header &header) { 1015 data.SetByteOrder(endian::InlHostByteOrder()); 1016 // Leave magic in the original byte order 1017 header.magic = data.GetU32(data_offset_ptr); 1018 bool can_parse = false; 1019 bool is_64_bit = false; 1020 switch (header.magic) { 1021 case MH_MAGIC: 1022 data.SetByteOrder(endian::InlHostByteOrder()); 1023 data.SetAddressByteSize(4); 1024 can_parse = true; 1025 break; 1026 1027 case MH_MAGIC_64: 1028 data.SetByteOrder(endian::InlHostByteOrder()); 1029 data.SetAddressByteSize(8); 1030 can_parse = true; 1031 is_64_bit = true; 1032 break; 1033 1034 case MH_CIGAM: 1035 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1036 ? eByteOrderLittle 1037 : eByteOrderBig); 1038 data.SetAddressByteSize(4); 1039 can_parse = true; 1040 break; 1041 1042 case MH_CIGAM_64: 1043 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1044 ? eByteOrderLittle 1045 : eByteOrderBig); 1046 data.SetAddressByteSize(8); 1047 is_64_bit = true; 1048 can_parse = true; 1049 break; 1050 1051 default: 1052 break; 1053 } 1054 1055 if (can_parse) { 1056 data.GetU32(data_offset_ptr, &header.cputype, 6); 1057 if (is_64_bit) 1058 *data_offset_ptr += 4; 1059 return true; 1060 } else { 1061 memset(&header, 0, sizeof(header)); 1062 } 1063 return false; 1064 } 1065 1066 bool ObjectFileMachO::ParseHeader() { 1067 ModuleSP module_sp(GetModule()); 1068 if (module_sp) { 1069 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1070 bool can_parse = false; 1071 lldb::offset_t offset = 0; 1072 m_data.SetByteOrder(endian::InlHostByteOrder()); 1073 // Leave magic in the original byte order 1074 m_header.magic = m_data.GetU32(&offset); 1075 switch (m_header.magic) { 1076 case MH_MAGIC: 1077 m_data.SetByteOrder(endian::InlHostByteOrder()); 1078 m_data.SetAddressByteSize(4); 1079 can_parse = true; 1080 break; 1081 1082 case MH_MAGIC_64: 1083 m_data.SetByteOrder(endian::InlHostByteOrder()); 1084 m_data.SetAddressByteSize(8); 1085 can_parse = true; 1086 break; 1087 1088 case MH_CIGAM: 1089 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1090 ? eByteOrderLittle 1091 : eByteOrderBig); 1092 m_data.SetAddressByteSize(4); 1093 can_parse = true; 1094 break; 1095 1096 case MH_CIGAM_64: 1097 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1098 ? eByteOrderLittle 1099 : eByteOrderBig); 1100 m_data.SetAddressByteSize(8); 1101 can_parse = true; 1102 break; 1103 1104 default: 1105 break; 1106 } 1107 1108 if (can_parse) { 1109 m_data.GetU32(&offset, &m_header.cputype, 6); 1110 1111 ArchSpec mach_arch; 1112 1113 if (GetArchitecture(mach_arch)) { 1114 // Check if the module has a required architecture 1115 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1116 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1117 return false; 1118 1119 if (SetModulesArchitecture(mach_arch)) { 1120 const size_t header_and_lc_size = 1121 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1122 if (m_data.GetByteSize() < header_and_lc_size) { 1123 DataBufferSP data_sp; 1124 ProcessSP process_sp(m_process_wp.lock()); 1125 if (process_sp) { 1126 data_sp = 1127 ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1128 } else { 1129 // Read in all only the load command data from the file on disk 1130 data_sp = DataBufferLLVM::CreateSliceFromPath( 1131 m_file.GetPath(), header_and_lc_size, m_file_offset); 1132 if (data_sp->GetByteSize() != header_and_lc_size) 1133 return false; 1134 } 1135 if (data_sp) 1136 m_data.SetData(data_sp); 1137 } 1138 } 1139 return true; 1140 } 1141 } else { 1142 memset(&m_header, 0, sizeof(struct mach_header)); 1143 } 1144 } 1145 return false; 1146 } 1147 1148 ByteOrder ObjectFileMachO::GetByteOrder() const { 1149 return m_data.GetByteOrder(); 1150 } 1151 1152 bool ObjectFileMachO::IsExecutable() const { 1153 return m_header.filetype == MH_EXECUTE; 1154 } 1155 1156 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1157 return m_data.GetAddressByteSize(); 1158 } 1159 1160 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1161 Symtab *symtab = GetSymtab(); 1162 if (symtab) { 1163 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1164 if (symbol) { 1165 if (symbol->ValueIsAddress()) { 1166 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1167 if (section_sp) { 1168 const lldb::SectionType section_type = section_sp->GetType(); 1169 switch (section_type) { 1170 case eSectionTypeInvalid: 1171 return eAddressClassUnknown; 1172 1173 case eSectionTypeCode: 1174 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1175 // For ARM we have a bit in the n_desc field of the symbol 1176 // that tells us ARM/Thumb which is bit 0x0008. 1177 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1178 return eAddressClassCodeAlternateISA; 1179 } 1180 return eAddressClassCode; 1181 1182 case eSectionTypeContainer: 1183 return eAddressClassUnknown; 1184 1185 case eSectionTypeData: 1186 case eSectionTypeDataCString: 1187 case eSectionTypeDataCStringPointers: 1188 case eSectionTypeDataSymbolAddress: 1189 case eSectionTypeData4: 1190 case eSectionTypeData8: 1191 case eSectionTypeData16: 1192 case eSectionTypeDataPointers: 1193 case eSectionTypeZeroFill: 1194 case eSectionTypeDataObjCMessageRefs: 1195 case eSectionTypeDataObjCCFStrings: 1196 case eSectionTypeGoSymtab: 1197 return eAddressClassData; 1198 1199 case eSectionTypeDebug: 1200 case eSectionTypeDWARFDebugAbbrev: 1201 case eSectionTypeDWARFDebugAddr: 1202 case eSectionTypeDWARFDebugAranges: 1203 case eSectionTypeDWARFDebugFrame: 1204 case eSectionTypeDWARFDebugInfo: 1205 case eSectionTypeDWARFDebugLine: 1206 case eSectionTypeDWARFDebugLoc: 1207 case eSectionTypeDWARFDebugMacInfo: 1208 case eSectionTypeDWARFDebugMacro: 1209 case eSectionTypeDWARFDebugPubNames: 1210 case eSectionTypeDWARFDebugPubTypes: 1211 case eSectionTypeDWARFDebugRanges: 1212 case eSectionTypeDWARFDebugStr: 1213 case eSectionTypeDWARFDebugStrOffsets: 1214 case eSectionTypeDWARFAppleNames: 1215 case eSectionTypeDWARFAppleTypes: 1216 case eSectionTypeDWARFAppleNamespaces: 1217 case eSectionTypeDWARFAppleObjC: 1218 return eAddressClassDebug; 1219 1220 case eSectionTypeEHFrame: 1221 case eSectionTypeARMexidx: 1222 case eSectionTypeARMextab: 1223 case eSectionTypeCompactUnwind: 1224 return eAddressClassRuntime; 1225 1226 case eSectionTypeAbsoluteAddress: 1227 case eSectionTypeELFSymbolTable: 1228 case eSectionTypeELFDynamicSymbols: 1229 case eSectionTypeELFRelocationEntries: 1230 case eSectionTypeELFDynamicLinkInfo: 1231 case eSectionTypeOther: 1232 return eAddressClassUnknown; 1233 } 1234 } 1235 } 1236 1237 const SymbolType symbol_type = symbol->GetType(); 1238 switch (symbol_type) { 1239 case eSymbolTypeAny: 1240 return eAddressClassUnknown; 1241 case eSymbolTypeAbsolute: 1242 return eAddressClassUnknown; 1243 1244 case eSymbolTypeCode: 1245 case eSymbolTypeTrampoline: 1246 case eSymbolTypeResolver: 1247 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1248 // For ARM we have a bit in the n_desc field of the symbol 1249 // that tells us ARM/Thumb which is bit 0x0008. 1250 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1251 return eAddressClassCodeAlternateISA; 1252 } 1253 return eAddressClassCode; 1254 1255 case eSymbolTypeData: 1256 return eAddressClassData; 1257 case eSymbolTypeRuntime: 1258 return eAddressClassRuntime; 1259 case eSymbolTypeException: 1260 return eAddressClassRuntime; 1261 case eSymbolTypeSourceFile: 1262 return eAddressClassDebug; 1263 case eSymbolTypeHeaderFile: 1264 return eAddressClassDebug; 1265 case eSymbolTypeObjectFile: 1266 return eAddressClassDebug; 1267 case eSymbolTypeCommonBlock: 1268 return eAddressClassDebug; 1269 case eSymbolTypeBlock: 1270 return eAddressClassDebug; 1271 case eSymbolTypeLocal: 1272 return eAddressClassData; 1273 case eSymbolTypeParam: 1274 return eAddressClassData; 1275 case eSymbolTypeVariable: 1276 return eAddressClassData; 1277 case eSymbolTypeVariableType: 1278 return eAddressClassDebug; 1279 case eSymbolTypeLineEntry: 1280 return eAddressClassDebug; 1281 case eSymbolTypeLineHeader: 1282 return eAddressClassDebug; 1283 case eSymbolTypeScopeBegin: 1284 return eAddressClassDebug; 1285 case eSymbolTypeScopeEnd: 1286 return eAddressClassDebug; 1287 case eSymbolTypeAdditional: 1288 return eAddressClassUnknown; 1289 case eSymbolTypeCompiler: 1290 return eAddressClassDebug; 1291 case eSymbolTypeInstrumentation: 1292 return eAddressClassDebug; 1293 case eSymbolTypeUndefined: 1294 return eAddressClassUnknown; 1295 case eSymbolTypeObjCClass: 1296 return eAddressClassRuntime; 1297 case eSymbolTypeObjCMetaClass: 1298 return eAddressClassRuntime; 1299 case eSymbolTypeObjCIVar: 1300 return eAddressClassRuntime; 1301 case eSymbolTypeReExported: 1302 return eAddressClassRuntime; 1303 } 1304 } 1305 } 1306 return eAddressClassUnknown; 1307 } 1308 1309 Symtab *ObjectFileMachO::GetSymtab() { 1310 ModuleSP module_sp(GetModule()); 1311 if (module_sp) { 1312 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1313 if (m_symtab_ap.get() == NULL) { 1314 m_symtab_ap.reset(new Symtab(this)); 1315 std::lock_guard<std::recursive_mutex> symtab_guard( 1316 m_symtab_ap->GetMutex()); 1317 ParseSymtab(); 1318 m_symtab_ap->Finalize(); 1319 } 1320 } 1321 return m_symtab_ap.get(); 1322 } 1323 1324 bool ObjectFileMachO::IsStripped() { 1325 if (m_dysymtab.cmd == 0) { 1326 ModuleSP module_sp(GetModule()); 1327 if (module_sp) { 1328 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1329 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1330 const lldb::offset_t load_cmd_offset = offset; 1331 1332 load_command lc; 1333 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 1334 break; 1335 if (lc.cmd == LC_DYSYMTAB) { 1336 m_dysymtab.cmd = lc.cmd; 1337 m_dysymtab.cmdsize = lc.cmdsize; 1338 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1339 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1340 NULL) { 1341 // Clear m_dysymtab if we were unable to read all items from the 1342 // load command 1343 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1344 } 1345 } 1346 offset = load_cmd_offset + lc.cmdsize; 1347 } 1348 } 1349 } 1350 if (m_dysymtab.cmd) 1351 return m_dysymtab.nlocalsym <= 1; 1352 return false; 1353 } 1354 1355 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1356 if (!m_sections_ap.get()) { 1357 m_sections_ap.reset(new SectionList()); 1358 1359 const bool is_dsym = (m_header.filetype == MH_DSYM); 1360 lldb::user_id_t segID = 0; 1361 lldb::user_id_t sectID = 0; 1362 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1363 uint32_t i; 1364 const bool is_core = GetType() == eTypeCoreFile; 1365 // bool dump_sections = false; 1366 ModuleSP module_sp(GetModule()); 1367 // First look up any LC_ENCRYPTION_INFO load commands 1368 typedef RangeArray<uint32_t, uint32_t, 8> EncryptedFileRanges; 1369 EncryptedFileRanges encrypted_file_ranges; 1370 encryption_info_command encryption_cmd; 1371 for (i = 0; i < m_header.ncmds; ++i) { 1372 const lldb::offset_t load_cmd_offset = offset; 1373 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1374 break; 1375 1376 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for 1377 // the 3 fields we care about, so treat them the same. 1378 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1379 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1380 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1381 if (encryption_cmd.cryptid != 0) { 1382 EncryptedFileRanges::Entry entry; 1383 entry.SetRangeBase(encryption_cmd.cryptoff); 1384 entry.SetByteSize(encryption_cmd.cryptsize); 1385 encrypted_file_ranges.Append(entry); 1386 } 1387 } 1388 } 1389 offset = load_cmd_offset + encryption_cmd.cmdsize; 1390 } 1391 1392 bool section_file_addresses_changed = false; 1393 1394 offset = MachHeaderSizeFromMagic(m_header.magic); 1395 1396 struct segment_command_64 load_cmd; 1397 for (i = 0; i < m_header.ncmds; ++i) { 1398 const lldb::offset_t load_cmd_offset = offset; 1399 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1400 break; 1401 1402 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) { 1403 if (m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) { 1404 bool add_section = true; 1405 bool add_to_unified = true; 1406 ConstString const_segname(load_cmd.segname, 1407 std::min<size_t>(strlen(load_cmd.segname), 1408 sizeof(load_cmd.segname))); 1409 1410 SectionSP unified_section_sp( 1411 unified_section_list.FindSectionByName(const_segname)); 1412 if (is_dsym && unified_section_sp) { 1413 if (const_segname == GetSegmentNameLINKEDIT()) { 1414 // We need to keep the __LINKEDIT segment private to this object 1415 // file only 1416 add_to_unified = false; 1417 } else { 1418 // This is the dSYM file and this section has already been created 1419 // by 1420 // the object file, no need to create it. 1421 add_section = false; 1422 } 1423 } 1424 load_cmd.vmaddr = m_data.GetAddress(&offset); 1425 load_cmd.vmsize = m_data.GetAddress(&offset); 1426 load_cmd.fileoff = m_data.GetAddress(&offset); 1427 load_cmd.filesize = m_data.GetAddress(&offset); 1428 if (m_length != 0 && load_cmd.filesize != 0) { 1429 if (load_cmd.fileoff > m_length) { 1430 // We have a load command that says it extends past the end of the 1431 // file. This is likely 1432 // a corrupt file. We don't have any way to return an error 1433 // condition here (this method 1434 // was likely invoked from something like 1435 // ObjectFile::GetSectionList()) -- all we can do 1436 // is null out the SectionList vector and if a process has been 1437 // set up, dump a message 1438 // to stdout. The most common case here is core file debugging 1439 // with a truncated file. 1440 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 1441 ? "LC_SEGMENT_64" 1442 : "LC_SEGMENT"; 1443 module_sp->ReportWarning( 1444 "load command %u %s has a fileoff (0x%" PRIx64 1445 ") that extends beyond the end of the file (0x%" PRIx64 1446 "), ignoring this section", 1447 i, lc_segment_name, load_cmd.fileoff, m_length); 1448 1449 load_cmd.fileoff = 0; 1450 load_cmd.filesize = 0; 1451 } 1452 1453 if (load_cmd.fileoff + load_cmd.filesize > m_length) { 1454 // We have a load command that says it extends past the end of the 1455 // file. This is likely 1456 // a corrupt file. We don't have any way to return an error 1457 // condition here (this method 1458 // was likely invoked from something like 1459 // ObjectFile::GetSectionList()) -- all we can do 1460 // is null out the SectionList vector and if a process has been 1461 // set up, dump a message 1462 // to stdout. The most common case here is core file debugging 1463 // with a truncated file. 1464 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 1465 ? "LC_SEGMENT_64" 1466 : "LC_SEGMENT"; 1467 GetModule()->ReportWarning( 1468 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1469 ") that extends beyond the end of the file (0x%" PRIx64 1470 "), the segment will be truncated to match", 1471 i, lc_segment_name, load_cmd.fileoff + load_cmd.filesize, 1472 m_length); 1473 1474 // Tuncase the length 1475 load_cmd.filesize = m_length - load_cmd.fileoff; 1476 } 1477 } 1478 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) { 1479 uint32_t segment_permissions = 0; 1480 if (load_cmd.initprot & VM_PROT_READ) 1481 segment_permissions |= ePermissionsReadable; 1482 if (load_cmd.initprot & VM_PROT_WRITE) 1483 segment_permissions |= ePermissionsWritable; 1484 if (load_cmd.initprot & VM_PROT_EXECUTE) 1485 segment_permissions |= ePermissionsExecutable; 1486 1487 const bool segment_is_encrypted = 1488 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1489 1490 // Keep a list of mach segments around in case we need to 1491 // get at data that isn't stored in the abstracted Sections. 1492 m_mach_segments.push_back(load_cmd); 1493 1494 // Use a segment ID of the segment index shifted left by 8 so they 1495 // never conflict with any of the sections. 1496 SectionSP segment_sp; 1497 if (add_section && (const_segname || is_core)) { 1498 segment_sp.reset(new Section( 1499 module_sp, // Module to which this section belongs 1500 this, // Object file to which this sections belongs 1501 ++segID << 8, // Section ID is the 1 based segment index 1502 // shifted right by 8 bits as not to collide 1503 // with any of the 256 section IDs that are 1504 // possible 1505 const_segname, // Name of this section 1506 eSectionTypeContainer, // This section is a container of other 1507 // sections. 1508 load_cmd.vmaddr, // File VM address == addresses as they are 1509 // found in the object file 1510 load_cmd.vmsize, // VM size in bytes of this section 1511 load_cmd.fileoff, // Offset to the data for this section in 1512 // the file 1513 load_cmd.filesize, // Size in bytes of this section as found 1514 // in the file 1515 0, // Segments have no alignment information 1516 load_cmd.flags)); // Flags for this section 1517 1518 segment_sp->SetIsEncrypted(segment_is_encrypted); 1519 m_sections_ap->AddSection(segment_sp); 1520 segment_sp->SetPermissions(segment_permissions); 1521 if (add_to_unified) 1522 unified_section_list.AddSection(segment_sp); 1523 } else if (unified_section_sp) { 1524 if (is_dsym && 1525 unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1526 // Check to see if the module was read from memory? 1527 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) { 1528 // We have a module that is in memory and needs to have its 1529 // file address adjusted. We need to do this because when we 1530 // load a file from memory, its addresses will be slid 1531 // already, 1532 // yet the addresses in the new symbol file will still be 1533 // unslid. 1534 // Since everything is stored as section offset, this 1535 // shouldn't 1536 // cause any problems. 1537 1538 // Make sure we've parsed the symbol table from the 1539 // ObjectFile before we go around changing its Sections. 1540 module_sp->GetObjectFile()->GetSymtab(); 1541 // eh_frame would present the same problems but we parse that 1542 // on 1543 // a per-function basis as-needed so it's more difficult to 1544 // remove its use of the Sections. Realistically, the 1545 // environments 1546 // where this code path will be taken will not have eh_frame 1547 // sections. 1548 1549 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1550 1551 // Notify the module that the section addresses have been 1552 // changed once 1553 // we're done so any file-address caches can be updated. 1554 section_file_addresses_changed = true; 1555 } 1556 } 1557 m_sections_ap->AddSection(unified_section_sp); 1558 } 1559 1560 struct section_64 sect64; 1561 ::memset(§64, 0, sizeof(sect64)); 1562 // Push a section into our mach sections for the section at 1563 // index zero (NO_SECT) if we don't have any mach sections yet... 1564 if (m_mach_sections.empty()) 1565 m_mach_sections.push_back(sect64); 1566 uint32_t segment_sect_idx; 1567 const lldb::user_id_t first_segment_sectID = sectID + 1; 1568 1569 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1570 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1571 ++segment_sect_idx) { 1572 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1573 sizeof(sect64.sectname)) == NULL) 1574 break; 1575 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1576 sizeof(sect64.segname)) == NULL) 1577 break; 1578 sect64.addr = m_data.GetAddress(&offset); 1579 sect64.size = m_data.GetAddress(&offset); 1580 1581 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1582 break; 1583 1584 // Keep a list of mach sections around in case we need to 1585 // get at data that isn't stored in the abstracted Sections. 1586 m_mach_sections.push_back(sect64); 1587 1588 if (add_section) { 1589 ConstString section_name( 1590 sect64.sectname, std::min<size_t>(strlen(sect64.sectname), 1591 sizeof(sect64.sectname))); 1592 if (!const_segname) { 1593 // We have a segment with no name so we need to conjure up 1594 // segments that correspond to the section's segname if there 1595 // isn't already such a section. If there is such a section, 1596 // we resize the section so that it spans all sections. 1597 // We also mark these sections as fake so address matches 1598 // don't 1599 // hit if they land in the gaps between the child sections. 1600 const_segname.SetTrimmedCStringWithLength( 1601 sect64.segname, sizeof(sect64.segname)); 1602 segment_sp = 1603 unified_section_list.FindSectionByName(const_segname); 1604 if (segment_sp.get()) { 1605 Section *segment = segment_sp.get(); 1606 // Grow the section size as needed. 1607 const lldb::addr_t sect64_min_addr = sect64.addr; 1608 const lldb::addr_t sect64_max_addr = 1609 sect64_min_addr + sect64.size; 1610 const lldb::addr_t curr_seg_byte_size = 1611 segment->GetByteSize(); 1612 const lldb::addr_t curr_seg_min_addr = 1613 segment->GetFileAddress(); 1614 const lldb::addr_t curr_seg_max_addr = 1615 curr_seg_min_addr + curr_seg_byte_size; 1616 if (sect64_min_addr >= curr_seg_min_addr) { 1617 const lldb::addr_t new_seg_byte_size = 1618 sect64_max_addr - curr_seg_min_addr; 1619 // Only grow the section size if needed 1620 if (new_seg_byte_size > curr_seg_byte_size) 1621 segment->SetByteSize(new_seg_byte_size); 1622 } else { 1623 // We need to change the base address of the segment and 1624 // adjust the child section offsets for all existing 1625 // children. 1626 const lldb::addr_t slide_amount = 1627 sect64_min_addr - curr_seg_min_addr; 1628 segment->Slide(slide_amount, false); 1629 segment->GetChildren().Slide(-slide_amount, false); 1630 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1631 } 1632 1633 // Grow the section size as needed. 1634 if (sect64.offset) { 1635 const lldb::addr_t segment_min_file_offset = 1636 segment->GetFileOffset(); 1637 const lldb::addr_t segment_max_file_offset = 1638 segment_min_file_offset + segment->GetFileSize(); 1639 1640 const lldb::addr_t section_min_file_offset = 1641 sect64.offset; 1642 const lldb::addr_t section_max_file_offset = 1643 section_min_file_offset + sect64.size; 1644 const lldb::addr_t new_file_offset = std::min( 1645 section_min_file_offset, segment_min_file_offset); 1646 const lldb::addr_t new_file_size = 1647 std::max(section_max_file_offset, 1648 segment_max_file_offset) - 1649 new_file_offset; 1650 segment->SetFileOffset(new_file_offset); 1651 segment->SetFileSize(new_file_size); 1652 } 1653 } else { 1654 // Create a fake section for the section's named segment 1655 segment_sp.reset(new Section( 1656 segment_sp, // Parent section 1657 module_sp, // Module to which this section belongs 1658 this, // Object file to which this section belongs 1659 ++segID << 8, // Section ID is the 1 based segment index 1660 // shifted right by 8 bits as not to 1661 // collide with any of the 256 section IDs 1662 // that are possible 1663 const_segname, // Name of this section 1664 eSectionTypeContainer, // This section is a container of 1665 // other sections. 