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 #include "llvm/ADT/StringRef.h" 11 12 #include "lldb/lldb-private-log.h" 13 #include "lldb/Core/ArchSpec.h" 14 #include "lldb/Core/DataBuffer.h" 15 #include "lldb/Core/Debugger.h" 16 #include "lldb/Core/Error.h" 17 #include "lldb/Core/FileSpecList.h" 18 #include "lldb/Core/Log.h" 19 #include "lldb/Core/Module.h" 20 #include "lldb/Core/ModuleSpec.h" 21 #include "lldb/Core/PluginManager.h" 22 #include "lldb/Core/RangeMap.h" 23 #include "lldb/Core/Section.h" 24 #include "lldb/Core/StreamFile.h" 25 #include "lldb/Core/StreamString.h" 26 #include "lldb/Core/Timer.h" 27 #include "lldb/Core/UUID.h" 28 #include "lldb/Host/Host.h" 29 #include "lldb/Host/FileSpec.h" 30 #include "lldb/Symbol/ClangNamespaceDecl.h" 31 #include "lldb/Symbol/DWARFCallFrameInfo.h" 32 #include "lldb/Symbol/ObjectFile.h" 33 #include "lldb/Target/Platform.h" 34 #include "lldb/Target/Process.h" 35 #include "lldb/Target/SectionLoadList.h" 36 #include "lldb/Target/Target.h" 37 #include "lldb/Target/Thread.h" 38 #include "lldb/Target/ThreadList.h" 39 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 40 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h" 41 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 42 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 43 44 #include "lldb/Utility/SafeMachO.h" 45 46 #include "ObjectFileMachO.h" 47 48 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 49 // GetLLDBSharedCacheUUID() needs to call dlsym() 50 #include <dlfcn.h> 51 #endif 52 53 #ifndef __APPLE__ 54 #include "Utility/UuidCompatibility.h" 55 #endif 56 57 using namespace lldb; 58 using namespace lldb_private; 59 using namespace llvm::MachO; 60 61 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 62 { 63 public: 64 RegisterContextDarwin_x86_64_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 65 RegisterContextDarwin_x86_64 (thread, 0) 66 { 67 SetRegisterDataFrom_LC_THREAD (data); 68 } 69 70 virtual void 71 InvalidateAllRegisters () 72 { 73 // Do nothing... registers are always valid... 74 } 75 76 void 77 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 78 { 79 lldb::offset_t offset = 0; 80 SetError (GPRRegSet, Read, -1); 81 SetError (FPURegSet, Read, -1); 82 SetError (EXCRegSet, Read, -1); 83 bool done = false; 84 85 while (!done) 86 { 87 int flavor = data.GetU32 (&offset); 88 if (flavor == 0) 89 done = true; 90 else 91 { 92 uint32_t i; 93 uint32_t count = data.GetU32 (&offset); 94 switch (flavor) 95 { 96 case GPRRegSet: 97 for (i=0; i<count; ++i) 98 (&gpr.rax)[i] = data.GetU64(&offset); 99 SetError (GPRRegSet, Read, 0); 100 done = true; 101 102 break; 103 case FPURegSet: 104 // TODO: fill in FPU regs.... 105 //SetError (FPURegSet, Read, -1); 106 done = true; 107 108 break; 109 case EXCRegSet: 110 exc.trapno = data.GetU32(&offset); 111 exc.err = data.GetU32(&offset); 112 exc.faultvaddr = data.GetU64(&offset); 113 SetError (EXCRegSet, Read, 0); 114 done = true; 115 break; 116 case 7: 117 case 8: 118 case 9: 119 // fancy flavors that encapsulate of the above 120 // flavors... 121 break; 122 123 default: 124 done = true; 125 break; 126 } 127 } 128 } 129 } 130 131 132 static size_t 133 WriteRegister (RegisterContext *reg_ctx, const char *name, const char *alt_name, size_t reg_byte_size, Stream &data) 134 { 135 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 136 if (reg_info == NULL) 137 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 138 if (reg_info) 139 { 140 lldb_private::RegisterValue reg_value; 141 if (reg_ctx->ReadRegister(reg_info, reg_value)) 142 { 143 if (reg_info->byte_size >= reg_byte_size) 144 data.Write(reg_value.GetBytes(), reg_byte_size); 145 else 146 { 147 data.Write(reg_value.GetBytes(), reg_info->byte_size); 148 for (size_t i=0, n = reg_byte_size - reg_info->byte_size; i<n; ++ i) 149 data.PutChar(0); 150 } 151 return reg_byte_size; 152 } 153 } 154 // Just write zeros if all else fails 155 for (size_t i=0; i<reg_byte_size; ++ i) 156 data.PutChar(0); 157 return reg_byte_size; 158 } 159 160 static bool 161 Create_LC_THREAD (Thread *thread, Stream &data) 162 { 163 RegisterContextSP reg_ctx_sp (thread->GetRegisterContext()); 164 if (reg_ctx_sp) 165 { 166 RegisterContext *reg_ctx = reg_ctx_sp.get(); 167 168 data.PutHex32 (GPRRegSet); // Flavor 169 data.PutHex32 (sizeof(GPR)/sizeof(uint64_t)); // Number of uint64_t values that follow 170 WriteRegister (reg_ctx, "rax", NULL, 8, data); 171 WriteRegister (reg_ctx, "rbx", NULL, 8, data); 172 WriteRegister (reg_ctx, "rcx", NULL, 8, data); 173 WriteRegister (reg_ctx, "rdx", NULL, 8, data); 174 WriteRegister (reg_ctx, "rdi", NULL, 8, data); 175 WriteRegister (reg_ctx, "rsi", NULL, 8, data); 176 WriteRegister (reg_ctx, "rbp", NULL, 8, data); 177 WriteRegister (reg_ctx, "rsp", NULL, 8, data); 178 WriteRegister (reg_ctx, "r8", NULL, 8, data); 179 WriteRegister (reg_ctx, "r9", NULL, 8, data); 180 WriteRegister (reg_ctx, "r10", NULL, 8, data); 181 WriteRegister (reg_ctx, "r11", NULL, 8, data); 182 WriteRegister (reg_ctx, "r12", NULL, 8, data); 183 WriteRegister (reg_ctx, "r13", NULL, 8, data); 184 WriteRegister (reg_ctx, "r14", NULL, 8, data); 185 WriteRegister (reg_ctx, "r15", NULL, 8, data); 186 WriteRegister (reg_ctx, "rip", NULL, 8, data); 187 WriteRegister (reg_ctx, "rflags", NULL, 8, data); 188 WriteRegister (reg_ctx, "cs", NULL, 8, data); 189 WriteRegister (reg_ctx, "fs", NULL, 8, data); 190 WriteRegister (reg_ctx, "gs", NULL, 8, data); 191 192 // // Write out the FPU registers 193 // const size_t fpu_byte_size = sizeof(FPU); 194 // size_t bytes_written = 0; 195 // data.PutHex32 (FPURegSet); 196 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 197 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 198 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 199 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, data); // uint16_t fcw; // "fctrl" 200 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, data); // uint16_t fsw; // "fstat" 201 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, data); // uint8_t ftw; // "ftag" 202 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 203 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, data); // uint16_t fop; // "fop" 204 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, data); // uint32_t ip; // "fioff" 205 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, data); // uint16_t cs; // "fiseg" 206 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 207 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, data); // uint32_t dp; // "fooff" 208 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, data); // uint16_t ds; // "foseg" 209 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 210 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, data); // uint32_t mxcsr; 211 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 4, data);// uint32_t mxcsrmask; 212 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, sizeof(MMSReg), data); 213 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, sizeof(MMSReg), data); 214 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, sizeof(MMSReg), data); 215 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, sizeof(MMSReg), data); 216 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, sizeof(MMSReg), data); 217 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, sizeof(MMSReg), data); 218 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, sizeof(MMSReg), data); 219 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, sizeof(MMSReg), data); 220 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, sizeof(XMMReg), data); 221 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, sizeof(XMMReg), data); 222 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, sizeof(XMMReg), data); 223 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, sizeof(XMMReg), data); 224 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, sizeof(XMMReg), data); 225 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, sizeof(XMMReg), data); 226 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, sizeof(XMMReg), data); 227 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, sizeof(XMMReg), data); 228 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, sizeof(XMMReg), data); 229 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, sizeof(XMMReg), data); 230 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, sizeof(XMMReg), data); 231 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, sizeof(XMMReg), data); 232 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, sizeof(XMMReg), data); 233 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, sizeof(XMMReg), data); 234 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, sizeof(XMMReg), data); 235 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, sizeof(XMMReg), data); 236 // 237 // // Fill rest with zeros 238 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ i) 239 // data.PutChar(0); 240 241 // Write out the EXC registers 242 data.PutHex32 (EXCRegSet); 243 data.PutHex32 (sizeof(EXC)/sizeof(uint64_t)); 244 WriteRegister (reg_ctx, "trapno", NULL, 4, data); 245 WriteRegister (reg_ctx, "err", NULL, 4, data); 246 WriteRegister (reg_ctx, "faultvaddr", NULL, 8, data); 247 return true; 248 } 249 return false; 250 } 251 252 protected: 253 virtual int 254 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 255 { 256 return 0; 257 } 258 259 virtual int 260 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 261 { 262 return 0; 263 } 264 265 virtual int 266 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 267 { 268 return 0; 269 } 270 271 virtual int 272 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 273 { 274 return 0; 275 } 276 277 virtual int 278 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 279 { 280 return 0; 281 } 282 283 virtual int 284 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 285 { 286 return 0; 287 } 288 }; 289 290 291 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 292 { 293 public: 294 RegisterContextDarwin_i386_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 295 RegisterContextDarwin_i386 (thread, 0) 296 { 297 SetRegisterDataFrom_LC_THREAD (data); 298 } 299 300 virtual void 301 InvalidateAllRegisters () 302 { 303 // Do nothing... registers are always valid... 304 } 305 306 void 307 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 308 { 309 lldb::offset_t offset = 0; 310 SetError (GPRRegSet, Read, -1); 311 SetError (FPURegSet, Read, -1); 312 SetError (EXCRegSet, Read, -1); 313 bool done = false; 314 315 while (!done) 316 { 317 int flavor = data.GetU32 (&offset); 318 if (flavor == 0) 319 done = true; 320 else 321 { 322 uint32_t i; 323 uint32_t count = data.GetU32 (&offset); 324 switch (flavor) 325 { 326 case GPRRegSet: 327 for (i=0; i<count; ++i) 328 (&gpr.eax)[i] = data.GetU32(&offset); 329 SetError (GPRRegSet, Read, 0); 330 done = true; 331 332 break; 333 case FPURegSet: 334 // TODO: fill in FPU regs.... 335 //SetError (FPURegSet, Read, -1); 336 done = true; 337 338 break; 339 case EXCRegSet: 340 exc.trapno = data.GetU32(&offset); 341 exc.err = data.GetU32(&offset); 342 exc.faultvaddr = data.GetU32(&offset); 343 SetError (EXCRegSet, Read, 0); 344 done = true; 345 break; 346 case 7: 347 case 8: 348 case 9: 349 // fancy flavors that encapsulate of the above 350 // flavors... 351 break; 352 353 default: 354 done = true; 355 break; 356 } 357 } 358 } 359 } 360 protected: 361 virtual int 362 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 363 { 364 return 0; 365 } 366 367 virtual int 368 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 369 { 370 return 0; 371 } 372 373 virtual int 374 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 375 { 376 return 0; 377 } 378 379 virtual int 380 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 381 { 382 return 0; 383 } 384 385 virtual int 386 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 387 { 388 return 0; 389 } 390 391 virtual int 392 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 393 { 394 return 0; 395 } 396 }; 397 398 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm 399 { 400 public: 401 RegisterContextDarwin_arm_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 402 RegisterContextDarwin_arm (thread, 0) 403 { 404 SetRegisterDataFrom_LC_THREAD (data); 405 } 406 407 virtual void 408 InvalidateAllRegisters () 409 { 410 // Do nothing... registers are always valid... 411 } 412 413 void 414 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 415 { 416 lldb::offset_t offset = 0; 417 SetError (GPRRegSet, Read, -1); 418 SetError (FPURegSet, Read, -1); 419 SetError (EXCRegSet, Read, -1); 420 bool done = false; 421 422 while (!done) 423 { 424 int flavor = data.GetU32 (&offset); 425 uint32_t count = data.GetU32 (&offset); 426 lldb::offset_t next_thread_state = offset + (count * 4); 427 switch (flavor) 428 { 429 case GPRRegSet: 430 for (uint32_t i=0; i<count; ++i) 431 { 432 gpr.r[i] = data.GetU32(&offset); 433 } 434 435 // Note that gpr.cpsr is also copied by the above loop; this loop technically extends 436 // one element past the end of the gpr.r[] array. 437 438 SetError (GPRRegSet, Read, 0); 439 offset = next_thread_state; 440 break; 441 442 case FPURegSet: 443 { 444 uint8_t *fpu_reg_buf = (uint8_t*) &fpu.floats.s[0]; 445 const int fpu_reg_buf_size = sizeof (fpu.floats); 446 if (data.ExtractBytes (offset, fpu_reg_buf_size, eByteOrderLittle, fpu_reg_buf) == fpu_reg_buf_size) 447 { 448 offset += fpu_reg_buf_size; 449 fpu.fpscr = data.GetU32(&offset); 450 SetError (FPURegSet, Read, 0); 451 } 452 else 453 { 454 done = true; 455 } 456 } 457 offset = next_thread_state; 458 break; 459 460 case EXCRegSet: 461 if (count == 3) 462 { 463 exc.exception = data.GetU32(&offset); 464 exc.fsr = data.GetU32(&offset); 465 exc.far = data.GetU32(&offset); 466 SetError (EXCRegSet, Read, 0); 467 } 468 done = true; 469 offset = next_thread_state; 470 break; 471 472 // Unknown register set flavor, stop trying to parse. 473 default: 474 done = true; 475 } 476 } 477 } 478 protected: 479 virtual int 480 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 481 { 482 return -1; 483 } 484 485 virtual int 486 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 487 { 488 return -1; 489 } 490 491 virtual int 492 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 493 { 494 return -1; 495 } 496 497 virtual int 498 DoReadDBG (lldb::tid_t tid, int flavor, DBG &dbg) 499 { 500 return -1; 501 } 502 503 virtual int 504 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 505 { 506 return 0; 507 } 508 509 virtual int 510 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 511 { 512 return 0; 513 } 514 515 virtual int 516 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 517 { 518 return 0; 519 } 520 521 virtual int 522 DoWriteDBG (lldb::tid_t tid, int flavor, const DBG &dbg) 523 { 524 return -1; 525 } 526 }; 527 528 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 529 { 530 public: 531 RegisterContextDarwin_arm64_Mach (lldb_private::Thread &thread, const DataExtractor &data) : 532 RegisterContextDarwin_arm64 (thread, 0) 533 { 534 SetRegisterDataFrom_LC_THREAD (data); 535 } 536 537 virtual void 538 InvalidateAllRegisters () 539 { 540 // Do nothing... registers are always valid... 541 } 542 543 void 544 SetRegisterDataFrom_LC_THREAD (const DataExtractor &data) 545 { 546 lldb::offset_t offset = 0; 547 SetError (GPRRegSet, Read, -1); 548 SetError (FPURegSet, Read, -1); 549 SetError (EXCRegSet, Read, -1); 550 bool done = false; 551 while (!done) 552 { 553 int flavor = data.GetU32 (&offset); 554 uint32_t count = data.GetU32 (&offset); 555 lldb::offset_t next_thread_state = offset + (count * 4); 556 switch (flavor) 557 { 558 case GPRRegSet: 559 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 32-bit register) 560 if (count >= (33 * 2) + 1) 561 { 562 for (uint32_t i=0; i<33; ++i) 563 gpr.x[i] = data.GetU64(&offset); 564 gpr.cpsr = data.GetU32(&offset); 565 SetError (GPRRegSet, Read, 0); 566 } 567 offset = next_thread_state; 568 break; 569 case FPURegSet: 570 { 571 uint8_t *fpu_reg_buf = (uint8_t*) &fpu.v[0]; 572 const int fpu_reg_buf_size = sizeof (fpu); 573 if (fpu_reg_buf_size == count 574 && data.ExtractBytes (offset, fpu_reg_buf_size, eByteOrderLittle, fpu_reg_buf) == fpu_reg_buf_size) 575 { 576 SetError (FPURegSet, Read, 0); 577 } 578 else 579 { 580 done = true; 581 } 582 } 583 offset = next_thread_state; 584 break; 585 case EXCRegSet: 586 if (count == 4) 587 { 588 exc.far = data.GetU64(&offset); 589 exc.esr = data.GetU32(&offset); 590 exc.exception = data.GetU32(&offset); 591 SetError (EXCRegSet, Read, 0); 592 } 593 offset = next_thread_state; 594 break; 595 default: 596 done = true; 597 break; 598 } 599 } 600 } 601 protected: 602 virtual int 603 DoReadGPR (lldb::tid_t tid, int flavor, GPR &gpr) 604 { 605 return -1; 606 } 607 608 virtual int 609 DoReadFPU (lldb::tid_t tid, int flavor, FPU &fpu) 610 { 611 return -1; 612 } 613 614 virtual int 615 DoReadEXC (lldb::tid_t tid, int flavor, EXC &exc) 616 { 617 return -1; 618 } 619 620 virtual int 621 DoReadDBG (lldb::tid_t tid, int flavor, DBG &dbg) 622 { 623 return -1; 624 } 625 626 virtual int 627 DoWriteGPR (lldb::tid_t tid, int flavor, const GPR &gpr) 628 { 629 return 0; 630 } 631 632 virtual int 633 DoWriteFPU (lldb::tid_t tid, int flavor, const FPU &fpu) 634 { 635 return 0; 636 } 637 638 virtual int 639 DoWriteEXC (lldb::tid_t tid, int flavor, const EXC &exc) 640 { 641 return 0; 642 } 643 644 virtual int 645 DoWriteDBG (lldb::tid_t tid, int flavor, const DBG &dbg) 646 { 647 return -1; 648 } 649 }; 650 651 static uint32_t 652 MachHeaderSizeFromMagic(uint32_t magic) 653 { 654 switch (magic) 655 { 656 case MH_MAGIC: 657 case MH_CIGAM: 658 return sizeof(struct mach_header); 659 660 case MH_MAGIC_64: 661 case MH_CIGAM_64: 662 return sizeof(struct mach_header_64); 663 break; 664 665 default: 666 break; 667 } 668 return 0; 669 } 670 671 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 672 673 void 674 ObjectFileMachO::Initialize() 675 { 676 PluginManager::RegisterPlugin (GetPluginNameStatic(), 677 GetPluginDescriptionStatic(), 678 CreateInstance, 679 CreateMemoryInstance, 680 GetModuleSpecifications, 681 SaveCore); 682 } 683 684 void 685 ObjectFileMachO::Terminate() 686 { 687 PluginManager::UnregisterPlugin (CreateInstance); 688 } 689 690 691 lldb_private::ConstString 692 ObjectFileMachO::GetPluginNameStatic() 693 { 694 static ConstString g_name("mach-o"); 695 return g_name; 696 } 697 698 const char * 699 ObjectFileMachO::GetPluginDescriptionStatic() 700 { 701 return "Mach-o object file reader (32 and 64 bit)"; 702 } 703 704 ObjectFile * 705 ObjectFileMachO::CreateInstance (const lldb::ModuleSP &module_sp, 706 DataBufferSP& data_sp, 707 lldb::offset_t data_offset, 708 const FileSpec* file, 709 lldb::offset_t file_offset, 710 lldb::offset_t length) 711 { 712 if (!data_sp) 713 { 714 data_sp = file->MemoryMapFileContents(file_offset, length); 715 data_offset = 0; 716 } 717 718 if (ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 719 { 720 // Update the data to contain the entire file if it doesn't already 721 if (data_sp->GetByteSize() < length) 722 { 723 data_sp = file->MemoryMapFileContents(file_offset, length); 724 data_offset = 0; 725 } 726 std::unique_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module_sp, data_sp, data_offset, file, file_offset, length)); 727 if (objfile_ap.get() && objfile_ap->ParseHeader()) 728 return objfile_ap.release(); 729 } 730 return NULL; 731 } 732 733 ObjectFile * 734 ObjectFileMachO::CreateMemoryInstance (const lldb::ModuleSP &module_sp, 735 DataBufferSP& data_sp, 736 const ProcessSP &process_sp, 737 lldb::addr_t header_addr) 738 { 739 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) 740 { 741 std::unique_ptr<ObjectFile> objfile_ap(new ObjectFileMachO (module_sp, data_sp, process_sp, header_addr)); 742 if (objfile_ap.get() && objfile_ap->ParseHeader()) 743 return objfile_ap.release(); 744 } 745 return NULL; 746 } 747 748 size_t 749 ObjectFileMachO::GetModuleSpecifications (const lldb_private::FileSpec& file, 750 lldb::DataBufferSP& data_sp, 751 lldb::offset_t data_offset, 752 lldb::offset_t file_offset, 753 lldb::offset_t length, 754 lldb_private::ModuleSpecList &specs) 755 { 756 const size_t initial_count = specs.GetSize(); 757 758 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) 759 { 760 DataExtractor data; 761 data.SetData(data_sp); 762 llvm::MachO::mach_header header; 763 if (ParseHeader (data, &data_offset, header)) 764 { 765 size_t header_and_load_cmds = header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 766 if (header_and_load_cmds >= data_sp->GetByteSize()) 767 { 768 data_sp = file.ReadFileContents(file_offset, header_and_load_cmds); 769 data.SetData(data_sp); 770 data_offset = MachHeaderSizeFromMagic(header.magic); 771 } 772 if (data_sp) 773 { 774 ModuleSpec spec; 775 spec.GetFileSpec() = file; 776 spec.SetObjectOffset(file_offset); 777 778 if (GetArchitecture (header, data, data_offset, spec.GetArchitecture())) 779 { 780 if (spec.GetArchitecture().IsValid()) 781 { 782 GetUUID (header, data, data_offset, spec.GetUUID()); 783 specs.Append(spec); 784 } 785 } 786 } 787 } 788 } 789 return specs.GetSize() - initial_count; 790 } 791 792 793 794 const ConstString & 795 ObjectFileMachO::GetSegmentNameTEXT() 796 { 797 static ConstString g_segment_name_TEXT ("__TEXT"); 798 return g_segment_name_TEXT; 799 } 800 801 const ConstString & 802 ObjectFileMachO::GetSegmentNameDATA() 803 { 804 static ConstString g_segment_name_DATA ("__DATA"); 805 return g_segment_name_DATA; 806 } 807 808 const ConstString & 809 ObjectFileMachO::GetSegmentNameOBJC() 810 { 811 static ConstString g_segment_name_OBJC ("__OBJC"); 812 return g_segment_name_OBJC; 813 } 814 815 const ConstString & 816 ObjectFileMachO::GetSegmentNameLINKEDIT() 817 { 818 static ConstString g_section_name_LINKEDIT ("__LINKEDIT"); 819 return g_section_name_LINKEDIT; 820 } 821 822 const ConstString & 823 ObjectFileMachO::GetSectionNameEHFrame() 824 { 825 static ConstString g_section_name_eh_frame ("__eh_frame"); 826 return g_section_name_eh_frame; 827 } 828 829 bool 830 