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__)) 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 the above 120 // falvors... 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 the above 350 // falvors... 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 hte 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 invokved 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 hte 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 invokved 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 the file 1349 load_cmd.flags)); // Flags for this section 1350 1351 segment_sp->SetIsEncrypted (segment_is_encrypted); 1352 m_sections_ap->AddSection(segment_sp); 1353 if (add_to_unified) 1354 unified_section_list.AddSection(segment_sp); 1355 } 1356 else if (unified_section_sp) 1357 { 1358 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) 1359 { 1360 // Check to see if the module was read from memory? 1361 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) 1362 { 1363 // We have a module that is in memory and needs to have its 1364 // file address adjusted. We need to do this because when we 1365 // load a file from memory, its addresses will be slid already, 1366 // yet the addresses in the new symbol file will still be unslid. 1367 // Since everything is stored as section offset, this shouldn't 1368 // cause any problems. 1369 1370 // Make sure we've parsed the symbol table from the 1371 // ObjectFile before we go around changing its Sections. 1372 module_sp->GetObjectFile()->GetSymtab(); 1373 // eh_frame would present the same problems but we parse that on 1374 // a per-function basis as-needed so it's more difficult to 1375 // remove its use of the Sections. Realistically, the environments 1376 // where this code path will be taken will not have eh_frame sections. 1377 1378 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1379 } 1380 } 1381 m_sections_ap->AddSection(unified_section_sp); 1382 } 1383 1384 struct section_64 sect64; 1385 ::memset (§64, 0, sizeof(sect64)); 1386 // Push a section into our mach sections for the section at 1387 // index zero (NO_SECT) if we don't have any mach sections yet... 1388 if (m_mach_sections.empty()) 1389 m_mach_sections.push_back(sect64); 1390 uint32_t segment_sect_idx; 1391 const lldb::user_id_t first_segment_sectID = sectID + 1; 1392 1393 1394 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1395 for (segment_sect_idx=0; segment_sect_idx<load_cmd.nsects; ++segment_sect_idx) 1396 { 1397 if (m_data.GetU8(&offset, (uint8_t*)sect64.sectname, sizeof(sect64.sectname)) == NULL) 1398 break; 1399 if (m_data.GetU8(&offset, (uint8_t*)sect64.segname, sizeof(sect64.segname)) == NULL) 1400 break; 1401 sect64.addr = m_data.GetAddress(&offset); 1402 sect64.size = m_data.GetAddress(&offset); 1403 1404 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1405 break; 1406 1407 // Keep a list of mach sections around in case we need to 1408 // get at data that isn't stored in the abstracted Sections. 1409 m_mach_sections.push_back (sect64); 1410 1411 if (add_section) 1412 { 1413 ConstString section_name (sect64.sectname, std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 1414 if (!const_segname) 1415 { 1416 // We have a segment with no name so we need to conjure up 1417 // segments that correspond to the section's segname if there 1418 // isn't already such a section. If there is such a section, 1419 // we resize the section so that it spans all sections. 1420 // We also mark these sections as fake so address matches don't 1421 // hit if they land in the gaps between the child sections. 1422 const_segname.SetTrimmedCStringWithLength(sect64.segname, sizeof(sect64.segname)); 1423 segment_sp = unified_section_list.FindSectionByName (const_segname); 1424 if (segment_sp.get()) 1425 { 1426 Section *segment = segment_sp.get(); 1427 // Grow the section size as needed. 1428 const lldb::addr_t sect64_min_addr = sect64.addr; 1429 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1430 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1431 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1432 const lldb::addr_t curr_seg_max_addr = curr_seg_min_addr + curr_seg_byte_size; 1433 if (sect64_min_addr >= curr_seg_min_addr) 1434 { 1435 const lldb::addr_t new_seg_byte_size = sect64_max_addr - curr_seg_min_addr; 1436 // Only grow the section size if needed 1437 if (new_seg_byte_size > curr_seg_byte_size) 1438 segment->SetByteSize (new_seg_byte_size); 1439 } 1440 else 1441 { 1442 // We need to change the base address of the segment and 1443 // adjust the child section offsets for all existing children. 