1666 sect64.addr, // File VM address == addresses as they are 1667 // found in the object file 1668 sect64.size, // VM size in bytes of this section 1669 sect64.offset, // Offset to the data for this section in 1670 // the file 1671 sect64.offset ? sect64.size : 0, // Size in bytes of 1672 // this section as 1673 // found in the file 1674 sect64.align, 1675 load_cmd.flags)); // Flags for this section 1676 segment_sp->SetIsFake(true); 1677 segment_sp->SetPermissions(segment_permissions); 1678 m_sections_ap->AddSection(segment_sp); 1679 if (add_to_unified) 1680 unified_section_list.AddSection(segment_sp); 1681 segment_sp->SetIsEncrypted(segment_is_encrypted); 1682 } 1683 } 1684 assert(segment_sp.get()); 1685 1686 lldb::SectionType sect_type = eSectionTypeOther; 1687 1688 if (sect64.flags & 1689 (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1690 sect_type = eSectionTypeCode; 1691 else { 1692 uint32_t mach_sect_type = sect64.flags & SECTION_TYPE; 1693 static ConstString g_sect_name_objc_data("__objc_data"); 1694 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1695 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1696 static ConstString g_sect_name_objc_classrefs( 1697 "__objc_classrefs"); 1698 static ConstString g_sect_name_objc_superrefs( 1699 "__objc_superrefs"); 1700 static ConstString g_sect_name_objc_const("__objc_const"); 1701 static ConstString g_sect_name_objc_classlist( 1702 "__objc_classlist"); 1703 static ConstString g_sect_name_cfstring("__cfstring"); 1704 1705 static ConstString g_sect_name_dwarf_debug_abbrev( 1706 "__debug_abbrev"); 1707 static ConstString g_sect_name_dwarf_debug_aranges( 1708 "__debug_aranges"); 1709 static ConstString g_sect_name_dwarf_debug_frame( 1710 "__debug_frame"); 1711 static ConstString g_sect_name_dwarf_debug_info( 1712 "__debug_info"); 1713 static ConstString g_sect_name_dwarf_debug_line( 1714 "__debug_line"); 1715 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1716 static ConstString g_sect_name_dwarf_debug_macinfo( 1717 "__debug_macinfo"); 1718 static ConstString g_sect_name_dwarf_debug_pubnames( 1719 "__debug_pubnames"); 1720 static ConstString g_sect_name_dwarf_debug_pubtypes( 1721 "__debug_pubtypes"); 1722 static ConstString g_sect_name_dwarf_debug_ranges( 1723 "__debug_ranges"); 1724 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1725 static ConstString g_sect_name_dwarf_apple_names( 1726 "__apple_names"); 1727 static ConstString g_sect_name_dwarf_apple_types( 1728 "__apple_types"); 1729 static ConstString g_sect_name_dwarf_apple_namespaces( 1730 "__apple_namespac"); 1731 static ConstString g_sect_name_dwarf_apple_objc( 1732 "__apple_objc"); 1733 static ConstString g_sect_name_eh_frame("__eh_frame"); 1734 static ConstString g_sect_name_compact_unwind( 1735 "__unwind_info"); 1736 static ConstString g_sect_name_text("__text"); 1737 static ConstString g_sect_name_data("__data"); 1738 static ConstString g_sect_name_go_symtab("__gosymtab"); 1739 1740 if (section_name == g_sect_name_dwarf_debug_abbrev) 1741 sect_type = eSectionTypeDWARFDebugAbbrev; 1742 else if (section_name == g_sect_name_dwarf_debug_aranges) 1743 sect_type = eSectionTypeDWARFDebugAranges; 1744 else if (section_name == g_sect_name_dwarf_debug_frame) 1745 sect_type = eSectionTypeDWARFDebugFrame; 1746 else if (section_name == g_sect_name_dwarf_debug_info) 1747 sect_type = eSectionTypeDWARFDebugInfo; 1748 else if (section_name == g_sect_name_dwarf_debug_line) 1749 sect_type = eSectionTypeDWARFDebugLine; 1750 else if (section_name == g_sect_name_dwarf_debug_loc) 1751 sect_type = eSectionTypeDWARFDebugLoc; 1752 else if (section_name == g_sect_name_dwarf_debug_macinfo) 1753 sect_type = eSectionTypeDWARFDebugMacInfo; 1754 else if (section_name == g_sect_name_dwarf_debug_pubnames) 1755 sect_type = eSectionTypeDWARFDebugPubNames; 1756 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 1757 sect_type = eSectionTypeDWARFDebugPubTypes; 1758 else if (section_name == g_sect_name_dwarf_debug_ranges) 1759 sect_type = eSectionTypeDWARFDebugRanges; 1760 else if (section_name == g_sect_name_dwarf_debug_str) 1761 sect_type = eSectionTypeDWARFDebugStr; 1762 else if (section_name == g_sect_name_dwarf_apple_names) 1763 sect_type = eSectionTypeDWARFAppleNames; 1764 else if (section_name == g_sect_name_dwarf_apple_types) 1765 sect_type = eSectionTypeDWARFAppleTypes; 1766 else if (section_name == g_sect_name_dwarf_apple_namespaces) 1767 sect_type = eSectionTypeDWARFAppleNamespaces; 1768 else if (section_name == g_sect_name_dwarf_apple_objc) 1769 sect_type = eSectionTypeDWARFAppleObjC; 1770 else if (section_name == g_sect_name_objc_selrefs) 1771 sect_type = eSectionTypeDataCStringPointers; 1772 else if (section_name == g_sect_name_objc_msgrefs) 1773 sect_type = eSectionTypeDataObjCMessageRefs; 1774 else if (section_name == g_sect_name_eh_frame) 1775 sect_type = eSectionTypeEHFrame; 1776 else if (section_name == g_sect_name_compact_unwind) 1777 sect_type = eSectionTypeCompactUnwind; 1778 else if (section_name == g_sect_name_cfstring) 1779 sect_type = eSectionTypeDataObjCCFStrings; 1780 else if (section_name == g_sect_name_go_symtab) 1781 sect_type = eSectionTypeGoSymtab; 1782 else if (section_name == g_sect_name_objc_data || 1783 section_name == g_sect_name_objc_classrefs || 1784 section_name == g_sect_name_objc_superrefs || 1785 section_name == g_sect_name_objc_const || 1786 section_name == g_sect_name_objc_classlist) { 1787 sect_type = eSectionTypeDataPointers; 1788 } 1789 1790 if (sect_type == eSectionTypeOther) { 1791 switch (mach_sect_type) { 1792 // TODO: categorize sections by other flags for regular 1793 // sections 1794 case S_REGULAR: 1795 if (section_name == g_sect_name_text) 1796 sect_type = eSectionTypeCode; 1797 else if (section_name == g_sect_name_data) 1798 sect_type = eSectionTypeData; 1799 else 1800 sect_type = eSectionTypeOther; 1801 break; 1802 case S_ZEROFILL: 1803 sect_type = eSectionTypeZeroFill; 1804 break; 1805 case S_CSTRING_LITERALS: 1806 sect_type = eSectionTypeDataCString; 1807 break; // section with only literal C strings 1808 case S_4BYTE_LITERALS: 1809 sect_type = eSectionTypeData4; 1810 break; // section with only 4 byte literals 1811 case S_8BYTE_LITERALS: 1812 sect_type = eSectionTypeData8; 1813 break; // section with only 8 byte literals 1814 case S_LITERAL_POINTERS: 1815 sect_type = eSectionTypeDataPointers; 1816 break; // section with only pointers to literals 1817 case S_NON_LAZY_SYMBOL_POINTERS: 1818 sect_type = eSectionTypeDataPointers; 1819 break; // section with only non-lazy symbol pointers 1820 case S_LAZY_SYMBOL_POINTERS: 1821 sect_type = eSectionTypeDataPointers; 1822 break; // section with only lazy symbol pointers 1823 case S_SYMBOL_STUBS: 1824 sect_type = eSectionTypeCode; 1825 break; // section with only symbol stubs, byte size of 1826 // stub in the reserved2 field 1827 case S_MOD_INIT_FUNC_POINTERS: 1828 sect_type = eSectionTypeDataPointers; 1829 break; // section with only function pointers for 1830 // initialization 1831 case S_MOD_TERM_FUNC_POINTERS: 1832 sect_type = eSectionTypeDataPointers; 1833 break; // section with only function pointers for 1834 // termination 1835 case S_COALESCED: 1836 sect_type = eSectionTypeOther; 1837 break; 1838 case S_GB_ZEROFILL: 1839 sect_type = eSectionTypeZeroFill; 1840 break; 1841 case S_INTERPOSING: 1842 sect_type = eSectionTypeCode; 1843 break; // section with only pairs of function pointers for 1844 // interposing 1845 case S_16BYTE_LITERALS: 1846 sect_type = eSectionTypeData16; 1847 break; // section with only 16 byte literals 1848 case S_DTRACE_DOF: 1849 sect_type = eSectionTypeDebug; 1850 break; 1851 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1852 sect_type = eSectionTypeDataPointers; 1853 break; 1854 default: 1855 break; 1856 } 1857 } 1858 } 1859 1860 SectionSP section_sp(new Section( 1861 segment_sp, module_sp, this, ++sectID, section_name, 1862 sect_type, sect64.addr - segment_sp->GetFileAddress(), 1863 sect64.size, sect64.offset, 1864 sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1865 sect64.flags)); 1866 // Set the section to be encrypted to match the segment 1867 1868 bool section_is_encrypted = false; 1869 if (!segment_is_encrypted && load_cmd.filesize != 0) 1870 section_is_encrypted = 1871 encrypted_file_ranges.FindEntryThatContains( 1872 sect64.offset) != NULL; 1873 1874 section_sp->SetIsEncrypted(segment_is_encrypted || 1875 section_is_encrypted); 1876 section_sp->SetPermissions(segment_permissions); 1877 segment_sp->GetChildren().AddSection(section_sp); 1878 1879 if (segment_sp->IsFake()) { 1880 segment_sp.reset(); 1881 const_segname.Clear(); 1882 } 1883 } 1884 } 1885 if (segment_sp && is_dsym) { 1886 if (first_segment_sectID <= sectID) { 1887 lldb::user_id_t sect_uid; 1888 for (sect_uid = first_segment_sectID; sect_uid <= sectID; 1889 ++sect_uid) { 1890 SectionSP curr_section_sp( 1891 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1892 SectionSP next_section_sp; 1893 if (sect_uid + 1 <= sectID) 1894 next_section_sp = 1895 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1896 1897 if (curr_section_sp.get()) { 1898 if (curr_section_sp->GetByteSize() == 0) { 1899 if (next_section_sp.get() != NULL) 1900 curr_section_sp->SetByteSize( 1901 next_section_sp->GetFileAddress() - 1902 curr_section_sp->GetFileAddress()); 1903 else 1904 curr_section_sp->SetByteSize(load_cmd.vmsize); 1905 } 1906 } 1907 } 1908 } 1909 } 1910 } 1911 } 1912 } else if (load_cmd.cmd == LC_DYSYMTAB) { 1913 m_dysymtab.cmd = load_cmd.cmd; 1914 m_dysymtab.cmdsize = load_cmd.cmdsize; 1915 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1916 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1917 } 1918 1919 offset = load_cmd_offset + load_cmd.cmdsize; 1920 } 1921 1922 if (section_file_addresses_changed && module_sp.get()) { 1923 module_sp->SectionFileAddressesChanged(); 1924 } 1925 } 1926 } 1927 1928 class MachSymtabSectionInfo { 1929 public: 1930 MachSymtabSectionInfo(SectionList *section_list) 1931 : m_section_list(section_list), m_section_infos() { 1932 // Get the number of sections down to a depth of 1 to include 1933 // all segments and their sections, but no other sections that 1934 // may be added for debug map or 1935 m_section_infos.resize(section_list->GetNumSections(1)); 1936 } 1937 1938 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1939 if (n_sect == 0) 1940 return SectionSP(); 1941 if (n_sect < m_section_infos.size()) { 1942 if (!m_section_infos[n_sect].section_sp) { 1943 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1944 m_section_infos[n_sect].section_sp = section_sp; 1945 if (section_sp) { 1946 m_section_infos[n_sect].vm_range.SetBaseAddress( 1947 section_sp->GetFileAddress()); 1948 m_section_infos[n_sect].vm_range.SetByteSize( 1949 section_sp->GetByteSize()); 1950 } else { 1951 Host::SystemLog(Host::eSystemLogError, 1952 "error: unable to find section for section %u\n", 1953 n_sect); 1954 } 1955 } 1956 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1957 // Symbol is in section. 1958 return m_section_infos[n_sect].section_sp; 1959 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1960 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1961 file_addr) { 1962 // Symbol is in section with zero size, but has the same start 1963 // address as the section. This can happen with linker symbols 1964 // (symbols that start with the letter 'l' or 'L'. 1965 return m_section_infos[n_sect].section_sp; 1966 } 1967 } 1968 return m_section_list->FindSectionContainingFileAddress(file_addr); 1969 } 1970 1971 protected: 1972 struct SectionInfo { 1973 SectionInfo() : vm_range(), section_sp() {} 1974 1975 VMRange vm_range; 1976 SectionSP section_sp; 1977 }; 1978 SectionList *m_section_list; 1979 std::vector<SectionInfo> m_section_infos; 1980 }; 1981 1982 struct TrieEntry { 1983 TrieEntry() 1984 : name(), address(LLDB_INVALID_ADDRESS), flags(0), other(0), 1985 import_name() {} 1986 1987 void Clear() { 1988 name.Clear(); 1989 address = LLDB_INVALID_ADDRESS; 1990 flags = 0; 1991 other = 0; 1992 import_name.Clear(); 1993 } 1994 1995 void Dump() const { 1996 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1997 static_cast<unsigned long long>(address), 1998 static_cast<unsigned long long>(flags), 1999 static_cast<unsigned long long>(other), name.GetCString()); 2000 if (import_name) 2001 printf(" -> \"%s\"\n", import_name.GetCString()); 2002 else 2003 printf("\n"); 2004 } 2005 ConstString name; 2006 uint64_t address; 2007 uint64_t flags; 2008 uint64_t other; 2009 ConstString import_name; 2010 }; 2011 2012 struct TrieEntryWithOffset { 2013 lldb::offset_t nodeOffset; 2014 TrieEntry entry; 2015 2016 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 2017 2018 void Dump(uint32_t idx) const { 2019 printf("[%3u] 0x%16.16llx: ", idx, 2020 static_cast<unsigned long long>(nodeOffset)); 2021 entry.Dump(); 2022 } 2023 2024 bool operator<(const TrieEntryWithOffset &other) const { 2025 return (nodeOffset < other.nodeOffset); 2026 } 2027 }; 2028 2029 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 2030 const bool is_arm, 2031 std::vector<llvm::StringRef> &nameSlices, 2032 std::set<lldb::addr_t> &resolver_addresses, 2033 std::vector<TrieEntryWithOffset> &output) { 2034 if (!data.ValidOffset(offset)) 2035 return true; 2036 2037 const uint64_t terminalSize = data.GetULEB128(&offset); 2038 lldb::offset_t children_offset = offset + terminalSize; 2039 if (terminalSize != 0) { 2040 TrieEntryWithOffset e(offset); 2041 e.entry.flags = data.GetULEB128(&offset); 2042 const char *import_name = NULL; 2043 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2044 e.entry.address = 0; 2045 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2046 import_name = data.GetCStr(&offset); 2047 } else { 2048 e.entry.address = data.GetULEB128(&offset); 2049 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2050 e.entry.other = data.GetULEB128(&offset); 2051 uint64_t resolver_addr = e.entry.other; 2052 if (is_arm) 2053 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2054 resolver_addresses.insert(resolver_addr); 2055 } else 2056 e.entry.other = 0; 2057 } 2058 // Only add symbols that are reexport symbols with a valid import name 2059 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && 2060 import_name[0]) { 2061 std::string name; 2062 if (!nameSlices.empty()) { 2063 for (auto name_slice : nameSlices) 2064 name.append(name_slice.data(), name_slice.size()); 2065 } 2066 if (name.size() > 1) { 2067 // Skip the leading '_' 2068 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2069 } 2070 if (import_name) { 2071 // Skip the leading '_' 2072 e.entry.import_name.SetCString(import_name + 1); 2073 } 2074 output.push_back(e); 2075 } 2076 } 2077 2078 const uint8_t childrenCount = data.GetU8(&children_offset); 2079 for (uint8_t i = 0; i < childrenCount; ++i) { 2080 const char *cstr = data.GetCStr(&children_offset); 2081 if (cstr) 2082 nameSlices.push_back(llvm::StringRef(cstr)); 2083 else 2084 return false; // Corrupt data 2085 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2086 if (childNodeOffset) { 2087 if (!ParseTrieEntries(data, childNodeOffset, is_arm, nameSlices, 2088 resolver_addresses, output)) { 2089 return false; 2090 } 2091 } 2092 nameSlices.pop_back(); 2093 } 2094 return true; 2095 } 2096 2097 // Read the UUID out of a dyld_shared_cache file on-disk. 2098 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2099 const ByteOrder byte_order, 2100 const uint32_t addr_byte_size) { 2101 UUID dsc_uuid; 2102 DataBufferSP DscData = DataBufferLLVM::CreateSliceFromPath( 2103 dyld_shared_cache.GetPath(), 2104 sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2105 if (!DscData) 2106 return dsc_uuid; 2107 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2108 2109 char version_str[7]; 2110 lldb::offset_t offset = 0; 2111 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2112 version_str[6] = '\0'; 2113 if (strcmp(version_str, "dyld_v") == 0) { 2114 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2115 uint8_t uuid_bytes[sizeof(uuid_t)]; 2116 memcpy(uuid_bytes, dsc_header_data.GetData(&offset, sizeof(uuid_t)), 2117 sizeof(uuid_t)); 2118 dsc_uuid.SetBytes(uuid_bytes); 2119 } 2120 return dsc_uuid; 2121 } 2122 2123 size_t ObjectFileMachO::ParseSymtab() { 2124 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION); 2125 Timer scoped_timer(func_cat, "ObjectFileMachO::ParseSymtab () module = %s", 2126 m_file.GetFilename().AsCString("")); 2127 ModuleSP module_sp(GetModule()); 2128 if (!module_sp) 2129 return 0; 2130 2131 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2132 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2133 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2134 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2135 FunctionStarts function_starts; 2136 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2137 uint32_t i; 2138 FileSpecList dylib_files; 2139 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2140 static const llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2141 static const llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2142 static const llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2143 2144 for (i = 0; i < m_header.ncmds; ++i) { 2145 const lldb::offset_t cmd_offset = offset; 2146 // Read in the load command and load command size 2147 struct load_command lc; 2148 if (m_data.GetU32(&offset, &lc, 2) == NULL) 2149 break; 2150 // Watch for the symbol table load command 2151 switch (lc.cmd) { 2152 case LC_SYMTAB: 2153 symtab_load_command.cmd = lc.cmd; 2154 symtab_load_command.cmdsize = lc.cmdsize; 2155 // Read in the rest of the symtab load command 2156 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2157 0) // fill in symoff, nsyms, stroff, strsize fields 2158 return 0; 2159 if (symtab_load_command.symoff == 0) { 2160 if (log) 2161 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2162 return 0; 2163 } 2164 2165 if (symtab_load_command.stroff == 0) { 2166 if (log) 2167 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2168 return 0; 2169 } 2170 2171 if (symtab_load_command.nsyms == 0) { 2172 if (log) 2173 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2174 return 0; 2175 } 2176 2177 if (symtab_load_command.strsize == 0) { 2178 if (log) 2179 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2180 return 0; 2181 } 2182 break; 2183 2184 case LC_DYLD_INFO: 2185 case LC_DYLD_INFO_ONLY: 2186 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2187 dyld_info.cmd = lc.cmd; 2188 dyld_info.cmdsize = lc.cmdsize; 2189 } else { 2190 memset(&dyld_info, 0, sizeof(dyld_info)); 2191 } 2192 break; 2193 2194 case LC_LOAD_DYLIB: 2195 case LC_LOAD_WEAK_DYLIB: 2196 case LC_REEXPORT_DYLIB: 2197 case LC_LOADFVMLIB: 2198 case LC_LOAD_UPWARD_DYLIB: { 2199 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2200 const char *path = m_data.PeekCStr(name_offset); 2201 if (path) { 2202 FileSpec file_spec(path, false); 2203 // Strip the path if there is @rpath, @executable, etc so we just use 2204 // the basename 2205 if (path[0] == '@') 2206 file_spec.GetDirectory().Clear(); 2207 2208 if (lc.cmd == LC_REEXPORT_DYLIB) { 2209 m_reexported_dylibs.AppendIfUnique(file_spec); 2210 } 2211 2212 dylib_files.Append(file_spec); 2213 } 2214 } break; 2215 2216 case LC_FUNCTION_STARTS: 2217 function_starts_load_command.cmd = lc.cmd; 2218 function_starts_load_command.cmdsize = lc.cmdsize; 2219 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2220 NULL) // fill in symoff, nsyms, stroff, strsize fields 2221 memset(&function_starts_load_command, 0, 2222 sizeof(function_starts_load_command)); 2223 break; 2224 2225 default: 2226 break; 2227 } 2228 offset = cmd_offset + lc.cmdsize; 2229 } 2230 2231 if (symtab_load_command.cmd) { 2232 Symtab *symtab = m_symtab_ap.get(); 2233 SectionList *section_list = GetSectionList(); 2234 if (section_list == NULL) 2235 return 0; 2236 2237 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2238 const ByteOrder byte_order = m_data.GetByteOrder(); 2239 bool bit_width_32 = addr_byte_size == 4; 2240 const size_t nlist_byte_size = 2241 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2242 2243 DataExtractor nlist_data(NULL, 0, byte_order, addr_byte_size); 2244 DataExtractor strtab_data(NULL, 0, byte_order, addr_byte_size); 2245 DataExtractor function_starts_data(NULL, 0, byte_order, addr_byte_size); 2246 DataExtractor indirect_symbol_index_data(NULL, 0, byte_order, 2247 addr_byte_size); 2248 DataExtractor dyld_trie_data(NULL, 0, byte_order, addr_byte_size); 2249 2250 const addr_t nlist_data_byte_size = 2251 symtab_load_command.nsyms * nlist_byte_size; 2252 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2253 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2254 2255 ProcessSP process_sp(m_process_wp.lock()); 2256 Process *process = process_sp.get(); 2257 2258 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2259 2260 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2261 Target &target = process->GetTarget(); 2262 2263 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2264 2265 SectionSP linkedit_section_sp( 2266 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2267 // Reading mach file from memory in a process or core file... 2268 2269 if (linkedit_section_sp) { 2270 addr_t linkedit_load_addr = 2271 linkedit_section_sp->GetLoadBaseAddress(&target); 2272 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2273 // We might be trying to access the symbol table before the 2274 // __LINKEDIT's load 2275 // address has been set in the target. We can't fail to read the 2276 // symbol table, 2277 // so calculate the right address manually 2278 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2279 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2280 } 2281 2282 const addr_t linkedit_file_offset = 2283 linkedit_section_sp->GetFileOffset(); 2284 const addr_t symoff_addr = linkedit_load_addr + 2285 symtab_load_command.symoff - 2286 linkedit_file_offset; 2287 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2288 linkedit_file_offset; 2289 2290 bool data_was_read = false; 2291 2292 #if defined(__APPLE__) && \ 2293 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2294 if (m_header.flags & 0x80000000u && 2295 process->GetAddressByteSize() == sizeof(void *)) { 2296 // This mach-o memory file is in the dyld shared cache. If this 2297 // program is not remote and this is iOS, then this process will 2298 // share the same shared cache as the process we are debugging and 2299 // we can read the entire __LINKEDIT from the address space in this 2300 // process. This is a needed optimization that is used for local iOS 2301 // debugging only since all shared libraries in the shared cache do 2302 // not have corresponding files that exist in the file system of the 2303 // device. They have been combined into a single file. This means we 2304 // always have to load these files from memory. All of the symbol and 2305 // string tables from all of the __LINKEDIT sections from the shared 2306 // libraries in the shared cache have been merged into a single large 2307 // symbol and string table. Reading all of this symbol and string 2308 // table 2309 // data across can slow down debug launch times, so we optimize this 2310 // by 2311 // reading the memory for the __LINKEDIT section from this process. 