ObjectFileMachO::MagicBytesMatch (DataBufferSP& data_sp, 831 lldb::addr_t data_offset, 832 lldb::addr_t data_length) 833 { 834 DataExtractor data; 835 data.SetData (data_sp, data_offset, data_length); 836 lldb::offset_t offset = 0; 837 uint32_t magic = data.GetU32(&offset); 838 return MachHeaderSizeFromMagic(magic) != 0; 839 } 840 841 842 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 843 DataBufferSP& data_sp, 844 lldb::offset_t data_offset, 845 const FileSpec* file, 846 lldb::offset_t file_offset, 847 lldb::offset_t length) : 848 ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 849 m_mach_segments(), 850 m_mach_sections(), 851 m_entry_point_address(), 852 m_thread_context_offsets(), 853 m_thread_context_offsets_valid(false) 854 { 855 ::memset (&m_header, 0, sizeof(m_header)); 856 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 857 } 858 859 ObjectFileMachO::ObjectFileMachO (const lldb::ModuleSP &module_sp, 860 lldb::DataBufferSP& header_data_sp, 861 const lldb::ProcessSP &process_sp, 862 lldb::addr_t header_addr) : 863 ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 864 m_mach_segments(), 865 m_mach_sections(), 866 m_entry_point_address(), 867 m_thread_context_offsets(), 868 m_thread_context_offsets_valid(false) 869 { 870 ::memset (&m_header, 0, sizeof(m_header)); 871 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 872 } 873 874 ObjectFileMachO::~ObjectFileMachO() 875 { 876 } 877 878 bool 879 ObjectFileMachO::ParseHeader (DataExtractor &data, 880 lldb::offset_t *data_offset_ptr, 881 llvm::MachO::mach_header &header) 882 { 883 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 884 // Leave magic in the original byte order 885 header.magic = data.GetU32(data_offset_ptr); 886 bool can_parse = false; 887 bool is_64_bit = false; 888 switch (header.magic) 889 { 890 case MH_MAGIC: 891 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 892 data.SetAddressByteSize(4); 893 can_parse = true; 894 break; 895 896 case MH_MAGIC_64: 897 data.SetByteOrder (lldb::endian::InlHostByteOrder()); 898 data.SetAddressByteSize(8); 899 can_parse = true; 900 is_64_bit = true; 901 break; 902 903 case MH_CIGAM: 904 data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 905 data.SetAddressByteSize(4); 906 can_parse = true; 907 break; 908 909 case MH_CIGAM_64: 910 data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 911 data.SetAddressByteSize(8); 912 is_64_bit = true; 913 can_parse = true; 914 break; 915 916 default: 917 break; 918 } 919 920 if (can_parse) 921 { 922 data.GetU32(data_offset_ptr, &header.cputype, 6); 923 if (is_64_bit) 924 *data_offset_ptr += 4; 925 return true; 926 } 927 else 928 { 929 memset(&header, 0, sizeof(header)); 930 } 931 return false; 932 } 933 934 bool 935 ObjectFileMachO::ParseHeader () 936 { 937 ModuleSP module_sp(GetModule()); 938 if (module_sp) 939 { 940 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 941 bool can_parse = false; 942 lldb::offset_t offset = 0; 943 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 944 // Leave magic in the original byte order 945 m_header.magic = m_data.GetU32(&offset); 946 switch (m_header.magic) 947 { 948 case MH_MAGIC: 949 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 950 m_data.SetAddressByteSize(4); 951 can_parse = true; 952 break; 953 954 case MH_MAGIC_64: 955 m_data.SetByteOrder (lldb::endian::InlHostByteOrder()); 956 m_data.SetAddressByteSize(8); 957 can_parse = true; 958 break; 959 960 case MH_CIGAM: 961 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 962 m_data.SetAddressByteSize(4); 963 can_parse = true; 964 break; 965 966 case MH_CIGAM_64: 967 m_data.SetByteOrder(lldb::endian::InlHostByteOrder() == eByteOrderBig ? eByteOrderLittle : eByteOrderBig); 968 m_data.SetAddressByteSize(8); 969 can_parse = true; 970 break; 971 972 default: 973 break; 974 } 975 976 if (can_parse) 977 { 978 m_data.GetU32(&offset, &m_header.cputype, 6); 979 980 981 ArchSpec mach_arch; 982 983 if (GetArchitecture (mach_arch)) 984 { 985 // Check if the module has a required architecture 986 const ArchSpec &module_arch = module_sp->GetArchitecture(); 987 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 988 return false; 989 990 if (SetModulesArchitecture (mach_arch)) 991 { 992 const size_t header_and_lc_size = m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 993 if (m_data.GetByteSize() < header_and_lc_size) 994 { 995 DataBufferSP data_sp; 996 ProcessSP process_sp (m_process_wp.lock()); 997 if (process_sp) 998 { 999 data_sp = ReadMemory (process_sp, m_memory_addr, header_and_lc_size); 1000 } 1001 else 1002 { 1003 // Read in all only the load command data from the file on disk 1004 data_sp = m_file.ReadFileContents(m_file_offset, header_and_lc_size); 1005 if (data_sp->GetByteSize() != header_and_lc_size) 1006 return false; 1007 } 1008 if (data_sp) 1009 m_data.SetData (data_sp); 1010 } 1011 } 1012 return true; 1013 } 1014 } 1015 else 1016 { 1017 memset(&m_header, 0, sizeof(struct mach_header)); 1018 } 1019 } 1020 return false; 1021 } 1022 1023 1024 ByteOrder 1025 ObjectFileMachO::GetByteOrder () const 1026 { 1027 return m_data.GetByteOrder (); 1028 } 1029 1030 bool 1031 ObjectFileMachO::IsExecutable() const 1032 { 1033 return m_header.filetype == MH_EXECUTE; 1034 } 1035 1036 uint32_t 1037 ObjectFileMachO::GetAddressByteSize () const 1038 { 1039 return m_data.GetAddressByteSize (); 1040 } 1041 1042 AddressClass 1043 ObjectFileMachO::GetAddressClass (lldb::addr_t file_addr) 1044 { 1045 Symtab *symtab = GetSymtab(); 1046 if (symtab) 1047 { 1048 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1049 if (symbol) 1050 { 1051 if (symbol->ValueIsAddress()) 1052 { 1053 SectionSP section_sp (symbol->GetAddress().GetSection()); 1054 if (section_sp) 1055 { 1056 const lldb::SectionType section_type = section_sp->GetType(); 1057 switch (section_type) 1058 { 1059 case eSectionTypeInvalid: return eAddressClassUnknown; 1060 case eSectionTypeCode: 1061 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) 1062 { 1063 // For ARM we have a bit in the n_desc field of the symbol 1064 // that tells us ARM/Thumb which is bit 0x0008. 1065 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1066 return eAddressClassCodeAlternateISA; 1067 } 1068 return eAddressClassCode; 1069 1070 case eSectionTypeContainer: return eAddressClassUnknown; 1071 case eSectionTypeData: 1072 case eSectionTypeDataCString: 1073 case eSectionTypeDataCStringPointers: 1074 case eSectionTypeDataSymbolAddress: 1075 case eSectionTypeData4: 1076 case eSectionTypeData8: 1077 case eSectionTypeData16: 1078 case eSectionTypeDataPointers: 1079 case eSectionTypeZeroFill: 1080 case eSectionTypeDataObjCMessageRefs: 1081 case eSectionTypeDataObjCCFStrings: 1082 return eAddressClassData; 1083 case eSectionTypeDebug: 1084 case eSectionTypeDWARFDebugAbbrev: 1085 case eSectionTypeDWARFDebugAranges: 1086 case eSectionTypeDWARFDebugFrame: 1087 case eSectionTypeDWARFDebugInfo: 1088 case eSectionTypeDWARFDebugLine: 1089 case eSectionTypeDWARFDebugLoc: 1090 case eSectionTypeDWARFDebugMacInfo: 1091 case eSectionTypeDWARFDebugPubNames: 1092 case eSectionTypeDWARFDebugPubTypes: 1093 case eSectionTypeDWARFDebugRanges: 1094 case eSectionTypeDWARFDebugStr: 1095 case eSectionTypeDWARFAppleNames: 1096 case eSectionTypeDWARFAppleTypes: 1097 case eSectionTypeDWARFAppleNamespaces: 1098 case eSectionTypeDWARFAppleObjC: 1099 return eAddressClassDebug; 1100 case eSectionTypeEHFrame: return eAddressClassRuntime; 1101 case eSectionTypeELFSymbolTable: 1102 case eSectionTypeELFDynamicSymbols: 1103 case eSectionTypeELFRelocationEntries: 1104 case eSectionTypeELFDynamicLinkInfo: 1105 case eSectionTypeOther: return eAddressClassUnknown; 1106 } 1107 } 1108 } 1109 1110 const SymbolType symbol_type = symbol->GetType(); 1111 switch (symbol_type) 1112 { 1113 case eSymbolTypeAny: return eAddressClassUnknown; 1114 case eSymbolTypeAbsolute: return eAddressClassUnknown; 1115 1116 case eSymbolTypeCode: 1117 case eSymbolTypeTrampoline: 1118 case eSymbolTypeResolver: 1119 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) 1120 { 1121 // For ARM we have a bit in the n_desc field of the symbol 1122 // that tells us ARM/Thumb which is bit 0x0008. 1123 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1124 return eAddressClassCodeAlternateISA; 1125 } 1126 return eAddressClassCode; 1127 1128 case eSymbolTypeData: return eAddressClassData; 1129 case eSymbolTypeRuntime: return eAddressClassRuntime; 1130 case eSymbolTypeException: return eAddressClassRuntime; 1131 case eSymbolTypeSourceFile: return eAddressClassDebug; 1132 case eSymbolTypeHeaderFile: return eAddressClassDebug; 1133 case eSymbolTypeObjectFile: return eAddressClassDebug; 1134 case eSymbolTypeCommonBlock: return eAddressClassDebug; 1135 case eSymbolTypeBlock: return eAddressClassDebug; 1136 case eSymbolTypeLocal: return eAddressClassData; 1137 case eSymbolTypeParam: return eAddressClassData; 1138 case eSymbolTypeVariable: return eAddressClassData; 1139 case eSymbolTypeVariableType: return eAddressClassDebug; 1140 case eSymbolTypeLineEntry: return eAddressClassDebug; 1141 case eSymbolTypeLineHeader: return eAddressClassDebug; 1142 case eSymbolTypeScopeBegin: return eAddressClassDebug; 1143 case eSymbolTypeScopeEnd: return eAddressClassDebug; 1144 case eSymbolTypeAdditional: return eAddressClassUnknown; 1145 case eSymbolTypeCompiler: return eAddressClassDebug; 1146 case eSymbolTypeInstrumentation:return eAddressClassDebug; 1147 case eSymbolTypeUndefined: return eAddressClassUnknown; 1148 case eSymbolTypeObjCClass: return eAddressClassRuntime; 1149 case eSymbolTypeObjCMetaClass: return eAddressClassRuntime; 1150 case eSymbolTypeObjCIVar: return eAddressClassRuntime; 1151 case eSymbolTypeReExported: return eAddressClassRuntime; 1152 } 1153 } 1154 } 1155 return eAddressClassUnknown; 1156 } 1157 1158 Symtab * 1159 ObjectFileMachO::GetSymtab() 1160 { 1161 ModuleSP module_sp(GetModule()); 1162 if (module_sp) 1163 { 1164 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 1165 if (m_symtab_ap.get() == NULL) 1166 { 1167 m_symtab_ap.reset(new Symtab(this)); 1168 Mutex::Locker symtab_locker (m_symtab_ap->GetMutex()); 1169 ParseSymtab (); 1170 m_symtab_ap->Finalize (); 1171 } 1172 } 1173 return m_symtab_ap.get(); 1174 } 1175 1176 bool 1177 ObjectFileMachO::IsStripped () 1178 { 1179 if (m_dysymtab.cmd == 0) 1180 { 1181 ModuleSP module_sp(GetModule()); 1182 if (module_sp) 1183 { 1184 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1185 for (uint32_t i=0; i<m_header.ncmds; ++i) 1186 { 1187 const lldb::offset_t load_cmd_offset = offset; 1188 1189 load_command lc; 1190 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 1191 break; 1192 if (lc.cmd == LC_DYSYMTAB) 1193 { 1194 m_dysymtab.cmd = lc.cmd; 1195 m_dysymtab.cmdsize = lc.cmdsize; 1196 if (m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == NULL) 1197 { 1198 // Clear m_dysymtab if we were unable to read all items from the load command 1199 ::memset (&m_dysymtab, 0, sizeof(m_dysymtab)); 1200 } 1201 } 1202 offset = load_cmd_offset + lc.cmdsize; 1203 } 1204 } 1205 } 1206 if (m_dysymtab.cmd) 1207 return m_dysymtab.nlocalsym <= 1; 1208 return false; 1209 } 1210 1211 void 1212 ObjectFileMachO::CreateSections (SectionList &unified_section_list) 1213 { 1214 if (!m_sections_ap.get()) 1215 { 1216 m_sections_ap.reset(new SectionList()); 1217 1218 const bool is_dsym = (m_header.filetype == MH_DSYM); 1219 lldb::user_id_t segID = 0; 1220 lldb::user_id_t sectID = 0; 1221 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1222 uint32_t i; 1223 const bool is_core = GetType() == eTypeCoreFile; 1224 //bool dump_sections = false; 1225 ModuleSP module_sp (GetModule()); 1226 // First look up any LC_ENCRYPTION_INFO load commands 1227 typedef RangeArray<uint32_t, uint32_t, 8> EncryptedFileRanges; 1228 EncryptedFileRanges encrypted_file_ranges; 1229 encryption_info_command encryption_cmd; 1230 for (i=0; i<m_header.ncmds; ++i) 1231 { 1232 const lldb::offset_t load_cmd_offset = offset; 1233 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1234 break; 1235 1236 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO) 1237 { 1238 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) 1239 { 1240 if (encryption_cmd.cryptid != 0) 1241 { 1242 EncryptedFileRanges::Entry entry; 1243 entry.SetRangeBase(encryption_cmd.cryptoff); 1244 entry.SetByteSize(encryption_cmd.cryptsize); 1245 encrypted_file_ranges.Append(entry); 1246 } 1247 } 1248 } 1249 offset = load_cmd_offset + encryption_cmd.cmdsize; 1250 } 1251 1252 bool section_file_addresses_changed = false; 1253 1254 offset = MachHeaderSizeFromMagic(m_header.magic); 1255 1256 struct segment_command_64 load_cmd; 1257 for (i=0; i<m_header.ncmds; ++i) 1258 { 1259 const lldb::offset_t load_cmd_offset = offset; 1260 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1261 break; 1262 1263 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1264 { 1265 if (m_data.GetU8(&offset, (uint8_t*)load_cmd.segname, 16)) 1266 { 1267 bool add_section = true; 1268 bool add_to_unified = true; 1269 ConstString const_segname (load_cmd.segname, std::min<size_t>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 1270 1271 SectionSP unified_section_sp(unified_section_list.FindSectionByName(const_segname)); 1272 if (is_dsym && unified_section_sp) 1273 { 1274 if (const_segname == GetSegmentNameLINKEDIT()) 1275 { 1276 // We need to keep the __LINKEDIT segment private to this object file only 1277 add_to_unified = false; 1278 } 1279 else 1280 { 1281 // This is the dSYM file and this section has already been created by 1282 // the object file, no need to create it. 1283 add_section = false; 1284 } 1285 } 1286 load_cmd.vmaddr = m_data.GetAddress(&offset); 1287 load_cmd.vmsize = m_data.GetAddress(&offset); 1288 load_cmd.fileoff = m_data.GetAddress(&offset); 1289 load_cmd.filesize = m_data.GetAddress(&offset); 1290 if (m_length != 0 && load_cmd.filesize != 0) 1291 { 1292 if (load_cmd.fileoff > m_length) 1293 { 1294 // We have a load command that says it extends past the end of the file. This is likely 1295 // a corrupt file. We don't have any way to return an error condition here (this method 1296 // was likely invoked from something like ObjectFile::GetSectionList()) -- all we can do 1297 // is null out the SectionList vector and if a process has been set up, dump a message 1298 // to stdout. The most common case here is core file debugging with a truncated file. 1299 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1300 module_sp->ReportWarning("load command %u %s has a fileoff (0x%" PRIx64 ") that extends beyond the end of the file (0x%" PRIx64 "), ignoring this section", 1301 i, 1302 lc_segment_name, 1303 load_cmd.fileoff, 1304 m_length); 1305 1306 load_cmd.fileoff = 0; 1307 load_cmd.filesize = 0; 1308 } 1309 1310 if (load_cmd.fileoff + load_cmd.filesize > m_length) 1311 { 1312 // We have a load command that says it extends past the end of the file. This is likely 1313 // a corrupt file. We don't have any way to return an error condition here (this method 1314 // was likely invoked from something like ObjectFile::GetSectionList()) -- all we can do 1315 // is null out the SectionList vector and if a process has been set up, dump a message 1316 // to stdout. The most common case here is core file debugging with a truncated file. 1317 const char *lc_segment_name = load_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1318 GetModule()->ReportWarning("load command %u %s has a fileoff + filesize (0x%" PRIx64 ") that extends beyond the end of the file (0x%" PRIx64 "), the segment will be truncated to match", 1319 i, 1320 lc_segment_name, 1321 load_cmd.fileoff + load_cmd.filesize, 1322 m_length); 1323 1324 // Tuncase the length 1325 load_cmd.filesize = m_length - load_cmd.fileoff; 1326 } 1327 } 1328 if (m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1329 { 1330 1331 const bool segment_is_encrypted = (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1332 1333 // Keep a list of mach segments around in case we need to 1334 // get at data that isn't stored in the abstracted Sections. 1335 m_mach_segments.push_back (load_cmd); 1336 1337 // Use a segment ID of the segment index shifted left by 8 so they 1338 // never conflict with any of the sections. 1339 SectionSP segment_sp; 1340 if (add_section && (const_segname || is_core)) 1341 { 1342 segment_sp.reset(new Section (module_sp, // Module to which this section belongs 1343 this, // Object file to which this sections belongs 1344 ++segID << 8, // Section ID is the 1 based segment index shifted right by 8 bits as not to collide with any of the 256 section IDs that are possible 1345 const_segname, // Name of this section 1346 eSectionTypeContainer, // This section is a container of other sections. 1347 load_cmd.vmaddr, // File VM address == addresses as they are found in the object file 1348 load_cmd.vmsize, // VM size in bytes of this section 1349 load_cmd.fileoff, // Offset to the data for this section in the file 1350 load_cmd.filesize, // Size in bytes of this section as found in the file 1351 0, // Segments have no alignment information 1352 load_cmd.flags)); // Flags for this section 1353 1354 segment_sp->SetIsEncrypted (segment_is_encrypted); 1355 m_sections_ap->AddSection(segment_sp); 1356 if (add_to_unified) 1357 unified_section_list.AddSection(segment_sp); 1358 } 1359 else if (unified_section_sp) 1360 { 1361 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) 1362 { 1363 // Check to see if the module was read from memory? 1364 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) 1365 { 1366 // We have a module that is in memory and needs to have its 1367 // file address adjusted. We need to do this because when we 1368 // load a file from memory, its addresses will be slid already, 1369 // yet the addresses in the new symbol file will still be unslid. 1370 // Since everything is stored as section offset, this shouldn't 1371 // cause any problems. 1372 1373 // Make sure we've parsed the symbol table from the 1374 // ObjectFile before we go around changing its Sections. 1375 module_sp->GetObjectFile()->GetSymtab(); 1376 // eh_frame would present the same problems but we parse that on 1377 // a per-function basis as-needed so it's more difficult to 1378 // remove its use of the Sections. Realistically, the environments 1379 // where this code path will be taken will not have eh_frame sections. 1380 1381 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1382 1383 // Notify the module that the section addresses have been changed once 1384 // we're done so any file-address caches can be updated. 1385 section_file_addresses_changed = true; 1386 } 1387 } 1388 m_sections_ap->AddSection(unified_section_sp); 1389 } 1390 1391 struct section_64 sect64; 1392 ::memset (§64, 0, sizeof(sect64)); 1393 // Push a section into our mach sections for the section at 1394 // index zero (NO_SECT) if we don't have any mach sections yet... 1395 if (m_mach_sections.empty()) 1396 m_mach_sections.push_back(sect64); 1397 uint32_t segment_sect_idx; 1398 const lldb::user_id_t first_segment_sectID = sectID + 1; 1399 1400 1401 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1402 for (segment_sect_idx=0; segment_sect_idx<load_cmd.nsects; ++segment_sect_idx) 1403 { 1404 if (m_data.GetU8(&offset, (uint8_t*)sect64.sectname, sizeof(sect64.sectname)) == NULL) 1405 break; 1406 if (m_data.GetU8(&offset, (uint8_t*)sect64.segname, sizeof(sect64.segname)) == NULL) 1407 break; 1408 sect64.addr = m_data.GetAddress(&offset); 1409 sect64.size = m_data.GetAddress(&offset); 1410 1411 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1412 break; 1413 1414 // Keep a list of mach sections around in case we need to 1415 // get at data that isn't stored in the abstracted Sections. 1416 m_mach_sections.push_back (sect64); 1417 1418 if (add_section) 1419 { 1420 ConstString section_name (sect64.sectname, std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 1421 if (!const_segname) 1422 { 1423 // We have a segment with no name so we need to conjure up 1424 // segments that correspond to the section's segname if there 1425 // isn't already such a section. If there is such a section, 1426 // we resize the section so that it spans all sections. 1427 // We also mark these sections as fake so address matches don't 1428 // hit if they land in the gaps between the child sections. 1429 const_segname.SetTrimmedCStringWithLength(sect64.segname, sizeof(sect64.segname)); 1430 segment_sp = unified_section_list.FindSectionByName (const_segname); 1431 if (segment_sp.get()) 1432 { 1433 Section *segment = segment_sp.get(); 1434 // Grow the section size as needed. 1435 const lldb::addr_t sect64_min_addr = sect64.addr; 1436 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1437 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1438 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1439 const lldb::addr_t curr_seg_max_addr = curr_seg_min_addr + curr_seg_byte_size; 1440 if (sect64_min_addr >= curr_seg_min_addr) 1441 { 1442 const lldb::addr_t new_seg_byte_size = sect64_max_addr - curr_seg_min_addr; 1443 // Only grow the section size if needed 1444 if (new_seg_byte_size > curr_seg_byte_size) 1445 segment->SetByteSize (new_seg_byte_size); 1446 } 1447 else 1448 { 1449 // We need to change the base address of the segment and 1450 // adjust the child section offsets for all existing children. 1451 const lldb::addr_t slide_amount = sect64_min_addr - curr_seg_min_addr; 1452 segment->Slide(slide_amount, false); 1453 segment->GetChildren().Slide(-slide_amount, false); 1454 segment->SetByteSize (curr_seg_max_addr - sect64_min_addr); 1455 } 1456 1457 // Grow the section size as needed. 