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