2312 2313 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 2314 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 2315 bool use_lldb_cache = true; 2316 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2317 lldb_shared_cache != process_shared_cache) { 2318 use_lldb_cache = false; 2319 ModuleSP module_sp(GetModule()); 2320 if (module_sp) 2321 module_sp->ReportWarning("shared cache in process does not match " 2322 "lldb's own shared cache, startup will " 2323 "be slow."); 2324 } 2325 2326 PlatformSP platform_sp(target.GetPlatform()); 2327 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2328 data_was_read = true; 2329 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2330 eByteOrderLittle); 2331 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2332 eByteOrderLittle); 2333 if (function_starts_load_command.cmd) { 2334 const addr_t func_start_addr = 2335 linkedit_load_addr + function_starts_load_command.dataoff - 2336 linkedit_file_offset; 2337 function_starts_data.SetData( 2338 (void *)func_start_addr, 2339 function_starts_load_command.datasize, eByteOrderLittle); 2340 } 2341 } 2342 } 2343 #endif 2344 2345 if (!data_was_read) { 2346 // Always load dyld - the dynamic linker - from memory if we didn't 2347 // find a binary anywhere else. 2348 // lldb will not register dylib/framework/bundle loads/unloads if we 2349 // don't have the dyld symbols, 2350 // we force dyld to load from memory despite the user's 2351 // target.memory-module-load-level setting. 2352 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2353 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2354 DataBufferSP nlist_data_sp( 2355 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2356 if (nlist_data_sp) 2357 nlist_data.SetData(nlist_data_sp, 0, 2358 nlist_data_sp->GetByteSize()); 2359 // Load strings individually from memory when loading from memory 2360 // since shared cache 2361 // string tables contain strings for all symbols from all shared 2362 // cached libraries 2363 // DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, 2364 // strtab_data_byte_size)); 2365 // if (strtab_data_sp) 2366 // strtab_data.SetData (strtab_data_sp, 0, 2367 // strtab_data_sp->GetByteSize()); 2368 if (m_dysymtab.nindirectsyms != 0) { 2369 const addr_t indirect_syms_addr = linkedit_load_addr + 2370 m_dysymtab.indirectsymoff - 2371 linkedit_file_offset; 2372 DataBufferSP indirect_syms_data_sp( 2373 ReadMemory(process_sp, indirect_syms_addr, 2374 m_dysymtab.nindirectsyms * 4)); 2375 if (indirect_syms_data_sp) 2376 indirect_symbol_index_data.SetData( 2377 indirect_syms_data_sp, 0, 2378 indirect_syms_data_sp->GetByteSize()); 2379 } 2380 } else if (memory_module_load_level >= 2381 eMemoryModuleLoadLevelPartial) { 2382 if (function_starts_load_command.cmd) { 2383 const addr_t func_start_addr = 2384 linkedit_load_addr + function_starts_load_command.dataoff - 2385 linkedit_file_offset; 2386 DataBufferSP func_start_data_sp( 2387 ReadMemory(process_sp, func_start_addr, 2388 function_starts_load_command.datasize)); 2389 if (func_start_data_sp) 2390 function_starts_data.SetData(func_start_data_sp, 0, 2391 func_start_data_sp->GetByteSize()); 2392 } 2393 } 2394 } 2395 } 2396 } else { 2397 nlist_data.SetData(m_data, symtab_load_command.symoff, 2398 nlist_data_byte_size); 2399 strtab_data.SetData(m_data, symtab_load_command.stroff, 2400 strtab_data_byte_size); 2401 2402 if (dyld_info.export_size > 0) { 2403 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2404 dyld_info.export_size); 2405 } 2406 2407 if (m_dysymtab.nindirectsyms != 0) { 2408 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2409 m_dysymtab.nindirectsyms * 4); 2410 } 2411 if (function_starts_load_command.cmd) { 2412 function_starts_data.SetData(m_data, 2413 function_starts_load_command.dataoff, 2414 function_starts_load_command.datasize); 2415 } 2416 } 2417 2418 if (nlist_data.GetByteSize() == 0 && 2419 memory_module_load_level == eMemoryModuleLoadLevelComplete) { 2420 if (log) 2421 module_sp->LogMessage(log, "failed to read nlist data"); 2422 return 0; 2423 } 2424 2425 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2426 if (!have_strtab_data) { 2427 if (process) { 2428 if (strtab_addr == LLDB_INVALID_ADDRESS) { 2429 if (log) 2430 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2431 return 0; 2432 } 2433 } else { 2434 if (log) 2435 module_sp->LogMessage(log, "failed to read strtab data"); 2436 return 0; 2437 } 2438 } 2439 2440 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2441 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2442 const ConstString &g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2443 const ConstString &g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2444 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2445 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2446 SectionSP text_section_sp( 2447 section_list->FindSectionByName(g_segment_name_TEXT)); 2448 SectionSP data_section_sp( 2449 section_list->FindSectionByName(g_segment_name_DATA)); 2450 SectionSP data_dirty_section_sp( 2451 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2452 SectionSP data_const_section_sp( 2453 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2454 SectionSP objc_section_sp( 2455 section_list->FindSectionByName(g_segment_name_OBJC)); 2456 SectionSP eh_frame_section_sp; 2457 if (text_section_sp.get()) 2458 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2459 g_section_name_eh_frame); 2460 else 2461 eh_frame_section_sp = 2462 section_list->FindSectionByName(g_section_name_eh_frame); 2463 2464 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2465 2466 // lldb works best if it knows the start address of all functions in a 2467 // module. 2468 // Linker symbols or debug info are normally the best source of information 2469 // for start addr / size but 2470 // they may be stripped in a released binary. 2471 // Two additional sources of information exist in Mach-O binaries: 2472 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2473 // function's start address in the 2474 // binary, relative to the text section. 2475 // eh_frame - the eh_frame FDEs have the start addr & size of 2476 // each function 2477 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2478 // all modern binaries. 2479 // Binaries built to run on older releases may need to use eh_frame 2480 // information. 2481 2482 if (text_section_sp && function_starts_data.GetByteSize()) { 2483 FunctionStarts::Entry function_start_entry; 2484 function_start_entry.data = false; 2485 lldb::offset_t function_start_offset = 0; 2486 function_start_entry.addr = text_section_sp->GetFileAddress(); 2487 uint64_t delta; 2488 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2489 0) { 2490 // Now append the current entry 2491 function_start_entry.addr += delta; 2492 function_starts.Append(function_start_entry); 2493 } 2494 } else { 2495 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2496 // load command claiming an eh_frame 2497 // but it doesn't actually have the eh_frame content. And if we have a 2498 // dSYM, we don't need to do any 2499 // of this fill-in-the-missing-symbols works anyway - the debug info 2500 // should give us all the functions in 2501 // the module. 2502 if (text_section_sp.get() && eh_frame_section_sp.get() && 2503 m_type != eTypeDebugInfo) { 2504 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2505 DWARFCallFrameInfo::EH); 2506 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2507 eh_frame.GetFunctionAddressAndSizeVector(functions); 2508 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2509 size_t count = functions.GetSize(); 2510 for (size_t i = 0; i < count; ++i) { 2511 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2512 functions.GetEntryAtIndex(i); 2513 if (func) { 2514 FunctionStarts::Entry function_start_entry; 2515 function_start_entry.addr = func->base - text_base_addr; 2516 function_starts.Append(function_start_entry); 2517 } 2518 } 2519 } 2520 } 2521 2522 const size_t function_starts_count = function_starts.GetSize(); 2523 2524 // For user process binaries (executables, dylibs, frameworks, bundles), if 2525 // we don't have 2526 // LC_FUNCTION_STARTS/eh_frame section in this binary, we're going to assume 2527 // the binary 2528 // has been stripped. Don't allow assembly language instruction emulation 2529 // because we don't 2530 // know proper function start boundaries. 2531 // 2532 // For all other types of binaries (kernels, stand-alone bare board 2533 // binaries, kexts), they 2534 // may not have LC_FUNCTION_STARTS / eh_frame sections - we should not make 2535 // any assumptions 2536 // about them based on that. 2537 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2538 m_allow_assembly_emulation_unwind_plans = false; 2539 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2540 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2541 2542 if (unwind_or_symbol_log) 2543 module_sp->LogMessage( 2544 unwind_or_symbol_log, 2545 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2546 } 2547 2548 const user_id_t TEXT_eh_frame_sectID = 2549 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2550 : static_cast<user_id_t>(NO_SECT); 2551 2552 lldb::offset_t nlist_data_offset = 0; 2553 2554 uint32_t N_SO_index = UINT32_MAX; 2555 2556 MachSymtabSectionInfo section_info(section_list); 2557 std::vector<uint32_t> N_FUN_indexes; 2558 std::vector<uint32_t> N_NSYM_indexes; 2559 std::vector<uint32_t> N_INCL_indexes; 2560 std::vector<uint32_t> N_BRAC_indexes; 2561 std::vector<uint32_t> N_COMM_indexes; 2562 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2563 typedef std::map<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2564 typedef std::map<const char *, uint32_t> ConstNameToSymbolIndexMap; 2565 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2566 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2567 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2568 // Any symbols that get merged into another will get an entry 2569 // in this map so we know 2570 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2571 uint32_t nlist_idx = 0; 2572 Symbol *symbol_ptr = NULL; 2573 2574 uint32_t sym_idx = 0; 2575 Symbol *sym = NULL; 2576 size_t num_syms = 0; 2577 std::string memory_symbol_name; 2578 uint32_t unmapped_local_symbols_found = 0; 2579 2580 std::vector<TrieEntryWithOffset> trie_entries; 2581 std::set<lldb::addr_t> resolver_addresses; 2582 2583 if (dyld_trie_data.GetByteSize() > 0) { 2584 std::vector<llvm::StringRef> nameSlices; 2585 ParseTrieEntries(dyld_trie_data, 0, is_arm, nameSlices, 2586 resolver_addresses, trie_entries); 2587 2588 ConstString text_segment_name("__TEXT"); 2589 SectionSP text_segment_sp = 2590 GetSectionList()->FindSectionByName(text_segment_name); 2591 if (text_segment_sp) { 2592 const lldb::addr_t text_segment_file_addr = 2593 text_segment_sp->GetFileAddress(); 2594 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) { 2595 for (auto &e : trie_entries) 2596 e.entry.address += text_segment_file_addr; 2597 } 2598 } 2599 } 2600 2601 typedef std::set<ConstString> IndirectSymbols; 2602 IndirectSymbols indirect_symbol_names; 2603 2604 #if defined(__APPLE__) && \ 2605 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2606 2607 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2608 // optimized by moving LOCAL 2609 // symbols out of the memory mapped portion of the DSC. The symbol 2610 // information has all been retained, 2611 // but it isn't available in the normal nlist data. However, there *are* 2612 // duplicate entries of *some* 2613 // LOCAL symbols in the normal nlist data. To handle this situation 2614 // correctly, we must first attempt 2615 // to parse any DSC unmapped symbol information. If we find any, we set a 2616 // flag that tells the normal 2617 // nlist parser to ignore all LOCAL symbols. 2618 2619 if (m_header.flags & 0x80000000u) { 2620 // Before we can start mapping the DSC, we need to make certain the target 2621 // process is actually 2622 // using the cache we can find. 2623 2624 // Next we need to determine the correct path for the dyld shared cache. 2625 2626 ArchSpec header_arch; 2627 GetArchitecture(header_arch); 2628 char dsc_path[PATH_MAX]; 2629 char dsc_path_development[PATH_MAX]; 2630 2631 snprintf( 2632 dsc_path, sizeof(dsc_path), "%s%s%s", 2633 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2634 */ 2635 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2636 header_arch.GetArchitectureName()); 2637 2638 snprintf( 2639 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2640 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2641 */ 2642 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2643 header_arch.GetArchitectureName(), ".development"); 2644 2645 FileSpec dsc_nondevelopment_filespec(dsc_path, false); 2646 FileSpec dsc_development_filespec(dsc_path_development, false); 2647 FileSpec dsc_filespec; 2648 2649 UUID dsc_uuid; 2650 UUID process_shared_cache_uuid; 2651 2652 if (process) { 2653 process_shared_cache_uuid = GetProcessSharedCacheUUID(process); 2654 } 2655 2656 // First see if we can find an exact match for the inferior process shared 2657 // cache UUID in 2658 // the development or non-development shared caches on disk. 2659 if (process_shared_cache_uuid.IsValid()) { 2660 if (dsc_development_filespec.Exists()) { 2661 UUID dsc_development_uuid = GetSharedCacheUUID( 2662 dsc_development_filespec, byte_order, addr_byte_size); 2663 if (dsc_development_uuid.IsValid() && 2664 dsc_development_uuid == process_shared_cache_uuid) { 2665 dsc_filespec = dsc_development_filespec; 2666 dsc_uuid = dsc_development_uuid; 2667 } 2668 } 2669 if (!dsc_uuid.IsValid() && dsc_nondevelopment_filespec.Exists()) { 2670 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2671 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2672 if (dsc_nondevelopment_uuid.IsValid() && 2673 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2674 dsc_filespec = dsc_nondevelopment_filespec; 2675 dsc_uuid = dsc_nondevelopment_uuid; 2676 } 2677 } 2678 } 2679 2680 // Failing a UUID match, prefer the development dyld_shared cache if both 2681 // are present. 2682 if (!dsc_filespec.Exists()) { 2683 if (dsc_development_filespec.Exists()) { 2684 dsc_filespec = dsc_development_filespec; 2685 } else { 2686 dsc_filespec = dsc_nondevelopment_filespec; 2687 } 2688 } 2689 2690 /* The dyld_cache_header has a pointer to the 2691 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2692 The dyld_cache_local_symbols_info structure gives us three things: 2693 1. The start and count of the nlist records in the dyld_shared_cache 2694 file 2695 2. The start and size of the strings for these nlist records 2696 3. The start and count of dyld_cache_local_symbols_entry entries 2697 2698 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2699 dyld shared cache. 2700 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2701 the dyld shared cache. 2702 The dyld_cache_local_symbols_entry also lists the start of this 2703 dylib/framework's nlist records 2704 and the count of how many nlist records there are for this 2705 dylib/framework. 2706 */ 2707 2708 // Process the dyld shared cache header to find the unmapped symbols 2709 2710 DataBufferSP dsc_data_sp = DataBufferLLVM::CreateSliceFromPath( 2711 dsc_filespec.GetPath(), sizeof(struct lldb_copy_dyld_cache_header_v1), 2712 0); 2713 if (!dsc_uuid.IsValid()) { 2714 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2715 } 2716 if (dsc_data_sp) { 2717 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2718 2719 bool uuid_match = true; 2720 if (dsc_uuid.IsValid() && process) { 2721 if (process_shared_cache_uuid.IsValid() && 2722 dsc_uuid != process_shared_cache_uuid) { 2723 // The on-disk dyld_shared_cache file is not the same as the one in 2724 // this 2725 // process' memory, don't use it. 2726 uuid_match = false; 2727 ModuleSP module_sp(GetModule()); 2728 if (module_sp) 2729 module_sp->ReportWarning("process shared cache does not match " 2730 "on-disk dyld_shared_cache file, some " 2731 "symbol names will be missing."); 2732 } 2733 } 2734 2735 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2736 2737 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2738 2739 // If the mappingOffset points to a location inside the header, we've 2740 // opened an old dyld shared cache, and should not proceed further. 2741 if (uuid_match && 2742 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2743 2744 DataBufferSP dsc_mapping_info_data_sp = 2745 DataBufferLLVM::CreateSliceFromPath( 2746 dsc_filespec.GetPath(), 2747 sizeof(struct lldb_copy_dyld_cache_mapping_info), 2748 mappingOffset); 2749 2750 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2751 byte_order, addr_byte_size); 2752 offset = 0; 2753 2754 // The File addresses (from the in-memory Mach-O load commands) for 2755 // the shared libraries 2756 // in the shared library cache need to be adjusted by an offset to 2757 // match up with the 2758 // dylibOffset identifying field in the 2759 // dyld_cache_local_symbol_entry's. This offset is 2760 // recorded in mapping_offset_value. 2761 const uint64_t mapping_offset_value = 2762 dsc_mapping_info_data.GetU64(&offset); 2763 2764 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, 2765 localSymbolsOffset); 2766 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2767 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2768 2769 if (localSymbolsOffset && localSymbolsSize) { 2770 // Map the local symbols 2771 DataBufferSP dsc_local_symbols_data_sp = 2772 DataBufferLLVM::CreateSliceFromPath(dsc_filespec.GetPath(), 2773 localSymbolsSize, 2774 localSymbolsOffset); 2775 2776 if (dsc_local_symbols_data_sp) { 2777 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2778 byte_order, addr_byte_size); 2779 2780 offset = 0; 2781 2782 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2783 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2784 UndefinedNameToDescMap undefined_name_to_desc; 2785 SymbolIndexToName reexport_shlib_needs_fixup; 2786 2787 // Read the local_symbols_infos struct in one shot 2788 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2789 dsc_local_symbols_data.GetU32(&offset, 2790 &local_symbols_info.nlistOffset, 6); 2791 2792 SectionSP text_section_sp( 2793 section_list->FindSectionByName(GetSegmentNameTEXT())); 2794 2795 uint32_t header_file_offset = 2796 (text_section_sp->GetFileAddress() - mapping_offset_value); 2797 2798 offset = local_symbols_info.entriesOffset; 2799 for (uint32_t entry_index = 0; 2800 entry_index < local_symbols_info.entriesCount; 2801 entry_index++) { 2802 struct lldb_copy_dyld_cache_local_symbols_entry 2803 local_symbols_entry; 2804 local_symbols_entry.dylibOffset = 2805 dsc_local_symbols_data.GetU32(&offset); 2806 local_symbols_entry.nlistStartIndex = 2807 dsc_local_symbols_data.GetU32(&offset); 2808 local_symbols_entry.nlistCount = 2809 dsc_local_symbols_data.GetU32(&offset); 2810 2811 if (header_file_offset == local_symbols_entry.dylibOffset) { 2812 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2813 2814 // The normal nlist code cannot correctly size the Symbols 2815 // array, we need to allocate it here. 2816 sym = symtab->Resize( 2817 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2818 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2819 num_syms = symtab->GetNumSymbols(); 2820 2821 nlist_data_offset = 2822 local_symbols_info.nlistOffset + 2823 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2824 uint32_t string_table_offset = 2825 local_symbols_info.stringsOffset; 2826 2827 for (uint32_t nlist_index = 0; 2828 nlist_index < local_symbols_entry.nlistCount; 2829 nlist_index++) { 2830 ///////////////////////////// 2831 { 2832 struct nlist_64 nlist; 2833 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize( 2834 nlist_data_offset, nlist_byte_size)) 2835 break; 2836 2837 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked( 2838 &nlist_data_offset); 2839 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked( 2840 &nlist_data_offset); 2841 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked( 2842 &nlist_data_offset); 2843 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked( 2844 &nlist_data_offset); 2845 nlist.n_value = 2846 dsc_local_symbols_data.GetAddress_unchecked( 2847 &nlist_data_offset); 2848 2849 SymbolType type = eSymbolTypeInvalid; 2850 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2851 string_table_offset + nlist.n_strx); 2852 2853 if (symbol_name == NULL) { 2854 // No symbol should be NULL, even the symbols with no 2855 // string values should have an offset zero which points 2856 // to an empty C-string 2857 Host::SystemLog( 2858 Host::eSystemLogError, 2859 "error: DSC unmapped local symbol[%u] has invalid " 2860 "string table offset 0x%x in %s, ignoring symbol\n", 2861 entry_index, nlist.n_strx, 2862 module_sp->GetFileSpec().GetPath().c_str()); 2863 continue; 2864 } 2865 if (symbol_name[0] == '\0') 2866 symbol_name = NULL; 2867 2868 const char *symbol_name_non_abi_mangled = NULL; 2869 2870 SectionSP symbol_section; 2871 uint32_t symbol_byte_size = 0; 2872 bool add_nlist = true; 2873 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2874 bool demangled_is_synthesized = false; 2875 bool is_gsym = false; 2876 bool set_value = true; 2877 2878 assert(sym_idx < num_syms); 2879 2880 sym[sym_idx].SetDebug(is_debug); 2881 2882 if (is_debug) { 2883 switch (nlist.n_type) { 2884 case N_GSYM: 2885 // global symbol: name,,NO_SECT,type,0 2886 // Sometimes the N_GSYM value contains the address. 