1458 if (sect64.offset) 1459 { 1460 const lldb::addr_t segment_min_file_offset = segment->GetFileOffset(); 1461 const lldb::addr_t segment_max_file_offset = segment_min_file_offset + segment->GetFileSize(); 1462 1463 const lldb::addr_t section_min_file_offset = sect64.offset; 1464 const lldb::addr_t section_max_file_offset = section_min_file_offset + sect64.size; 1465 const lldb::addr_t new_file_offset = std::min (section_min_file_offset, segment_min_file_offset); 1466 const lldb::addr_t new_file_size = std::max (section_max_file_offset, segment_max_file_offset) - new_file_offset; 1467 segment->SetFileOffset (new_file_offset); 1468 segment->SetFileSize (new_file_size); 1469 } 1470 } 1471 else 1472 { 1473 // Create a fake section for the section's named segment 1474 segment_sp.reset(new Section (segment_sp, // Parent section 1475 module_sp, // Module to which this section belongs 1476 this, // Object file to which this section belongs 1477 ++segID << 8, // Section ID is the 1 based segment index shifted right by 8 bits as not to collide with any of the 256 section IDs that are possible 1478 const_segname, // Name of this section 1479 eSectionTypeContainer, // This section is a container of other sections. 1480 sect64.addr, // File VM address == addresses as they are found in the object file 1481 sect64.size, // VM size in bytes of this section 1482 sect64.offset, // Offset to the data for this section in the file 1483 sect64.offset ? sect64.size : 0, // Size in bytes of this section as found in the file 1484 sect64.align, 1485 load_cmd.flags)); // Flags for this section 1486 segment_sp->SetIsFake(true); 1487 1488 m_sections_ap->AddSection(segment_sp); 1489 if (add_to_unified) 1490 unified_section_list.AddSection(segment_sp); 1491 segment_sp->SetIsEncrypted (segment_is_encrypted); 1492 } 1493 } 1494 assert (segment_sp.get()); 1495 1496 lldb::SectionType sect_type = eSectionTypeOther; 1497 1498 if (sect64.flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1499 sect_type = eSectionTypeCode; 1500 else 1501 { 1502 uint32_t mach_sect_type = sect64.flags & SECTION_TYPE; 1503 static ConstString g_sect_name_objc_data ("__objc_data"); 1504 static ConstString g_sect_name_objc_msgrefs ("__objc_msgrefs"); 1505 static ConstString g_sect_name_objc_selrefs ("__objc_selrefs"); 1506 static ConstString g_sect_name_objc_classrefs ("__objc_classrefs"); 1507 static ConstString g_sect_name_objc_superrefs ("__objc_superrefs"); 1508 static ConstString g_sect_name_objc_const ("__objc_const"); 1509 static ConstString g_sect_name_objc_classlist ("__objc_classlist"); 1510 static ConstString g_sect_name_cfstring ("__cfstring"); 1511 1512 static ConstString g_sect_name_dwarf_debug_abbrev ("__debug_abbrev"); 1513 static ConstString g_sect_name_dwarf_debug_aranges ("__debug_aranges"); 1514 static ConstString g_sect_name_dwarf_debug_frame ("__debug_frame"); 1515 static ConstString g_sect_name_dwarf_debug_info ("__debug_info"); 1516 static ConstString g_sect_name_dwarf_debug_line ("__debug_line"); 1517 static ConstString g_sect_name_dwarf_debug_loc ("__debug_loc"); 1518 static ConstString g_sect_name_dwarf_debug_macinfo ("__debug_macinfo"); 1519 static ConstString g_sect_name_dwarf_debug_pubnames ("__debug_pubnames"); 1520 static ConstString g_sect_name_dwarf_debug_pubtypes ("__debug_pubtypes"); 1521 static ConstString g_sect_name_dwarf_debug_ranges ("__debug_ranges"); 1522 static ConstString g_sect_name_dwarf_debug_str ("__debug_str"); 1523 static ConstString g_sect_name_dwarf_apple_names ("__apple_names"); 1524 static ConstString g_sect_name_dwarf_apple_types ("__apple_types"); 1525 static ConstString g_sect_name_dwarf_apple_namespaces ("__apple_namespac"); 1526 static ConstString g_sect_name_dwarf_apple_objc ("__apple_objc"); 1527 static ConstString g_sect_name_eh_frame ("__eh_frame"); 1528 static ConstString g_sect_name_text ("__text"); 1529 static ConstString g_sect_name_data ("__data"); 1530 1531 1532 if (section_name == g_sect_name_dwarf_debug_abbrev) 1533 sect_type = eSectionTypeDWARFDebugAbbrev; 1534 else if (section_name == g_sect_name_dwarf_debug_aranges) 1535 sect_type = eSectionTypeDWARFDebugAranges; 1536 else if (section_name == g_sect_name_dwarf_debug_frame) 1537 sect_type = eSectionTypeDWARFDebugFrame; 1538 else if (section_name == g_sect_name_dwarf_debug_info) 1539 sect_type = eSectionTypeDWARFDebugInfo; 1540 else if (section_name == g_sect_name_dwarf_debug_line) 1541 sect_type = eSectionTypeDWARFDebugLine; 1542 else if (section_name == g_sect_name_dwarf_debug_loc) 1543 sect_type = eSectionTypeDWARFDebugLoc; 1544 else if (section_name == g_sect_name_dwarf_debug_macinfo) 1545 sect_type = eSectionTypeDWARFDebugMacInfo; 1546 else if (section_name == g_sect_name_dwarf_debug_pubnames) 1547 sect_type = eSectionTypeDWARFDebugPubNames; 1548 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 1549 sect_type = eSectionTypeDWARFDebugPubTypes; 1550 else if (section_name == g_sect_name_dwarf_debug_ranges) 1551 sect_type = eSectionTypeDWARFDebugRanges; 1552 else if (section_name == g_sect_name_dwarf_debug_str) 1553 sect_type = eSectionTypeDWARFDebugStr; 1554 else if (section_name == g_sect_name_dwarf_apple_names) 1555 sect_type = eSectionTypeDWARFAppleNames; 1556 else if (section_name == g_sect_name_dwarf_apple_types) 1557 sect_type = eSectionTypeDWARFAppleTypes; 1558 else if (section_name == g_sect_name_dwarf_apple_namespaces) 1559 sect_type = eSectionTypeDWARFAppleNamespaces; 1560 else if (section_name == g_sect_name_dwarf_apple_objc) 1561 sect_type = eSectionTypeDWARFAppleObjC; 1562 else if (section_name == g_sect_name_objc_selrefs) 1563 sect_type = eSectionTypeDataCStringPointers; 1564 else if (section_name == g_sect_name_objc_msgrefs) 1565 sect_type = eSectionTypeDataObjCMessageRefs; 1566 else if (section_name == g_sect_name_eh_frame) 1567 sect_type = eSectionTypeEHFrame; 1568 else if (section_name == g_sect_name_cfstring) 1569 sect_type = eSectionTypeDataObjCCFStrings; 1570 else if (section_name == g_sect_name_objc_data || 1571 section_name == g_sect_name_objc_classrefs || 1572 section_name == g_sect_name_objc_superrefs || 1573 section_name == g_sect_name_objc_const || 1574 section_name == g_sect_name_objc_classlist) 1575 { 1576 sect_type = eSectionTypeDataPointers; 1577 } 1578 1579 if (sect_type == eSectionTypeOther) 1580 { 1581 switch (mach_sect_type) 1582 { 1583 // TODO: categorize sections by other flags for regular sections 1584 case S_REGULAR: 1585 if (section_name == g_sect_name_text) 1586 sect_type = eSectionTypeCode; 1587 else if (section_name == g_sect_name_data) 1588 sect_type = eSectionTypeData; 1589 else 1590 sect_type = eSectionTypeOther; 1591 break; 1592 case S_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1593 case S_CSTRING_LITERALS: sect_type = eSectionTypeDataCString; break; // section with only literal C strings 1594 case S_4BYTE_LITERALS: sect_type = eSectionTypeData4; break; // section with only 4 byte literals 1595 case S_8BYTE_LITERALS: sect_type = eSectionTypeData8; break; // section with only 8 byte literals 1596 case S_LITERAL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only pointers to literals 1597 case S_NON_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only non-lazy symbol pointers 1598 case S_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only lazy symbol pointers 1599 case S_SYMBOL_STUBS: sect_type = eSectionTypeCode; break; // section with only symbol stubs, byte size of stub in the reserved2 field 1600 case S_MOD_INIT_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for initialization 1601 case S_MOD_TERM_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for termination 1602 case S_COALESCED: sect_type = eSectionTypeOther; break; 1603 case S_GB_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1604 case S_INTERPOSING: sect_type = eSectionTypeCode; break; // section with only pairs of function pointers for interposing 1605 case S_16BYTE_LITERALS: sect_type = eSectionTypeData16; break; // section with only 16 byte literals 1606 case S_DTRACE_DOF: sect_type = eSectionTypeDebug; break; 1607 case S_LAZY_DYLIB_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; 1608 default: break; 1609 } 1610 } 1611 } 1612 1613 SectionSP section_sp(new Section (segment_sp, 1614 module_sp, 1615 this, 1616 ++sectID, 1617 section_name, 1618 sect_type, 1619 sect64.addr - segment_sp->GetFileAddress(), 1620 sect64.size, 1621 sect64.offset, 1622 sect64.offset == 0 ? 0 : sect64.size, 1623 sect64.align, 1624 sect64.flags)); 1625 // Set the section to be encrypted to match the segment 1626 1627 bool section_is_encrypted = false; 1628 if (!segment_is_encrypted && load_cmd.filesize != 0) 1629 section_is_encrypted = encrypted_file_ranges.FindEntryThatContains(sect64.offset) != NULL; 1630 1631 section_sp->SetIsEncrypted (segment_is_encrypted || section_is_encrypted); 1632 segment_sp->GetChildren().AddSection(section_sp); 1633 1634 if (segment_sp->IsFake()) 1635 { 1636 segment_sp.reset(); 1637 const_segname.Clear(); 1638 } 1639 } 1640 } 1641 if (segment_sp && is_dsym) 1642 { 1643 if (first_segment_sectID <= sectID) 1644 { 1645 lldb::user_id_t sect_uid; 1646 for (sect_uid = first_segment_sectID; sect_uid <= sectID; ++sect_uid) 1647 { 1648 SectionSP curr_section_sp(segment_sp->GetChildren().FindSectionByID (sect_uid)); 1649 SectionSP next_section_sp; 1650 if (sect_uid + 1 <= sectID) 1651 next_section_sp = segment_sp->GetChildren().FindSectionByID (sect_uid+1); 1652 1653 if (curr_section_sp.get()) 1654 { 1655 if (curr_section_sp->GetByteSize() == 0) 1656 { 1657 if (next_section_sp.get() != NULL) 1658 curr_section_sp->SetByteSize ( next_section_sp->GetFileAddress() - curr_section_sp->GetFileAddress() ); 1659 else 1660 curr_section_sp->SetByteSize ( load_cmd.vmsize ); 1661 } 1662 } 1663 } 1664 } 1665 } 1666 } 1667 } 1668 } 1669 else if (load_cmd.cmd == LC_DYSYMTAB) 1670 { 1671 m_dysymtab.cmd = load_cmd.cmd; 1672 m_dysymtab.cmdsize = load_cmd.cmdsize; 1673 m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1674 } 1675 1676 offset = load_cmd_offset + load_cmd.cmdsize; 1677 } 1678 1679 1680 if (section_file_addresses_changed && module_sp.get()) 1681 { 1682 module_sp->SectionFileAddressesChanged(); 1683 } 1684 } 1685 } 1686 1687 class MachSymtabSectionInfo 1688 { 1689 public: 1690 1691 MachSymtabSectionInfo (SectionList *section_list) : 1692 m_section_list (section_list), 1693 m_section_infos() 1694 { 1695 // Get the number of sections down to a depth of 1 to include 1696 // all segments and their sections, but no other sections that 1697 // may be added for debug map or 1698 m_section_infos.resize(section_list->GetNumSections(1)); 1699 } 1700 1701 1702 SectionSP 1703 GetSection (uint8_t n_sect, addr_t file_addr) 1704 { 1705 if (n_sect == 0) 1706 return SectionSP(); 1707 if (n_sect < m_section_infos.size()) 1708 { 1709 if (!m_section_infos[n_sect].section_sp) 1710 { 1711 SectionSP section_sp (m_section_list->FindSectionByID (n_sect)); 1712 m_section_infos[n_sect].section_sp = section_sp; 1713 if (section_sp) 1714 { 1715 m_section_infos[n_sect].vm_range.SetBaseAddress (section_sp->GetFileAddress()); 1716 m_section_infos[n_sect].vm_range.SetByteSize (section_sp->GetByteSize()); 1717 } 1718 else 1719 { 1720 Host::SystemLog (Host::eSystemLogError, "error: unable to find section for section %u\n", n_sect); 1721 } 1722 } 1723 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) 1724 { 1725 // Symbol is in section. 1726 return m_section_infos[n_sect].section_sp; 1727 } 1728 else if (m_section_infos[n_sect].vm_range.GetByteSize () == 0 && 1729 m_section_infos[n_sect].vm_range.GetBaseAddress() == file_addr) 1730 { 1731 // Symbol is in section with zero size, but has the same start 1732 // address as the section. This can happen with linker symbols 1733 // (symbols that start with the letter 'l' or 'L'. 1734 return m_section_infos[n_sect].section_sp; 1735 } 1736 } 1737 return m_section_list->FindSectionContainingFileAddress(file_addr); 1738 } 1739 1740 protected: 1741 struct SectionInfo 1742 { 1743 SectionInfo () : 1744 vm_range(), 1745 section_sp () 1746 { 1747 } 1748 1749 VMRange vm_range; 1750 SectionSP section_sp; 1751 }; 1752 SectionList *m_section_list; 1753 std::vector<SectionInfo> m_section_infos; 1754 }; 1755 1756 struct TrieEntry 1757 { 1758 TrieEntry () : 1759 name(), 1760 address(LLDB_INVALID_ADDRESS), 1761 flags (0), 1762 other(0), 1763 import_name() 1764 { 1765 } 1766 1767 void 1768 Clear () 1769 { 1770 name.Clear(); 1771 address = LLDB_INVALID_ADDRESS; 1772 flags = 0; 1773 other = 0; 1774 import_name.Clear(); 1775 } 1776 1777 void 1778 Dump () const 1779 { 1780 printf ("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1781 static_cast<unsigned long long>(address), 1782 static_cast<unsigned long long>(flags), 1783 static_cast<unsigned long long>(other), name.GetCString()); 1784 if (import_name) 1785 printf (" -> \"%s\"\n", import_name.GetCString()); 1786 else 1787 printf ("\n"); 1788 } 1789 ConstString name; 1790 uint64_t address; 1791 uint64_t flags; 1792 uint64_t other; 1793 ConstString import_name; 1794 }; 1795 1796 struct TrieEntryWithOffset 1797 { 1798 lldb::offset_t nodeOffset; 1799 TrieEntry entry; 1800 1801 TrieEntryWithOffset (lldb::offset_t offset) : 1802 nodeOffset (offset), 1803 entry() 1804 { 1805 } 1806 1807 void 1808 Dump (uint32_t idx) const 1809 { 1810 printf ("[%3u] 0x%16.16llx: ", idx, 1811 static_cast<unsigned long long>(nodeOffset)); 1812 entry.Dump(); 1813 } 1814 1815 bool 1816 operator<(const TrieEntryWithOffset& other) const 1817 { 1818 return ( nodeOffset < other.nodeOffset ); 1819 } 1820 }; 1821 1822 static void 1823 ParseTrieEntries (DataExtractor &data, 1824 lldb::offset_t offset, 1825 std::vector<llvm::StringRef> &nameSlices, 1826 std::set<lldb::addr_t> &resolver_addresses, 1827 std::vector<TrieEntryWithOffset>& output) 1828 { 1829 if (!data.ValidOffset(offset)) 1830 return; 1831 1832 const uint64_t terminalSize = data.GetULEB128(&offset); 1833 lldb::offset_t children_offset = offset + terminalSize; 1834 if ( terminalSize != 0 ) { 1835 TrieEntryWithOffset e (offset); 1836 e.entry.flags = data.GetULEB128(&offset); 1837 const char *import_name = NULL; 1838 if ( e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT ) { 1839 e.entry.address = 0; 1840 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 1841 import_name = data.GetCStr(&offset); 1842 } 1843 else { 1844 e.entry.address = data.GetULEB128(&offset); 1845 if ( e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER ) 1846 { 1847 //resolver_addresses.insert(e.entry.address); 1848 e.entry.other = data.GetULEB128(&offset); 1849 resolver_addresses.insert(e.entry.other); 1850 } 1851 else 1852 e.entry.other = 0; 1853 } 1854 // Only add symbols that are reexport symbols with a valid import name 1855 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && import_name[0]) 1856 { 1857 std::string name; 1858 if (!nameSlices.empty()) 1859 { 1860 for (auto name_slice: nameSlices) 1861 name.append(name_slice.data(), name_slice.size()); 1862 } 1863 if (name.size() > 1) 1864 { 1865 // Skip the leading '_' 1866 e.entry.name.SetCStringWithLength(name.c_str() + 1,name.size() - 1); 1867 } 1868 if (import_name) 1869 { 1870 // Skip the leading '_' 1871 e.entry.import_name.SetCString(import_name+1); 1872 } 1873 output.push_back(e); 1874 } 1875 } 1876 1877 const uint8_t childrenCount = data.GetU8(&children_offset); 1878 for (uint8_t i=0; i < childrenCount; ++i) { 1879 nameSlices.push_back(data.GetCStr(&children_offset)); 1880 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 1881 if (childNodeOffset) 1882 { 1883 ParseTrieEntries(data, 1884 childNodeOffset, 1885 nameSlices, 1886 resolver_addresses, 1887 output); 1888 } 1889 nameSlices.pop_back(); 1890 } 1891 } 1892 1893 size_t 1894 ObjectFileMachO::ParseSymtab () 1895 { 1896 Timer scoped_timer(__PRETTY_FUNCTION__, 1897 "ObjectFileMachO::ParseSymtab () module = %s", 1898 m_file.GetFilename().AsCString("")); 1899 ModuleSP module_sp (GetModule()); 1900 if (!module_sp) 1901 return 0; 1902 1903 struct symtab_command symtab_load_command = { 0, 0, 0, 0, 0, 0 }; 1904 struct linkedit_data_command function_starts_load_command = { 0, 0, 0, 0 }; 1905 struct dyld_info_command dyld_info = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }; 1906 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 1907 FunctionStarts function_starts; 1908 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1909 uint32_t i; 1910 FileSpecList dylib_files; 1911 Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_SYMBOLS)); 1912 1913 for (i=0; i<m_header.ncmds; ++i) 1914 { 1915 const lldb::offset_t cmd_offset = offset; 1916 // Read in the load command and load command size 1917 struct load_command lc; 1918 if (m_data.GetU32(&offset, &lc, 2) == NULL) 1919 break; 1920 // Watch for the symbol table load command 1921 switch (lc.cmd) 1922 { 1923 case LC_SYMTAB: 1924 symtab_load_command.cmd = lc.cmd; 1925 symtab_load_command.cmdsize = lc.cmdsize; 1926 // Read in the rest of the symtab load command 1927 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 0) // fill in symoff, nsyms, stroff, strsize fields 1928 return 0; 1929 if (symtab_load_command.symoff == 0) 1930 { 1931 if (log) 1932 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 1933 return 0; 1934 } 1935 1936 if (symtab_load_command.stroff == 0) 1937 { 1938 if (log) 1939 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 1940 return 0; 1941 } 1942 1943 if (symtab_load_command.nsyms == 0) 1944 { 1945 if (log) 1946 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 1947 return 0; 1948 } 1949 1950 if (symtab_load_command.strsize == 0) 1951 { 1952 if (log) 1953 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 1954 return 0; 1955 } 1956 break; 1957 1958 case LC_DYLD_INFO: 1959 case LC_DYLD_INFO_ONLY: 1960 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) 1961 { 1962 dyld_info.cmd = lc.cmd; 1963 dyld_info.cmdsize = lc.cmdsize; 1964 } 1965 else 1966 { 1967 memset (&dyld_info, 0, sizeof(dyld_info)); 1968 } 1969 break; 1970 1971 case LC_LOAD_DYLIB: 1972 case LC_LOAD_WEAK_DYLIB: 1973 case LC_REEXPORT_DYLIB: 1974 case LC_LOADFVMLIB: 1975 case LC_LOAD_UPWARD_DYLIB: 1976 { 1977 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 1978 const char *path = m_data.PeekCStr(name_offset); 1979 if (path) 1980 { 1981 FileSpec file_spec(path, false); 1982 // Strip the path if there is @rpath, @executable, etc so we just use the basename 1983 if (path[0] == '@') 1984 file_spec.GetDirectory().Clear(); 1985 1986 if (lc.cmd == LC_REEXPORT_DYLIB) 1987 { 1988 m_reexported_dylibs.AppendIfUnique(file_spec); 1989 } 1990 1991 dylib_files.Append(file_spec); 1992 } 1993 } 1994 break; 1995 1996 case LC_FUNCTION_STARTS: 1997 function_starts_load_command.cmd = lc.cmd; 1998 function_starts_load_command.cmdsize = lc.cmdsize; 1999 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == NULL) // fill in symoff, nsyms, stroff, strsize fields 2000 memset (&function_starts_load_command, 0, sizeof(function_starts_load_command)); 2001 break; 2002 2003 default: 2004 break; 2005 } 2006 offset = cmd_offset + lc.cmdsize; 2007 } 2008 2009 if (symtab_load_command.cmd) 2010 { 2011 Symtab *symtab = m_symtab_ap.get(); 2012 SectionList *section_list = GetSectionList(); 2013 if (section_list == NULL) 2014 return 0; 2015 2016 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2017 const ByteOrder byte_order = m_data.GetByteOrder(); 2018 bool bit_width_32 = addr_byte_size == 4; 2019 const size_t nlist_byte_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2020 2021 DataExtractor nlist_data (NULL, 0, byte_order, addr_byte_size); 2022 DataExtractor strtab_data (NULL, 0, byte_order, addr_byte_size); 2023 DataExtractor function_starts_data (NULL, 0, byte_order, addr_byte_size); 2024 DataExtractor indirect_symbol_index_data (NULL, 0, byte_order, addr_byte_size); 2025 DataExtractor dyld_trie_data (NULL, 0, byte_order, addr_byte_size); 2026 2027 const addr_t nlist_data_byte_size = symtab_load_command.nsyms * nlist_byte_size; 2028 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2029 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2030 2031 ProcessSP process_sp (m_process_wp.lock()); 2032 Process *process = process_sp.get(); 2033 2034 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2035 2036 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) 2037 { 2038 Target &target = process->GetTarget(); 2039 2040 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2041 2042 SectionSP linkedit_section_sp(section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2043 // Reading mach file from memory in a process or core file... 2044 2045 if (linkedit_section_sp) 2046 { 2047 const addr_t linkedit_load_addr = linkedit_section_sp->GetLoadBaseAddress(&target); 2048 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2049 const addr_t symoff_addr = linkedit_load_addr + symtab_load_command.symoff - linkedit_file_offset; 2050 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - linkedit_file_offset; 2051 2052 bool data_was_read = false; 2053 2054 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 2055 if (m_header.flags & 0x80000000u && process->GetAddressByteSize() == sizeof (void*)) 2056 { 2057 // This mach-o memory file is in the dyld shared cache. If this 2058 // program is not remote and this is iOS, then this process will 2059 // share the same shared cache as the process we are debugging and 2060 // we can read the entire __LINKEDIT from the address space in this 2061 // process. This is a needed optimization that is used for local iOS 2062 // debugging only since all shared libraries in the shared cache do 2063 // not have corresponding files that exist in the file system of the 2064 // device. They have been combined into a single file. This means we 2065 // always have to load these files from memory. All of the symbol and 2066 // string tables from all of the __LINKEDIT sections from the shared 2067 // libraries in the shared cache have been merged into a single large 2068 // symbol and string table. Reading all of this symbol and string table 2069 // data across can slow down debug launch times, so we optimize this by 2070 // reading the memory for the __LINKEDIT section from this process. 