2887 2888 // FIXME: In the .o files, we have a GSYM and a debug 2889 // symbol for all the ObjC data. They 2890 // have the same address, but we want to ensure that 2891 // we always find only the real symbol, 2892 // 'cause we don't currently correctly attribute the 2893 // GSYM one to the ObjCClass/Ivar/MetaClass 2894 // symbol type. This is a temporary hack to make sure 2895 // the ObjectiveC symbols get treated 2896 // correctly. To do this right, we should coalesce 2897 // all the GSYM & global symbols that have the 2898 // same address. 2899 2900 is_gsym = true; 2901 sym[sym_idx].SetExternal(true); 2902 2903 if (symbol_name && symbol_name[0] == '_' && 2904 symbol_name[1] == 'O') { 2905 llvm::StringRef symbol_name_ref(symbol_name); 2906 if (symbol_name_ref.startswith( 2907 g_objc_v2_prefix_class)) { 2908 symbol_name_non_abi_mangled = symbol_name + 1; 2909 symbol_name = 2910 symbol_name + g_objc_v2_prefix_class.size(); 2911 type = eSymbolTypeObjCClass; 2912 demangled_is_synthesized = true; 2913 2914 } else if (symbol_name_ref.startswith( 2915 g_objc_v2_prefix_metaclass)) { 2916 symbol_name_non_abi_mangled = symbol_name + 1; 2917 symbol_name = symbol_name + 2918 g_objc_v2_prefix_metaclass.size(); 2919 type = eSymbolTypeObjCMetaClass; 2920 demangled_is_synthesized = true; 2921 } else if (symbol_name_ref.startswith( 2922 g_objc_v2_prefix_ivar)) { 2923 symbol_name_non_abi_mangled = symbol_name + 1; 2924 symbol_name = 2925 symbol_name + g_objc_v2_prefix_ivar.size(); 2926 type = eSymbolTypeObjCIVar; 2927 demangled_is_synthesized = true; 2928 } 2929 } else { 2930 if (nlist.n_value != 0) 2931 symbol_section = section_info.GetSection( 2932 nlist.n_sect, nlist.n_value); 2933 type = eSymbolTypeData; 2934 } 2935 break; 2936 2937 case N_FNAME: 2938 // procedure name (f77 kludge): name,,NO_SECT,0,0 2939 type = eSymbolTypeCompiler; 2940 break; 2941 2942 case N_FUN: 2943 // procedure: name,,n_sect,linenumber,address 2944 if (symbol_name) { 2945 type = eSymbolTypeCode; 2946 symbol_section = section_info.GetSection( 2947 nlist.n_sect, nlist.n_value); 2948 2949 N_FUN_addr_to_sym_idx.insert( 2950 std::make_pair(nlist.n_value, sym_idx)); 2951 // We use the current number of symbols in the 2952 // symbol table in lieu of 2953 // using nlist_idx in case we ever start trimming 2954 // entries out 2955 N_FUN_indexes.push_back(sym_idx); 2956 } else { 2957 type = eSymbolTypeCompiler; 2958 2959 if (!N_FUN_indexes.empty()) { 2960 // Copy the size of the function into the original 2961 // STAB entry so we don't have 2962 // to hunt for it later 2963 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2964 ->SetByteSize(nlist.n_value); 2965 N_FUN_indexes.pop_back(); 2966 // We don't really need the end function STAB as 2967 // it contains the size which 2968 // we already placed with the original symbol, so 2969 // don't add it if we want a 2970 // minimal symbol table 2971 add_nlist = false; 2972 } 2973 } 2974 break; 2975 2976 case N_STSYM: 2977 // static symbol: name,,n_sect,type,address 2978 N_STSYM_addr_to_sym_idx.insert( 2979 std::make_pair(nlist.n_value, sym_idx)); 2980 symbol_section = section_info.GetSection( 2981 nlist.n_sect, nlist.n_value); 2982 if (symbol_name && symbol_name[0]) { 2983 type = ObjectFile::GetSymbolTypeFromName( 2984 symbol_name + 1, eSymbolTypeData); 2985 } 2986 break; 2987 2988 case N_LCSYM: 2989 // .lcomm symbol: name,,n_sect,type,address 2990 symbol_section = section_info.GetSection( 2991 nlist.n_sect, nlist.n_value); 2992 type = eSymbolTypeCommonBlock; 2993 break; 2994 2995 case N_BNSYM: 2996 // We use the current number of symbols in the symbol 2997 // table in lieu of 2998 // using nlist_idx in case we ever start trimming 2999 // entries out 3000 // Skip these if we want minimal symbol tables 3001 add_nlist = false; 3002 break; 3003 3004 case N_ENSYM: 3005 // Set the size of the N_BNSYM to the terminating 3006 // index of this N_ENSYM 3007 // so that we can always skip the entire symbol if we 3008 // need to navigate 3009 // more quickly at the source level when parsing STABS 3010 // Skip these if we want minimal symbol tables 3011 add_nlist = false; 3012 break; 3013 3014 case N_OPT: 3015 // emitted with gcc2_compiled and in gcc source 3016 type = eSymbolTypeCompiler; 3017 break; 3018 3019 case N_RSYM: 3020 // register sym: name,,NO_SECT,type,register 3021 type = eSymbolTypeVariable; 3022 break; 3023 3024 case N_SLINE: 3025 // src line: 0,,n_sect,linenumber,address 3026 symbol_section = section_info.GetSection( 3027 nlist.n_sect, nlist.n_value); 3028 type = eSymbolTypeLineEntry; 3029 break; 3030 3031 case N_SSYM: 3032 // structure elt: name,,NO_SECT,type,struct_offset 3033 type = eSymbolTypeVariableType; 3034 break; 3035 3036 case N_SO: 3037 // source file name 3038 type = eSymbolTypeSourceFile; 3039 if (symbol_name == NULL) { 3040 add_nlist = false; 3041 if (N_SO_index != UINT32_MAX) { 3042 // Set the size of the N_SO to the terminating 3043 // index of this N_SO 3044 // so that we can always skip the entire N_SO if 3045 // we need to navigate 3046 // more quickly at the source level when parsing 3047 // STABS 3048 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3049 symbol_ptr->SetByteSize(sym_idx); 3050 symbol_ptr->SetSizeIsSibling(true); 3051 } 3052 N_NSYM_indexes.clear(); 3053 N_INCL_indexes.clear(); 3054 N_BRAC_indexes.clear(); 3055 N_COMM_indexes.clear(); 3056 N_FUN_indexes.clear(); 3057 N_SO_index = UINT32_MAX; 3058 } else { 3059 // We use the current number of symbols in the 3060 // symbol table in lieu of 3061 // using nlist_idx in case we ever start trimming 3062 // entries out 3063 const bool N_SO_has_full_path = 3064 symbol_name[0] == '/'; 3065 if (N_SO_has_full_path) { 3066 if ((N_SO_index == sym_idx - 1) && 3067 ((sym_idx - 1) < num_syms)) { 3068 // We have two consecutive N_SO entries where 3069 // the first contains a directory 3070 // and the second contains a full path. 3071 sym[sym_idx - 1].GetMangled().SetValue( 3072 ConstString(symbol_name), false); 3073 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3074 add_nlist = false; 3075 } else { 3076 // This is the first entry in a N_SO that 3077 // contains a directory or 3078 // a full path to the source file 3079 N_SO_index = sym_idx; 3080 } 3081 } else if ((N_SO_index == sym_idx - 1) && 3082 ((sym_idx - 1) < num_syms)) { 3083 // This is usually the second N_SO entry that 3084 // contains just the filename, 3085 // so here we combine it with the first one if we 3086 // are minimizing the symbol table 3087 const char *so_path = 3088 sym[sym_idx - 1] 3089 .GetMangled() 3090 .GetDemangledName( 3091 lldb::eLanguageTypeUnknown) 3092 .AsCString(); 3093 if (so_path && so_path[0]) { 3094 std::string full_so_path(so_path); 3095 const size_t double_slash_pos = 3096 full_so_path.find("//"); 3097 if (double_slash_pos != std::string::npos) { 3098 // The linker has been generating bad N_SO 3099 // entries with doubled up paths 3100 // in the format "%s%s" where the first string 3101 // in the DW_AT_comp_dir, 3102 // and the second is the directory for the 3103 // source file so you end up with 3104 // a path that looks like "/tmp/src//tmp/src/" 3105 FileSpec so_dir(so_path, false); 3106 if (!so_dir.Exists()) { 3107 so_dir.SetFile( 3108 &full_so_path[double_slash_pos + 1], 3109 false); 3110 if (so_dir.Exists()) { 3111 // Trim off the incorrect path 3112 full_so_path.erase(0, 3113 double_slash_pos + 1); 3114 } 3115 } 3116 } 3117 if (*full_so_path.rbegin() != '/') 3118 full_so_path += '/'; 3119 full_so_path += symbol_name; 3120 sym[sym_idx - 1].GetMangled().SetValue( 3121 ConstString(full_so_path.c_str()), false); 3122 add_nlist = false; 3123 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3124 } 3125 } else { 3126 // This could be a relative path to a N_SO 3127 N_SO_index = sym_idx; 3128 } 3129 } 3130 break; 3131 3132 case N_OSO: 3133 // object file name: name,,0,0,st_mtime 3134 type = eSymbolTypeObjectFile; 3135 break; 3136 3137 case N_LSYM: 3138 // local sym: name,,NO_SECT,type,offset 3139 type = eSymbolTypeLocal; 3140 break; 3141 3142 //---------------------------------------------------------------------- 3143 // INCL scopes 3144 //---------------------------------------------------------------------- 3145 case N_BINCL: 3146 // include file beginning: name,,NO_SECT,0,sum 3147 // We use the current number of symbols in the symbol 3148 // table in lieu of 3149 // using nlist_idx in case we ever start trimming 3150 // entries out 3151 N_INCL_indexes.push_back(sym_idx); 3152 type = eSymbolTypeScopeBegin; 3153 break; 3154 3155 case N_EINCL: 3156 // include file end: name,,NO_SECT,0,0 3157 // Set the size of the N_BINCL to the terminating 3158 // index of this N_EINCL 3159 // so that we can always skip the entire symbol if we 3160 // need to navigate 3161 // more quickly at the source level when parsing STABS 3162 if (!N_INCL_indexes.empty()) { 3163 symbol_ptr = 3164 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3165 symbol_ptr->SetByteSize(sym_idx + 1); 3166 symbol_ptr->SetSizeIsSibling(true); 3167 N_INCL_indexes.pop_back(); 3168 } 3169 type = eSymbolTypeScopeEnd; 3170 break; 3171 3172 case N_SOL: 3173 // #included file name: name,,n_sect,0,address 3174 type = eSymbolTypeHeaderFile; 3175 3176 // We currently don't use the header files on darwin 3177 add_nlist = false; 3178 break; 3179 3180 case N_PARAMS: 3181 // compiler parameters: name,,NO_SECT,0,0 3182 type = eSymbolTypeCompiler; 3183 break; 3184 3185 case N_VERSION: 3186 // compiler version: name,,NO_SECT,0,0 3187 type = eSymbolTypeCompiler; 3188 break; 3189 3190 case N_OLEVEL: 3191 // compiler -O level: name,,NO_SECT,0,0 3192 type = eSymbolTypeCompiler; 3193 break; 3194 3195 case N_PSYM: 3196 // parameter: name,,NO_SECT,type,offset 3197 type = eSymbolTypeVariable; 3198 break; 3199 3200 case N_ENTRY: 3201 // alternate entry: name,,n_sect,linenumber,address 3202 symbol_section = section_info.GetSection( 3203 nlist.n_sect, nlist.n_value); 3204 type = eSymbolTypeLineEntry; 3205 break; 3206 3207 //---------------------------------------------------------------------- 3208 // Left and Right Braces 3209 //---------------------------------------------------------------------- 3210 case N_LBRAC: 3211 // left bracket: 0,,NO_SECT,nesting level,address 3212 // We use the current number of symbols in the symbol 3213 // table in lieu of 3214 // using nlist_idx in case we ever start trimming 3215 // entries out 3216 symbol_section = section_info.GetSection( 3217 nlist.n_sect, nlist.n_value); 3218 N_BRAC_indexes.push_back(sym_idx); 3219 type = eSymbolTypeScopeBegin; 3220 break; 3221 3222 case N_RBRAC: 3223 // right bracket: 0,,NO_SECT,nesting level,address 3224 // Set the size of the N_LBRAC to the terminating 3225 // index of this N_RBRAC 3226 // so that we can always skip the entire symbol if we 3227 // need to navigate 3228 // more quickly at the source level when parsing STABS 3229 symbol_section = section_info.GetSection( 3230 nlist.n_sect, nlist.n_value); 3231 if (!N_BRAC_indexes.empty()) { 3232 symbol_ptr = 3233 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3234 symbol_ptr->SetByteSize(sym_idx + 1); 3235 symbol_ptr->SetSizeIsSibling(true); 3236 N_BRAC_indexes.pop_back(); 3237 } 3238 type = eSymbolTypeScopeEnd; 3239 break; 3240 3241 case N_EXCL: 3242 // deleted include file: name,,NO_SECT,0,sum 3243 type = eSymbolTypeHeaderFile; 3244 break; 3245 3246 //---------------------------------------------------------------------- 3247 // COMM scopes 3248 //---------------------------------------------------------------------- 3249 case N_BCOMM: 3250 // begin common: name,,NO_SECT,0,0 3251 // We use the current number of symbols in the symbol 3252 // table in lieu of 3253 // using nlist_idx in case we ever start trimming 3254 // entries out 3255 type = eSymbolTypeScopeBegin; 3256 N_COMM_indexes.push_back(sym_idx); 3257 break; 3258 3259 case N_ECOML: 3260 // end common (local name): 0,,n_sect,0,address 3261 symbol_section = section_info.GetSection( 3262 nlist.n_sect, nlist.n_value); 3263 // Fall through 3264 3265 case N_ECOMM: 3266 // end common: name,,n_sect,0,0 3267 // Set the size of the N_BCOMM to the terminating 3268 // index of this N_ECOMM/N_ECOML 3269 // so that we can always skip the entire symbol if we 3270 // need to navigate 3271 // more quickly at the source level when parsing STABS 3272 if (!N_COMM_indexes.empty()) { 3273 symbol_ptr = 3274 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3275 symbol_ptr->SetByteSize(sym_idx + 1); 3276 symbol_ptr->SetSizeIsSibling(true); 3277 N_COMM_indexes.pop_back(); 3278 } 3279 type = eSymbolTypeScopeEnd; 3280 break; 3281 3282 case N_LENG: 3283 // second stab entry with length information 3284 type = eSymbolTypeAdditional; 3285 break; 3286 3287 default: 3288 break; 3289 } 3290 } else { 3291 // uint8_t n_pext = N_PEXT & nlist.n_type; 3292 uint8_t n_type = N_TYPE & nlist.n_type; 3293 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3294 3295 switch (n_type) { 3296 case N_INDR: { 3297 const char *reexport_name_cstr = 3298 strtab_data.PeekCStr(nlist.n_value); 3299 if (reexport_name_cstr && reexport_name_cstr[0]) { 3300 type = eSymbolTypeReExported; 3301 ConstString reexport_name( 3302 reexport_name_cstr + 3303 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3304 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3305 set_value = false; 3306 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3307 indirect_symbol_names.insert( 3308 ConstString(symbol_name + 3309 ((symbol_name[0] == '_') ? 1 : 0))); 3310 } else 3311 type = eSymbolTypeUndefined; 3312 } break; 3313 3314 case N_UNDF: 3315 if (symbol_name && symbol_name[0]) { 3316 ConstString undefined_name( 3317 symbol_name + 3318 ((symbol_name[0] == '_') ? 1 : 0)); 3319 undefined_name_to_desc[undefined_name] = 3320 nlist.n_desc; 3321 } 3322 // Fall through 3323 case N_PBUD: 3324 type = eSymbolTypeUndefined; 3325 break; 3326 3327 case N_ABS: 3328 type = eSymbolTypeAbsolute; 3329 break; 3330 3331 case N_SECT: { 3332 symbol_section = section_info.GetSection( 3333 nlist.n_sect, nlist.n_value); 3334 3335 if (symbol_section == NULL) { 3336 // TODO: warn about this? 3337 add_nlist = false; 3338 break; 3339 } 3340 3341 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3342 type = eSymbolTypeException; 3343 } else { 3344 uint32_t section_type = 3345 symbol_section->Get() & SECTION_TYPE; 3346 3347 switch (section_type) { 3348 case S_CSTRING_LITERALS: 3349 type = eSymbolTypeData; 3350 break; // section with only literal C strings 3351 case S_4BYTE_LITERALS: 3352 type = eSymbolTypeData; 3353 break; // section with only 4 byte literals 3354 case S_8BYTE_LITERALS: 3355 type = eSymbolTypeData; 3356 break; // section with only 8 byte literals 3357 case S_LITERAL_POINTERS: 3358 type = eSymbolTypeTrampoline; 3359 break; // section with only pointers to literals 3360 case S_NON_LAZY_SYMBOL_POINTERS: 3361 type = eSymbolTypeTrampoline; 3362 break; // section with only non-lazy symbol 3363 // pointers 3364 case S_LAZY_SYMBOL_POINTERS: 3365 type = eSymbolTypeTrampoline; 3366 break; // section with only lazy symbol pointers 3367 case S_SYMBOL_STUBS: 3368 type = eSymbolTypeTrampoline; 3369 break; // section with only symbol stubs, byte 3370 // size of stub in the reserved2 field 3371 case S_MOD_INIT_FUNC_POINTERS: 3372 type = eSymbolTypeCode; 3373 break; // section with only function pointers for 3374 // initialization 3375 case S_MOD_TERM_FUNC_POINTERS: 3376 type = eSymbolTypeCode; 3377 break; // section with only function pointers for 3378 // termination 3379 case S_INTERPOSING: 3380 type = eSymbolTypeTrampoline; 3381 break; // section with only pairs of function 3382 // pointers for interposing 3383 case S_16BYTE_LITERALS: 3384 type = eSymbolTypeData; 3385 break; // section with only 16 byte literals 3386 case S_DTRACE_DOF: 3387 type = eSymbolTypeInstrumentation; 3388 break; 3389 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3390 type = eSymbolTypeTrampoline; 3391 break; 3392 default: 3393 switch (symbol_section->GetType()) { 3394 case lldb::eSectionTypeCode: 3395 type = eSymbolTypeCode; 3396 break; 3397 case eSectionTypeData: 3398 case eSectionTypeDataCString: // Inlined C string 3399 // data 3400 case eSectionTypeDataCStringPointers: // Pointers 3401 // to C 3402 // string 3403 // data 3404 case eSectionTypeDataSymbolAddress: // Address of 3405 // a symbol in 3406 // the symbol 3407 // table 3408 case eSectionTypeData4: 3409 case eSectionTypeData8: 3410 case eSectionTypeData16: 3411 type = eSymbolTypeData; 3412 break; 3413 default: 3414 break; 3415 } 3416 break; 3417 } 3418 3419 if (type == eSymbolTypeInvalid) { 3420 const char *symbol_sect_name = 3421 symbol_section->GetName().AsCString(); 3422 if (symbol_section->IsDescendant( 3423 text_section_sp.get())) { 3424 if (symbol_section->IsClear( 3425 S_ATTR_PURE_INSTRUCTIONS | 3426 S_ATTR_SELF_MODIFYING_CODE | 3427 S_ATTR_SOME_INSTRUCTIONS)) 3428 type = eSymbolTypeData; 3429 else 3430 type = eSymbolTypeCode; 3431 } else if (symbol_section->IsDescendant( 3432 data_section_sp.get()) || 3433 symbol_section->IsDescendant( 3434 data_dirty_section_sp.get()) || 3435 symbol_section->IsDescendant( 3436 data_const_section_sp.get())) { 3437 if (symbol_sect_name && 3438 ::strstr(symbol_sect_name, "__objc") == 3439 symbol_sect_name) { 3440 type = eSymbolTypeRuntime; 3441 3442 if (symbol_name) { 3443 llvm::StringRef symbol_name_ref( 3444 symbol_name); 3445 if (symbol_name_ref.startswith("_OBJC_")) { 3446 static const llvm::StringRef 3447 g_objc_v2_prefix_class( 3448 "_OBJC_CLASS_$_"); 3449 static const llvm::StringRef 3450 g_objc_v2_prefix_metaclass( 3451 "_OBJC_METACLASS_$_"); 3452 static const llvm::StringRef 3453 g_objc_v2_prefix_ivar( 3454 "_OBJC_IVAR_$_"); 3455 if (symbol_name_ref.startswith( 3456 g_objc_v2_prefix_class)) { 3457 symbol_name_non_abi_mangled = 3458 symbol_name + 1; 3459 symbol_name = 3460 symbol_name + 3461 g_objc_v2_prefix_class.size(); 3462 type = eSymbolTypeObjCClass; 3463 demangled_is_synthesized = true; 3464 } else if ( 3465 symbol_name_ref.startswith( 3466 g_objc_v2_prefix_metaclass)) { 3467 symbol_name_non_abi_mangled = 3468 symbol_name + 1; 3469 symbol_name = 3470 symbol_name + 3471 g_objc_v2_prefix_metaclass.size(); 3472 type = eSymbolTypeObjCMetaClass; 3473 demangled_is_synthesized = true; 3474 } else if (symbol_name_ref.startswith( 3475 g_objc_v2_prefix_ivar)) { 3476 symbol_name_non_abi_mangled = 3477 symbol_name + 1; 3478 symbol_name = 3479 symbol_name + 3480 g_objc_v2_prefix_ivar.size(); 3481 type = eSymbolTypeObjCIVar; 3482 demangled_is_synthesized = true; 3483 } 3484 } 3485 } 3486 } else if (symbol_sect_name && 3487 ::strstr(symbol_sect_name, 3488 "__gcc_except_tab") == 3489 symbol_sect_name) { 3490 type = eSymbolTypeException; 3491 } else { 3492 type = eSymbolTypeData; 3493 } 3494 } else if (symbol_sect_name && 3495 ::strstr(symbol_sect_name, 3496 "__IMPORT") == 3497 symbol_sect_name) { 3498 type = eSymbolTypeTrampoline; 3499 } else if (symbol_section->IsDescendant( 3500 objc_section_sp.get())) { 3501 type = eSymbolTypeRuntime; 3502 if (symbol_name && symbol_name[0] == '.') { 3503 llvm::StringRef symbol_name_ref(symbol_name); 3504 static const llvm::StringRef 3505 g_objc_v1_prefix_class( 3506 ".objc_class_name_"); 3507 if (symbol_name_ref.startswith( 3508 g_objc_v1_prefix_class)) { 3509 symbol_name_non_abi_mangled = symbol_name; 3510 symbol_name = symbol_name + 3511 g_objc_v1_prefix_class.size(); 3512 type = eSymbolTypeObjCClass; 3513 demangled_is_synthesized = true; 3514 } 3515 } 3516 } 3517 } 3518 } 3519 } break; 3520 } 3521 } 3522 3523 if (add_nlist) { 3524 uint64_t symbol_value = nlist.n_value; 3525 if (symbol_name_non_abi_mangled) { 3526 sym[sym_idx].GetMangled().SetMangledName( 3527 ConstString(symbol_name_non_abi_mangled)); 3528 sym[sym_idx].GetMangled().SetDemangledName( 3529 ConstString(symbol_name)); 3530 } else { 3531 bool symbol_name_is_mangled = false; 3532 3533 if (symbol_name && symbol_name[0] == '_') { 3534 symbol_name_is_mangled = symbol_name[1] == '_'; 3535 symbol_name++; // Skip the leading underscore 3536 } 3537 3538 if (symbol_name) { 3539 ConstString const_symbol_name(symbol_name); 3540 sym[sym_idx].GetMangled().SetValue( 3541 const_symbol_name, symbol_name_is_mangled); 3542 if (is_gsym && is_debug) { 3543 const char *gsym_name = 3544 sym[sym_idx] 3545 .GetMangled() 3546 .GetName(lldb::eLanguageTypeUnknown, 3547 Mangled::ePreferMangled) 3548 .GetCString(); 3549 if (gsym_name) 3550 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3551 } 3552 } 3553 } 3554 if (symbol_section) { 3555 const addr_t section_file_addr = 3556 symbol_section->GetFileAddress(); 3557 if (symbol_byte_size == 0 && 3558 function_starts_count > 0) { 3559 addr_t symbol_lookup_file_addr = nlist.n_value; 3560 // Do an exact address match for non-ARM addresses, 3561 // else get the closest since 3562 // the symbol might be a thumb symbol which has an 3563 // address with bit zero set 3564 FunctionStarts::Entry *func_start_entry = 3565 function_starts.FindEntry( 3566 symbol_lookup_file_addr, !is_arm); 3567 if (is_arm && func_start_entry) { 3568 // Verify that the function start address is the 3569 // symbol address (ARM) 3570 // or the symbol address + 1 (thumb) 3571 if (func_start_entry->addr != 3572 symbol_lookup_file_addr && 3573 func_start_entry->addr != 3574 (symbol_lookup_file_addr + 1)) { 3575 // Not the right entry, NULL it out... 3576 func_start_entry = NULL; 3577 } 3578 } 3579 if (func_start_entry) { 3580 func_start_entry->data = true; 3581 3582 addr_t symbol_file_addr = func_start_entry->addr; 3583 uint32_t symbol_flags = 0; 3584 if (is_arm) { 3585 if (symbol_file_addr & 1) 3586 symbol_flags = 3587 MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3588 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3589 } 3590 3591 const FunctionStarts::Entry 3592 *next_func_start_entry = 3593 function_starts.FindNextEntry( 3594 func_start_entry); 3595 const addr_t section_end_file_addr = 3596 section_file_addr + 3597 symbol_section->GetByteSize(); 3598 if (next_func_start_entry) { 3599 addr_t next_symbol_file_addr = 3600 next_func_start_entry->addr; 3601 // Be sure the clear the Thumb address bit when 3602 // we calculate the size 3603 // from the current and next address 3604 if (is_arm) 3605 next_symbol_file_addr &= 3606 THUMB_ADDRESS_BIT_MASK; 3607 symbol_byte_size = std::min<lldb::addr_t>( 3608 next_symbol_file_addr - symbol_file_addr, 3609 section_end_file_addr - symbol_file_addr); 3610 } else { 3611 symbol_byte_size = 3612 section_end_file_addr - symbol_file_addr; 3613 } 3614 } 3615 } 3616 symbol_value -= section_file_addr; 3617 } 3618 3619 if (is_debug == false) { 3620 if (type == eSymbolTypeCode) { 3621 // See if we can find a N_FUN entry for any code 3622 // symbols. 3623 // If we do find a match, and the name matches, then 3624 // we 3625 // can merge the two into just the function symbol 3626 // to avoid 3627 // duplicate entries in the symbol table 3628 std::pair<ValueToSymbolIndexMap::const_iterator, 3629 ValueToSymbolIndexMap::const_iterator> 3630 range; 3631 range = N_FUN_addr_to_sym_idx.equal_range( 3632 nlist.n_value); 3633 if (range.first != range.second) { 3634 bool found_it = false; 3635 for (ValueToSymbolIndexMap::const_iterator pos = 3636 range.first; 3637 pos != range.second; ++pos) { 3638 if (sym[sym_idx].GetMangled().GetName( 3639 lldb::eLanguageTypeUnknown, 3640 Mangled::ePreferMangled) == 3641 sym[pos->second].GetMangled().GetName( 3642 lldb::eLanguageTypeUnknown, 3643 Mangled::ePreferMangled)) { 3644 m_nlist_idx_to_sym_idx[nlist_idx] = 3645 pos->second; 3646 // We just need the flags from the linker 3647 // symbol, so put these flags 3648 // into the N_FUN flags to avoid duplicate 3649 // symbols in the symbol table 3650 sym[pos->second].SetExternal( 3651 sym[sym_idx].IsExternal()); 3652 sym[pos->second].SetFlags(nlist.n_type << 16 | 3653 nlist.n_desc); 3654 if (resolver_addresses.find(nlist.n_value) != 3655 resolver_addresses.end()) 3656 sym[pos->second].SetType( 3657 eSymbolTypeResolver); 3658 sym[sym_idx].Clear(); 3659 found_it = true; 3660 break; 3661 } 3662 } 3663 if (found_it) 3664 continue; 3665 } else { 3666 if (resolver_addresses.find(nlist.n_value) != 3667 resolver_addresses.end()) 3668 type = eSymbolTypeResolver; 3669 } 3670 } else if (type == eSymbolTypeData || 3671 type == eSymbolTypeObjCClass || 3672 type == eSymbolTypeObjCMetaClass || 3673 type == eSymbolTypeObjCIVar) { 3674 // See if we can find a N_STSYM entry for any data 3675 // symbols. 3676 // If we do find a match, and the name matches, then 3677 // we 3678 // can merge the two into just the Static symbol to 3679 // avoid 3680 // duplicate entries in the symbol table 3681 std::pair<ValueToSymbolIndexMap::const_iterator, 3682 ValueToSymbolIndexMap::const_iterator> 3683 range; 3684 range = N_STSYM_addr_to_sym_idx.equal_range( 3685 nlist.n_value); 3686 if (range.first != range.second) { 3687 bool found_it = false; 3688 for (ValueToSymbolIndexMap::const_iterator pos = 3689 range.first; 3690 pos != range.second; ++pos) { 3691 if (sym[sym_idx].GetMangled().GetName( 3692 lldb::eLanguageTypeUnknown, 3693 Mangled::ePreferMangled) == 3694 sym[pos->second].GetMangled().GetName( 3695 lldb::eLanguageTypeUnknown, 3696 Mangled::ePreferMangled)) { 3697 m_nlist_idx_to_sym_idx[nlist_idx] = 3698 pos->second; 3699 // We just need the flags from the linker 3700 // symbol, so put these flags 3701 // into the N_STSYM flags to avoid duplicate 3702 // symbols in the symbol table 3703 sym[pos->second].SetExternal( 3704 sym[sym_idx].IsExternal()); 3705 sym[pos->second].SetFlags(nlist.n_type << 16 | 3706 nlist.n_desc); 3707 sym[sym_idx].Clear(); 3708 found_it = true; 3709 break; 3710 } 3711 } 3712 if (found_it) 3713 continue; 3714 } else { 3715 const char *gsym_name = 3716 sym[sym_idx] 3717 .GetMangled() 3718 .GetName(lldb::eLanguageTypeUnknown, 3719 Mangled::ePreferMangled) 3720 .GetCString(); 3721 if (gsym_name) { 3722 // Combine N_GSYM stab entries with the non stab 3723 // symbol 3724 ConstNameToSymbolIndexMap::const_iterator pos = 3725 N_GSYM_name_to_sym_idx.find(gsym_name); 3726 if (pos != N_GSYM_name_to_sym_idx.end()) { 3727 const uint32_t GSYM_sym_idx = pos->second; 3728 m_nlist_idx_to_sym_idx[nlist_idx] = 3729 GSYM_sym_idx; 3730 // Copy the address, because often the N_GSYM 3731 // address has an invalid address of zero 3732 // when the global is a common symbol 3733 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3734 symbol_section); 3735 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3736 symbol_value); 3737 // We just need the flags from the linker 3738 // symbol, so put these flags 3739 // into the N_GSYM flags to avoid duplicate 3740 // symbols in the symbol table 3741 sym[GSYM_sym_idx].SetFlags( 3742 nlist.n_type << 16 | nlist.n_desc); 3743 sym[sym_idx].Clear(); 3744 continue; 3745 } 3746 } 3747 } 3748 } 3749 } 3750 3751 sym[sym_idx].SetID(nlist_idx); 3752 sym[sym_idx].SetType(type); 3753 if (set_value) { 3754 sym[sym_idx].GetAddressRef().SetSection( 3755 symbol_section); 3756 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3757 } 3758 sym[sym_idx].SetFlags(nlist.n_type << 16 | 3759 nlist.n_desc); 3760 3761 if (symbol_byte_size > 0) 3762 sym[sym_idx].SetByteSize(symbol_byte_size); 3763 3764 if (demangled_is_synthesized) 3765 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3766 ++sym_idx; 3767 } else { 3768 sym[sym_idx].Clear(); 3769 } 3770 } 3771 ///////////////////////////// 3772 } 3773 break; // No more entries to consider 3774 } 3775 } 3776 3777 for (const auto &pos : reexport_shlib_needs_fixup) { 3778 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3779 if (undef_pos != undefined_name_to_desc.end()) { 3780 const uint8_t dylib_ordinal = 3781 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3782 if (dylib_ordinal > 0 && 3783 dylib_ordinal < dylib_files.GetSize()) 3784 sym[pos.first].SetReExportedSymbolSharedLibrary( 3785 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3786 } 3787 } 3788 } 3789 } 3790 } 3791 } 3792 } 3793 3794 // Must reset this in case it was mutated above! 3795 nlist_data_offset = 0; 3796 #endif 3797 3798 if (nlist_data.GetByteSize() > 0) { 3799 3800 // If the sym array was not created while parsing the DSC unmapped 3801 // symbols, create it now. 3802 if (sym == NULL) { 3803 sym = symtab->Resize(symtab_load_command.nsyms + 3804 m_dysymtab.nindirectsyms); 3805 num_syms = symtab->GetNumSymbols(); 3806 } 3807 3808 if (unmapped_local_symbols_found) { 3809 assert(m_dysymtab.ilocalsym == 0); 3810 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3811 nlist_idx = m_dysymtab.nlocalsym; 3812 } else { 3813 nlist_idx = 0; 3814 } 3815 3816 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3817 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3818 UndefinedNameToDescMap undefined_name_to_desc; 3819 SymbolIndexToName reexport_shlib_needs_fixup; 3820 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 3821 struct nlist_64 nlist; 3822 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, 3823 nlist_byte_size)) 3824 break; 3825 3826 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3827 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 3828 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 3829 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 3830 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 3831 3832 SymbolType type = eSymbolTypeInvalid; 3833 const char *symbol_name = NULL; 3834 3835 if (have_strtab_data) { 3836 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3837 3838 if (symbol_name == NULL) { 3839 // No symbol should be NULL, even the symbols with no 3840 // string values should have an offset zero which points 3841 // to an empty C-string 3842 Host::SystemLog(Host::eSystemLogError, 3843 "error: symbol[%u] has invalid string table offset " 3844 "0x%x in %s, ignoring symbol\n", 3845 nlist_idx, nlist.n_strx, 3846 module_sp->GetFileSpec().GetPath().c_str()); 3847 continue; 3848 } 3849 if (symbol_name[0] == '\0') 3850 symbol_name = NULL; 3851 } else { 3852 const addr_t str_addr = strtab_addr + nlist.n_strx; 3853 Status str_error; 3854 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3855 str_error)) 3856 symbol_name = memory_symbol_name.c_str(); 3857 } 3858 const char *symbol_name_non_abi_mangled = NULL; 3859 3860 SectionSP symbol_section; 3861 lldb::addr_t symbol_byte_size = 0; 3862 bool add_nlist = true; 3863 bool is_gsym = false; 3864 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3865 bool demangled_is_synthesized = false; 3866 bool set_value = true; 3867 assert(sym_idx < num_syms); 3868 3869 sym[sym_idx].SetDebug(is_debug); 3870 3871 if (is_debug) { 3872 switch (nlist.n_type) { 3873 case N_GSYM: 3874 // global symbol: name,,NO_SECT,type,0 3875 // Sometimes the N_GSYM value contains the address. 3876 3877 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3878 // the ObjC data. They 3879 // have the same address, but we want to ensure that we always find 3880 // only the real symbol, 3881 // 'cause we don't currently correctly attribute the GSYM one to the 3882 // ObjCClass/Ivar/MetaClass 3883 // symbol type. This is a temporary hack to make sure the 3884 // ObjectiveC symbols get treated 3885 // correctly. To do this right, we should coalesce all the GSYM & 3886 // global symbols that have the 3887 // same address. 3888 is_gsym = true; 3889 sym[sym_idx].SetExternal(true); 3890 3891 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3892 llvm::StringRef symbol_name_ref(symbol_name); 3893 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3894 symbol_name_non_abi_mangled = symbol_name + 1; 3895 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3896 type = eSymbolTypeObjCClass; 3897 demangled_is_synthesized = true; 3898 3899 } else if (symbol_name_ref.startswith( 3900 g_objc_v2_prefix_metaclass)) { 3901 symbol_name_non_abi_mangled = symbol_name + 1; 3902 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3903 type = eSymbolTypeObjCMetaClass; 3904 demangled_is_synthesized = true; 3905 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3906 symbol_name_non_abi_mangled = symbol_name + 1; 3907 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3908 type = eSymbolTypeObjCIVar; 3909 demangled_is_synthesized = true; 3910 } 3911 } else { 3912 if (nlist.n_value != 0) 3913 symbol_section = 3914 section_info.GetSection(nlist.n_sect, nlist.n_value); 3915 type = eSymbolTypeData; 3916 } 3917 break; 3918 3919 case N_FNAME: 3920 // procedure name (f77 kludge): name,,NO_SECT,0,0 3921 type = eSymbolTypeCompiler; 3922 break; 3923 3924 case N_FUN: 3925 // procedure: name,,n_sect,linenumber,address 3926 if (symbol_name) { 3927 type = eSymbolTypeCode; 3928 symbol_section = 3929 section_info.GetSection(nlist.n_sect, nlist.n_value); 3930 3931 N_FUN_addr_to_sym_idx.insert( 3932 std::make_pair(nlist.n_value, sym_idx)); 3933 // We use the current number of symbols in the symbol table in 3934 // lieu of 3935 // using nlist_idx in case we ever start trimming entries out 3936 N_FUN_indexes.push_back(sym_idx); 3937 } else { 3938 type = eSymbolTypeCompiler; 3939 3940 if (!N_FUN_indexes.empty()) { 3941 // Copy the size of the function into the original STAB entry so 3942 // we don't have 3943 // to hunt for it later 3944 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3945 ->SetByteSize(nlist.n_value); 3946 N_FUN_indexes.pop_back(); 3947 // We don't really need the end function STAB as it contains the 3948 // size which 3949 // we already placed with the original symbol, so don't add it 3950 // if we want a 3951 // minimal symbol table 3952 add_nlist = false; 3953 } 3954 } 3955 break; 3956 3957 case N_STSYM: 3958 // static symbol: name,,n_sect,type,address 3959 N_STSYM_addr_to_sym_idx.insert( 3960 std::make_pair(nlist.n_value, sym_idx)); 3961 symbol_section = 3962 section_info.GetSection(nlist.n_sect, nlist.n_value); 3963 if (symbol_name && symbol_name[0]) { 3964 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3965 eSymbolTypeData); 3966 } 3967 break; 3968 3969 case N_LCSYM: 3970 // .lcomm symbol: name,,n_sect,type,address 3971 symbol_section = 3972 section_info.GetSection(nlist.n_sect, nlist.n_value); 3973 type = eSymbolTypeCommonBlock; 3974 break; 3975 3976 case N_BNSYM: 3977 // We use the current number of symbols in the symbol table in lieu 3978 // of 3979 // using nlist_idx in case we ever start trimming entries out 3980 // Skip these if we want minimal symbol tables 3981 add_nlist = false; 3982 break; 3983 3984 case N_ENSYM: 3985 // Set the size of the N_BNSYM to the terminating index of this 3986 // N_ENSYM 3987 // so that we can always skip the entire symbol if we need to 3988 // navigate 3989 // more quickly at the source level when parsing STABS 3990 // Skip these if we want minimal symbol tables 3991 add_nlist = false; 3992 break; 3993 3994 case N_OPT: 3995 // emitted with gcc2_compiled and in gcc source 3996 type = eSymbolTypeCompiler; 3997 break; 3998 3999 case N_RSYM: 4000 // register sym: name,,NO_SECT,type,register 4001 type = eSymbolTypeVariable; 4002 break; 4003 4004 case N_SLINE: 4005 // src line: 0,,n_sect,linenumber,address 4006 symbol_section = 4007 section_info.GetSection(nlist.n_sect, nlist.n_value); 4008 type = eSymbolTypeLineEntry; 4009 break; 4010 4011 case N_SSYM: 4012 // structure elt: name,,NO_SECT,type,struct_offset 4013 type = eSymbolTypeVariableType; 4014 break; 4015 4016 case N_SO: 4017 // source file name 4018 type = eSymbolTypeSourceFile; 4019 if (symbol_name == NULL) { 4020 add_nlist = false; 4021 if (N_SO_index != UINT32_MAX) { 4022 // Set the size of the N_SO to the terminating index of this 4023 // N_SO 4024 // so that we can always skip the entire N_SO if we need to 4025 // navigate 4026 // more quickly at the source level when parsing STABS 4027 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 4028 symbol_ptr->SetByteSize(sym_idx); 4029 symbol_ptr->SetSizeIsSibling(true); 4030 } 4031 N_NSYM_indexes.clear(); 4032 N_INCL_indexes.clear(); 4033 N_BRAC_indexes.clear(); 4034 N_COMM_indexes.clear(); 4035 N_FUN_indexes.clear(); 4036 N_SO_index = UINT32_MAX; 4037 } else { 4038 // We use the current number of symbols in the symbol table in 4039 // lieu of 4040 // using nlist_idx in case we ever start trimming entries out 4041 const bool N_SO_has_full_path = symbol_name[0] == '/'; 4042 if (N_SO_has_full_path) { 4043 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 4044 // We have two consecutive N_SO entries where the first 4045 // contains a directory 4046 // and the second contains a full path. 4047 sym[sym_idx - 1].GetMangled().SetValue( 4048 ConstString(symbol_name), false); 4049 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4050 add_nlist = false; 4051 } else { 4052 // This is the first entry in a N_SO that contains a directory 4053 // or 4054 // a full path to the source file 4055 N_SO_index = sym_idx; 4056 } 4057 } else if ((N_SO_index == sym_idx - 1) && 4058 ((sym_idx - 1) < num_syms)) { 4059 // This is usually the second N_SO entry that contains just the 4060 // filename, 4061 // so here we combine it with the first one if we are minimizing 4062 // the symbol table 4063 const char *so_path = 4064 sym[sym_idx - 1] 4065 .GetMangled() 4066 .GetDemangledName(lldb::eLanguageTypeUnknown) 4067 .AsCString(); 4068 if (so_path && so_path[0]) { 4069 std::string full_so_path(so_path); 4070 const size_t double_slash_pos = full_so_path.find("//"); 4071 if (double_slash_pos != std::string::npos) { 4072 // The linker has been generating bad N_SO entries with 4073 // doubled up paths 4074 // in the format "%s%s" where the first string in the 4075 // DW_AT_comp_dir, 4076 // and the second is the directory for the source file so 4077 // you end up with 4078 // a path that looks like "/tmp/src//tmp/src/" 4079 FileSpec so_dir(so_path, false); 4080 if (!so_dir.Exists()) { 4081 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 4082 false); 4083 if (so_dir.Exists()) { 4084 // Trim off the incorrect path 4085 full_so_path.erase(0, double_slash_pos + 1); 4086 } 4087 } 4088 } 4089 if (*full_so_path.rbegin() != '/') 4090 full_so_path += '/'; 4091 full_so_path += symbol_name; 4092 sym[sym_idx - 1].GetMangled().SetValue( 4093 ConstString(full_so_path.c_str()), false); 4094 add_nlist = false; 4095 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4096 } 4097 } else { 4098 // This could be a relative path to a N_SO 4099 N_SO_index = sym_idx; 4100 } 4101 } 4102 break; 4103 4104 case N_OSO: 4105 // object file name: name,,0,0,st_mtime 4106 type = eSymbolTypeObjectFile; 4107 break; 4108 4109 case N_LSYM: 4110 // local sym: name,,NO_SECT,type,offset 4111 type = eSymbolTypeLocal; 4112 break; 4113 4114 //---------------------------------------------------------------------- 4115 // INCL scopes 4116 //---------------------------------------------------------------------- 4117 case N_BINCL: 4118 // include file beginning: name,,NO_SECT,0,sum 4119 // We use the current number of symbols in the symbol table in lieu 4120 // of 4121 // using nlist_idx in case we ever start trimming entries out 4122 N_INCL_indexes.push_back(sym_idx); 4123 type = eSymbolTypeScopeBegin; 4124 break; 4125 4126 case N_EINCL: 4127 // include file end: name,,NO_SECT,0,0 4128 // Set the size of the N_BINCL to the terminating index of this 4129 // N_EINCL 4130 // so that we can always skip the entire symbol if we need to 4131 // navigate 4132 // more quickly at the source level when parsing STABS 4133 if (!N_INCL_indexes.empty()) { 4134 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4135 symbol_ptr->SetByteSize(sym_idx + 1); 4136 symbol_ptr->SetSizeIsSibling(true); 4137 N_INCL_indexes.pop_back(); 4138 } 4139 type = eSymbolTypeScopeEnd; 4140 break; 4141 4142 case N_SOL: 4143 // #included file name: name,,n_sect,0,address 4144 type = eSymbolTypeHeaderFile; 4145 4146 // We currently don't use the header files on darwin 4147 add_nlist = false; 4148 break; 4149 4150 case N_PARAMS: 4151 // compiler parameters: name,,NO_SECT,0,0 4152 type = eSymbolTypeCompiler; 4153 break; 4154 4155 case N_VERSION: 4156 // compiler version: name,,NO_SECT,0,0 4157 type = eSymbolTypeCompiler; 4158 break; 4159 4160 case N_OLEVEL: 4161 // compiler -O level: name,,NO_SECT,0,0 4162 type = eSymbolTypeCompiler; 4163 break; 4164 4165 case N_PSYM: 4166 // parameter: name,,NO_SECT,type,offset 4167 type = eSymbolTypeVariable; 4168 break; 4169 4170 case N_ENTRY: 4171 // alternate entry: name,,n_sect,linenumber,address 4172 symbol_section = 4173 section_info.GetSection(nlist.n_sect, nlist.n_value); 4174 type = eSymbolTypeLineEntry; 4175 break; 4176 4177 //---------------------------------------------------------------------- 4178 // Left and Right Braces 4179 //---------------------------------------------------------------------- 4180 case N_LBRAC: 4181 // left bracket: 0,,NO_SECT,nesting level,address 4182 // We use the current number of symbols in the symbol table in lieu 4183 // of 4184 // using nlist_idx in case we ever start trimming entries out 4185 symbol_section = 4186 section_info.GetSection(nlist.n_sect, nlist.n_value); 4187 N_BRAC_indexes.push_back(sym_idx); 4188 type = eSymbolTypeScopeBegin; 4189 break; 4190 4191 case N_RBRAC: 4192 // right bracket: 0,,NO_SECT,nesting level,address 4193 // Set the size of the N_LBRAC to the terminating index of this 4194 // N_RBRAC 4195 // so that we can always skip the entire symbol if we need to 4196 // navigate 4197 // more quickly at the source level when parsing STABS 4198 symbol_section = 4199 section_info.GetSection(nlist.n_sect, nlist.n_value); 4200 if (!N_BRAC_indexes.empty()) { 4201 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4202 symbol_ptr->SetByteSize(sym_idx + 1); 4203 symbol_ptr->SetSizeIsSibling(true); 4204 N_BRAC_indexes.pop_back(); 4205 } 4206 type = eSymbolTypeScopeEnd; 4207 break; 4208 4209 case N_EXCL: 4210 // deleted include file: name,,NO_SECT,0,sum 4211 type = eSymbolTypeHeaderFile; 4212 break; 4213 4214 //---------------------------------------------------------------------- 4215 // COMM scopes 4216 //---------------------------------------------------------------------- 4217 case N_BCOMM: 4218 // begin common: name,,NO_SECT,0,0 4219 // We use the current number of symbols in the symbol table in lieu 4220 // of 4221 // using nlist_idx in case we ever start trimming entries out 4222 type = eSymbolTypeScopeBegin; 4223 N_COMM_indexes.push_back(sym_idx); 4224 break; 4225 4226 case N_ECOML: 4227 // end common (local name): 0,,n_sect,0,address 4228 symbol_section = 4229 section_info.GetSection(nlist.n_sect, nlist.n_value); 4230 LLVM_FALLTHROUGH; 4231 4232 case N_ECOMM: 4233 // end common: name,,n_sect,0,0 4234 // Set the size of the N_BCOMM to the terminating index of this 4235 // N_ECOMM/N_ECOML 4236 // so that we can always skip the entire symbol if we need to 4237 // navigate 4238 // more quickly at the source level when parsing STABS 4239 if (!N_COMM_indexes.empty()) { 4240 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4241 symbol_ptr->SetByteSize(sym_idx + 1); 4242 symbol_ptr->SetSizeIsSibling(true); 4243 N_COMM_indexes.pop_back(); 4244 } 4245 type = eSymbolTypeScopeEnd; 4246 break; 4247 4248 case N_LENG: 4249 // second stab entry with length information 4250 type = eSymbolTypeAdditional; 4251 break; 4252 4253 default: 4254 break; 4255 } 4256 } else { 4257 // uint8_t n_pext = N_PEXT & nlist.n_type; 4258 uint8_t n_type = N_TYPE & nlist.n_type; 4259 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4260 4261 switch (n_type) { 4262 case N_INDR: { 4263 const char *reexport_name_cstr = 4264 strtab_data.PeekCStr(nlist.n_value); 4265 if (reexport_name_cstr && reexport_name_cstr[0]) { 4266 type = eSymbolTypeReExported; 4267 ConstString reexport_name( 4268 reexport_name_cstr + 4269 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4270 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4271 set_value = false; 4272 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4273 indirect_symbol_names.insert( 4274 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4275 } else 4276 type = eSymbolTypeUndefined; 4277 } break; 4278 4279 case N_UNDF: 4280 if (symbol_name && symbol_name[0]) { 4281 ConstString undefined_name(symbol_name + 4282 ((symbol_name[0] == '_') ? 