2071 2072 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 2073 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 2074 bool use_lldb_cache = true; 2075 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && lldb_shared_cache != process_shared_cache) 2076 { 2077 use_lldb_cache = false; 2078 ModuleSP module_sp (GetModule()); 2079 if (module_sp) 2080 module_sp->ReportWarning ("shared cache in process does not match lldb's own shared cache, startup will be slow."); 2081 2082 } 2083 2084 PlatformSP platform_sp (target.GetPlatform()); 2085 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) 2086 { 2087 data_was_read = true; 2088 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, eByteOrderLittle); 2089 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, eByteOrderLittle); 2090 if (function_starts_load_command.cmd) 2091 { 2092 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 2093 function_starts_data.SetData ((void *)func_start_addr, function_starts_load_command.datasize, eByteOrderLittle); 2094 } 2095 } 2096 } 2097 #endif 2098 2099 if (!data_was_read) 2100 { 2101 if (memory_module_load_level == eMemoryModuleLoadLevelComplete) 2102 { 2103 DataBufferSP nlist_data_sp (ReadMemory (process_sp, symoff_addr, nlist_data_byte_size)); 2104 if (nlist_data_sp) 2105 nlist_data.SetData (nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2106 // Load strings individually from memory when loading from memory since shared cache 2107 // string tables contain strings for all symbols from all shared cached libraries 2108 //DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, strtab_data_byte_size)); 2109 //if (strtab_data_sp) 2110 // strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 2111 if (m_dysymtab.nindirectsyms != 0) 2112 { 2113 const addr_t indirect_syms_addr = linkedit_load_addr + m_dysymtab.indirectsymoff - linkedit_file_offset; 2114 DataBufferSP indirect_syms_data_sp (ReadMemory (process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 2115 if (indirect_syms_data_sp) 2116 indirect_symbol_index_data.SetData (indirect_syms_data_sp, 0, indirect_syms_data_sp->GetByteSize()); 2117 } 2118 } 2119 2120 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) 2121 { 2122 if (function_starts_load_command.cmd) 2123 { 2124 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 2125 DataBufferSP func_start_data_sp (ReadMemory (process_sp, func_start_addr, function_starts_load_command.datasize)); 2126 if (func_start_data_sp) 2127 function_starts_data.SetData (func_start_data_sp, 0, func_start_data_sp->GetByteSize()); 2128 } 2129 } 2130 } 2131 } 2132 } 2133 else 2134 { 2135 nlist_data.SetData (m_data, 2136 symtab_load_command.symoff, 2137 nlist_data_byte_size); 2138 strtab_data.SetData (m_data, 2139 symtab_load_command.stroff, 2140 strtab_data_byte_size); 2141 2142 if (dyld_info.export_size > 0) 2143 { 2144 dyld_trie_data.SetData (m_data, 2145 dyld_info.export_off, 2146 dyld_info.export_size); 2147 } 2148 2149 if (m_dysymtab.nindirectsyms != 0) 2150 { 2151 indirect_symbol_index_data.SetData (m_data, 2152 m_dysymtab.indirectsymoff, 2153 m_dysymtab.nindirectsyms * 4); 2154 } 2155 if (function_starts_load_command.cmd) 2156 { 2157 function_starts_data.SetData (m_data, 2158 function_starts_load_command.dataoff, 2159 function_starts_load_command.datasize); 2160 } 2161 } 2162 2163 if (nlist_data.GetByteSize() == 0 && memory_module_load_level == eMemoryModuleLoadLevelComplete) 2164 { 2165 if (log) 2166 module_sp->LogMessage(log, "failed to read nlist data"); 2167 return 0; 2168 } 2169 2170 2171 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2172 if (!have_strtab_data) 2173 { 2174 if (process) 2175 { 2176 if (strtab_addr == LLDB_INVALID_ADDRESS) 2177 { 2178 if (log) 2179 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2180 return 0; 2181 } 2182 } 2183 else 2184 { 2185 if (log) 2186 module_sp->LogMessage(log, "failed to read strtab data"); 2187 return 0; 2188 } 2189 } 2190 2191 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2192 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2193 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2194 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2195 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT)); 2196 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA)); 2197 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC)); 2198 SectionSP eh_frame_section_sp; 2199 if (text_section_sp.get()) 2200 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame); 2201 else 2202 eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame); 2203 2204 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2205 2206 // lldb works best if it knows the start address of all functions in a module. 2207 // Linker symbols or debug info are normally the best source of information for start addr / size but 2208 // they may be stripped in a released binary. 2209 // Two additional sources of information exist in Mach-O binaries: 2210 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each function's start address in the 2211 // binary, relative to the text section. 2212 // eh_frame - the eh_frame FDEs have the start addr & size of each function 2213 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on all modern binaries. 2214 // Binaries built to run on older releases may need to use eh_frame information. 2215 2216 if (text_section_sp && function_starts_data.GetByteSize()) 2217 { 2218 FunctionStarts::Entry function_start_entry; 2219 function_start_entry.data = false; 2220 lldb::offset_t function_start_offset = 0; 2221 function_start_entry.addr = text_section_sp->GetFileAddress(); 2222 uint64_t delta; 2223 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 0) 2224 { 2225 // Now append the current entry 2226 function_start_entry.addr += delta; 2227 function_starts.Append(function_start_entry); 2228 } 2229 } 2230 else 2231 { 2232 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the load command claiming an eh_frame 2233 // but it doesn't actually have the eh_frame content. And if we have a dSYM, we don't need to do any 2234 // of this fill-in-the-missing-symbols works anyway - the debug info should give us all the functions in 2235 // the module. 2236 if (text_section_sp.get() && eh_frame_section_sp.get() && m_type != eTypeDebugInfo) 2237 { 2238 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, eRegisterKindGCC, true); 2239 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2240 eh_frame.GetFunctionAddressAndSizeVector (functions); 2241 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2242 size_t count = functions.GetSize(); 2243 for (size_t i = 0; i < count; ++i) 2244 { 2245 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = functions.GetEntryAtIndex (i); 2246 if (func) 2247 { 2248 FunctionStarts::Entry function_start_entry; 2249 function_start_entry.addr = func->base - text_base_addr; 2250 function_starts.Append(function_start_entry); 2251 } 2252 } 2253 } 2254 } 2255 2256 const size_t function_starts_count = function_starts.GetSize(); 2257 2258 const user_id_t TEXT_eh_frame_sectID = 2259 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2260 : static_cast<user_id_t>(NO_SECT); 2261 2262 lldb::offset_t nlist_data_offset = 0; 2263 2264 uint32_t N_SO_index = UINT32_MAX; 2265 2266 MachSymtabSectionInfo section_info (section_list); 2267 std::vector<uint32_t> N_FUN_indexes; 2268 std::vector<uint32_t> N_NSYM_indexes; 2269 std::vector<uint32_t> N_INCL_indexes; 2270 std::vector<uint32_t> N_BRAC_indexes; 2271 std::vector<uint32_t> N_COMM_indexes; 2272 typedef std::multimap <uint64_t, uint32_t> ValueToSymbolIndexMap; 2273 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2274 typedef std::map <const char *, uint32_t> ConstNameToSymbolIndexMap; 2275 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2276 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2277 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2278 // Any symbols that get merged into another will get an entry 2279 // in this map so we know 2280 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2281 uint32_t nlist_idx = 0; 2282 Symbol *symbol_ptr = NULL; 2283 2284 uint32_t sym_idx = 0; 2285 Symbol *sym = NULL; 2286 size_t num_syms = 0; 2287 std::string memory_symbol_name; 2288 uint32_t unmapped_local_symbols_found = 0; 2289 2290 std::vector<TrieEntryWithOffset> trie_entries; 2291 std::set<lldb::addr_t> resolver_addresses; 2292 2293 if (dyld_trie_data.GetByteSize() > 0) 2294 { 2295 std::vector<llvm::StringRef> nameSlices; 2296 ParseTrieEntries (dyld_trie_data, 2297 0, 2298 nameSlices, 2299 resolver_addresses, 2300 trie_entries); 2301 2302 ConstString text_segment_name ("__TEXT"); 2303 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 2304 if (text_segment_sp) 2305 { 2306 const lldb::addr_t text_segment_file_addr = text_segment_sp->GetFileAddress(); 2307 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) 2308 { 2309 for (auto &e : trie_entries) 2310 e.entry.address += text_segment_file_addr; 2311 } 2312 } 2313 } 2314 2315 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 2316 2317 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been optimized by moving LOCAL 2318 // symbols out of the memory mapped portion of the DSC. The symbol information has all been retained, 2319 // but it isn't available in the normal nlist data. However, there *are* duplicate entries of *some* 2320 // LOCAL symbols in the normal nlist data. To handle this situation correctly, we must first attempt 2321 // to parse any DSC unmapped symbol information. If we find any, we set a flag that tells the normal 2322 // nlist parser to ignore all LOCAL symbols. 2323 2324 if (m_header.flags & 0x80000000u) 2325 { 2326 // Before we can start mapping the DSC, we need to make certain the target process is actually 2327 // using the cache we can find. 2328 2329 // Next we need to determine the correct path for the dyld shared cache. 2330 2331 ArchSpec header_arch; 2332 GetArchitecture(header_arch); 2333 char dsc_path[PATH_MAX]; 2334 2335 snprintf(dsc_path, sizeof(dsc_path), "%s%s%s", 2336 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR */ 2337 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2338 header_arch.GetArchitectureName()); 2339 2340 FileSpec dsc_filespec(dsc_path, false); 2341 2342 // We need definitions of two structures in the on-disk DSC, copy them here manually 2343 struct lldb_copy_dyld_cache_header_v0 2344 { 2345 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 2346 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 2347 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 2348 uint32_t imagesOffset; 2349 uint32_t imagesCount; 2350 uint64_t dyldBaseAddress; 2351 uint64_t codeSignatureOffset; 2352 uint64_t codeSignatureSize; 2353 uint64_t slideInfoOffset; 2354 uint64_t slideInfoSize; 2355 uint64_t localSymbolsOffset; // file offset of where local symbols are stored 2356 uint64_t localSymbolsSize; // size of local symbols information 2357 }; 2358 struct lldb_copy_dyld_cache_header_v1 2359 { 2360 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 2361 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 2362 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 2363 uint32_t imagesOffset; 2364 uint32_t imagesCount; 2365 uint64_t dyldBaseAddress; 2366 uint64_t codeSignatureOffset; 2367 uint64_t codeSignatureSize; 2368 uint64_t slideInfoOffset; 2369 uint64_t slideInfoSize; 2370 uint64_t localSymbolsOffset; 2371 uint64_t localSymbolsSize; 2372 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 and later 2373 }; 2374 2375 struct lldb_copy_dyld_cache_mapping_info 2376 { 2377 uint64_t address; 2378 uint64_t size; 2379 uint64_t fileOffset; 2380 uint32_t maxProt; 2381 uint32_t initProt; 2382 }; 2383 2384 struct lldb_copy_dyld_cache_local_symbols_info 2385 { 2386 uint32_t nlistOffset; 2387 uint32_t nlistCount; 2388 uint32_t stringsOffset; 2389 uint32_t stringsSize; 2390 uint32_t entriesOffset; 2391 uint32_t entriesCount; 2392 }; 2393 struct lldb_copy_dyld_cache_local_symbols_entry 2394 { 2395 uint32_t dylibOffset; 2396 uint32_t nlistStartIndex; 2397 uint32_t nlistCount; 2398 }; 2399 2400 /* The dyld_cache_header has a pointer to the dyld_cache_local_symbols_info structure (localSymbolsOffset). 2401 The dyld_cache_local_symbols_info structure gives us three things: 2402 1. The start and count of the nlist records in the dyld_shared_cache file 2403 2. The start and size of the strings for these nlist records 2404 3. The start and count of dyld_cache_local_symbols_entry entries 2405 2406 There is one dyld_cache_local_symbols_entry per dylib/framework in the dyld shared cache. 2407 The "dylibOffset" field is the Mach-O header of this dylib/framework in the dyld shared cache. 2408 The dyld_cache_local_symbols_entry also lists the start of this dylib/framework's nlist records 2409 and the count of how many nlist records there are for this dylib/framework. 2410 */ 2411 2412 // Process the dsc header to find the unmapped symbols 2413 // 2414 // Save some VM space, do not map the entire cache in one shot. 2415 2416 DataBufferSP dsc_data_sp; 2417 dsc_data_sp = dsc_filespec.MemoryMapFileContents(0, sizeof(struct lldb_copy_dyld_cache_header_v1)); 2418 2419 if (dsc_data_sp) 2420 { 2421 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2422 2423 char version_str[17]; 2424 int version = -1; 2425 lldb::offset_t offset = 0; 2426 memcpy (version_str, dsc_header_data.GetData (&offset, 16), 16); 2427 version_str[16] = '\0'; 2428 if (strncmp (version_str, "dyld_v", 6) == 0 && isdigit (version_str[6])) 2429 { 2430 int v; 2431 if (::sscanf (version_str + 6, "%d", &v) == 1) 2432 { 2433 version = v; 2434 } 2435 } 2436 2437 UUID dsc_uuid; 2438 if (version >= 1) 2439 { 2440 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, uuid); 2441 uint8_t uuid_bytes[sizeof (uuid_t)]; 2442 memcpy (uuid_bytes, dsc_header_data.GetData (&offset, sizeof (uuid_t)), sizeof (uuid_t)); 2443 dsc_uuid.SetBytes (uuid_bytes); 2444 } 2445 2446 bool uuid_match = true; 2447 if (dsc_uuid.IsValid() && process) 2448 { 2449 UUID shared_cache_uuid(GetProcessSharedCacheUUID(process)); 2450 2451 if (shared_cache_uuid.IsValid() && dsc_uuid != shared_cache_uuid) 2452 { 2453 // The on-disk dyld_shared_cache file is not the same as the one in this 2454 // process' memory, don't use it. 2455 uuid_match = false; 2456 ModuleSP module_sp (GetModule()); 2457 if (module_sp) 2458 module_sp->ReportWarning ("process shared cache does not match on-disk dyld_shared_cache file, some symbol names will be missing."); 2459 } 2460 } 2461 2462 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2463 2464 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2465 2466 // If the mappingOffset points to a location inside the header, we've 2467 // opened an old dyld shared cache, and should not proceed further. 2468 if (uuid_match && mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v0)) 2469 { 2470 2471 DataBufferSP dsc_mapping_info_data_sp = dsc_filespec.MemoryMapFileContents(mappingOffset, sizeof (struct lldb_copy_dyld_cache_mapping_info)); 2472 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, byte_order, addr_byte_size); 2473 offset = 0; 2474 2475 // The File addresses (from the in-memory Mach-O load commands) for the shared libraries 2476 // in the shared library cache need to be adjusted by an offset to match up with the 2477 // dylibOffset identifying field in the dyld_cache_local_symbol_entry's. This offset is 2478 // recorded in mapping_offset_value. 2479 const uint64_t mapping_offset_value = dsc_mapping_info_data.GetU64(&offset); 2480 2481 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2482 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2483 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2484 2485 if (localSymbolsOffset && localSymbolsSize) 2486 { 2487 // Map the local symbols 2488 if (DataBufferSP dsc_local_symbols_data_sp = dsc_filespec.MemoryMapFileContents(localSymbolsOffset, localSymbolsSize)) 2489 { 2490 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, byte_order, addr_byte_size); 2491 2492 offset = 0; 2493 2494 // Read the local_symbols_infos struct in one shot 2495 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2496 dsc_local_symbols_data.GetU32(&offset, &local_symbols_info.nlistOffset, 6); 2497 2498 SectionSP text_section_sp(section_list->FindSectionByName(GetSegmentNameTEXT())); 2499 2500 uint32_t header_file_offset = (text_section_sp->GetFileAddress() - mapping_offset_value); 2501 2502 offset = local_symbols_info.entriesOffset; 2503 for (uint32_t entry_index = 0; entry_index < local_symbols_info.entriesCount; entry_index++) 2504 { 2505 struct lldb_copy_dyld_cache_local_symbols_entry local_symbols_entry; 2506 local_symbols_entry.dylibOffset = dsc_local_symbols_data.GetU32(&offset); 2507 local_symbols_entry.nlistStartIndex = dsc_local_symbols_data.GetU32(&offset); 2508 local_symbols_entry.nlistCount = dsc_local_symbols_data.GetU32(&offset); 2509 2510 if (header_file_offset == local_symbols_entry.dylibOffset) 2511 { 2512 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2513 2514 // The normal nlist code cannot correctly size the Symbols array, we need to allocate it here. 2515 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms + unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2516 num_syms = symtab->GetNumSymbols(); 2517 2518 nlist_data_offset = local_symbols_info.nlistOffset + (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2519 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2520 2521 for (uint32_t nlist_index = 0; nlist_index < local_symbols_entry.nlistCount; nlist_index++) 2522 { 2523 ///////////////////////////// 2524 { 2525 struct nlist_64 nlist; 2526 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2527 break; 2528 2529 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked(&nlist_data_offset); 2530 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2531 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2532 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked (&nlist_data_offset); 2533 nlist.n_value = dsc_local_symbols_data.GetAddress_unchecked (&nlist_data_offset); 2534 2535 SymbolType type = eSymbolTypeInvalid; 2536 const char *symbol_name = dsc_local_symbols_data.PeekCStr(string_table_offset + nlist.n_strx); 2537 2538 if (symbol_name == NULL) 2539 { 2540 // No symbol should be NULL, even the symbols with no 2541 // string values should have an offset zero which points 2542 // to an empty C-string 2543 Host::SystemLog (Host::eSystemLogError, 2544 "error: DSC unmapped local symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 2545 entry_index, 2546 nlist.n_strx, 2547 module_sp->GetFileSpec().GetPath().c_str()); 2548 continue; 2549 } 2550 if (symbol_name[0] == '\0') 2551 symbol_name = NULL; 2552 2553 const char *symbol_name_non_abi_mangled = NULL; 2554 2555 SectionSP symbol_section; 2556 uint32_t symbol_byte_size = 0; 2557 bool add_nlist = true; 2558 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2559 bool demangled_is_synthesized = false; 2560 bool is_gsym = false; 2561 2562 assert (sym_idx < num_syms); 2563 2564 sym[sym_idx].SetDebug (is_debug); 2565 2566 if (is_debug) 2567 { 2568 switch (nlist.n_type) 2569 { 2570 case N_GSYM: 2571 // global symbol: name,,NO_SECT,type,0 2572 // Sometimes the N_GSYM value contains the address. 2573 2574 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 2575 // have the same address, but we want to ensure that we always find only the real symbol, 2576 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 2577 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 2578 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 2579 // same address. 