1 : 0)); 4283 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4284 } 4285 LLVM_FALLTHROUGH; 4286 4287 case N_PBUD: 4288 type = eSymbolTypeUndefined; 4289 break; 4290 4291 case N_ABS: 4292 type = eSymbolTypeAbsolute; 4293 break; 4294 4295 case N_SECT: { 4296 symbol_section = 4297 section_info.GetSection(nlist.n_sect, nlist.n_value); 4298 4299 if (!symbol_section) { 4300 // TODO: warn about this? 4301 add_nlist = false; 4302 break; 4303 } 4304 4305 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4306 type = eSymbolTypeException; 4307 } else { 4308 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4309 4310 switch (section_type) { 4311 case S_CSTRING_LITERALS: 4312 type = eSymbolTypeData; 4313 break; // section with only literal C strings 4314 case S_4BYTE_LITERALS: 4315 type = eSymbolTypeData; 4316 break; // section with only 4 byte literals 4317 case S_8BYTE_LITERALS: 4318 type = eSymbolTypeData; 4319 break; // section with only 8 byte literals 4320 case S_LITERAL_POINTERS: 4321 type = eSymbolTypeTrampoline; 4322 break; // section with only pointers to literals 4323 case S_NON_LAZY_SYMBOL_POINTERS: 4324 type = eSymbolTypeTrampoline; 4325 break; // section with only non-lazy symbol pointers 4326 case S_LAZY_SYMBOL_POINTERS: 4327 type = eSymbolTypeTrampoline; 4328 break; // section with only lazy symbol pointers 4329 case S_SYMBOL_STUBS: 4330 type = eSymbolTypeTrampoline; 4331 break; // section with only symbol stubs, byte size of stub in 4332 // the reserved2 field 4333 case S_MOD_INIT_FUNC_POINTERS: 4334 type = eSymbolTypeCode; 4335 break; // section with only function pointers for initialization 4336 case S_MOD_TERM_FUNC_POINTERS: 4337 type = eSymbolTypeCode; 4338 break; // section with only function pointers for termination 4339 case S_INTERPOSING: 4340 type = eSymbolTypeTrampoline; 4341 break; // section with only pairs of function pointers for 4342 // interposing 4343 case S_16BYTE_LITERALS: 4344 type = eSymbolTypeData; 4345 break; // section with only 16 byte literals 4346 case S_DTRACE_DOF: 4347 type = eSymbolTypeInstrumentation; 4348 break; 4349 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4350 type = eSymbolTypeTrampoline; 4351 break; 4352 default: 4353 switch (symbol_section->GetType()) { 4354 case lldb::eSectionTypeCode: 4355 type = eSymbolTypeCode; 4356 break; 4357 case eSectionTypeData: 4358 case eSectionTypeDataCString: // Inlined C string data 4359 case eSectionTypeDataCStringPointers: // Pointers to C string 4360 // data 4361 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4362 // the symbol table 4363 case eSectionTypeData4: 4364 case eSectionTypeData8: 4365 case eSectionTypeData16: 4366 type = eSymbolTypeData; 4367 break; 4368 default: 4369 break; 4370 } 4371 break; 4372 } 4373 4374 if (type == eSymbolTypeInvalid) { 4375 const char *symbol_sect_name = 4376 symbol_section->GetName().AsCString(); 4377 if (symbol_section->IsDescendant(text_section_sp.get())) { 4378 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4379 S_ATTR_SELF_MODIFYING_CODE | 4380 S_ATTR_SOME_INSTRUCTIONS)) 4381 type = eSymbolTypeData; 4382 else 4383 type = eSymbolTypeCode; 4384 } else if (symbol_section->IsDescendant( 4385 data_section_sp.get()) || 4386 symbol_section->IsDescendant( 4387 data_dirty_section_sp.get()) || 4388 symbol_section->IsDescendant( 4389 data_const_section_sp.get())) { 4390 if (symbol_sect_name && 4391 ::strstr(symbol_sect_name, "__objc") == 4392 symbol_sect_name) { 4393 type = eSymbolTypeRuntime; 4394 4395 if (symbol_name) { 4396 llvm::StringRef symbol_name_ref(symbol_name); 4397 if (symbol_name_ref.startswith("_OBJC_")) { 4398 static const llvm::StringRef g_objc_v2_prefix_class( 4399 "_OBJC_CLASS_$_"); 4400 static const llvm::StringRef g_objc_v2_prefix_metaclass( 4401 "_OBJC_METACLASS_$_"); 4402 static const llvm::StringRef g_objc_v2_prefix_ivar( 4403 "_OBJC_IVAR_$_"); 4404 if (symbol_name_ref.startswith( 4405 g_objc_v2_prefix_class)) { 4406 symbol_name_non_abi_mangled = symbol_name + 1; 4407 symbol_name = 4408 symbol_name + g_objc_v2_prefix_class.size(); 4409 type = eSymbolTypeObjCClass; 4410 demangled_is_synthesized = true; 4411 } else if (symbol_name_ref.startswith( 4412 g_objc_v2_prefix_metaclass)) { 4413 symbol_name_non_abi_mangled = symbol_name + 1; 4414 symbol_name = 4415 symbol_name + g_objc_v2_prefix_metaclass.size(); 4416 type = eSymbolTypeObjCMetaClass; 4417 demangled_is_synthesized = true; 4418 } else if (symbol_name_ref.startswith( 4419 g_objc_v2_prefix_ivar)) { 4420 symbol_name_non_abi_mangled = symbol_name + 1; 4421 symbol_name = 4422 symbol_name + g_objc_v2_prefix_ivar.size(); 4423 type = eSymbolTypeObjCIVar; 4424 demangled_is_synthesized = true; 4425 } 4426 } 4427 } 4428 } else if (symbol_sect_name && 4429 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4430 symbol_sect_name) { 4431 type = eSymbolTypeException; 4432 } else { 4433 type = eSymbolTypeData; 4434 } 4435 } else if (symbol_sect_name && 4436 ::strstr(symbol_sect_name, "__IMPORT") == 4437 symbol_sect_name) { 4438 type = eSymbolTypeTrampoline; 4439 } else if (symbol_section->IsDescendant( 4440 objc_section_sp.get())) { 4441 type = eSymbolTypeRuntime; 4442 if (symbol_name && symbol_name[0] == '.') { 4443 llvm::StringRef symbol_name_ref(symbol_name); 4444 static const llvm::StringRef g_objc_v1_prefix_class( 4445 ".objc_class_name_"); 4446 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4447 symbol_name_non_abi_mangled = symbol_name; 4448 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4449 type = eSymbolTypeObjCClass; 4450 demangled_is_synthesized = true; 4451 } 4452 } 4453 } 4454 } 4455 } 4456 } break; 4457 } 4458 } 4459 4460 if (add_nlist) { 4461 uint64_t symbol_value = nlist.n_value; 4462 4463 if (symbol_name_non_abi_mangled) { 4464 sym[sym_idx].GetMangled().SetMangledName( 4465 ConstString(symbol_name_non_abi_mangled)); 4466 sym[sym_idx].GetMangled().SetDemangledName( 4467 ConstString(symbol_name)); 4468 } else { 4469 bool symbol_name_is_mangled = false; 4470 4471 if (symbol_name && symbol_name[0] == '_') { 4472 symbol_name_is_mangled = symbol_name[1] == '_'; 4473 symbol_name++; // Skip the leading underscore 4474 } 4475 4476 if (symbol_name) { 4477 ConstString const_symbol_name(symbol_name); 4478 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4479 symbol_name_is_mangled); 4480 } 4481 } 4482 4483 if (is_gsym) { 4484 const char *gsym_name = sym[sym_idx] 4485 .GetMangled() 4486 .GetName(lldb::eLanguageTypeUnknown, 4487 Mangled::ePreferMangled) 4488 .GetCString(); 4489 if (gsym_name) 4490 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4491 } 4492 4493 if (symbol_section) { 4494 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4495 if (symbol_byte_size == 0 && function_starts_count > 0) { 4496 addr_t symbol_lookup_file_addr = nlist.n_value; 4497 // Do an exact address match for non-ARM addresses, else get the 4498 // closest since 4499 // the symbol might be a thumb symbol which has an address with 4500 // bit zero set 4501 FunctionStarts::Entry *func_start_entry = 4502 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4503 if (is_arm && func_start_entry) { 4504 // Verify that the function start address is the symbol address 4505 // (ARM) 4506 // or the symbol address + 1 (thumb) 4507 if (func_start_entry->addr != symbol_lookup_file_addr && 4508 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4509 // Not the right entry, NULL it out... 4510 func_start_entry = NULL; 4511 } 4512 } 4513 if (func_start_entry) { 4514 func_start_entry->data = true; 4515 4516 addr_t symbol_file_addr = func_start_entry->addr; 4517 if (is_arm) 4518 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4519 4520 const FunctionStarts::Entry *next_func_start_entry = 4521 function_starts.FindNextEntry(func_start_entry); 4522 const addr_t section_end_file_addr = 4523 section_file_addr + symbol_section->GetByteSize(); 4524 if (next_func_start_entry) { 4525 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4526 // Be sure the clear the Thumb address bit when we calculate 4527 // the size 4528 // from the current and next address 4529 if (is_arm) 4530 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4531 symbol_byte_size = std::min<lldb::addr_t>( 4532 next_symbol_file_addr - symbol_file_addr, 4533 section_end_file_addr - symbol_file_addr); 4534 } else { 4535 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4536 } 4537 } 4538 } 4539 symbol_value -= section_file_addr; 4540 } 4541 4542 if (is_debug == false) { 4543 if (type == eSymbolTypeCode) { 4544 // See if we can find a N_FUN entry for any code symbols. 4545 // If we do find a match, and the name matches, then we 4546 // can merge the two into just the function symbol to avoid 4547 // duplicate entries in the symbol table 4548 std::pair<ValueToSymbolIndexMap::const_iterator, 4549 ValueToSymbolIndexMap::const_iterator> 4550 range; 4551 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4552 if (range.first != range.second) { 4553 bool found_it = false; 4554 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4555 pos != range.second; ++pos) { 4556 if (sym[sym_idx].GetMangled().GetName( 4557 lldb::eLanguageTypeUnknown, 4558 Mangled::ePreferMangled) == 4559 sym[pos->second].GetMangled().GetName( 4560 lldb::eLanguageTypeUnknown, 4561 Mangled::ePreferMangled)) { 4562 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4563 // We just need the flags from the linker symbol, so put 4564 // these flags 4565 // into the N_FUN flags to avoid duplicate symbols in the 4566 // symbol table 4567 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4568 sym[pos->second].SetFlags(nlist.n_type << 16 | 4569 nlist.n_desc); 4570 if (resolver_addresses.find(nlist.n_value) != 4571 resolver_addresses.end()) 4572 sym[pos->second].SetType(eSymbolTypeResolver); 4573 sym[sym_idx].Clear(); 4574 found_it = true; 4575 break; 4576 } 4577 } 4578 if (found_it) 4579 continue; 4580 } else { 4581 if (resolver_addresses.find(nlist.n_value) != 4582 resolver_addresses.end()) 4583 type = eSymbolTypeResolver; 4584 } 4585 } else if (type == eSymbolTypeData || 4586 type == eSymbolTypeObjCClass || 4587 type == eSymbolTypeObjCMetaClass || 4588 type == eSymbolTypeObjCIVar) { 4589 // See if we can find a N_STSYM entry for any data symbols. 4590 // If we do find a match, and the name matches, then we 4591 // can merge the two into just the Static symbol to avoid 4592 // duplicate entries in the symbol table 4593 std::pair<ValueToSymbolIndexMap::const_iterator, 4594 ValueToSymbolIndexMap::const_iterator> 4595 range; 4596 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4597 if (range.first != range.second) { 4598 bool found_it = false; 4599 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4600 pos != range.second; ++pos) { 4601 if (sym[sym_idx].GetMangled().GetName( 4602 lldb::eLanguageTypeUnknown, 4603 Mangled::ePreferMangled) == 4604 sym[pos->second].GetMangled().GetName( 4605 lldb::eLanguageTypeUnknown, 4606 Mangled::ePreferMangled)) { 4607 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4608 // We just need the flags from the linker symbol, so put 4609 // these flags 4610 // into the N_STSYM flags to avoid duplicate symbols in the 4611 // symbol table 4612 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4613 sym[pos->second].SetFlags(nlist.n_type << 16 | 4614 nlist.n_desc); 4615 sym[sym_idx].Clear(); 4616 found_it = true; 4617 break; 4618 } 4619 } 4620 if (found_it) 4621 continue; 4622 } else { 4623 // Combine N_GSYM stab entries with the non stab symbol 4624 const char *gsym_name = sym[sym_idx] 4625 .GetMangled() 4626 .GetName(lldb::eLanguageTypeUnknown, 4627 Mangled::ePreferMangled) 4628 .GetCString(); 4629 if (gsym_name) { 4630 ConstNameToSymbolIndexMap::const_iterator pos = 4631 N_GSYM_name_to_sym_idx.find(gsym_name); 4632 if (pos != N_GSYM_name_to_sym_idx.end()) { 4633 const uint32_t GSYM_sym_idx = pos->second; 4634 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4635 // Copy the address, because often the N_GSYM address has an 4636 // invalid address of zero 4637 // when the global is a common symbol 4638 sym[GSYM_sym_idx].GetAddressRef().SetSection( 4639 symbol_section); 4640 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4641 // We just need the flags from the linker symbol, so put 4642 // these flags 4643 // into the N_GSYM flags to avoid duplicate symbols in the 4644 // symbol table 4645 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 4646 nlist.n_desc); 4647 sym[sym_idx].Clear(); 4648 continue; 4649 } 4650 } 4651 } 4652 } 4653 } 4654 4655 sym[sym_idx].SetID(nlist_idx); 4656 sym[sym_idx].SetType(type); 4657 if (set_value) { 4658 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4659 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4660 } 4661 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4662 4663 if (symbol_byte_size > 0) 4664 sym[sym_idx].SetByteSize(symbol_byte_size); 4665 4666 if (demangled_is_synthesized) 4667 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4668 4669 ++sym_idx; 4670 } else { 4671 sym[sym_idx].Clear(); 4672 } 4673 } 4674 4675 for (const auto &pos : reexport_shlib_needs_fixup) { 4676 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4677 if (undef_pos != undefined_name_to_desc.end()) { 4678 const uint8_t dylib_ordinal = 4679 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4680 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4681 sym[pos.first].SetReExportedSymbolSharedLibrary( 4682 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4683 } 4684 } 4685 } 4686 4687 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4688 4689 if (function_starts_count > 0) { 4690 uint32_t num_synthetic_function_symbols = 0; 4691 for (i = 0; i < function_starts_count; ++i) { 4692 if (function_starts.GetEntryRef(i).data == false) 4693 ++num_synthetic_function_symbols; 4694 } 4695 4696 if (num_synthetic_function_symbols > 0) { 4697 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4698 num_syms = sym_idx + num_synthetic_function_symbols; 4699 sym = symtab->Resize(num_syms); 4700 } 4701 for (i = 0; i < function_starts_count; ++i) { 4702 const FunctionStarts::Entry *func_start_entry = 4703 function_starts.GetEntryAtIndex(i); 4704 if (func_start_entry->data == false) { 4705 addr_t symbol_file_addr = func_start_entry->addr; 4706 uint32_t symbol_flags = 0; 4707 if (is_arm) { 4708 if (symbol_file_addr & 1) 4709 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4710 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4711 } 4712 Address symbol_addr; 4713 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4714 SectionSP symbol_section(symbol_addr.GetSection()); 4715 uint32_t symbol_byte_size = 0; 4716 if (symbol_section) { 4717 const addr_t section_file_addr = 4718 symbol_section->GetFileAddress(); 4719 const FunctionStarts::Entry *next_func_start_entry = 4720 function_starts.FindNextEntry(func_start_entry); 4721 const addr_t section_end_file_addr = 4722 section_file_addr + symbol_section->GetByteSize(); 4723 if (next_func_start_entry) { 4724 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4725 if (is_arm) 4726 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4727 symbol_byte_size = std::min<lldb::addr_t>( 4728 next_symbol_file_addr - symbol_file_addr, 4729 section_end_file_addr - symbol_file_addr); 4730 } else { 4731 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4732 } 4733 sym[sym_idx].SetID(synthetic_sym_id++); 4734 sym[sym_idx].GetMangled().SetDemangledName( 4735 GetNextSyntheticSymbolName()); 4736 sym[sym_idx].SetType(eSymbolTypeCode); 4737 sym[sym_idx].SetIsSynthetic(true); 4738 sym[sym_idx].GetAddressRef() = symbol_addr; 4739 if (symbol_flags) 4740 sym[sym_idx].SetFlags(symbol_flags); 4741 if (symbol_byte_size) 4742 sym[sym_idx].SetByteSize(symbol_byte_size); 4743 ++sym_idx; 4744 } 4745 } 4746 } 4747 } 4748 } 4749 } 4750 4751 // Trim our symbols down to just what we ended up with after 4752 // removing any symbols. 4753 if (sym_idx < num_syms) { 4754 num_syms = sym_idx; 4755 sym = symtab->Resize(num_syms); 4756 } 4757 4758 // Now synthesize indirect symbols 4759 if (m_dysymtab.nindirectsyms != 0) { 4760 if (indirect_symbol_index_data.GetByteSize()) { 4761 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4762 m_nlist_idx_to_sym_idx.end(); 4763 4764 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4765 ++sect_idx) { 4766 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4767 S_SYMBOL_STUBS) { 4768 uint32_t symbol_stub_byte_size = 4769 m_mach_sections[sect_idx].reserved2; 4770 if (symbol_stub_byte_size == 0) 4771 continue; 4772 4773 const uint32_t num_symbol_stubs = 4774 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4775 4776 if (num_symbol_stubs == 0) 4777 continue; 4778 4779 const uint32_t symbol_stub_index_offset = 4780 m_mach_sections[sect_idx].reserved1; 4781 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; 4782 ++stub_idx) { 4783 const uint32_t symbol_stub_index = 4784 symbol_stub_index_offset + stub_idx; 4785 const lldb::addr_t symbol_stub_addr = 4786 m_mach_sections[sect_idx].addr + 4787 (stub_idx * symbol_stub_byte_size); 4788 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4789 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4790 symbol_stub_offset, 4)) { 4791 const uint32_t stub_sym_id = 4792 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4793 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4794 continue; 4795 4796 NListIndexToSymbolIndexMap::const_iterator index_pos = 4797 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4798 Symbol *stub_symbol = NULL; 4799 if (index_pos != end_index_pos) { 4800 // We have a remapping from the original nlist index to 4801 // a current symbol index, so just look this up by index 4802 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4803 } else { 4804 // We need to lookup a symbol using the original nlist 4805 // symbol index since this index is coming from the 4806 // S_SYMBOL_STUBS 4807 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4808 } 4809 4810 if (stub_symbol) { 4811 Address so_addr(symbol_stub_addr, section_list); 4812 4813 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4814 // Change the external symbol into a trampoline that makes 4815 // sense 4816 // These symbols were N_UNDF N_EXT, and are useless to us, 4817 // so we 4818 // can re-use them so we don't have to make up a synthetic 4819 // symbol 4820 // for no good reason. 4821 if (resolver_addresses.find(symbol_stub_addr) == 4822 resolver_addresses.end()) 4823 stub_symbol->SetType(eSymbolTypeTrampoline); 4824 else 4825 stub_symbol->SetType(eSymbolTypeResolver); 4826 stub_symbol->SetExternal(false); 4827 stub_symbol->GetAddressRef() = so_addr; 4828 stub_symbol->SetByteSize(symbol_stub_byte_size); 4829 } else { 4830 // Make a synthetic symbol to describe the trampoline stub 4831 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4832 if (sym_idx >= num_syms) { 4833 sym = symtab->Resize(++num_syms); 4834 stub_symbol = NULL; // this pointer no longer valid 4835 } 4836 sym[sym_idx].SetID(synthetic_sym_id++); 4837 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4838 if (resolver_addresses.find(symbol_stub_addr) == 4839 resolver_addresses.end()) 4840 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4841 else 4842 sym[sym_idx].SetType(eSymbolTypeResolver); 4843 sym[sym_idx].SetIsSynthetic(true); 4844 sym[sym_idx].GetAddressRef() = so_addr; 4845 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4846 ++sym_idx; 4847 } 4848 } else { 4849 if (log) 4850 log->Warning("symbol stub referencing symbol table symbol " 4851 "%u that isn't in our minimal symbol table, " 4852 "fix this!!!", 4853 stub_sym_id); 4854 } 4855 } 4856 } 4857 } 4858 } 4859 } 4860 } 4861 4862 if (!trie_entries.empty()) { 4863 for (const auto &e : trie_entries) { 4864 if (e.entry.import_name) { 4865 // Only add indirect symbols from the Trie entries if we 4866 // didn't have a N_INDR nlist entry for this already 4867 if (indirect_symbol_names.find(e.entry.name) == 4868 indirect_symbol_names.end()) { 4869 // Make a synthetic symbol to describe re-exported symbol. 4870 if (sym_idx >= num_syms) 4871 sym = symtab->Resize(++num_syms); 4872 sym[sym_idx].SetID(synthetic_sym_id++); 4873 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4874 sym[sym_idx].SetType(eSymbolTypeReExported); 4875 sym[sym_idx].SetIsSynthetic(true); 4876 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4877 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4878 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4879 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4880 } 4881 ++sym_idx; 4882 } 4883 } 4884 } 4885 } 4886 4887 // StreamFile s(stdout, false); 4888 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4889 // symtab->Dump(&s, NULL, eSortOrderNone); 4890 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4891 // sizes 4892 symtab->CalculateSymbolSizes(); 4893 4894 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4895 // symtab->Dump(&s, NULL, eSortOrderNone); 4896 4897 return symtab->GetNumSymbols(); 4898 } 4899 return 0; 4900 } 4901 4902 void ObjectFileMachO::Dump(Stream *s) { 4903 ModuleSP module_sp(GetModule()); 4904 if (module_sp) { 4905 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4906 s->Printf("%p: ", static_cast<void *>(this)); 4907 s->Indent(); 4908 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4909 s->PutCString("ObjectFileMachO64"); 4910 else 4911 s->PutCString("ObjectFileMachO32"); 4912 4913 ArchSpec header_arch; 4914 GetArchitecture(header_arch); 4915 4916 *s << ", file = '" << m_file 4917 << "', arch = " << header_arch.GetArchitectureName() << "\n"; 4918 4919 SectionList *sections = GetSectionList(); 4920 if (sections) 4921 sections->Dump(s, NULL, true, UINT32_MAX); 4922 4923 if (m_symtab_ap.get()) 4924 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4925 } 4926 } 4927 4928 bool ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4929 const lldb_private::DataExtractor &data, 4930 lldb::offset_t lc_offset, 4931 lldb_private::UUID &uuid) { 4932 uint32_t i; 4933 struct uuid_command load_cmd; 4934 4935 lldb::offset_t offset = lc_offset; 4936 for (i = 0; i < header.ncmds; ++i) { 4937 const lldb::offset_t cmd_offset = offset; 4938 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4939 break; 4940 4941 if (load_cmd.cmd == LC_UUID) { 4942 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4943 4944 if (uuid_bytes) { 4945 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4946 // We pretend these object files have no UUID to prevent crashing. 