2580 2581 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 2582 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 2583 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 2584 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 2585 add_nlist = false; 2586 else 2587 { 2588 is_gsym = true; 2589 sym[sym_idx].SetExternal(true); 2590 if (nlist.n_value != 0) 2591 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2592 type = eSymbolTypeData; 2593 } 2594 break; 2595 2596 case N_FNAME: 2597 // procedure name (f77 kludge): name,,NO_SECT,0,0 2598 type = eSymbolTypeCompiler; 2599 break; 2600 2601 case N_FUN: 2602 // procedure: name,,n_sect,linenumber,address 2603 if (symbol_name) 2604 { 2605 type = eSymbolTypeCode; 2606 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2607 2608 N_FUN_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 2609 // We use the current number of symbols in the symbol table in lieu of 2610 // using nlist_idx in case we ever start trimming entries out 2611 N_FUN_indexes.push_back(sym_idx); 2612 } 2613 else 2614 { 2615 type = eSymbolTypeCompiler; 2616 2617 if ( !N_FUN_indexes.empty() ) 2618 { 2619 // Copy the size of the function into the original STAB entry so we don't have 2620 // to hunt for it later 2621 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 2622 N_FUN_indexes.pop_back(); 2623 // We don't really need the end function STAB as it contains the size which 2624 // we already placed with the original symbol, so don't add it if we want a 2625 // minimal symbol table 2626 add_nlist = false; 2627 } 2628 } 2629 break; 2630 2631 case N_STSYM: 2632 // static symbol: name,,n_sect,type,address 2633 N_STSYM_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 2634 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2635 type = eSymbolTypeData; 2636 break; 2637 2638 case N_LCSYM: 2639 // .lcomm symbol: name,,n_sect,type,address 2640 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2641 type = eSymbolTypeCommonBlock; 2642 break; 2643 2644 case N_BNSYM: 2645 // We use the current number of symbols in the symbol table in lieu of 2646 // using nlist_idx in case we ever start trimming entries out 2647 // Skip these if we want minimal symbol tables 2648 add_nlist = false; 2649 break; 2650 2651 case N_ENSYM: 2652 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 2653 // so that we can always skip the entire symbol if we need to navigate 2654 // more quickly at the source level when parsing STABS 2655 // Skip these if we want minimal symbol tables 2656 add_nlist = false; 2657 break; 2658 2659 2660 case N_OPT: 2661 // emitted with gcc2_compiled and in gcc source 2662 type = eSymbolTypeCompiler; 2663 break; 2664 2665 case N_RSYM: 2666 // register sym: name,,NO_SECT,type,register 2667 type = eSymbolTypeVariable; 2668 break; 2669 2670 case N_SLINE: 2671 // src line: 0,,n_sect,linenumber,address 2672 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2673 type = eSymbolTypeLineEntry; 2674 break; 2675 2676 case N_SSYM: 2677 // structure elt: name,,NO_SECT,type,struct_offset 2678 type = eSymbolTypeVariableType; 2679 break; 2680 2681 case N_SO: 2682 // source file name 2683 type = eSymbolTypeSourceFile; 2684 if (symbol_name == NULL) 2685 { 2686 add_nlist = false; 2687 if (N_SO_index != UINT32_MAX) 2688 { 2689 // Set the size of the N_SO to the terminating index of this N_SO 2690 // so that we can always skip the entire N_SO if we need to navigate 2691 // more quickly at the source level when parsing STABS 2692 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2693 symbol_ptr->SetByteSize(sym_idx); 2694 symbol_ptr->SetSizeIsSibling(true); 2695 } 2696 N_NSYM_indexes.clear(); 2697 N_INCL_indexes.clear(); 2698 N_BRAC_indexes.clear(); 2699 N_COMM_indexes.clear(); 2700 N_FUN_indexes.clear(); 2701 N_SO_index = UINT32_MAX; 2702 } 2703 else 2704 { 2705 // We use the current number of symbols in the symbol table in lieu of 2706 // using nlist_idx in case we ever start trimming entries out 2707 const bool N_SO_has_full_path = symbol_name[0] == '/'; 2708 if (N_SO_has_full_path) 2709 { 2710 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2711 { 2712 // We have two consecutive N_SO entries where the first contains a directory 2713 // and the second contains a full path. 2714 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 2715 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2716 add_nlist = false; 2717 } 2718 else 2719 { 2720 // This is the first entry in a N_SO that contains a directory or 2721 // a full path to the source file 2722 N_SO_index = sym_idx; 2723 } 2724 } 2725 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2726 { 2727 // This is usually the second N_SO entry that contains just the filename, 2728 // so here we combine it with the first one if we are minimizing the symbol table 2729 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 2730 if (so_path && so_path[0]) 2731 { 2732 std::string full_so_path (so_path); 2733 const size_t double_slash_pos = full_so_path.find("//"); 2734 if (double_slash_pos != std::string::npos) 2735 { 2736 // The linker has been generating bad N_SO entries with doubled up paths 2737 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 2738 // and the second is the directory for the source file so you end up with 2739 // a path that looks like "/tmp/src//tmp/src/" 2740 FileSpec so_dir(so_path, false); 2741 if (!so_dir.Exists()) 2742 { 2743 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 2744 if (so_dir.Exists()) 2745 { 2746 // Trim off the incorrect path 2747 full_so_path.erase(0, double_slash_pos + 1); 2748 } 2749 } 2750 } 2751 if (*full_so_path.rbegin() != '/') 2752 full_so_path += '/'; 2753 full_so_path += symbol_name; 2754 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 2755 add_nlist = false; 2756 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2757 } 2758 } 2759 else 2760 { 2761 // This could be a relative path to a N_SO 2762 N_SO_index = sym_idx; 2763 } 2764 } 2765 break; 2766 2767 case N_OSO: 2768 // object file name: name,,0,0,st_mtime 2769 type = eSymbolTypeObjectFile; 2770 break; 2771 2772 case N_LSYM: 2773 // local sym: name,,NO_SECT,type,offset 2774 type = eSymbolTypeLocal; 2775 break; 2776 2777 //---------------------------------------------------------------------- 2778 // INCL scopes 2779 //---------------------------------------------------------------------- 2780 case N_BINCL: 2781 // include file beginning: name,,NO_SECT,0,sum 2782 // We use the current number of symbols in the symbol table in lieu of 2783 // using nlist_idx in case we ever start trimming entries out 2784 N_INCL_indexes.push_back(sym_idx); 2785 type = eSymbolTypeScopeBegin; 2786 break; 2787 2788 case N_EINCL: 2789 // include file end: name,,NO_SECT,0,0 2790 // Set the size of the N_BINCL to the terminating index of this N_EINCL 2791 // so that we can always skip the entire symbol if we need to navigate 2792 // more quickly at the source level when parsing STABS 2793 if ( !N_INCL_indexes.empty() ) 2794 { 2795 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 2796 symbol_ptr->SetByteSize(sym_idx + 1); 2797 symbol_ptr->SetSizeIsSibling(true); 2798 N_INCL_indexes.pop_back(); 2799 } 2800 type = eSymbolTypeScopeEnd; 2801 break; 2802 2803 case N_SOL: 2804 // #included file name: name,,n_sect,0,address 2805 type = eSymbolTypeHeaderFile; 2806 2807 // We currently don't use the header files on darwin 2808 add_nlist = false; 2809 break; 2810 2811 case N_PARAMS: 2812 // compiler parameters: name,,NO_SECT,0,0 2813 type = eSymbolTypeCompiler; 2814 break; 2815 2816 case N_VERSION: 2817 // compiler version: name,,NO_SECT,0,0 2818 type = eSymbolTypeCompiler; 2819 break; 2820 2821 case N_OLEVEL: 2822 // compiler -O level: name,,NO_SECT,0,0 2823 type = eSymbolTypeCompiler; 2824 break; 2825 2826 case N_PSYM: 2827 // parameter: name,,NO_SECT,type,offset 2828 type = eSymbolTypeVariable; 2829 break; 2830 2831 case N_ENTRY: 2832 // alternate entry: name,,n_sect,linenumber,address 2833 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2834 type = eSymbolTypeLineEntry; 2835 break; 2836 2837 //---------------------------------------------------------------------- 2838 // Left and Right Braces 2839 //---------------------------------------------------------------------- 2840 case N_LBRAC: 2841 // left bracket: 0,,NO_SECT,nesting level,address 2842 // We use the current number of symbols in the symbol table in lieu of 2843 // using nlist_idx in case we ever start trimming entries out 2844 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2845 N_BRAC_indexes.push_back(sym_idx); 2846 type = eSymbolTypeScopeBegin; 2847 break; 2848 2849 case N_RBRAC: 2850 // right bracket: 0,,NO_SECT,nesting level,address 2851 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 2852 // so that we can always skip the entire symbol if we need to navigate 2853 // more quickly at the source level when parsing STABS 2854 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2855 if ( !N_BRAC_indexes.empty() ) 2856 { 2857 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 2858 symbol_ptr->SetByteSize(sym_idx + 1); 2859 symbol_ptr->SetSizeIsSibling(true); 2860 N_BRAC_indexes.pop_back(); 2861 } 2862 type = eSymbolTypeScopeEnd; 2863 break; 2864 2865 case N_EXCL: 2866 // deleted include file: name,,NO_SECT,0,sum 2867 type = eSymbolTypeHeaderFile; 2868 break; 2869 2870 //---------------------------------------------------------------------- 2871 // COMM scopes 2872 //---------------------------------------------------------------------- 2873 case N_BCOMM: 2874 // begin common: name,,NO_SECT,0,0 2875 // We use the current number of symbols in the symbol table in lieu of 2876 // using nlist_idx in case we ever start trimming entries out 2877 type = eSymbolTypeScopeBegin; 2878 N_COMM_indexes.push_back(sym_idx); 2879 break; 2880 2881 case N_ECOML: 2882 // end common (local name): 0,,n_sect,0,address 2883 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2884 // Fall through 2885 2886 case N_ECOMM: 2887 // end common: name,,n_sect,0,0 2888 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 2889 // so that we can always skip the entire symbol if we need to navigate 2890 // more quickly at the source level when parsing STABS 2891 if ( !N_COMM_indexes.empty() ) 2892 { 2893 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 2894 symbol_ptr->SetByteSize(sym_idx + 1); 2895 symbol_ptr->SetSizeIsSibling(true); 2896 N_COMM_indexes.pop_back(); 2897 } 2898 type = eSymbolTypeScopeEnd; 2899 break; 2900 2901 case N_LENG: 2902 // second stab entry with length information 2903 type = eSymbolTypeAdditional; 2904 break; 2905 2906 default: break; 2907 } 2908 } 2909 else 2910 { 2911 //uint8_t n_pext = N_PEXT & nlist.n_type; 2912 uint8_t n_type = N_TYPE & nlist.n_type; 2913 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 2914 2915 switch (n_type) 2916 { 2917 case N_INDR: // Fall through 2918 case N_PBUD: // Fall through 2919 case N_UNDF: 2920 type = eSymbolTypeUndefined; 2921 break; 2922 2923 case N_ABS: 2924 type = eSymbolTypeAbsolute; 2925 break; 2926 2927 case N_SECT: 2928 { 2929 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2930 2931 if (symbol_section == NULL) 2932 { 2933 // TODO: warn about this? 2934 add_nlist = false; 2935 break; 2936 } 2937 2938 if (TEXT_eh_frame_sectID == nlist.n_sect) 2939 { 2940 type = eSymbolTypeException; 2941 } 2942 else 2943 { 2944 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 2945 2946 switch (section_type) 2947 { 2948 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 2949 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 2950 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 2951 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 2952 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 2953 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 2954 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 2955 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 2956 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 2957 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 2958 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 2959 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 2960 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 2961 default: 2962 switch (symbol_section->GetType()) 2963 { 2964 case lldb::eSectionTypeCode: 2965 type = eSymbolTypeCode; 2966 break; 2967 case eSectionTypeData: 2968 case eSectionTypeDataCString: // Inlined C string data 2969 case eSectionTypeDataCStringPointers: // Pointers to C string data 2970 case eSectionTypeDataSymbolAddress: // Address of a symbol in the symbol table 2971 case eSectionTypeData4: 2972 case eSectionTypeData8: 2973 case eSectionTypeData16: 2974 type = eSymbolTypeData; 2975 break; 2976 default: 2977 break; 2978 } 2979 break; 2980 } 2981 2982 if (type == eSymbolTypeInvalid) 2983 { 2984 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 2985 if (symbol_section->IsDescendant (text_section_sp.get())) 2986 { 2987 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 2988 S_ATTR_SELF_MODIFYING_CODE | 2989 S_ATTR_SOME_INSTRUCTIONS)) 2990 type = eSymbolTypeData; 2991 else 2992 type = eSymbolTypeCode; 2993 } 2994 else if (symbol_section->IsDescendant(data_section_sp.get())) 2995 { 2996 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 2997 { 2998 type = eSymbolTypeRuntime; 2999 3000 if (symbol_name && 3001 symbol_name[0] == '_' && 3002 symbol_name[1] == 'O' && 3003 symbol_name[2] == 'B') 3004 { 3005 llvm::StringRef symbol_name_ref(symbol_name); 3006 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 3007 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 3008 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 3009 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 3010 { 3011 symbol_name_non_abi_mangled = symbol_name + 1; 3012 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3013 type = eSymbolTypeObjCClass; 3014 demangled_is_synthesized = true; 3015 } 3016 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 3017 { 3018 symbol_name_non_abi_mangled = symbol_name + 1; 3019 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3020 type = eSymbolTypeObjCMetaClass; 3021 demangled_is_synthesized = true; 3022 } 3023 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 3024 { 3025 symbol_name_non_abi_mangled = symbol_name + 1; 3026 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3027 type = eSymbolTypeObjCIVar; 3028 demangled_is_synthesized = true; 3029 } 3030 } 3031 } 3032 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 3033 { 3034 type = eSymbolTypeException; 3035 } 3036 else 3037 { 3038 type = eSymbolTypeData; 3039 } 3040 } 3041 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 3042 { 3043 type = eSymbolTypeTrampoline; 3044 } 3045 else if (symbol_section->IsDescendant(objc_section_sp.get())) 3046 { 3047 type = eSymbolTypeRuntime; 3048 if (symbol_name && symbol_name[0] == '.') 3049 { 3050 llvm::StringRef symbol_name_ref(symbol_name); 3051 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 3052 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 3053 { 3054 symbol_name_non_abi_mangled = symbol_name; 3055 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 3056 type = eSymbolTypeObjCClass; 3057 demangled_is_synthesized = true; 3058 } 3059 } 3060 } 3061 } 3062 } 3063 } 3064 break; 3065 } 3066 } 3067 3068 if (add_nlist) 3069 { 3070 uint64_t symbol_value = nlist.n_value; 3071 if (symbol_name_non_abi_mangled) 3072 { 3073 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 3074 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 3075 } 3076 else 3077 { 3078 bool symbol_name_is_mangled = false; 3079 3080 if (symbol_name && symbol_name[0] == '_') 3081 { 3082 symbol_name_is_mangled = symbol_name[1] == '_'; 3083 symbol_name++; // Skip the leading underscore 3084 } 3085 3086 if (symbol_name) 3087 { 3088 ConstString const_symbol_name(symbol_name); 3089 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 3090 if (is_gsym && is_debug) 3091 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 3092 } 3093 } 3094 if (symbol_section) 3095 { 3096 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3097 if (symbol_byte_size == 0 && function_starts_count > 0) 3098 { 3099 addr_t symbol_lookup_file_addr = nlist.n_value; 3100 // Do an exact address match for non-ARM addresses, else get the closest since 3101 // the symbol might be a thumb symbol which has an address with bit zero set 3102 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 3103 if (is_arm && func_start_entry) 3104 { 3105 // Verify that the function start address is the symbol address (ARM) 3106 // or the symbol address + 1 (thumb) 3107 if (func_start_entry->addr != symbol_lookup_file_addr && 3108 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 3109 { 3110 // Not the right entry, NULL it out... 3111 func_start_entry = NULL; 3112 } 3113 } 3114 if (func_start_entry) 3115 { 3116 func_start_entry->data = true; 3117 3118 addr_t symbol_file_addr = func_start_entry->addr; 3119 uint32_t symbol_flags = 0; 3120 if (is_arm) 3121 { 3122 if (symbol_file_addr & 1) 3123 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3124 symbol_file_addr &= 0xfffffffffffffffeull; 3125 } 3126 3127 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3128 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3129 if (next_func_start_entry) 3130 { 3131 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3132 // Be sure the clear the Thumb address bit when we calculate the size 3133 // from the current and next address 3134 if (is_arm) 3135 next_symbol_file_addr &= 0xfffffffffffffffeull; 3136 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3137 } 3138 else 3139 { 3140 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3141 } 3142 } 3143 } 3144 symbol_value -= section_file_addr; 3145 } 3146 3147 if (is_debug == false) 3148 { 3149 if (type == eSymbolTypeCode) 3150 { 3151 // See if we can find a N_FUN entry for any code symbols. 3152 // If we do find a match, and the name matches, then we 3153 // can merge the two into just the function symbol to avoid 3154 // duplicate entries in the symbol table 3155 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3156 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3157 if (range.first != range.second) 3158 { 3159 bool found_it = false; 3160 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3161 { 3162 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3163 { 3164 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3165 // We just need the flags from the linker symbol, so put these flags 3166 // into the N_FUN flags to avoid duplicate symbols in the symbol table 3167 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3168 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3169 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3170 sym[pos->second].SetType (eSymbolTypeResolver); 3171 sym[sym_idx].Clear(); 3172 found_it = true; 3173 break; 3174 } 3175 } 3176 if (found_it) 3177 continue; 3178 } 3179 else 3180 { 3181 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3182 type = eSymbolTypeResolver; 3183 } 3184 } 3185 else if (type == eSymbolTypeData) 3186 { 3187 // See if we can find a N_STSYM entry for any data symbols. 3188 // If we do find a match, and the name matches, then we 3189 // can merge the two into just the Static symbol to avoid 3190 // duplicate entries in the symbol table 3191 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3192 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 3193 if (range.first != range.second) 3194 { 3195 bool found_it = false; 3196 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3197 { 3198 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3199 { 3200 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3201 // We just need the flags from the linker symbol, so put these flags 3202 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3203 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3204 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3205 sym[sym_idx].Clear(); 3206 found_it = true; 3207 break; 3208 } 3209 } 3210 if (found_it) 3211 continue; 3212 } 3213 else 3214 { 3215 // Combine N_GSYM stab entries with the non stab symbol 3216 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 3217 if (pos != N_GSYM_name_to_sym_idx.end()) 3218 { 3219 const uint32_t GSYM_sym_idx = pos->second; 3220 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 3221 // Copy the address, because often the N_GSYM address has an invalid address of zero 3222 // when the global is a common symbol 3223 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 3224 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 3225 // We just need the flags from the linker symbol, so put these flags 3226 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3227 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3228 sym[sym_idx].Clear(); 3229 continue; 3230 } 3231 } 3232 } 3233 } 3234 3235 sym[sym_idx].SetID (nlist_idx); 3236 sym[sym_idx].SetType (type); 3237 sym[sym_idx].GetAddress().SetSection (symbol_section); 3238 sym[sym_idx].GetAddress().SetOffset (symbol_value); 3239 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3240 3241 if (symbol_byte_size > 0) 3242 sym[sym_idx].SetByteSize(symbol_byte_size); 3243 3244 if (demangled_is_synthesized) 3245 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3246 ++sym_idx; 3247 } 3248 else 3249 { 3250 sym[sym_idx].Clear(); 3251 } 3252 3253 } 3254 ///////////////////////////// 3255 } 3256 break; // No more entries to consider 3257 } 3258 } 3259 } 3260 } 3261 } 3262 } 3263 } 3264 3265 // Must reset this in case it was mutated above! 3266 nlist_data_offset = 0; 3267 #endif 3268 3269 if (nlist_data.GetByteSize() > 0) 3270 { 3271 3272 // If the sym array was not created while parsing the DSC unmapped 3273 // symbols, create it now. 3274 if (sym == NULL) 3275 { 3276 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3277 num_syms = symtab->GetNumSymbols(); 3278 } 3279 3280 if (unmapped_local_symbols_found) 3281 { 3282 assert(m_dysymtab.ilocalsym == 0); 3283 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3284 nlist_idx = m_dysymtab.nlocalsym; 3285 } 3286 else 3287 { 3288 nlist_idx = 0; 3289 } 3290 3291 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) 3292 { 3293 struct nlist_64 nlist; 3294 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 3295 break; 3296 3297 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3298 nlist.n_type = nlist_data.GetU8_unchecked (&nlist_data_offset); 3299 nlist.n_sect = nlist_data.GetU8_unchecked (&nlist_data_offset); 3300 nlist.n_desc = nlist_data.GetU16_unchecked (&nlist_data_offset); 3301 nlist.n_value = nlist_data.GetAddress_unchecked (&nlist_data_offset); 3302 3303 SymbolType type = eSymbolTypeInvalid; 3304 const char *symbol_name = NULL; 3305 3306 if (have_strtab_data) 3307 { 3308 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3309 3310 if (symbol_name == NULL) 3311 { 3312 // No symbol should be NULL, even the symbols with no 3313 // string values should have an offset zero which points 3314 // to an empty C-string 3315 Host::SystemLog (Host::eSystemLogError, 3316 "error: symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 3317 nlist_idx, 3318 nlist.n_strx, 3319 module_sp->GetFileSpec().GetPath().c_str()); 3320 continue; 3321 } 3322 if (symbol_name[0] == '\0') 3323 symbol_name = NULL; 3324 } 3325 else 3326 { 3327 const addr_t str_addr = strtab_addr + nlist.n_strx; 3328 Error str_error; 3329 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, str_error)) 3330 symbol_name = memory_symbol_name.c_str(); 3331 } 3332 const char *symbol_name_non_abi_mangled = NULL; 3333 3334 SectionSP symbol_section; 3335 lldb::addr_t symbol_byte_size = 0; 3336 bool add_nlist = true; 3337 bool is_gsym = false; 3338 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3339 bool demangled_is_synthesized = false; 3340 3341 assert (sym_idx < num_syms); 3342 3343 sym[sym_idx].SetDebug (is_debug); 3344 3345 if (is_debug) 3346 { 3347 switch (nlist.n_type) 3348 { 3349 case N_GSYM: 3350 // global symbol: name,,NO_SECT,type,0 3351 // Sometimes the N_GSYM value contains the address. 3352 3353 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 3354 // have the same address, but we want to ensure that we always find only the real symbol, 3355 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 3356 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 3357 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 3358 // same address. 3359 3360 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 3361 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 3362 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 3363 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 3364 add_nlist = false; 3365 else 3366 { 3367 is_gsym = true; 3368 sym[sym_idx].SetExternal(true); 3369 if (nlist.n_value != 0) 3370 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3371 type = eSymbolTypeData; 3372 } 3373 break; 3374 3375 case N_FNAME: 3376 // procedure name (f77 kludge): name,,NO_SECT,0,0 3377 type = eSymbolTypeCompiler; 3378 break; 3379 3380 case N_FUN: 3381 // procedure: name,,n_sect,linenumber,address 3382 if (symbol_name) 3383 { 3384 type = eSymbolTypeCode; 3385 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3386 3387 N_FUN_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3388 // We use the current number of symbols in the symbol table in lieu of 3389 // using nlist_idx in case we ever start trimming entries out 3390 N_FUN_indexes.push_back(sym_idx); 3391 } 3392 else 3393 { 3394 type = eSymbolTypeCompiler; 3395 3396 if ( !N_FUN_indexes.empty() ) 3397 { 3398 // Copy the size of the function into the original STAB entry so we don't have 3399 // to hunt for it later 3400 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 3401 N_FUN_indexes.pop_back(); 3402 // We don't really need the end function STAB as it contains the size which 3403 // we already placed with the original symbol, so don't add it if we want a 3404 // minimal symbol table 3405 add_nlist = false; 3406 } 3407 } 3408 break; 3409 3410 case N_STSYM: 3411 // static symbol: name,,n_sect,type,address 3412 N_STSYM_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3413 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3414 type = eSymbolTypeData; 3415 break; 3416 3417 case N_LCSYM: 3418 // .lcomm symbol: name,,n_sect,type,address 3419 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3420 type = eSymbolTypeCommonBlock; 3421 break; 3422 3423 case N_BNSYM: 3424 // We use the current number of symbols in the symbol table in lieu of 3425 // using nlist_idx in case we ever start trimming entries out 3426 // Skip these if we want minimal symbol tables 3427 add_nlist = false; 3428 break; 3429 3430 case N_ENSYM: 3431 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 3432 // so that we can always skip the entire symbol if we need to navigate 3433 // more quickly at the source level when parsing STABS 3434 // Skip these if we want minimal symbol tables 3435 add_nlist = false; 3436 break; 3437 3438 3439 case N_OPT: 3440 // emitted with gcc2_compiled and in gcc source 3441 type = eSymbolTypeCompiler; 3442 break; 3443 3444 case N_RSYM: 3445 // register sym: name,,NO_SECT,type,register 3446 type = eSymbolTypeVariable; 3447 break; 3448 3449 case N_SLINE: 3450 // src line: 0,,n_sect,linenumber,address 3451 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3452 type = eSymbolTypeLineEntry; 3453 break; 3454 3455 case N_SSYM: 3456 // structure elt: name,,NO_SECT,type,struct_offset 3457 type = eSymbolTypeVariableType; 3458 break; 3459 3460 case N_SO: 3461 // source file name 3462 type = eSymbolTypeSourceFile; 3463 if (symbol_name == NULL) 3464 { 3465 add_nlist = false; 3466 if (N_SO_index != UINT32_MAX) 3467 { 3468 // Set the size of the N_SO to the terminating index of this N_SO 3469 // so that we can always skip the entire N_SO if we need to navigate 3470 // more quickly at the source level when parsing STABS 3471 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3472 symbol_ptr->SetByteSize(sym_idx); 3473 symbol_ptr->SetSizeIsSibling(true); 3474 } 3475 N_NSYM_indexes.clear(); 3476 N_INCL_indexes.clear(); 3477 N_BRAC_indexes.clear(); 3478 N_COMM_indexes.clear(); 3479 N_FUN_indexes.clear(); 3480 N_SO_index = UINT32_MAX; 3481 } 3482 else 3483 { 3484 // We use the current number of symbols in the symbol table in lieu of 3485 // using nlist_idx in case we ever start trimming entries out 3486 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3487 if (N_SO_has_full_path) 3488 { 3489 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3490 { 3491 // We have two consecutive N_SO entries where the first contains a directory 3492 // and the second contains a full path. 3493 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 3494 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3495 add_nlist = false; 3496 } 3497 else 3498 { 3499 // This is the first entry in a N_SO that contains a directory or 3500 // a full path to the source file 3501 N_SO_index = sym_idx; 3502 } 3503 } 3504 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3505 { 3506 // This is usually the second N_SO entry that contains just the filename, 3507 // so here we combine it with the first one if we are minimizing the symbol table 3508 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3509 if (so_path && so_path[0]) 3510 { 3511 std::string full_so_path (so_path); 3512 const size_t double_slash_pos = full_so_path.find("//"); 3513 if (double_slash_pos != std::string::npos) 3514 { 3515 // The linker has been generating bad N_SO entries with doubled up paths 3516 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 3517 // and the second is the directory for the source file so you end up with 3518 // a path that looks like "/tmp/src//tmp/src/" 3519 FileSpec so_dir(so_path, false); 3520 if (!so_dir.Exists()) 3521 { 3522 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 3523 if (so_dir.Exists()) 3524 { 3525 // Trim off the incorrect path 3526 full_so_path.erase(0, double_slash_pos + 1); 3527 } 3528 } 3529 } 3530 if (*full_so_path.rbegin() != '/') 3531 full_so_path += '/'; 3532 full_so_path += symbol_name; 3533 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 3534 add_nlist = false; 3535 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3536 } 3537 } 3538 else 3539 { 3540 // This could be a relative path to a N_SO 3541 N_SO_index = sym_idx; 3542 } 3543 } 3544 3545 break; 3546 3547 case N_OSO: 3548 // object file name: name,,0,0,st_mtime 3549 type = eSymbolTypeObjectFile; 3550 break; 3551 3552 case N_LSYM: 3553 // local sym: name,,NO_SECT,type,offset 3554 type = eSymbolTypeLocal; 3555 break; 3556 3557 //---------------------------------------------------------------------- 3558 // INCL scopes 3559 //---------------------------------------------------------------------- 3560 case N_BINCL: 3561 // include file beginning: name,,NO_SECT,0,sum 3562 // We use the current number of symbols in the symbol table in lieu of 3563 // using nlist_idx in case we ever start trimming entries out 3564 N_INCL_indexes.push_back(sym_idx); 3565 type = eSymbolTypeScopeBegin; 3566 break; 3567 3568 case N_EINCL: 3569 // include file end: name,,NO_SECT,0,0 3570 // Set the size of the N_BINCL to the terminating index of this N_EINCL 3571 // so that we can always skip the entire symbol if we need to navigate 3572 // more quickly at the source level when parsing STABS 3573 if ( !N_INCL_indexes.empty() ) 3574 { 3575 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3576 symbol_ptr->SetByteSize(sym_idx + 1); 3577 symbol_ptr->SetSizeIsSibling(true); 3578 N_INCL_indexes.pop_back(); 3579 } 3580 type = eSymbolTypeScopeEnd; 3581 break; 3582 3583 case N_SOL: 3584 // #included file name: name,,n_sect,0,address 3585 type = eSymbolTypeHeaderFile; 3586 3587 // We currently don't use the header files on darwin 3588 add_nlist = false; 3589 break; 3590 3591 case N_PARAMS: 3592 // compiler parameters: name,,NO_SECT,0,0 3593 type = eSymbolTypeCompiler; 3594 break; 3595 3596 case N_VERSION: 3597 // compiler version: name,,NO_SECT,0,0 3598 type = eSymbolTypeCompiler; 3599 break; 3600 3601 case N_OLEVEL: 3602 // compiler -O level: name,,NO_SECT,0,0 3603 type = eSymbolTypeCompiler; 3604 break; 3605 3606 case N_PSYM: 3607 // parameter: name,,NO_SECT,type,offset 3608 type = eSymbolTypeVariable; 3609 break; 3610 3611 case N_ENTRY: 3612 // alternate entry: name,,n_sect,linenumber,address 3613 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3614 type = eSymbolTypeLineEntry; 3615 break; 3616 3617 //---------------------------------------------------------------------- 3618 // Left and Right Braces 3619 //---------------------------------------------------------------------- 3620 case N_LBRAC: 3621 // left bracket: 0,,NO_SECT,nesting level,address 3622 // We use the current number of symbols in the symbol table in lieu of 3623 // using nlist_idx in case we ever start trimming entries out 3624 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3625 N_BRAC_indexes.push_back(sym_idx); 3626 type = eSymbolTypeScopeBegin; 3627 break; 3628 3629 case N_RBRAC: 3630 // right bracket: 0,,NO_SECT,nesting level,address 3631 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 3632 // so that we can always skip the entire symbol if we need to navigate 3633 // more quickly at the source level when parsing STABS 3634 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3635 if ( !N_BRAC_indexes.empty() ) 3636 { 3637 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3638 symbol_ptr->SetByteSize(sym_idx + 1); 3639 symbol_ptr->SetSizeIsSibling(true); 3640 N_BRAC_indexes.pop_back(); 3641 } 3642 type = eSymbolTypeScopeEnd; 3643 break; 3644 3645 case N_EXCL: 3646 // deleted include file: name,,NO_SECT,0,sum 3647 type = eSymbolTypeHeaderFile; 3648 break; 3649 3650 //---------------------------------------------------------------------- 3651 // COMM scopes 3652 //---------------------------------------------------------------------- 3653 case N_BCOMM: 3654 // begin common: name,,NO_SECT,0,0 3655 // We use the current number of symbols in the symbol table in lieu of 3656 // using nlist_idx in case we ever start trimming entries out 3657 type = eSymbolTypeScopeBegin; 3658 N_COMM_indexes.push_back(sym_idx); 3659 break; 3660 3661 case N_ECOML: 3662 // end common (local name): 0,,n_sect,0,address 3663 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3664 // Fall through 3665 3666 case N_ECOMM: 3667 // end common: name,,n_sect,0,0 3668 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 3669 // so that we can always skip the entire symbol if we need to navigate 3670 // more quickly at the source level when parsing STABS 3671 if ( !N_COMM_indexes.empty() ) 3672 { 3673 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 3674 symbol_ptr->SetByteSize(sym_idx + 1); 3675 symbol_ptr->SetSizeIsSibling(true); 3676 N_COMM_indexes.pop_back(); 3677 } 3678 type = eSymbolTypeScopeEnd; 3679 break; 3680 3681 case N_LENG: 3682 // second stab entry with length information 3683 type = eSymbolTypeAdditional; 3684 break; 3685 3686 default: break; 3687 } 3688 } 3689 else 3690 { 3691 //uint8_t n_pext = N_PEXT & nlist.n_type; 3692 uint8_t n_type = N_TYPE & nlist.n_type; 3693 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3694 3695 switch (n_type) 3696 { 3697 case N_INDR:// Fall through 3698 case N_PBUD:// Fall through 3699 case N_UNDF: 3700 type = eSymbolTypeUndefined; 3701 break; 3702 3703 case N_ABS: 3704 type = eSymbolTypeAbsolute; 3705 break; 3706 3707 case N_SECT: 3708 { 3709 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3710 3711 if (!symbol_section) 3712 { 3713 // TODO: warn about this? 3714 add_nlist = false; 3715 break; 3716 } 3717 3718 if (TEXT_eh_frame_sectID == nlist.n_sect) 3719 { 3720 type = eSymbolTypeException; 3721 } 3722 else 3723 { 3724 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 3725 3726 switch (section_type) 3727 { 3728 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 3729 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 3730 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 3731 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 3732 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 3733 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 3734 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 3735 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 3736 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 3737 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 3738 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 3739 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 3740 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 3741 default: 3742 switch (symbol_section->GetType()) 3743 { 3744 case lldb::eSectionTypeCode: 3745 type = eSymbolTypeCode; 3746 break; 3747 case eSectionTypeData: 3748 case eSectionTypeDataCString: // Inlined C string data 3749 case eSectionTypeDataCStringPointers: // Pointers to C string data 3750 case eSectionTypeDataSymbolAddress: // Address of a symbol in the symbol table 3751 case eSectionTypeData4: 3752 case eSectionTypeData8: 3753 case eSectionTypeData16: 3754 type = eSymbolTypeData; 3755 break; 3756 default: 3757 break; 3758 } 3759 break; 3760 } 3761 3762 if (type == eSymbolTypeInvalid) 3763 { 3764 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 3765 if (symbol_section->IsDescendant (text_section_sp.get())) 3766 { 3767 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 3768 S_ATTR_SELF_MODIFYING_CODE | 3769 S_ATTR_SOME_INSTRUCTIONS)) 3770 type = eSymbolTypeData; 3771 else 3772 type = eSymbolTypeCode; 3773 } 3774 else 3775 if (symbol_section->IsDescendant(data_section_sp.get())) 3776 { 3777 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 3778 { 3779 type = eSymbolTypeRuntime; 3780 3781 if (symbol_name && 3782 symbol_name[0] == '_' && 3783 symbol_name[1] == 'O' && 3784 symbol_name[2] == 'B') 3785 { 3786 llvm::StringRef symbol_name_ref(symbol_name); 3787 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 3788 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 3789 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 3790 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 3791 { 3792 symbol_name_non_abi_mangled = symbol_name + 1; 3793 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3794 type = eSymbolTypeObjCClass; 3795 demangled_is_synthesized = true; 3796 } 3797 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 3798 { 3799 symbol_name_non_abi_mangled = symbol_name + 1; 3800 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3801 type = eSymbolTypeObjCMetaClass; 3802 demangled_is_synthesized = true; 3803 } 3804 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 3805 { 3806 symbol_name_non_abi_mangled = symbol_name + 1; 3807 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3808 type = eSymbolTypeObjCIVar; 3809 demangled_is_synthesized = true; 3810 } 3811 } 3812 } 3813 else 3814 if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 3815 { 3816 type = eSymbolTypeException; 3817 } 3818 else 3819 { 3820 type = eSymbolTypeData; 3821 } 3822 } 3823 else 3824 if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 3825 { 3826 type = eSymbolTypeTrampoline; 3827 } 3828 else 3829 if (symbol_section->IsDescendant(objc_section_sp.get())) 3830 { 3831 type = eSymbolTypeRuntime; 3832 if (symbol_name && symbol_name[0] == '.') 3833 { 3834 llvm::StringRef symbol_name_ref(symbol_name); 3835 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 3836 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 3837 { 3838 symbol_name_non_abi_mangled = symbol_name; 3839 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 3840 type = eSymbolTypeObjCClass; 3841 demangled_is_synthesized = true; 3842 } 3843 } 3844 } 3845 } 3846 } 3847 } 3848 break; 3849 } 3850 } 3851 3852 if (add_nlist) 3853 { 3854 uint64_t symbol_value = nlist.n_value; 3855 3856 if (symbol_name_non_abi_mangled) 3857 { 3858 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 3859 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 3860 } 3861 else 3862 { 3863 bool symbol_name_is_mangled = false; 3864 3865 if (symbol_name && symbol_name[0] == '_') 3866 { 3867 symbol_name_is_mangled = symbol_name[1] == '_'; 3868 symbol_name++; // Skip the leading underscore 3869 } 3870 3871 if (symbol_name) 3872 { 3873 ConstString const_symbol_name(symbol_name); 3874 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 3875 if (is_gsym && is_debug) 3876 { 3877 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 3878 } 3879 } 3880 } 3881 if (symbol_section) 3882 { 3883 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3884 if (symbol_byte_size == 0 && function_starts_count > 0) 3885 { 3886 addr_t symbol_lookup_file_addr = nlist.n_value; 3887 // Do an exact address match for non-ARM addresses, else get the closest since 3888 // the symbol might be a thumb symbol which has an address with bit zero set 3889 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 3890 if (is_arm && func_start_entry) 3891 { 3892 // Verify that the function start address is the symbol address (ARM) 3893 // or the symbol address + 1 (thumb) 3894 if (func_start_entry->addr != symbol_lookup_file_addr && 3895 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 3896 { 3897 // Not the right entry, NULL it out... 3898 func_start_entry = NULL; 3899 } 3900 } 3901 if (func_start_entry) 3902 { 3903 func_start_entry->data = true; 3904 3905 addr_t symbol_file_addr = func_start_entry->addr; 3906 if (is_arm) 3907 symbol_file_addr &= 0xfffffffffffffffeull; 3908 3909 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3910 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3911 if (next_func_start_entry) 3912 { 3913 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3914 // Be sure the clear the Thumb address bit when we calculate the size 3915 // from the current and next address 3916 if (is_arm) 3917 next_symbol_file_addr &= 0xfffffffffffffffeull; 3918 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3919 } 3920 else 3921 { 3922 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3923 } 3924 } 3925 } 3926 symbol_value -= section_file_addr; 3927 } 3928 3929 if (is_debug == false) 3930 { 3931 if (type == eSymbolTypeCode) 3932 { 3933 // See if we can find a N_FUN entry for any code symbols. 3934 // If we do find a match, and the name matches, then we 3935 // can merge the two into just the function symbol to avoid 3936 // duplicate entries in the symbol table 3937 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3938 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3939 if (range.first != range.second) 3940 { 3941 bool found_it = false; 3942 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3943 { 3944 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3945 { 3946 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3947 // We just need the flags from the linker symbol, so put these flags 3948 // into the N_FUN flags to avoid duplicate symbols in the symbol table 3949 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3950 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3951 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3952 sym[pos->second].SetType (eSymbolTypeResolver); 3953 sym[sym_idx].Clear(); 3954 found_it = true; 3955 break; 3956 } 3957 } 3958 if (found_it) 3959 continue; 3960 } 3961 else 3962 { 3963 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3964 type = eSymbolTypeResolver; 3965 } 3966 } 3967 else if (type == eSymbolTypeData) 3968 { 3969 // See if we can find a N_STSYM entry for any data symbols. 3970 // If we do find a match, and the name matches, then we 3971 // can merge the two into just the Static symbol to avoid 3972 // duplicate entries in the symbol table 3973 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3974 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 3975 if (range.first != range.second) 3976 { 3977 bool found_it = false; 3978 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3979 { 3980 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3981 { 3982 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3983 // We just need the flags from the linker symbol, so put these flags 3984 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3985 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3986 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3987 sym[sym_idx].Clear(); 3988 found_it = true; 3989 break; 3990 } 3991 } 3992 if (found_it) 3993 continue; 3994 } 3995 else 3996 { 3997 // Combine N_GSYM stab entries with the non stab symbol 3998 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 3999 if (pos != N_GSYM_name_to_sym_idx.end()) 4000 { 4001 const uint32_t GSYM_sym_idx = pos->second; 4002 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4003 // Copy the address, because often the N_GSYM address has an invalid address of zero 4004 // when the global is a common symbol 4005 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 4006 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 4007 // We just need the flags from the linker symbol, so put these flags 4008 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 4009 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4010 sym[sym_idx].Clear(); 4011 continue; 4012 } 4013 } 4014 } 4015 } 4016 4017 sym[sym_idx].SetID (nlist_idx); 4018 sym[sym_idx].SetType (type); 4019 sym[sym_idx].GetAddress().SetSection (symbol_section); 4020 sym[sym_idx].GetAddress().SetOffset (symbol_value); 4021 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 4022 4023 if (symbol_byte_size > 0) 4024 sym[sym_idx].SetByteSize(symbol_byte_size); 4025 4026 if (demangled_is_synthesized) 4027 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4028 4029 ++sym_idx; 4030 } 4031 else 4032 { 4033 sym[sym_idx].Clear(); 4034 } 4035 } 4036 } 4037 4038 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4039 4040 if (function_starts_count > 0) 4041 { 4042 char synthetic_function_symbol[PATH_MAX]; 4043 uint32_t num_synthetic_function_symbols = 0; 4044 for (i=0; i<function_starts_count; ++i) 4045 { 4046 if (function_starts.GetEntryRef (i).data == false) 4047 ++num_synthetic_function_symbols; 4048 } 4049 4050 if (num_synthetic_function_symbols > 0) 4051 { 4052 if (num_syms < sym_idx + num_synthetic_function_symbols) 4053 { 4054 num_syms = sym_idx + num_synthetic_function_symbols; 4055 sym = symtab->Resize (num_syms); 4056 } 4057 uint32_t synthetic_function_symbol_idx = 0; 4058 for (i=0; i<function_starts_count; ++i) 4059 { 4060 const FunctionStarts::Entry *func_start_entry = function_starts.GetEntryAtIndex (i); 4061 if (func_start_entry->data == false) 4062 { 4063 addr_t symbol_file_addr = func_start_entry->addr; 4064 uint32_t symbol_flags = 0; 4065 if (is_arm) 4066 { 4067 if (symbol_file_addr & 1) 4068 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4069 symbol_file_addr &= 0xfffffffffffffffeull; 4070 } 4071 Address symbol_addr; 4072 if (module_sp->ResolveFileAddress (symbol_file_addr, symbol_addr)) 4073 { 4074 SectionSP symbol_section (symbol_addr.GetSection()); 4075 uint32_t symbol_byte_size = 0; 4076 if (symbol_section) 4077 { 4078 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4079 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 4080 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 4081 if (next_func_start_entry) 4082 { 4083 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4084 if (is_arm) 4085 next_symbol_file_addr &= 0xfffffffffffffffeull; 4086 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 4087 } 4088 else 4089 { 4090 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4091 } 4092 snprintf (synthetic_function_symbol, 4093 sizeof(synthetic_function_symbol), 4094 "___lldb_unnamed_function%u$$%s", 4095 ++synthetic_function_symbol_idx, 4096 module_sp->GetFileSpec().GetFilename().GetCString()); 4097 sym[sym_idx].SetID (synthetic_sym_id++); 4098 sym[sym_idx].GetMangled().SetDemangledName(ConstString(synthetic_function_symbol)); 4099 sym[sym_idx].SetType (eSymbolTypeCode); 4100 sym[sym_idx].SetIsSynthetic (true); 4101 sym[sym_idx].GetAddress() = symbol_addr; 4102 if (symbol_flags) 4103 sym[sym_idx].SetFlags (symbol_flags); 4104 if (symbol_byte_size) 4105 sym[sym_idx].SetByteSize (symbol_byte_size); 4106 ++sym_idx; 4107 } 4108 } 4109 } 4110 } 4111 } 4112 } 4113 4114 // Trim our symbols down to just what we ended up with after 4115 // removing any symbols. 