4947 4948 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4949 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4950 0xbb, 0x14, 0xf0, 0x0d}; 4951 4952 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4953 return false; 4954 4955 uuid.SetBytes(uuid_bytes); 4956 return true; 4957 } 4958 return false; 4959 } 4960 offset = cmd_offset + load_cmd.cmdsize; 4961 } 4962 return false; 4963 } 4964 4965 bool ObjectFileMachO::GetArchitecture(const llvm::MachO::mach_header &header, 4966 const lldb_private::DataExtractor &data, 4967 lldb::offset_t lc_offset, 4968 ArchSpec &arch) { 4969 arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4970 4971 if (arch.IsValid()) { 4972 llvm::Triple &triple = arch.GetTriple(); 4973 4974 // Set OS to an unspecified unknown or a "*" so it can match any OS 4975 triple.setOS(llvm::Triple::UnknownOS); 4976 triple.setOSName(llvm::StringRef()); 4977 4978 if (header.filetype == MH_PRELOAD) { 4979 if (header.cputype == CPU_TYPE_ARM) { 4980 // If this is a 32-bit arm binary, and it's a standalone binary, 4981 // force the Vendor to Apple so we don't accidentally pick up 4982 // the generic armv7 ABI at runtime. Apple's armv7 ABI always uses 4983 // r7 for the frame pointer register; most other armv7 ABIs use a 4984 // combination of r7 and r11. 4985 triple.setVendor(llvm::Triple::Apple); 4986 } else { 4987 // Set vendor to an unspecified unknown or a "*" so it can match any 4988 // vendor 4989 // This is required for correct behavior of EFI debugging on x86_64 4990 triple.setVendor(llvm::Triple::UnknownVendor); 4991 triple.setVendorName(llvm::StringRef()); 4992 } 4993 return true; 4994 } else { 4995 struct load_command load_cmd; 4996 4997 lldb::offset_t offset = lc_offset; 4998 for (uint32_t i = 0; i < header.ncmds; ++i) { 4999 const lldb::offset_t cmd_offset = offset; 5000 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 5001 break; 5002 5003 switch (load_cmd.cmd) { 5004 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 5005 triple.setOS(llvm::Triple::IOS); 5006 return true; 5007 5008 case llvm::MachO::LC_VERSION_MIN_MACOSX: 5009 triple.setOS(llvm::Triple::MacOSX); 5010 return true; 5011 5012 case llvm::MachO::LC_VERSION_MIN_TVOS: 5013 triple.setOS(llvm::Triple::TvOS); 5014 return true; 5015 5016 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 5017 triple.setOS(llvm::Triple::WatchOS); 5018 return true; 5019 5020 default: 5021 break; 5022 } 5023 5024 offset = cmd_offset + load_cmd.cmdsize; 5025 } 5026 5027 if (header.filetype != MH_KEXT_BUNDLE) { 5028 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 5029 // so lets not say our Vendor is Apple, leave it as an unspecified 5030 // unknown 5031 triple.setVendor(llvm::Triple::UnknownVendor); 5032 triple.setVendorName(llvm::StringRef()); 5033 } 5034 } 5035 } 5036 return arch.IsValid(); 5037 } 5038 5039 bool ObjectFileMachO::GetUUID(lldb_private::UUID *uuid) { 5040 ModuleSP module_sp(GetModule()); 5041 if (module_sp) { 5042 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5043 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5044 return GetUUID(m_header, m_data, offset, *uuid); 5045 } 5046 return false; 5047 } 5048 5049 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 5050 uint32_t count = 0; 5051 ModuleSP module_sp(GetModule()); 5052 if (module_sp) { 5053 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5054 struct load_command load_cmd; 5055 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5056 std::vector<std::string> rpath_paths; 5057 std::vector<std::string> rpath_relative_paths; 5058 std::vector<std::string> at_exec_relative_paths; 5059 const bool resolve_path = false; // Don't resolve the dependent file paths 5060 // since they may not reside on this system 5061 uint32_t i; 5062 for (i = 0; i < m_header.ncmds; ++i) { 5063 const uint32_t cmd_offset = offset; 5064 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5065 break; 5066 5067 switch (load_cmd.cmd) { 5068 case LC_RPATH: 5069 case LC_LOAD_DYLIB: 5070 case LC_LOAD_WEAK_DYLIB: 5071 case LC_REEXPORT_DYLIB: 5072 case LC_LOAD_DYLINKER: 5073 case LC_LOADFVMLIB: 5074 case LC_LOAD_UPWARD_DYLIB: { 5075 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5076 const char *path = m_data.PeekCStr(name_offset); 5077 if (path) { 5078 if (load_cmd.cmd == LC_RPATH) 5079 rpath_paths.push_back(path); 5080 else { 5081 if (path[0] == '@') { 5082 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5083 rpath_relative_paths.push_back(path + strlen("@rpath")); 5084 else if (strncmp(path, "@executable_path", 5085 strlen("@executable_path")) == 0) 5086 at_exec_relative_paths.push_back(path 5087 + strlen("@executable_path")); 5088 } else { 5089 FileSpec file_spec(path, resolve_path); 5090 if (files.AppendIfUnique(file_spec)) 5091 count++; 5092 } 5093 } 5094 } 5095 } break; 5096 5097 default: 5098 break; 5099 } 5100 offset = cmd_offset + load_cmd.cmdsize; 5101 } 5102 5103 FileSpec this_file_spec(m_file); 5104 this_file_spec.ResolvePath(); 5105 5106 if (!rpath_paths.empty()) { 5107 // Fixup all LC_RPATH values to be absolute paths 5108 std::string loader_path("@loader_path"); 5109 std::string executable_path("@executable_path"); 5110 for (auto &rpath : rpath_paths) { 5111 if (rpath.find(loader_path) == 0) { 5112 rpath.erase(0, loader_path.size()); 5113 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5114 } else if (rpath.find(executable_path) == 0) { 5115 rpath.erase(0, executable_path.size()); 5116 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5117 } 5118 } 5119 5120 for (const auto &rpath_relative_path : rpath_relative_paths) { 5121 for (const auto &rpath : rpath_paths) { 5122 std::string path = rpath; 5123 path += rpath_relative_path; 5124 // It is OK to resolve this path because we must find a file on 5125 // disk for us to accept it anyway if it is rpath relative. 5126 FileSpec file_spec(path, true); 5127 // Remove any redundant parts of the path (like "../foo") since 5128 // LC_RPATH values often contain "..". 5129 file_spec = file_spec.GetNormalizedPath(); 5130 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) { 5131 count++; 5132 break; 5133 } 5134 } 5135 } 5136 } 5137 5138 // We may have @executable_paths but no RPATHS. Figure those out here. 5139 // Only do this if this object file is the executable. We have no way to 5140 // get back to the actual executable otherwise, so we won't get the right 5141 // path. 5142 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5143 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5144 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5145 FileSpec file_spec = 5146 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5147 file_spec = file_spec.GetNormalizedPath(); 5148 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) { 5149 count++; 5150 break; 5151 } 5152 } 5153 } 5154 } 5155 return count; 5156 } 5157 5158 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5159 // If the object file is not an executable it can't hold the entry point. 5160 // m_entry_point_address 5161 // is initialized to an invalid address, so we can just return that. 5162 // If m_entry_point_address is valid it means we've found it already, so 5163 // return the cached value. 5164 5165 if (!IsExecutable() || m_entry_point_address.IsValid()) 5166 return m_entry_point_address; 5167 5168 // Otherwise, look for the UnixThread or Thread command. The data for the 5169 // Thread command is given in 5170 // /usr/include/mach-o.h, but it is basically: 5171 // 5172 // uint32_t flavor - this is the flavor argument you would pass to 5173 // thread_get_state 5174 // uint32_t count - this is the count of longs in the thread state data 5175 // struct XXX_thread_state state - this is the structure from 5176 // <machine/thread_status.h> corresponding to the flavor. 5177 // <repeat this trio> 5178 // 5179 // So we just keep reading the various register flavors till we find the GPR 5180 // one, then read the PC out of there. 5181 // FIXME: We will need to have a "RegisterContext data provider" class at some 5182 // point that can get all the registers 5183 // out of data in this form & attach them to a given thread. That should 5184 // underlie the MacOS X User process plugin, 5185 // and we'll also need it for the MacOS X Core File process plugin. When we 5186 // have that we can also use it here. 5187 // 5188 // For now we hard-code the offsets and flavors we need: 5189 // 5190 // 5191 5192 ModuleSP module_sp(GetModule()); 5193 if (module_sp) { 5194 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5195 struct load_command load_cmd; 5196 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5197 uint32_t i; 5198 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5199 bool done = false; 5200 5201 for (i = 0; i < m_header.ncmds; ++i) { 5202 const lldb::offset_t cmd_offset = offset; 5203 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5204 break; 5205 5206 switch (load_cmd.cmd) { 5207 case LC_UNIXTHREAD: 5208 case LC_THREAD: { 5209 while (offset < cmd_offset + load_cmd.cmdsize) { 5210 uint32_t flavor = m_data.GetU32(&offset); 5211 uint32_t count = m_data.GetU32(&offset); 5212 if (count == 0) { 5213 // We've gotten off somehow, log and exit; 5214 return m_entry_point_address; 5215 } 5216 5217 switch (m_header.cputype) { 5218 case llvm::MachO::CPU_TYPE_ARM: 5219 if (flavor == 1 || 5220 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 from 5221 // mach/arm/thread_status.h 5222 { 5223 offset += 60; // This is the offset of pc in the GPR thread state 5224 // data structure. 5225 start_address = m_data.GetU32(&offset); 5226 done = true; 5227 } 5228 break; 5229 case llvm::MachO::CPU_TYPE_ARM64: 5230 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5231 { 5232 offset += 256; // This is the offset of pc in the GPR thread state 5233 // data structure. 5234 start_address = m_data.GetU64(&offset); 5235 done = true; 5236 } 5237 break; 5238 case llvm::MachO::CPU_TYPE_I386: 5239 if (flavor == 5240 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5241 { 5242 offset += 40; // This is the offset of eip in the GPR thread state 5243 // data structure. 5244 start_address = m_data.GetU32(&offset); 5245 done = true; 5246 } 5247 break; 5248 case llvm::MachO::CPU_TYPE_X86_64: 5249 if (flavor == 5250 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5251 { 5252 offset += 16 * 8; // This is the offset of rip in the GPR thread 5253 // state data structure. 5254 start_address = m_data.GetU64(&offset); 5255 done = true; 5256 } 5257 break; 5258 default: 5259 return m_entry_point_address; 5260 } 5261 // Haven't found the GPR flavor yet, skip over the data for this 5262 // flavor: 5263 if (done) 5264 break; 5265 offset += count * 4; 5266 } 5267 } break; 5268 case LC_MAIN: { 5269 ConstString text_segment_name("__TEXT"); 5270 uint64_t entryoffset = m_data.GetU64(&offset); 5271 SectionSP text_segment_sp = 5272 GetSectionList()->FindSectionByName(text_segment_name); 5273 if (text_segment_sp) { 5274 done = true; 5275 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5276 } 5277 } break; 5278 5279 default: 5280 break; 5281 } 5282 if (done) 5283 break; 5284 5285 // Go to the next load command: 5286 offset = cmd_offset + load_cmd.cmdsize; 5287 } 5288 5289 if (start_address != LLDB_INVALID_ADDRESS) { 5290 // We got the start address from the load commands, so now resolve that 5291 // address in the sections 5292 // of this ObjectFile: 5293 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5294 start_address, GetSectionList())) { 5295 m_entry_point_address.Clear(); 5296 } 5297 } else { 5298 // We couldn't read the UnixThread load command - maybe it wasn't there. 5299 // As a fallback look for the 5300 // "start" symbol in the main executable. 5301 5302 ModuleSP module_sp(GetModule()); 5303 5304 if (module_sp) { 5305 SymbolContextList contexts; 5306 SymbolContext context; 5307 if (module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5308 eSymbolTypeCode, contexts)) { 5309 if (contexts.GetContextAtIndex(0, context)) 5310 m_entry_point_address = context.symbol->GetAddress(); 5311 } 5312 } 5313 } 5314 } 5315 5316 return m_entry_point_address; 5317 } 5318 5319 lldb_private::Address ObjectFileMachO::GetHeaderAddress() { 5320 lldb_private::Address header_addr; 5321 SectionList *section_list = GetSectionList(); 5322 if (section_list) { 5323 SectionSP text_segment_sp( 5324 section_list->FindSectionByName(GetSegmentNameTEXT())); 5325 if (text_segment_sp) { 5326 header_addr.SetSection(text_segment_sp); 5327 header_addr.SetOffset(0); 5328 } 5329 } 5330 return header_addr; 5331 } 5332 5333 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5334 ModuleSP module_sp(GetModule()); 5335 if (module_sp) { 5336 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5337 if (!m_thread_context_offsets_valid) { 5338 m_thread_context_offsets_valid = true; 5339 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5340 FileRangeArray::Entry file_range; 5341 thread_command thread_cmd; 5342 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5343 const uint32_t cmd_offset = offset; 5344 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 5345 break; 5346 5347 if (thread_cmd.cmd == LC_THREAD) { 5348 file_range.SetRangeBase(offset); 5349 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5350 m_thread_context_offsets.Append(file_range); 5351 } 5352 offset = cmd_offset + thread_cmd.cmdsize; 5353 } 5354 } 5355 } 5356 return m_thread_context_offsets.GetSize(); 5357 } 5358 5359 std::string ObjectFileMachO::GetIdentifierString() { 5360 std::string result; 5361 ModuleSP module_sp(GetModule()); 5362 if (module_sp) { 5363 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5364 5365 // First, look over the load commands for an LC_NOTE load command 5366 // with data_owner string "kern ver str" & use that if found. 5367 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5368 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5369 const uint32_t cmd_offset = offset; 5370 load_command lc; 5371 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5372 break; 5373 if (lc.cmd == LC_NOTE) 5374 { 5375 char data_owner[17]; 5376 m_data.CopyData (offset, 16, data_owner); 5377 data_owner[16] = '\0'; 5378 offset += 16; 5379 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5380 uint64_t size = m_data.GetU64_unchecked (&offset); 5381 5382 // "kern ver str" has a uint32_t version and then a 5383 // nul terminated c-string. 5384 if (strcmp ("kern ver str", data_owner) == 0) 5385 { 5386 offset = fileoff; 5387 uint32_t version; 5388 if (m_data.GetU32 (&offset, &version, 1) != nullptr) 5389 { 5390 if (version == 1) 5391 { 5392 uint32_t strsize = size - sizeof (uint32_t); 5393 char *buf = (char*) malloc (strsize); 5394 if (buf) 5395 { 5396 m_data.CopyData (offset, strsize, buf); 5397 buf[strsize - 1] = '\0'; 5398 result = buf; 5399 if (buf) 5400 free (buf); 5401 return result; 5402 } 5403 } 5404 } 5405 } 5406 } 5407 offset = cmd_offset + lc.cmdsize; 5408 } 5409 5410 // Second, make a pass over the load commands looking for an 5411 // obsolete LC_IDENT load command. 5412 offset = MachHeaderSizeFromMagic(m_header.magic); 5413 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5414 const uint32_t cmd_offset = offset; 5415 struct ident_command ident_command; 5416 if (m_data.GetU32(&offset, &ident_command, 2) == NULL) 5417 break; 5418 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5419 char *buf = (char *) malloc (ident_command.cmdsize); 5420 if (buf != nullptr 5421 && m_data.CopyData (offset, ident_command.cmdsize, buf) == ident_command.cmdsize) { 5422 buf[ident_command.cmdsize - 1] = '\0'; 5423 result = buf; 5424 } 5425 if (buf) 5426 free (buf); 5427 } 5428 offset = cmd_offset + ident_command.cmdsize; 5429 } 5430 5431 } 5432 return result; 5433 } 5434 5435 bool ObjectFileMachO::GetCorefileMainBinaryInfo (addr_t &address, UUID &uuid) { 5436 address = LLDB_INVALID_ADDRESS; 5437 uuid.Clear(); 5438 ModuleSP module_sp(GetModule()); 5439 if (module_sp) { 5440 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5441 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5442 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5443 const uint32_t cmd_offset = offset; 5444 load_command lc; 5445 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5446 break; 5447 if (lc.cmd == LC_NOTE) 5448 { 5449 char data_owner[17]; 5450 memset (data_owner, 0, sizeof (data_owner)); 5451 m_data.CopyData (offset, 16, data_owner); 5452 offset += 16; 5453 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5454 uint64_t size = m_data.GetU64_unchecked (&offset); 5455 5456 // "main bin spec" (main binary specification) data payload is formatted: 5457 // uint32_t version [currently 1] 5458 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user process] 5459 // uint64_t address [ UINT64_MAX if address not specified ] 5460 // uuid_t uuid [ all zero's if uuid not specified ] 5461 // uint32_t log2_pagesize [ process page size in log base 2, e.g. 4k pages are 12. 0 for unspecified ] 5462 5463 if (strcmp ("main bin spec", data_owner) == 0 && size >= 32) 5464 { 5465 offset = fileoff; 5466 uint32_t version; 5467 if (m_data.GetU32 (&offset, &version, 1) != nullptr && version == 1) 5468 { 5469 uint32_t type = 0; 5470 uuid_t raw_uuid; 5471 memset (raw_uuid, 0, sizeof (uuid_t)); 5472 5473 if (m_data.GetU32 (&offset, &type, 1) 5474 && m_data.GetU64 (&offset, &address, 1) 5475 && m_data.CopyData (offset, sizeof (uuid_t), raw_uuid) != 0 5476 && uuid.SetBytes (raw_uuid, sizeof (uuid_t))) 5477 { 5478 return true; 5479 } 5480 } 5481 } 5482 } 5483 offset = cmd_offset + lc.cmdsize; 5484 } 5485 } 5486 return false; 5487 } 5488 5489 lldb::RegisterContextSP 5490 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5491 lldb_private::Thread &thread) { 5492 lldb::RegisterContextSP reg_ctx_sp; 5493 5494 ModuleSP module_sp(GetModule()); 5495 if (module_sp) { 5496 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5497 if (!m_thread_context_offsets_valid) 5498 GetNumThreadContexts(); 5499 5500 const FileRangeArray::Entry *thread_context_file_range = 5501 m_thread_context_offsets.GetEntryAtIndex(idx); 5502 if (thread_context_file_range) { 5503 5504 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5505 thread_context_file_range->GetByteSize()); 5506 5507 switch (m_header.cputype) { 5508 case llvm::MachO::CPU_TYPE_ARM64: 5509 reg_ctx_sp.reset(new RegisterContextDarwin_arm64_Mach(thread, data)); 5510 break; 5511 5512 case llvm::MachO::CPU_TYPE_ARM: 5513 reg_ctx_sp.reset(new RegisterContextDarwin_arm_Mach(thread, data)); 5514 break; 5515 5516 case llvm::MachO::CPU_TYPE_I386: 5517 reg_ctx_sp.reset(new RegisterContextDarwin_i386_Mach(thread, data)); 5518 break; 5519 5520 case llvm::MachO::CPU_TYPE_X86_64: 5521 reg_ctx_sp.reset(new RegisterContextDarwin_x86_64_Mach(thread, data)); 5522 break; 5523 } 5524 } 5525 } 5526 return reg_ctx_sp; 5527 } 5528 5529 ObjectFile::Type ObjectFileMachO::CalculateType() { 5530 switch (m_header.filetype) { 5531 case MH_OBJECT: // 0x1u 5532 if (GetAddressByteSize() == 4) { 5533 // 32 bit kexts are just object files, but they do have a valid 5534 // UUID load command. 5535 UUID uuid; 5536 if (GetUUID(&uuid)) { 5537 // this checking for the UUID load command is not enough 5538 // we could eventually look for the symbol named 5539 // "OSKextGetCurrentIdentifier" as this is required of kexts 5540 if (m_strata == eStrataInvalid) 5541 m_strata = eStrataKernel; 5542 return eTypeSharedLibrary; 5543 } 5544 } 5545 return eTypeObjectFile; 5546 5547 case MH_EXECUTE: 5548 return eTypeExecutable; // 0x2u 5549 case MH_FVMLIB: 5550 return eTypeSharedLibrary; // 0x3u 5551 case MH_CORE: 5552 return eTypeCoreFile; // 0x4u 5553 case MH_PRELOAD: 5554 return eTypeSharedLibrary; // 0x5u 5555 case MH_DYLIB: 5556 return eTypeSharedLibrary; // 0x6u 5557 case MH_DYLINKER: 5558 return eTypeDynamicLinker; // 0x7u 5559 case MH_BUNDLE: 5560 return eTypeSharedLibrary; // 0x8u 5561 case MH_DYLIB_STUB: 5562 return eTypeStubLibrary; // 0x9u 5563 case MH_DSYM: 5564 return eTypeDebugInfo; // 0xAu 5565 case MH_KEXT_BUNDLE: 5566 return eTypeSharedLibrary; // 0xBu 5567 default: 5568 break; 5569 } 5570 return eTypeUnknown; 5571 } 5572 5573 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5574 switch (m_header.filetype) { 5575 case MH_OBJECT: // 0x1u 5576 { 5577 // 32 bit kexts are just object files, but they do have a valid 5578 // UUID load command. 5579 UUID uuid; 5580 if (GetUUID(&uuid)) { 5581 // this checking for the UUID load command is not enough 5582 // we could eventually look for the symbol named 5583 // "OSKextGetCurrentIdentifier" as this is required of kexts 5584 if (m_type == eTypeInvalid) 5585 m_type = eTypeSharedLibrary; 5586 5587 return eStrataKernel; 5588 } 5589 } 5590 return eStrataUnknown; 5591 5592 case MH_EXECUTE: // 0x2u 5593 // Check for the MH_DYLDLINK bit in the flags 5594 if (m_header.flags & MH_DYLDLINK) { 5595 return eStrataUser; 5596 } else { 5597 SectionList *section_list = GetSectionList(); 5598 if (section_list) { 5599 static ConstString g_kld_section_name("__KLD"); 5600 if (section_list->FindSectionByName(g_kld_section_name)) 5601 return eStrataKernel; 5602 } 5603 } 5604 return eStrataRawImage; 5605 5606 case MH_FVMLIB: 5607 return eStrataUser; // 0x3u 5608 case MH_CORE: 5609 return eStrataUnknown; // 0x4u 5610 case MH_PRELOAD: 5611 return eStrataRawImage; // 0x5u 5612 case MH_DYLIB: 5613 return eStrataUser; // 0x6u 5614 case MH_DYLINKER: 5615 return eStrataUser; // 0x7u 5616 case MH_BUNDLE: 5617 return eStrataUser; // 0x8u 5618 case MH_DYLIB_STUB: 5619 return eStrataUser; // 0x9u 5620 case MH_DSYM: 5621 return eStrataUnknown; // 0xAu 5622 case MH_KEXT_BUNDLE: 5623 return eStrataKernel; // 0xBu 5624 default: 5625 break; 5626 } 5627 return eStrataUnknown; 5628 } 5629 5630 uint32_t ObjectFileMachO::GetVersion(uint32_t *versions, 5631 uint32_t num_versions) { 5632 ModuleSP module_sp(GetModule()); 5633 if (module_sp) { 5634 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5635 struct dylib_command load_cmd; 5636 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5637 uint32_t version_cmd = 0; 5638 uint64_t version = 0; 5639 uint32_t i; 5640 for (i = 0; i < m_header.ncmds; ++i) { 5641 const lldb::offset_t cmd_offset = offset; 5642 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5643 break; 5644 5645 if (load_cmd.cmd == LC_ID_DYLIB) { 5646 if (version_cmd == 0) { 5647 version_cmd = load_cmd.cmd; 5648 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 5649 break; 5650 version = load_cmd.dylib.current_version; 5651 } 5652 break; // Break for now unless there is another more complete version 5653 // number load command in the future. 5654 } 5655 offset = cmd_offset + load_cmd.cmdsize; 5656 } 5657 5658 if (version_cmd == LC_ID_DYLIB) { 5659 if (versions != NULL && num_versions > 0) { 5660 if (num_versions > 0) 5661 versions[0] = (version & 0xFFFF0000ull) >> 16; 5662 if (num_versions > 1) 5663 versions[1] = (version & 0x0000FF00ull) >> 8; 5664 if (num_versions > 2) 5665 versions[2] = (version & 0x000000FFull); 5666 // Fill in an remaining version numbers with invalid values 5667 for (i = 3; i < num_versions; ++i) 5668 versions[i] = UINT32_MAX; 5669 } 5670 // The LC_ID_DYLIB load command has a version with 3 version numbers 5671 // in it, so always return 3 5672 return 3; 5673 } 5674 } 5675 return false; 5676 } 5677 5678 bool ObjectFileMachO::GetArchitecture(ArchSpec &arch) { 5679 ModuleSP module_sp(GetModule()); 5680 if (module_sp) { 5681 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5682 return GetArchitecture(m_header, m_data, 5683 MachHeaderSizeFromMagic(m_header.magic), arch); 5684 } 5685 return false; 5686 } 5687 5688 UUID ObjectFileMachO::GetProcessSharedCacheUUID(Process *process) { 5689 UUID uuid; 5690 if (process && process->GetDynamicLoader()) { 5691 DynamicLoader *dl = process->GetDynamicLoader(); 5692 addr_t load_address; 5693 LazyBool using_shared_cache; 5694 LazyBool private_shared_cache; 5695 dl->GetSharedCacheInformation(load_address, uuid, using_shared_cache, 5696 private_shared_cache); 5697 } 5698 return uuid; 5699 } 5700 5701 UUID ObjectFileMachO::GetLLDBSharedCacheUUID() { 5702 UUID uuid; 5703 #if defined(__APPLE__) && \ 5704 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5705 uint8_t *(*dyld_get_all_image_infos)(void); 5706 dyld_get_all_image_infos = 5707 (uint8_t * (*)())dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5708 if (dyld_get_all_image_infos) { 5709 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5710 if (dyld_all_image_infos_address) { 5711 uint32_t *version = (uint32_t *) 5712 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5713 if (*version >= 13) { 5714 uuid_t *sharedCacheUUID_address = 0; 5715 int wordsize = sizeof(uint8_t *); 5716 if (wordsize == 8) { 5717 sharedCacheUUID_address = 5718 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5719 160); // sharedCacheUUID <mach-o/dyld_images.h> 5720 } else { 5721 sharedCacheUUID_address = 5722 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5723 84); // sharedCacheUUID <mach-o/dyld_images.h> 5724 } 5725 uuid.SetBytes(sharedCacheUUID_address); 5726 } 5727 } 5728 } 5729 #endif 5730 return uuid; 5731 } 5732 5733 uint32_t ObjectFileMachO::GetMinimumOSVersion(uint32_t *versions, 5734 uint32_t num_versions) { 5735 if (m_min_os_versions.empty()) { 5736 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5737 bool success = false; 5738 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5739 const lldb::offset_t load_cmd_offset = offset; 5740 5741 version_min_command lc; 5742 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5743 break; 5744 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5745 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5746 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5747 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5748 if (m_data.GetU32(&offset, &lc.version, 5749 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5750 const uint32_t xxxx = lc.version >> 16; 5751 const uint32_t yy = (lc.version >> 8) & 0xffu; 5752 const uint32_t zz = lc.version & 0xffu; 5753 if (xxxx) { 5754 m_min_os_versions.push_back(xxxx); 5755 m_min_os_versions.push_back(yy); 5756 m_min_os_versions.push_back(zz); 5757 } 5758 success = true; 5759 } 5760 } 5761 offset = load_cmd_offset + lc.cmdsize; 5762 } 5763 5764 if (success == false) { 5765 // Push an invalid value so we don't keep trying to 5766 m_min_os_versions.push_back(UINT32_MAX); 5767 } 5768 } 5769 5770 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) { 5771 if (versions != NULL && num_versions > 0) { 5772 for (size_t i = 0; i < num_versions; ++i) { 5773 if (i < m_min_os_versions.size()) 5774 versions[i] = m_min_os_versions[i]; 5775 else 5776 versions[i] = 0; 5777 } 5778 } 5779 return m_min_os_versions.size(); 5780 } 5781 // Call the superclasses version that will empty out the data 5782 return ObjectFile::GetMinimumOSVersion(versions, num_versions); 5783 } 5784 5785 uint32_t ObjectFileMachO::GetSDKVersion(uint32_t *versions, 5786 uint32_t num_versions) { 5787 if (m_sdk_versions.empty()) { 5788 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5789 bool success = false; 5790 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5791 const lldb::offset_t load_cmd_offset = offset; 5792 5793 version_min_command lc; 5794 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5795 break; 5796 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5797 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5798 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5799 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5800 if (m_data.GetU32(&offset, &lc.version, 5801 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5802 const uint32_t xxxx = lc.sdk >> 16; 5803 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5804 const uint32_t zz = lc.sdk & 0xffu; 5805 if (xxxx) { 5806 m_sdk_versions.push_back(xxxx); 5807 m_sdk_versions.push_back(yy); 5808 m_sdk_versions.push_back(zz); 5809 } 5810 success = true; 5811 } 5812 } 5813 offset = load_cmd_offset + lc.cmdsize; 5814 } 5815 5816 if (success == false) { 5817 // Push an invalid value so we don't keep trying to 5818 m_sdk_versions.push_back(UINT32_MAX); 5819 } 5820 } 5821 5822 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) { 5823 if (versions != NULL && num_versions > 0) { 5824 for (size_t i = 0; i < num_versions; ++i) { 5825 if (i < m_sdk_versions.size()) 5826 versions[i] = m_sdk_versions[i]; 5827 else 5828 versions[i] = 0; 5829 } 5830 } 5831 return m_sdk_versions.size(); 5832 } 5833 // Call the superclasses version that will empty out the data 5834 return ObjectFile::GetSDKVersion(versions, num_versions); 5835 } 5836 5837 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5838 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5839 } 5840 5841 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5842 return m_allow_assembly_emulation_unwind_plans; 5843 } 5844 5845 //------------------------------------------------------------------ 5846 // PluginInterface protocol 5847 //------------------------------------------------------------------ 5848 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5849 return GetPluginNameStatic(); 5850 } 5851 5852 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5853 5854 Section *ObjectFileMachO::GetMachHeaderSection() { 5855 // Find the first address of the mach header which is the first non-zero 5856 // file sized section whose file offset is zero. This is the base file address 5857 // of the mach-o file which can be subtracted from the vmaddr of the other 5858 // segments found in memory and added to the load address 5859 ModuleSP module_sp = GetModule(); 5860 if (module_sp) { 5861 SectionList *section_list = GetSectionList(); 5862 if (section_list) { 5863 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5864 const size_t num_sections = section_list->GetSize(); 5865 5866 for (size_t sect_idx = 0; sect_idx < num_sections && 5867 mach_base_file_addr == LLDB_INVALID_ADDRESS; 5868 ++sect_idx) { 5869 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5870 if (section && section->GetFileSize() > 0 && 5871 section->GetFileOffset() == 0 && 5872 section->IsThreadSpecific() == false && 5873 module_sp.get() == section->GetModule().get()) { 5874 return section; 5875 } 5876 } 5877 } 5878 } 5879 return nullptr; 5880 } 5881 5882 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 5883 lldb::addr_t mach_header_load_address, const Section *mach_header_section, 5884 const Section *section) { 5885 ModuleSP module_sp = GetModule(); 5886 if (module_sp && mach_header_section && section && 5887 mach_header_load_address != LLDB_INVALID_ADDRESS) { 5888 lldb::addr_t mach_header_file_addr = mach_header_section->GetFileAddress(); 5889 if (mach_header_file_addr != LLDB_INVALID_ADDRESS) { 5890 if (section && section->GetFileSize() > 0 && 5891 section->IsThreadSpecific() == false && 5892 module_sp.get() == section->GetModule().get()) { 5893 // Ignore __LINKEDIT and __DWARF segments 5894 if (section->GetName() == GetSegmentNameLINKEDIT()) { 5895 // Only map __LINKEDIT if we have an in memory image and this isn't 5896 // a kernel binary like a kext or mach_kernel. 5897 const bool is_memory_image = (bool)m_process_wp.lock(); 5898 const Strata strata = GetStrata(); 5899 if (is_memory_image == false || strata == eStrataKernel) 5900 return LLDB_INVALID_ADDRESS; 5901 } 5902 return section->GetFileAddress() - mach_header_file_addr + 5903 mach_header_load_address; 5904 } 5905 } 5906 } 5907 return LLDB_INVALID_ADDRESS; 5908 } 5909 5910 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 5911 bool value_is_offset) { 5912 ModuleSP module_sp = GetModule(); 5913 if (module_sp) { 5914 size_t num_loaded_sections = 0; 5915 SectionList *section_list = GetSectionList(); 5916 if (section_list) { 5917 const size_t num_sections = section_list->GetSize(); 5918 5919 if (value_is_offset) { 5920 // "value" is an offset to apply to each top level segment 5921 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5922 // Iterate through the object file sections to find all 5923 // of the sections that size on disk (to avoid __PAGEZERO) 5924 // and load them 5925 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5926 if (section_sp && section_sp->GetFileSize() > 0 && 5927 section_sp->IsThreadSpecific() == false && 5928 module_sp.get() == section_sp->GetModule().get()) { 5929 // Ignore __LINKEDIT and __DWARF segments 5930 if (section_sp->GetName() == GetSegmentNameLINKEDIT()) { 5931 // Only map __LINKEDIT if we have an in memory image and this 5932 // isn't 5933 // a kernel binary like a kext or mach_kernel. 5934 const bool is_memory_image = (bool)m_process_wp.lock(); 5935 const Strata strata = GetStrata(); 5936 if (is_memory_image == false || strata == eStrataKernel) 5937 continue; 5938 } 5939 if (target.GetSectionLoadList().SetSectionLoadAddress( 5940 section_sp, section_sp->GetFileAddress() + value)) 5941 ++num_loaded_sections; 5942 } 5943 } 5944 } else { 5945 // "value" is the new base address of the mach_header, adjust each 5946 // section accordingly 5947 5948 Section *mach_header_section = GetMachHeaderSection(); 5949 if (mach_header_section) { 5950 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5951 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5952 5953 lldb::addr_t section_load_addr = 5954 CalculateSectionLoadAddressForMemoryImage( 5955 value, mach_header_section, section_sp.get()); 5956 if (section_load_addr != LLDB_INVALID_ADDRESS) { 5957 if (target.GetSectionLoadList().SetSectionLoadAddress( 5958 section_sp, section_load_addr)) 5959 ++num_loaded_sections; 5960 } 5961 } 5962 } 5963 } 5964 } 5965 return num_loaded_sections > 0; 5966 } 5967 return false; 5968 } 5969 5970 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 5971 const FileSpec &outfile, Status &error) { 5972 if (process_sp) { 5973 Target &target = process_sp->GetTarget(); 5974 const ArchSpec target_arch = target.GetArchitecture(); 5975 const llvm::Triple &target_triple = target_arch.GetTriple(); 5976 if (target_triple.getVendor() == llvm::Triple::Apple && 5977 (target_triple.getOS() == llvm::Triple::MacOSX || 5978 target_triple.getOS() == llvm::Triple::IOS || 5979 target_triple.getOS() == llvm::Triple::WatchOS || 5980 target_triple.getOS() == llvm::Triple::TvOS)) { 5981 bool make_core = false; 5982 switch (target_arch.GetMachine()) { 5983 case llvm::Triple::aarch64: 5984 case llvm::Triple::arm: 5985 case llvm::Triple::thumb: 5986 case llvm::Triple::x86: 5987 case llvm::Triple::x86_64: 5988 make_core = true; 5989 break; 5990 default: 5991 error.SetErrorStringWithFormat("unsupported core architecture: %s", 5992 target_triple.str().c_str()); 5993 break; 5994 } 5995 5996 if (make_core) { 5997 std::vector<segment_command_64> segment_load_commands; 5998 // uint32_t range_info_idx = 0; 5999 MemoryRegionInfo range_info; 6000 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 6001 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 6002 const ByteOrder byte_order = target_arch.GetByteOrder(); 6003 if (range_error.Success()) { 6004 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 6005 const addr_t addr = range_info.GetRange().GetRangeBase(); 6006 const addr_t size = range_info.GetRange().GetByteSize(); 6007 6008 if (size == 0) 6009 break; 6010 6011 // Calculate correct protections 6012 uint32_t prot = 0; 6013 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 6014 prot |= VM_PROT_READ; 6015 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 6016 prot |= VM_PROT_WRITE; 6017 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 6018 prot |= VM_PROT_EXECUTE; 6019 6020 // printf ("[%3u] [0x%16.16" PRIx64 " - 6021 // 0x%16.16" PRIx64 ") %c%c%c\n", 6022 // range_info_idx, 6023 // addr, 6024 // size, 6025 // (prot & VM_PROT_READ ) ? 'r' : 6026 // '-', 6027 // (prot & VM_PROT_WRITE ) ? 'w' : 6028 // '-', 6029 // (prot & VM_PROT_EXECUTE) ? 'x' : 6030 // '-'); 6031 6032 if (prot != 0) { 6033 uint32_t cmd_type = LC_SEGMENT_64; 6034 uint32_t segment_size = sizeof(segment_command_64); 6035 if (addr_byte_size == 4) { 6036 cmd_type = LC_SEGMENT; 6037 segment_size = sizeof(segment_command); 6038 } 6039 segment_command_64 segment = { 6040 cmd_type, // uint32_t cmd; 6041 segment_size, // uint32_t cmdsize; 6042 {0}, // char segname[16]; 6043 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6044 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6045 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6046 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6047 prot, // uint32_t maxprot; 6048 prot, // uint32_t initprot; 6049 0, // uint32_t nsects; 6050 0}; // uint32_t flags; 6051 segment_load_commands.push_back(segment); 6052 } else { 6053 // No protections and a size of 1 used to be returned from old 6054 // debugservers when we asked about a region that was past the 6055 // last memory region and it indicates the end... 6056 if (size == 1) 6057 break; 6058 } 6059 6060 range_error = process_sp->GetMemoryRegionInfo( 6061 range_info.GetRange().GetRangeEnd(), range_info); 6062 if (range_error.Fail()) 6063 break; 6064 } 6065 6066 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6067 6068 mach_header_64 mach_header; 6069 if (addr_byte_size == 8) { 6070 mach_header.magic = MH_MAGIC_64; 6071 } else { 6072 mach_header.magic = MH_MAGIC; 6073 } 6074 mach_header.cputype = target_arch.GetMachOCPUType(); 6075 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6076 mach_header.filetype = MH_CORE; 6077 mach_header.ncmds = segment_load_commands.size(); 6078 mach_header.flags = 0; 6079 mach_header.reserved = 0; 6080 ThreadList &thread_list = process_sp->GetThreadList(); 6081 const uint32_t num_threads = thread_list.GetSize(); 6082 6083 // Make an array of LC_THREAD data items. Each one contains 6084 // the contents of the LC_THREAD load command. The data doesn't 6085 // contain the load command + load command size, we will 6086 // add the load command and load command size as we emit the data. 6087 std::vector<StreamString> LC_THREAD_datas(num_threads); 6088 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6089 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6090 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6091 LC_THREAD_data.SetByteOrder(byte_order); 6092 } 6093 for (uint32_t thread_idx = 0; thread_idx < num_threads; 6094 ++thread_idx) { 6095 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6096 if (thread_sp) { 6097 switch (mach_header.cputype) { 6098 case llvm::MachO::CPU_TYPE_ARM64: 6099 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6100 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6101 break; 6102 6103 case llvm::MachO::CPU_TYPE_ARM: 6104 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6105 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6106 break; 6107 6108 case llvm::MachO::CPU_TYPE_I386: 6109 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6110 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6111 break; 6112 6113 case llvm::MachO::CPU_TYPE_X86_64: 6114 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6115 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6116 break; 6117 } 6118 } 6119 } 6120 6121 // The size of the load command is the size of the segments... 6122 if (addr_byte_size == 8) { 6123 mach_header.sizeofcmds = segment_load_commands.size() * 6124 sizeof(struct segment_command_64); 6125 } else { 6126 mach_header.sizeofcmds = 6127 segment_load_commands.size() * sizeof(struct segment_command); 6128 } 6129 6130 // and the size of all LC_THREAD load command 6131 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6132 ++mach_header.ncmds; 6133 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6134 } 6135 6136 printf("mach_header: 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x " 6137 "0x%8.8x 0x%8.8x\n", 6138 mach_header.magic, mach_header.cputype, mach_header.cpusubtype, 6139 mach_header.filetype, mach_header.ncmds, 6140 mach_header.sizeofcmds, mach_header.flags, 6141 mach_header.reserved); 6142 6143 // Write the mach header 6144 buffer.PutHex32(mach_header.magic); 6145 buffer.PutHex32(mach_header.cputype); 6146 buffer.PutHex32(mach_header.cpusubtype); 6147 buffer.PutHex32(mach_header.filetype); 6148 buffer.PutHex32(mach_header.ncmds); 6149 buffer.PutHex32(mach_header.sizeofcmds); 6150 buffer.PutHex32(mach_header.flags); 6151 if (addr_byte_size == 8) { 6152 buffer.PutHex32(mach_header.reserved); 6153 } 6154 6155 // Skip the mach header and all load commands and align to the next 6156 // 0x1000 byte boundary 6157 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6158 if (file_offset & 0x00000fff) { 6159 file_offset += 0x00001000ull; 6160 file_offset &= (~0x00001000ull + 1); 6161 } 6162 6163 for (auto &segment : segment_load_commands) { 6164 segment.fileoff = file_offset; 6165 file_offset += segment.filesize; 6166 } 6167 6168 // Write out all of the LC_THREAD load commands 6169 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6170 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6171 buffer.PutHex32(LC_THREAD); 6172 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6173 buffer.Write(LC_THREAD_data.GetString().data(), 6174 LC_THREAD_data_size); 6175 } 6176 6177 // Write out all of the segment load commands 6178 for (const auto &segment : segment_load_commands) { 6179 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6180 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6181 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6182 segment.cmd, segment.cmdsize, segment.vmaddr, 6183 segment.vmaddr + segment.vmsize, segment.fileoff, 6184 segment.filesize, segment.maxprot, segment.initprot, 6185 segment.nsects, segment.flags); 6186 6187 buffer.PutHex32(segment.cmd); 6188 buffer.PutHex32(segment.cmdsize); 6189 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6190 if (addr_byte_size == 8) { 6191 buffer.PutHex64(segment.vmaddr); 6192 buffer.PutHex64(segment.vmsize); 6193 buffer.PutHex64(segment.fileoff); 6194 buffer.PutHex64(segment.filesize); 6195 } else { 6196 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6197 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6198 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6199 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6200 } 6201 buffer.PutHex32(segment.maxprot); 6202 buffer.PutHex32(segment.initprot); 6203 buffer.PutHex32(segment.nsects); 6204 buffer.PutHex32(segment.flags); 6205 } 6206 6207 File core_file; 6208 std::string core_file_path(outfile.GetPath()); 6209 error = core_file.Open(core_file_path.c_str(), 6210 File::eOpenOptionWrite | 6211 File::eOpenOptionTruncate | 6212 File::eOpenOptionCanCreate); 6213 if (error.Success()) { 6214 // Read 1 page at a time 6215 uint8_t bytes[0x1000]; 6216 // Write the mach header and load commands out to the core file 6217 size_t bytes_written = buffer.GetString().size(); 6218 error = core_file.Write(buffer.GetString().data(), bytes_written); 6219 if (error.Success()) { 6220 // Now write the file data for all memory segments in the process 6221 for (const auto &segment : segment_load_commands) { 6222 if (core_file.SeekFromStart(segment.fileoff) == -1) { 6223 error.SetErrorStringWithFormat( 6224 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6225 segment.fileoff, core_file_path.c_str()); 6226 break; 6227 } 6228 6229 printf("Saving %" PRId64 6230 " bytes of data for memory region at 0x%" PRIx64 "\n", 6231 segment.vmsize, segment.vmaddr); 6232 addr_t bytes_left = segment.vmsize; 6233 addr_t addr = segment.vmaddr; 6234 Status memory_read_error; 6235 while (bytes_left > 0 && error.Success()) { 6236 const size_t bytes_to_read = 6237 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6238 const size_t bytes_read = process_sp->ReadMemory( 6239 addr, bytes, bytes_to_read, memory_read_error); 6240 if (bytes_read == bytes_to_read) { 6241 size_t bytes_written = bytes_read; 6242 error = core_file.Write(bytes, bytes_written); 6243 bytes_left -= bytes_read; 6244 addr += bytes_read; 6245 } else { 6246 // Some pages within regions are not readable, those 6247 // should be zero filled 6248 memset(bytes, 0, bytes_to_read); 6249 size_t bytes_written = bytes_to_read; 6250 error = core_file.Write(bytes, bytes_written); 6251 bytes_left -= bytes_to_read; 6252 addr += bytes_to_read; 6253 } 6254 } 6255 } 6256 } 6257 } 6258 } else { 6259 error.SetErrorString( 6260 "process doesn't support getting memory region info"); 6261 } 6262 } 6263 return true; // This is the right plug to handle saving core files for 6264 // this process 6265 } 6266 } 6267 return false; 6268 } 6269