4116 if (sym_idx < num_syms) 4117 { 4118 num_syms = sym_idx; 4119 sym = symtab->Resize (num_syms); 4120 } 4121 4122 // Now synthesize indirect symbols 4123 if (m_dysymtab.nindirectsyms != 0) 4124 { 4125 if (indirect_symbol_index_data.GetByteSize()) 4126 { 4127 NListIndexToSymbolIndexMap::const_iterator end_index_pos = m_nlist_idx_to_sym_idx.end(); 4128 4129 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); ++sect_idx) 4130 { 4131 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == S_SYMBOL_STUBS) 4132 { 4133 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4134 if (symbol_stub_byte_size == 0) 4135 continue; 4136 4137 const uint32_t num_symbol_stubs = m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4138 4139 if (num_symbol_stubs == 0) 4140 continue; 4141 4142 const uint32_t symbol_stub_index_offset = m_mach_sections[sect_idx].reserved1; 4143 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) 4144 { 4145 const uint32_t symbol_stub_index = symbol_stub_index_offset + stub_idx; 4146 const lldb::addr_t symbol_stub_addr = m_mach_sections[sect_idx].addr + (stub_idx * symbol_stub_byte_size); 4147 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4148 if (indirect_symbol_index_data.ValidOffsetForDataOfSize(symbol_stub_offset, 4)) 4149 { 4150 const uint32_t stub_sym_id = indirect_symbol_index_data.GetU32 (&symbol_stub_offset); 4151 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4152 continue; 4153 4154 NListIndexToSymbolIndexMap::const_iterator index_pos = m_nlist_idx_to_sym_idx.find (stub_sym_id); 4155 Symbol *stub_symbol = NULL; 4156 if (index_pos != end_index_pos) 4157 { 4158 // We have a remapping from the original nlist index to 4159 // a current symbol index, so just look this up by index 4160 stub_symbol = symtab->SymbolAtIndex (index_pos->second); 4161 } 4162 else 4163 { 4164 // We need to lookup a symbol using the original nlist 4165 // symbol index since this index is coming from the 4166 // S_SYMBOL_STUBS 4167 stub_symbol = symtab->FindSymbolByID (stub_sym_id); 4168 } 4169 4170 if (stub_symbol) 4171 { 4172 Address so_addr(symbol_stub_addr, section_list); 4173 4174 if (stub_symbol->GetType() == eSymbolTypeUndefined) 4175 { 4176 // Change the external symbol into a trampoline that makes sense 4177 // These symbols were N_UNDF N_EXT, and are useless to us, so we 4178 // can re-use them so we don't have to make up a synthetic symbol 4179 // for no good reason. 4180 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 4181 stub_symbol->SetType (eSymbolTypeTrampoline); 4182 else 4183 stub_symbol->SetType (eSymbolTypeResolver); 4184 stub_symbol->SetExternal (false); 4185 stub_symbol->GetAddress() = so_addr; 4186 stub_symbol->SetByteSize (symbol_stub_byte_size); 4187 } 4188 else 4189 { 4190 // Make a synthetic symbol to describe the trampoline stub 4191 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4192 if (sym_idx >= num_syms) 4193 { 4194 sym = symtab->Resize (++num_syms); 4195 stub_symbol = NULL; // this pointer no longer valid 4196 } 4197 sym[sym_idx].SetID (synthetic_sym_id++); 4198 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4199 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 4200 sym[sym_idx].SetType (eSymbolTypeTrampoline); 4201 else 4202 sym[sym_idx].SetType (eSymbolTypeResolver); 4203 sym[sym_idx].SetIsSynthetic (true); 4204 sym[sym_idx].GetAddress() = so_addr; 4205 sym[sym_idx].SetByteSize (symbol_stub_byte_size); 4206 ++sym_idx; 4207 } 4208 } 4209 else 4210 { 4211 if (log) 4212 log->Warning ("symbol stub referencing symbol table symbol %u that isn't in our minimal symbol table, fix this!!!", stub_sym_id); 4213 } 4214 } 4215 } 4216 } 4217 } 4218 } 4219 } 4220 4221 4222 if (!trie_entries.empty()) 4223 { 4224 for (const auto &e : trie_entries) 4225 { 4226 if (e.entry.import_name) 4227 { 4228 // Make a synthetic symbol to describe re-exported symbol. 4229 if (sym_idx >= num_syms) 4230 sym = symtab->Resize (++num_syms); 4231 sym[sym_idx].SetID (synthetic_sym_id++); 4232 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4233 sym[sym_idx].SetType (eSymbolTypeReExported); 4234 sym[sym_idx].SetIsSynthetic (true); 4235 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4236 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) 4237 { 4238 sym[sym_idx].SetReExportedSymbolSharedLibrary(dylib_files.GetFileSpecAtIndex(e.entry.other-1)); 4239 } 4240 ++sym_idx; 4241 } 4242 } 4243 } 4244 4245 4246 4247 // StreamFile s(stdout, false); 4248 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4249 // symtab->Dump(&s, NULL, eSortOrderNone); 4250 // Set symbol byte sizes correctly since mach-o nlist entries don't have sizes 4251 symtab->CalculateSymbolSizes(); 4252 4253 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4254 // symtab->Dump(&s, NULL, eSortOrderNone); 4255 4256 return symtab->GetNumSymbols(); 4257 } 4258 return 0; 4259 } 4260 4261 4262 void 4263 ObjectFileMachO::Dump (Stream *s) 4264 { 4265 ModuleSP module_sp(GetModule()); 4266 if (module_sp) 4267 { 4268 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4269 s->Printf("%p: ", static_cast<void*>(this)); 4270 s->Indent(); 4271 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4272 s->PutCString("ObjectFileMachO64"); 4273 else 4274 s->PutCString("ObjectFileMachO32"); 4275 4276 ArchSpec header_arch; 4277 GetArchitecture(header_arch); 4278 4279 *s << ", file = '" << m_file << "', arch = " << header_arch.GetArchitectureName() << "\n"; 4280 4281 SectionList *sections = GetSectionList(); 4282 if (sections) 4283 sections->Dump(s, NULL, true, UINT32_MAX); 4284 4285 if (m_symtab_ap.get()) 4286 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4287 } 4288 } 4289 4290 bool 4291 ObjectFileMachO::GetUUID (const llvm::MachO::mach_header &header, 4292 const lldb_private::DataExtractor &data, 4293 lldb::offset_t lc_offset, 4294 lldb_private::UUID& uuid) 4295 { 4296 uint32_t i; 4297 struct uuid_command load_cmd; 4298 4299 lldb::offset_t offset = lc_offset; 4300 for (i=0; i<header.ncmds; ++i) 4301 { 4302 const lldb::offset_t cmd_offset = offset; 4303 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4304 break; 4305 4306 if (load_cmd.cmd == LC_UUID) 4307 { 4308 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4309 4310 if (uuid_bytes) 4311 { 4312 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4313 // We pretend these object files have no UUID to prevent crashing. 4314 4315 const uint8_t opencl_uuid[] = { 0x8c, 0x8e, 0xb3, 0x9b, 4316 0x3b, 0xa8, 4317 0x4b, 0x16, 4318 0xb6, 0xa4, 4319 0x27, 0x63, 0xbb, 0x14, 0xf0, 0x0d }; 4320 4321 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4322 return false; 4323 4324 uuid.SetBytes (uuid_bytes); 4325 return true; 4326 } 4327 return false; 4328 } 4329 offset = cmd_offset + load_cmd.cmdsize; 4330 } 4331 return false; 4332 } 4333 4334 4335 bool 4336 ObjectFileMachO::GetArchitecture (const llvm::MachO::mach_header &header, 4337 const lldb_private::DataExtractor &data, 4338 lldb::offset_t lc_offset, 4339 ArchSpec &arch) 4340 { 4341 arch.SetArchitecture (eArchTypeMachO, header.cputype, header.cpusubtype); 4342 4343 if (arch.IsValid()) 4344 { 4345 llvm::Triple &triple = arch.GetTriple(); 4346 if (header.filetype == MH_PRELOAD) 4347 { 4348 // Set OS to "unknown" - this is a standalone binary with no dyld et al 4349 triple.setOS(llvm::Triple::UnknownOS); 4350 return true; 4351 } 4352 else 4353 { 4354 struct load_command load_cmd; 4355 4356 lldb::offset_t offset = lc_offset; 4357 for (uint32_t i=0; i<header.ncmds; ++i) 4358 { 4359 const lldb::offset_t cmd_offset = offset; 4360 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4361 break; 4362 4363 switch (load_cmd.cmd) 4364 { 4365 case LC_VERSION_MIN_IPHONEOS: 4366 triple.setOS (llvm::Triple::IOS); 4367 return true; 4368 4369 case LC_VERSION_MIN_MACOSX: 4370 triple.setOS (llvm::Triple::MacOSX); 4371 return true; 4372 4373 default: 4374 break; 4375 } 4376 4377 offset = cmd_offset + load_cmd.cmdsize; 4378 } 4379 4380 // Only set the OS to iOS for ARM, we don't want to set it for x86 and x86_64. 4381 // We do this because we now have MacOSX or iOS as the OS value for x86 and 4382 // x86_64 for normal desktop (MacOSX) and simulator (iOS) binaries. And if 4383 // we compare a "x86_64-apple-ios" to a "x86_64-apple-" triple, it will say 4384 // it is compatible (because the OS is unspecified in the second one and will 4385 // match anything in the first 4386 if (header.cputype == CPU_TYPE_ARM || header.cputype == CPU_TYPE_ARM64) 4387 triple.setOS (llvm::Triple::IOS); 4388 } 4389 } 4390 return arch.IsValid(); 4391 } 4392 4393 bool 4394 ObjectFileMachO::GetUUID (lldb_private::UUID* uuid) 4395 { 4396 ModuleSP module_sp(GetModule()); 4397 if (module_sp) 4398 { 4399 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4400 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4401 return GetUUID (m_header, m_data, offset, *uuid); 4402 } 4403 return false; 4404 } 4405 4406 4407 uint32_t 4408 ObjectFileMachO::GetDependentModules (FileSpecList& files) 4409 { 4410 uint32_t count = 0; 4411 ModuleSP module_sp(GetModule()); 4412 if (module_sp) 4413 { 4414 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4415 struct load_command load_cmd; 4416 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4417 const bool resolve_path = false; // Don't resolve the dependent file paths since they may not reside on this system 4418 uint32_t i; 4419 for (i=0; i<m_header.ncmds; ++i) 4420 { 4421 const uint32_t cmd_offset = offset; 4422 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4423 break; 4424 4425 switch (load_cmd.cmd) 4426 { 4427 case LC_LOAD_DYLIB: 4428 case LC_LOAD_WEAK_DYLIB: 4429 case LC_REEXPORT_DYLIB: 4430 case LC_LOAD_DYLINKER: 4431 case LC_LOADFVMLIB: 4432 case LC_LOAD_UPWARD_DYLIB: 4433 { 4434 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 4435 const char *path = m_data.PeekCStr(name_offset); 4436 // Skip any path that starts with '@' since these are usually: 4437 // @executable_path/.../file 4438 // @rpath/.../file 4439 if (path && path[0] != '@') 4440 { 4441 FileSpec file_spec(path, resolve_path); 4442 if (files.AppendIfUnique(file_spec)) 4443 count++; 4444 } 4445 } 4446 break; 4447 4448 default: 4449 break; 4450 } 4451 offset = cmd_offset + load_cmd.cmdsize; 4452 } 4453 } 4454 return count; 4455 } 4456 4457 lldb_private::Address 4458 ObjectFileMachO::GetEntryPointAddress () 4459 { 4460 // If the object file is not an executable it can't hold the entry point. m_entry_point_address 4461 // is initialized to an invalid address, so we can just return that. 4462 // If m_entry_point_address is valid it means we've found it already, so return the cached value. 4463 4464 if (!IsExecutable() || m_entry_point_address.IsValid()) 4465 return m_entry_point_address; 4466 4467 // Otherwise, look for the UnixThread or Thread command. The data for the Thread command is given in 4468 // /usr/include/mach-o.h, but it is basically: 4469 // 4470 // uint32_t flavor - this is the flavor argument you would pass to thread_get_state 4471 // uint32_t count - this is the count of longs in the thread state data 4472 // struct XXX_thread_state state - this is the structure from <machine/thread_status.h> corresponding to the flavor. 4473 // <repeat this trio> 4474 // 4475 // So we just keep reading the various register flavors till we find the GPR one, then read the PC out of there. 4476 // FIXME: We will need to have a "RegisterContext data provider" class at some point that can get all the registers 4477 // out of data in this form & attach them to a given thread. That should underlie the MacOS X User process plugin, 4478 // and we'll also need it for the MacOS X Core File process plugin. When we have that we can also use it here. 4479 // 4480 // For now we hard-code the offsets and flavors we need: 4481 // 4482 // 4483 4484 ModuleSP module_sp(GetModule()); 4485 if (module_sp) 4486 { 4487 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4488 struct load_command load_cmd; 4489 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4490 uint32_t i; 4491 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 4492 bool done = false; 4493 4494 for (i=0; i<m_header.ncmds; ++i) 4495 { 4496 const lldb::offset_t cmd_offset = offset; 4497 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4498 break; 4499 4500 switch (load_cmd.cmd) 4501 { 4502 case LC_UNIXTHREAD: 4503 case LC_THREAD: 4504 { 4505 while (offset < cmd_offset + load_cmd.cmdsize) 4506 { 4507 uint32_t flavor = m_data.GetU32(&offset); 4508 uint32_t count = m_data.GetU32(&offset); 4509 if (count == 0) 4510 { 4511 // We've gotten off somehow, log and exit; 4512 return m_entry_point_address; 4513 } 4514 4515 switch (m_header.cputype) 4516 { 4517 case llvm::MachO::CPU_TYPE_ARM: 4518 if (flavor == 1) // ARM_THREAD_STATE from mach/arm/thread_status.h 4519 { 4520 offset += 60; // This is the offset of pc in the GPR thread state data structure. 4521 start_address = m_data.GetU32(&offset); 4522 done = true; 4523 } 4524 break; 4525 case llvm::MachO::CPU_TYPE_ARM64: 4526 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 4527 { 4528 offset += 256; // This is the offset of pc in the GPR thread state data structure. 4529 start_address = m_data.GetU64(&offset); 4530 done = true; 4531 } 4532 break; 4533 case llvm::MachO::CPU_TYPE_I386: 4534 if (flavor == 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 4535 { 4536 offset += 40; // This is the offset of eip in the GPR thread state data structure. 4537 start_address = m_data.GetU32(&offset); 4538 done = true; 4539 } 4540 break; 4541 case llvm::MachO::CPU_TYPE_X86_64: 4542 if (flavor == 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 4543 { 4544 offset += 16 * 8; // This is the offset of rip in the GPR thread state data structure. 4545 start_address = m_data.GetU64(&offset); 4546 done = true; 4547 } 4548 break; 4549 default: 4550 return m_entry_point_address; 4551 } 4552 // Haven't found the GPR flavor yet, skip over the data for this flavor: 4553 if (done) 4554 break; 4555 offset += count * 4; 4556 } 4557 } 4558 break; 4559 case LC_MAIN: 4560 { 4561 ConstString text_segment_name ("__TEXT"); 4562 uint64_t entryoffset = m_data.GetU64(&offset); 4563 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 4564 if (text_segment_sp) 4565 { 4566 done = true; 4567 start_address = text_segment_sp->GetFileAddress() + entryoffset; 4568 } 4569 } 4570 4571 default: 4572 break; 4573 } 4574 if (done) 4575 break; 4576 4577 // Go to the next load command: 4578 offset = cmd_offset + load_cmd.cmdsize; 4579 } 4580 4581 if (start_address != LLDB_INVALID_ADDRESS) 4582 { 4583 // We got the start address from the load commands, so now resolve that address in the sections 4584 // of this ObjectFile: 4585 if (!m_entry_point_address.ResolveAddressUsingFileSections (start_address, GetSectionList())) 4586 { 4587 m_entry_point_address.Clear(); 4588 } 4589 } 4590 else 4591 { 4592 // We couldn't read the UnixThread load command - maybe it wasn't there. As a fallback look for the 4593 // "start" symbol in the main executable. 4594 4595 ModuleSP module_sp (GetModule()); 4596 4597 if (module_sp) 4598 { 4599 SymbolContextList contexts; 4600 SymbolContext context; 4601 if (module_sp->FindSymbolsWithNameAndType(ConstString ("start"), eSymbolTypeCode, contexts)) 4602 { 4603 if (contexts.GetContextAtIndex(0, context)) 4604 m_entry_point_address = context.symbol->GetAddress(); 4605 } 4606 } 4607 } 4608 } 4609 4610 return m_entry_point_address; 4611 4612 } 4613 4614 lldb_private::Address 4615 ObjectFileMachO::GetHeaderAddress () 4616 { 4617 lldb_private::Address header_addr; 4618 SectionList *section_list = GetSectionList(); 4619 if (section_list) 4620 { 4621 SectionSP text_segment_sp (section_list->FindSectionByName (GetSegmentNameTEXT())); 4622 if (text_segment_sp) 4623 { 4624 header_addr.SetSection (text_segment_sp); 4625 header_addr.SetOffset (0); 4626 } 4627 } 4628 return header_addr; 4629 } 4630 4631 uint32_t 4632 ObjectFileMachO::GetNumThreadContexts () 4633 { 4634 ModuleSP module_sp(GetModule()); 4635 if (module_sp) 4636 { 4637 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4638 if (!m_thread_context_offsets_valid) 4639 { 4640 m_thread_context_offsets_valid = true; 4641 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4642 FileRangeArray::Entry file_range; 4643 thread_command thread_cmd; 4644 for (uint32_t i=0; i<m_header.ncmds; ++i) 4645 { 4646 const uint32_t cmd_offset = offset; 4647 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 4648 break; 4649 4650 if (thread_cmd.cmd == LC_THREAD) 4651 { 4652 file_range.SetRangeBase (offset); 4653 file_range.SetByteSize (thread_cmd.cmdsize - 8); 4654 m_thread_context_offsets.Append (file_range); 4655 } 4656 offset = cmd_offset + thread_cmd.cmdsize; 4657 } 4658 } 4659 } 4660 return m_thread_context_offsets.GetSize(); 4661 } 4662 4663 lldb::RegisterContextSP 4664 ObjectFileMachO::GetThreadContextAtIndex (uint32_t idx, lldb_private::Thread &thread) 4665 { 4666 lldb::RegisterContextSP reg_ctx_sp; 4667 4668 ModuleSP module_sp(GetModule()); 4669 if (module_sp) 4670 { 4671 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4672 if (!m_thread_context_offsets_valid) 4673 GetNumThreadContexts (); 4674 4675 const FileRangeArray::Entry *thread_context_file_range = m_thread_context_offsets.GetEntryAtIndex (idx); 4676 if (thread_context_file_range) 4677 { 4678 4679 DataExtractor data (m_data, 4680 thread_context_file_range->GetRangeBase(), 4681 thread_context_file_range->GetByteSize()); 4682 4683 switch (m_header.cputype) 4684 { 4685 case llvm::MachO::CPU_TYPE_ARM64: 4686 reg_ctx_sp.reset (new RegisterContextDarwin_arm64_Mach (thread, data)); 4687 break; 4688 4689 case llvm::MachO::CPU_TYPE_ARM: 4690 reg_ctx_sp.reset (new RegisterContextDarwin_arm_Mach (thread, data)); 4691 break; 4692 4693 case llvm::MachO::CPU_TYPE_I386: 4694 reg_ctx_sp.reset (new RegisterContextDarwin_i386_Mach (thread, data)); 4695 break; 4696 4697 case llvm::MachO::CPU_TYPE_X86_64: 4698 reg_ctx_sp.reset (new RegisterContextDarwin_x86_64_Mach (thread, data)); 4699 break; 4700 } 4701 } 4702 } 4703 return reg_ctx_sp; 4704 } 4705 4706 4707 ObjectFile::Type 4708 ObjectFileMachO::CalculateType() 4709 { 4710 switch (m_header.filetype) 4711 { 4712 case MH_OBJECT: // 0x1u 4713 if (GetAddressByteSize () == 4) 4714 { 4715 // 32 bit kexts are just object files, but they do have a valid 4716 // UUID load command. 4717 UUID uuid; 4718 if (GetUUID(&uuid)) 4719 { 4720 // this checking for the UUID load command is not enough 4721 // we could eventually look for the symbol named 4722 // "OSKextGetCurrentIdentifier" as this is required of kexts 4723 if (m_strata == eStrataInvalid) 4724 m_strata = eStrataKernel; 4725 return eTypeSharedLibrary; 4726 } 4727 } 4728 return eTypeObjectFile; 4729 4730 case MH_EXECUTE: return eTypeExecutable; // 0x2u 4731 case MH_FVMLIB: return eTypeSharedLibrary; // 0x3u 4732 case MH_CORE: return eTypeCoreFile; // 0x4u 4733 case MH_PRELOAD: return eTypeSharedLibrary; // 0x5u 4734 case MH_DYLIB: return eTypeSharedLibrary; // 0x6u 4735 case MH_DYLINKER: return eTypeDynamicLinker; // 0x7u 4736 case MH_BUNDLE: return eTypeSharedLibrary; // 0x8u 4737 case MH_DYLIB_STUB: return eTypeStubLibrary; // 0x9u 4738 case MH_DSYM: return eTypeDebugInfo; // 0xAu 4739 case MH_KEXT_BUNDLE: return eTypeSharedLibrary; // 0xBu 4740 default: 4741 break; 4742 } 4743 return eTypeUnknown; 4744 } 4745 4746 ObjectFile::Strata 4747 ObjectFileMachO::CalculateStrata() 4748 { 4749 switch (m_header.filetype) 4750 { 4751 case MH_OBJECT: // 0x1u 4752 { 4753 // 32 bit kexts are just object files, but they do have a valid 4754 // UUID load command. 4755 UUID uuid; 4756 if (GetUUID(&uuid)) 4757 { 4758 // this checking for the UUID load command is not enough 4759 // we could eventually look for the symbol named 4760 // "OSKextGetCurrentIdentifier" as this is required of kexts 4761 if (m_type == eTypeInvalid) 4762 m_type = eTypeSharedLibrary; 4763 4764 return eStrataKernel; 4765 } 4766 } 4767 return eStrataUnknown; 4768 4769 case MH_EXECUTE: // 0x2u 4770 // Check for the MH_DYLDLINK bit in the flags 4771 if (m_header.flags & MH_DYLDLINK) 4772 { 4773 return eStrataUser; 4774 } 4775 else 4776 { 4777 SectionList *section_list = GetSectionList(); 4778 if (section_list) 4779 { 4780 static ConstString g_kld_section_name ("__KLD"); 4781 if (section_list->FindSectionByName(g_kld_section_name)) 4782 return eStrataKernel; 4783 } 4784 } 4785 return eStrataRawImage; 4786 4787 case MH_FVMLIB: return eStrataUser; // 0x3u 4788 case MH_CORE: return eStrataUnknown; // 0x4u 4789 case MH_PRELOAD: return eStrataRawImage; // 0x5u 4790 case MH_DYLIB: return eStrataUser; // 0x6u 4791 case MH_DYLINKER: return eStrataUser; // 0x7u 4792 case MH_BUNDLE: return eStrataUser; // 0x8u 4793 case MH_DYLIB_STUB: return eStrataUser; // 0x9u 4794 case MH_DSYM: return eStrataUnknown; // 0xAu 4795 case MH_KEXT_BUNDLE: return eStrataKernel; // 0xBu 4796 default: 4797 break; 4798 } 4799 return eStrataUnknown; 4800 } 4801 4802 4803 uint32_t 4804 ObjectFileMachO::GetVersion (uint32_t *versions, uint32_t num_versions) 4805 { 4806 ModuleSP module_sp(GetModule()); 4807 if (module_sp) 4808 { 4809 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4810 struct dylib_command load_cmd; 4811 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4812 uint32_t version_cmd = 0; 4813 uint64_t version = 0; 4814 uint32_t i; 4815 for (i=0; i<m_header.ncmds; ++i) 4816 { 4817 const lldb::offset_t cmd_offset = offset; 4818 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4819 break; 4820 4821 if (load_cmd.cmd == LC_ID_DYLIB) 4822 { 4823 if (version_cmd == 0) 4824 { 4825 version_cmd = load_cmd.cmd; 4826 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 4827 break; 4828 version = load_cmd.dylib.current_version; 4829 } 4830 break; // Break for now unless there is another more complete version 4831 // number load command in the future. 4832 } 4833 offset = cmd_offset + load_cmd.cmdsize; 4834 } 4835 4836 if (version_cmd == LC_ID_DYLIB) 4837 { 4838 if (versions != NULL && num_versions > 0) 4839 { 4840 if (num_versions > 0) 4841 versions[0] = (version & 0xFFFF0000ull) >> 16; 4842 if (num_versions > 1) 4843 versions[1] = (version & 0x0000FF00ull) >> 8; 4844 if (num_versions > 2) 4845 versions[2] = (version & 0x000000FFull); 4846 // Fill in an remaining version numbers with invalid values 4847 for (i=3; i<num_versions; ++i) 4848 versions[i] = UINT32_MAX; 4849 } 4850 // The LC_ID_DYLIB load command has a version with 3 version numbers 4851 // in it, so always return 3 4852 return 3; 4853 } 4854 } 4855 return false; 4856 } 4857 4858 bool 4859 ObjectFileMachO::GetArchitecture (ArchSpec &arch) 4860 { 4861 ModuleSP module_sp(GetModule()); 4862 if (module_sp) 4863 { 4864 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4865 return GetArchitecture (m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), arch); 4866 } 4867 return false; 4868 } 4869 4870 4871 UUID 4872 ObjectFileMachO::GetProcessSharedCacheUUID (Process *process) 4873 { 4874 UUID uuid; 4875 if (process) 4876 { 4877 addr_t all_image_infos = process->GetImageInfoAddress(); 4878 4879 // The address returned by GetImageInfoAddress may be the address of dyld (don't want) 4880 // or it may be the address of the dyld_all_image_infos structure (want). The first four 4881 // bytes will be either the version field (all_image_infos) or a Mach-O file magic constant. 4882 // Version 13 and higher of dyld_all_image_infos is required to get the sharedCacheUUID field. 4883 4884 Error err; 4885 uint32_t version_or_magic = process->ReadUnsignedIntegerFromMemory (all_image_infos, 4, -1, err); 4886 if (version_or_magic != static_cast<uint32_t>(-1) 4887 && version_or_magic != MH_MAGIC 4888 && version_or_magic != MH_CIGAM 4889 && version_or_magic != MH_MAGIC_64 4890 && version_or_magic != MH_CIGAM_64 4891 && version_or_magic >= 13) 4892 { 4893 addr_t sharedCacheUUID_address = LLDB_INVALID_ADDRESS; 4894 int wordsize = process->GetAddressByteSize(); 4895 if (wordsize == 8) 4896 { 4897 sharedCacheUUID_address = all_image_infos + 160; // sharedCacheUUID <mach-o/dyld_images.h> 4898 } 4899 if (wordsize == 4) 4900 { 4901 sharedCacheUUID_address = all_image_infos + 84; // sharedCacheUUID <mach-o/dyld_images.h> 4902 } 4903 if (sharedCacheUUID_address != LLDB_INVALID_ADDRESS) 4904 { 4905 uuid_t shared_cache_uuid; 4906 if (process->ReadMemory (sharedCacheUUID_address, shared_cache_uuid, sizeof (uuid_t), err) == sizeof (uuid_t)) 4907 { 4908 uuid.SetBytes (shared_cache_uuid); 4909 } 4910 } 4911 } 4912 } 4913 return uuid; 4914 } 4915 4916 UUID 4917 ObjectFileMachO::GetLLDBSharedCacheUUID () 4918 { 4919 UUID uuid; 4920 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__) || defined (__aarch64__)) 4921 uint8_t *(*dyld_get_all_image_infos)(void); 4922 dyld_get_all_image_infos = (uint8_t*(*)()) dlsym (RTLD_DEFAULT, "_dyld_get_all_image_infos"); 4923 if (dyld_get_all_image_infos) 4924 { 4925 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 4926 if (dyld_all_image_infos_address) 4927 { 4928 uint32_t *version = (uint32_t*) dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 4929 if (*version >= 13) 4930 { 4931 uuid_t *sharedCacheUUID_address = 0; 4932 int wordsize = sizeof (uint8_t *); 4933 if (wordsize == 8) 4934 { 4935 sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 160); // sharedCacheUUID <mach-o/dyld_images.h> 4936 } 4937 else 4938 { 4939 sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 84); // sharedCacheUUID <mach-o/dyld_images.h> 4940 } 4941 uuid.SetBytes (sharedCacheUUID_address); 4942 } 4943 } 4944 } 4945 #endif 4946 return uuid; 4947 } 4948 4949 uint32_t 4950 ObjectFileMachO::GetMinimumOSVersion (uint32_t *versions, uint32_t num_versions) 4951 { 4952 if (m_min_os_versions.empty()) 4953 { 4954 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4955 bool success = false; 4956 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 4957 { 4958 const lldb::offset_t load_cmd_offset = offset; 4959 4960 version_min_command lc; 4961 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 4962 break; 4963 if (lc.cmd == LC_VERSION_MIN_MACOSX || lc.cmd == LC_VERSION_MIN_IPHONEOS) 4964 { 4965 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 4966 { 4967 const uint32_t xxxx = lc.version >> 16; 4968 const uint32_t yy = (lc.version >> 8) & 0xffu; 4969 const uint32_t zz = lc.version & 0xffu; 4970 if (xxxx) 4971 { 4972 m_min_os_versions.push_back(xxxx); 4973 if (yy) 4974 { 4975 m_min_os_versions.push_back(yy); 4976 if (zz) 4977 m_min_os_versions.push_back(zz); 4978 } 4979 } 4980 success = true; 4981 } 4982 } 4983 offset = load_cmd_offset + lc.cmdsize; 4984 } 4985 4986 if (success == false) 4987 { 4988 // Push an invalid value so we don't keep trying to 4989 m_min_os_versions.push_back(UINT32_MAX); 4990 } 4991 } 4992 4993 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) 4994 { 4995 if (versions != NULL && num_versions > 0) 4996 { 4997 for (size_t i=0; i<num_versions; ++i) 4998 { 4999 if (i < m_min_os_versions.size()) 5000 versions[i] = m_min_os_versions[i]; 5001 else 5002 versions[i] = 0; 5003 } 5004 } 5005 return m_min_os_versions.size(); 5006 } 5007 // Call the superclasses version that will empty out the data 5008 return ObjectFile::GetMinimumOSVersion (versions, num_versions); 5009 } 5010 5011 uint32_t 5012 ObjectFileMachO::GetSDKVersion(uint32_t *versions, uint32_t num_versions) 5013 { 5014 if (m_sdk_versions.empty()) 5015 { 5016 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5017 bool success = false; 5018 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 5019 { 5020 const lldb::offset_t load_cmd_offset = offset; 5021 5022 version_min_command lc; 5023 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5024 break; 5025 if (lc.cmd == LC_VERSION_MIN_MACOSX || lc.cmd == LC_VERSION_MIN_IPHONEOS) 5026 { 5027 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 5028 { 5029 const uint32_t xxxx = lc.sdk >> 16; 5030 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5031 const uint32_t zz = lc.sdk & 0xffu; 5032 if (xxxx) 5033 { 5034 m_sdk_versions.push_back(xxxx); 5035 if (yy) 5036 { 5037 m_sdk_versions.push_back(yy); 5038 if (zz) 5039 m_sdk_versions.push_back(zz); 5040 } 5041 } 5042 success = true; 5043 } 5044 } 5045 offset = load_cmd_offset + lc.cmdsize; 5046 } 5047 5048 if (success == false) 5049 { 5050 // Push an invalid value so we don't keep trying to 5051 m_sdk_versions.push_back(UINT32_MAX); 5052 } 5053 } 5054 5055 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) 5056 { 5057 if (versions != NULL && num_versions > 0) 5058 { 5059 for (size_t i=0; i<num_versions; ++i) 5060 { 5061 if (i < m_sdk_versions.size()) 5062 versions[i] = m_sdk_versions[i]; 5063 else 5064 versions[i] = 0; 5065 } 5066 } 5067 return m_sdk_versions.size(); 5068 } 5069 // Call the superclasses version that will empty out the data 5070 return ObjectFile::GetSDKVersion (versions, num_versions); 5071 } 5072 5073 5074 //------------------------------------------------------------------ 5075 // PluginInterface protocol 5076 //------------------------------------------------------------------ 5077 lldb_private::ConstString 5078 ObjectFileMachO::GetPluginName() 5079 { 5080 return GetPluginNameStatic(); 5081 } 5082 5083 uint32_t 5084 ObjectFileMachO::GetPluginVersion() 5085 { 5086 return 1; 5087 } 5088 5089 5090 bool 5091 ObjectFileMachO::SetLoadAddress (Target &target, 5092 lldb::addr_t value, 5093 bool value_is_offset) 5094 { 5095 ModuleSP module_sp = GetModule(); 5096 if (module_sp) 5097 { 5098 size_t num_loaded_sections = 0; 5099 SectionList *section_list = GetSectionList (); 5100 if (section_list) 5101 { 5102 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5103 const size_t num_sections = section_list->GetSize(); 5104 5105 const bool is_memory_image = (bool)m_process_wp.lock(); 5106 const Strata strata = GetStrata(); 5107 static ConstString g_linkedit_segname ("__LINKEDIT"); 5108 if (value_is_offset) 5109 { 5110 // "value" is an offset to apply to each top level segment 5111 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) 5112 { 5113 // Iterate through the object file sections to find all 5114 // of the sections that size on disk (to avoid __PAGEZERO) 5115 // and load them 5116 SectionSP section_sp (section_list->GetSectionAtIndex (sect_idx)); 5117 if (section_sp && 5118 section_sp->GetFileSize() > 0 && 5119 section_sp->IsThreadSpecific() == false && 5120 module_sp.get() == section_sp->GetModule().get()) 5121 { 5122 // Ignore __LINKEDIT and __DWARF segments 5123 if (section_sp->GetName() == g_linkedit_segname) 5124 { 5125 // Only map __LINKEDIT if we have an in memory image and this isn't 5126 // a kernel binary like a kext or mach_kernel. 5127 if (is_memory_image == false || strata == eStrataKernel) 5128 continue; 5129 } 5130 if (target.GetSectionLoadList().SetSectionLoadAddress (section_sp, section_sp->GetFileAddress() + value)) 5131 ++num_loaded_sections; 5132 } 5133 } 5134 } 5135 else 5136 { 5137 // "value" is the new base address of the mach_header, adjust each 5138 // section accordingly 5139 5140 // First find the address of the mach header which is the first non-zero 5141 // file sized section whose file offset is zero as this will be subtracted 5142 // from each other valid section's vmaddr and then get "base_addr" added to 5143 // it when loading the module in the target 5144 for (size_t sect_idx = 0; 5145 sect_idx < num_sections && mach_base_file_addr == LLDB_INVALID_ADDRESS; 5146 ++sect_idx) 5147 { 5148 // Iterate through the object file sections to find all 5149 // of the sections that size on disk (to avoid __PAGEZERO) 5150 // and load them 5151 Section *section = section_list->GetSectionAtIndex (sect_idx).get(); 5152 if (section && 5153 section->GetFileSize() > 0 && 5154 section->GetFileOffset() == 0 && 5155 section->IsThreadSpecific() == false && 5156 module_sp.get() == section->GetModule().get()) 5157 { 5158 // Ignore __LINKEDIT and __DWARF segments 5159 if (section->GetName() == g_linkedit_segname) 5160 { 5161 // Only map __LINKEDIT if we have an in memory image and this isn't 5162 // a kernel binary like a kext or mach_kernel. 5163 if (is_memory_image == false || strata == eStrataKernel) 5164 continue; 5165 } 5166 mach_base_file_addr = section->GetFileAddress(); 5167 } 5168 } 5169 5170 if (mach_base_file_addr != LLDB_INVALID_ADDRESS) 5171 { 5172 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) 5173 { 5174 // Iterate through the object file sections to find all 5175 // of the sections that size on disk (to avoid __PAGEZERO) 5176 // and load them 5177 SectionSP section_sp (section_list->GetSectionAtIndex (sect_idx)); 5178 if (section_sp && 5179 section_sp->GetFileSize() > 0 && 5180 section_sp->IsThreadSpecific() == false && 5181 module_sp.get() == section_sp->GetModule().get()) 5182 { 5183 // Ignore __LINKEDIT and __DWARF segments 5184 if (section_sp->GetName() == g_linkedit_segname) 5185 { 5186 // Only map __LINKEDIT if we have an in memory image and this isn't 5187 // a kernel binary like a kext or mach_kernel. 5188 if (is_memory_image == false || strata == eStrataKernel) 5189 continue; 5190 } 5191 if (target.GetSectionLoadList().SetSectionLoadAddress (section_sp, section_sp->GetFileAddress() - mach_base_file_addr + value)) 5192 ++num_loaded_sections; 5193 } 5194 } 5195 } 5196 } 5197 } 5198 return num_loaded_sections > 0; 5199 } 5200 return false; 5201 } 5202 5203 bool 5204 ObjectFileMachO::SaveCore (const lldb::ProcessSP &process_sp, 5205 const FileSpec &outfile, 5206 Error &error) 5207 { 5208 if (process_sp) 5209 { 5210 Target &target = process_sp->GetTarget(); 5211 const ArchSpec target_arch = target.GetArchitecture(); 5212 const llvm::Triple &target_triple = target_arch.GetTriple(); 5213 if (target_triple.getVendor() == llvm::Triple::Apple && 5214 (target_triple.getOS() == llvm::Triple::MacOSX || 5215 target_triple.getOS() == llvm::Triple::IOS)) 5216 { 5217 bool make_core = false; 5218 switch (target_arch.GetMachine()) 5219 { 5220 case llvm::Triple::arm: 5221 case llvm::Triple::x86: 5222 case llvm::Triple::x86_64: 5223 make_core = true; 5224 break; 5225 default: 5226 error.SetErrorStringWithFormat ("unsupported core architecture: %s", target_triple.str().c_str()); 5227 break; 5228 } 5229 5230 if (make_core) 5231 { 5232 std::vector<segment_command_64> segment_load_commands; 5233 // uint32_t range_info_idx = 0; 5234 MemoryRegionInfo range_info; 5235 Error range_error = process_sp->GetMemoryRegionInfo(0, range_info); 5236 if (range_error.Success()) 5237 { 5238 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) 5239 { 5240 const addr_t addr = range_info.GetRange().GetRangeBase(); 5241 const addr_t size = range_info.GetRange().GetByteSize(); 5242 5243 if (size == 0) 5244 break; 5245 5246 // Calculate correct protections 5247 uint32_t prot = 0; 5248 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 5249 prot |= VM_PROT_READ; 5250 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 5251 prot |= VM_PROT_WRITE; 5252 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 5253 prot |= VM_PROT_EXECUTE; 5254 5255 // printf ("[%3u] [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 ") %c%c%c\n", 5256 // range_info_idx, 5257 // addr, 5258 // size, 5259 // (prot & VM_PROT_READ ) ? 'r' : '-', 5260 // (prot & VM_PROT_WRITE ) ? 'w' : '-', 5261 // (prot & VM_PROT_EXECUTE) ? 'x' : '-'); 5262 5263 if (prot != 0) 5264 { 5265 segment_command_64 segment = { 5266 LC_SEGMENT_64, // uint32_t cmd; 5267 sizeof(segment), // uint32_t cmdsize; 5268 {0}, // char segname[16]; 5269 addr, // uint64_t vmaddr; 5270 size, // uint64_t vmsize; 5271 0, // uint64_t fileoff; 5272 size, // uint64_t filesize; 5273 prot, // uint32_t maxprot; 5274 prot, // uint32_t initprot; 5275 0, // uint32_t nsects; 5276 0 }; // uint32_t flags; 5277 segment_load_commands.push_back(segment); 5278 } 5279 else 5280 { 5281 // No protections and a size of 1 used to be returned from old 5282 // debugservers when we asked about a region that was past the 5283 // last memory region and it indicates the end... 5284 if (size == 1) 5285 break; 5286 } 5287 5288 range_error = process_sp->GetMemoryRegionInfo(range_info.GetRange().GetRangeEnd(), range_info); 5289 if (range_error.Fail()) 5290 break; 5291 } 5292 5293 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 5294 const ByteOrder byte_order = target_arch.GetByteOrder(); 5295 StreamString buffer (Stream::eBinary, 5296 addr_byte_size, 5297 byte_order); 5298 5299 mach_header_64 mach_header; 5300 mach_header.magic = MH_MAGIC_64; 5301 mach_header.cputype = target_arch.GetMachOCPUType(); 5302 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 5303 mach_header.filetype = MH_CORE; 5304 mach_header.ncmds = segment_load_commands.size(); 5305 mach_header.flags = 0; 5306 mach_header.reserved = 0; 5307 ThreadList &thread_list = process_sp->GetThreadList(); 5308 const uint32_t num_threads = thread_list.GetSize(); 5309 5310 // Make an array of LC_THREAD data items. Each one contains 5311 // the contents of the LC_THREAD load command. The data doesn't 5312 // contain the load command + load command size, we will 5313 // add the load command and load command size as we emit the data. 5314 std::vector<StreamString> LC_THREAD_datas(num_threads); 5315 for (auto &LC_THREAD_data : LC_THREAD_datas) 5316 { 5317 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 5318 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 5319 LC_THREAD_data.SetByteOrder(byte_order); 5320 } 5321 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) 5322 { 5323 ThreadSP thread_sp (thread_list.GetThreadAtIndex(thread_idx)); 5324 if (thread_sp) 5325 { 5326 switch (mach_header.cputype) 5327 { 5328 case llvm::MachO::CPU_TYPE_ARM: 5329 //RegisterContextDarwin_arm_Mach::Create_LC_THREAD (thread_sp.get(), thread_load_commands); 5330 break; 5331 5332 case llvm::MachO::CPU_TYPE_I386: 5333 //RegisterContextDarwin_i386_Mach::Create_LC_THREAD (thread_sp.get(), thread_load_commands); 5334 break; 5335 5336 case llvm::MachO::CPU_TYPE_X86_64: 5337 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5338 break; 5339 } 5340 5341 } 5342 } 5343 5344 // The size of the load command is the size of the segments... 5345 mach_header.sizeofcmds = segment_load_commands.size() * segment_load_commands[0].cmdsize; 5346 5347 // and the size of all LC_THREAD load command 5348 for (const auto &LC_THREAD_data : LC_THREAD_datas) 5349 { 5350 ++mach_header.ncmds; 5351 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 5352 } 5353 5354 printf ("mach_header: 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x\n", 5355 mach_header.magic, 5356 mach_header.cputype, 5357 mach_header.cpusubtype, 5358 mach_header.filetype, 5359 mach_header.ncmds, 5360 mach_header.sizeofcmds, 5361 mach_header.flags, 5362 mach_header.reserved); 5363 5364 // Write the mach header 5365 buffer.PutHex32(mach_header.magic); 5366 buffer.PutHex32(mach_header.cputype); 5367 buffer.PutHex32(mach_header.cpusubtype); 5368 buffer.PutHex32(mach_header.filetype); 5369 buffer.PutHex32(mach_header.ncmds); 5370 buffer.PutHex32(mach_header.sizeofcmds); 5371 buffer.PutHex32(mach_header.flags); 5372 buffer.PutHex32(mach_header.reserved); 5373 5374 // Skip the mach header and all load commands and align to the next 5375 // 0x1000 byte boundary 5376 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 5377 if (file_offset & 0x00000fff) 5378 { 5379 file_offset += 0x00001000ull; 5380 file_offset &= (~0x00001000ull + 1); 5381 } 5382 5383 for (auto &segment : segment_load_commands) 5384 { 5385 segment.fileoff = file_offset; 5386 file_offset += segment.filesize; 5387 } 5388 5389 // Write out all of the LC_THREAD load commands 5390 for (const auto &LC_THREAD_data : LC_THREAD_datas) 5391 { 5392 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 5393 buffer.PutHex32(LC_THREAD); 5394 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 5395 buffer.Write(LC_THREAD_data.GetData(), LC_THREAD_data_size); 5396 } 5397 5398 // Write out all of the segment load commands 5399 for (const auto &segment : segment_load_commands) 5400 { 5401 printf ("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 5402 segment.cmd, 5403 segment.cmdsize, 5404 segment.vmaddr, 5405 segment.vmaddr + segment.vmsize, 5406 segment.fileoff, 5407 segment.filesize, 5408 segment.maxprot, 5409 segment.initprot, 5410 segment.nsects, 5411 segment.flags); 5412 5413 buffer.PutHex32(segment.cmd); 5414 buffer.PutHex32(segment.cmdsize); 5415 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 5416 buffer.PutHex64(segment.vmaddr); 5417 buffer.PutHex64(segment.vmsize); 5418 buffer.PutHex64(segment.fileoff); 5419 buffer.PutHex64(segment.filesize); 5420 buffer.PutHex32(segment.maxprot); 5421 buffer.PutHex32(segment.initprot); 5422 buffer.PutHex32(segment.nsects); 5423 buffer.PutHex32(segment.flags); 5424 } 5425 5426 File core_file; 5427 std::string core_file_path(outfile.GetPath()); 5428 error = core_file.Open(core_file_path.c_str(), 5429 File::eOpenOptionWrite | 5430 File::eOpenOptionTruncate | 5431 File::eOpenOptionCanCreate); 5432 if (error.Success()) 5433 { 5434 // Read 1 page at a time 5435 uint8_t bytes[0x1000]; 5436 // Write the mach header and load commands out to the core file 5437 size_t bytes_written = buffer.GetString().size(); 5438 error = core_file.Write(buffer.GetString().data(), bytes_written); 5439 if (error.Success()) 5440 { 5441 // Now write the file data for all memory segments in the process 5442 for (const auto &segment : segment_load_commands) 5443 { 5444 if (core_file.SeekFromStart(segment.fileoff) == -1) 5445 { 5446 error.SetErrorStringWithFormat("unable to seek to offset 0x%" PRIx64 " in '%s'", segment.fileoff, core_file_path.c_str()); 5447 break; 5448 } 5449 5450 printf ("Saving data for segment at 0x%" PRIx64 "\n", segment.vmaddr); 5451 addr_t bytes_left = segment.vmsize; 5452 addr_t addr = segment.vmaddr; 5453 Error memory_read_error; 5454 while (bytes_left > 0 && error.Success()) 5455 { 5456 const size_t bytes_to_read = bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 5457 const size_t bytes_read = process_sp->ReadMemory(addr, bytes, bytes_to_read, memory_read_error); 5458 if (bytes_read == bytes_to_read) 5459 { 5460 size_t bytes_written = bytes_read; 5461 error = core_file.Write(bytes, bytes_written); 5462 bytes_left -= bytes_read; 5463 addr += bytes_read; 5464 } 5465 else 5466 { 5467 // Some pages within regions are not readable, those 5468 // should be zero filled 5469 memset (bytes, 0, bytes_to_read); 5470 size_t bytes_written = bytes_to_read; 5471 error = core_file.Write(bytes, bytes_written); 5472 bytes_left -= bytes_to_read; 5473 addr += bytes_to_read; 5474 } 5475 } 5476 } 5477 } 5478 } 5479 } 5480 else 5481 { 5482 error.SetErrorString("process doesn't support getting memory region info"); 5483 } 5484 } 5485 return true; // This is the right plug to handle saving core files for this process 5486 } 5487 } 5488 return false; 5489 } 5490 5491