1444 const lldb::addr_t slide_amount = sect64_min_addr - curr_seg_min_addr; 1445 segment->Slide(slide_amount, false); 1446 segment->GetChildren().Slide(-slide_amount, false); 1447 segment->SetByteSize (curr_seg_max_addr - sect64_min_addr); 1448 } 1449 1450 // Grow the section size as needed. 1451 if (sect64.offset) 1452 { 1453 const lldb::addr_t segment_min_file_offset = segment->GetFileOffset(); 1454 const lldb::addr_t segment_max_file_offset = segment_min_file_offset + segment->GetFileSize(); 1455 1456 const lldb::addr_t section_min_file_offset = sect64.offset; 1457 const lldb::addr_t section_max_file_offset = section_min_file_offset + sect64.size; 1458 const lldb::addr_t new_file_offset = std::min (section_min_file_offset, segment_min_file_offset); 1459 const lldb::addr_t new_file_size = std::max (section_max_file_offset, segment_max_file_offset) - new_file_offset; 1460 segment->SetFileOffset (new_file_offset); 1461 segment->SetFileSize (new_file_size); 1462 } 1463 } 1464 else 1465 { 1466 // Create a fake section for the section's named segment 1467 segment_sp.reset(new Section (segment_sp, // Parent section 1468 module_sp, // Module to which this section belongs 1469 this, // Object file to which this section belongs 1470 ++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 1471 const_segname, // Name of this section 1472 eSectionTypeContainer, // This section is a container of other sections. 1473 sect64.addr, // File VM address == addresses as they are found in the object file 1474 sect64.size, // VM size in bytes of this section 1475 sect64.offset, // Offset to the data for this section in the file 1476 sect64.offset ? sect64.size : 0, // Size in bytes of this section as found in the the file 1477 load_cmd.flags)); // Flags for this section 1478 segment_sp->SetIsFake(true); 1479 1480 m_sections_ap->AddSection(segment_sp); 1481 if (add_to_unified) 1482 unified_section_list.AddSection(segment_sp); 1483 segment_sp->SetIsEncrypted (segment_is_encrypted); 1484 } 1485 } 1486 assert (segment_sp.get()); 1487 1488 uint32_t mach_sect_type = sect64.flags & SECTION_TYPE; 1489 static ConstString g_sect_name_objc_data ("__objc_data"); 1490 static ConstString g_sect_name_objc_msgrefs ("__objc_msgrefs"); 1491 static ConstString g_sect_name_objc_selrefs ("__objc_selrefs"); 1492 static ConstString g_sect_name_objc_classrefs ("__objc_classrefs"); 1493 static ConstString g_sect_name_objc_superrefs ("__objc_superrefs"); 1494 static ConstString g_sect_name_objc_const ("__objc_const"); 1495 static ConstString g_sect_name_objc_classlist ("__objc_classlist"); 1496 static ConstString g_sect_name_cfstring ("__cfstring"); 1497 1498 static ConstString g_sect_name_dwarf_debug_abbrev ("__debug_abbrev"); 1499 static ConstString g_sect_name_dwarf_debug_aranges ("__debug_aranges"); 1500 static ConstString g_sect_name_dwarf_debug_frame ("__debug_frame"); 1501 static ConstString g_sect_name_dwarf_debug_info ("__debug_info"); 1502 static ConstString g_sect_name_dwarf_debug_line ("__debug_line"); 1503 static ConstString g_sect_name_dwarf_debug_loc ("__debug_loc"); 1504 static ConstString g_sect_name_dwarf_debug_macinfo ("__debug_macinfo"); 1505 static ConstString g_sect_name_dwarf_debug_pubnames ("__debug_pubnames"); 1506 static ConstString g_sect_name_dwarf_debug_pubtypes ("__debug_pubtypes"); 1507 static ConstString g_sect_name_dwarf_debug_ranges ("__debug_ranges"); 1508 static ConstString g_sect_name_dwarf_debug_str ("__debug_str"); 1509 static ConstString g_sect_name_dwarf_apple_names ("__apple_names"); 1510 static ConstString g_sect_name_dwarf_apple_types ("__apple_types"); 1511 static ConstString g_sect_name_dwarf_apple_namespaces ("__apple_namespac"); 1512 static ConstString g_sect_name_dwarf_apple_objc ("__apple_objc"); 1513 static ConstString g_sect_name_eh_frame ("__eh_frame"); 1514 static ConstString g_sect_name_DATA ("__DATA"); 1515 static ConstString g_sect_name_TEXT ("__TEXT"); 1516 1517 lldb::SectionType sect_type = eSectionTypeOther; 1518 1519 if (section_name == g_sect_name_dwarf_debug_abbrev) 1520 sect_type = eSectionTypeDWARFDebugAbbrev; 1521 else if (section_name == g_sect_name_dwarf_debug_aranges) 1522 sect_type = eSectionTypeDWARFDebugAranges; 1523 else if (section_name == g_sect_name_dwarf_debug_frame) 1524 sect_type = eSectionTypeDWARFDebugFrame; 1525 else if (section_name == g_sect_name_dwarf_debug_info) 1526 sect_type = eSectionTypeDWARFDebugInfo; 1527 else if (section_name == g_sect_name_dwarf_debug_line) 1528 sect_type = eSectionTypeDWARFDebugLine; 1529 else if (section_name == g_sect_name_dwarf_debug_loc) 1530 sect_type = eSectionTypeDWARFDebugLoc; 1531 else if (section_name == g_sect_name_dwarf_debug_macinfo) 1532 sect_type = eSectionTypeDWARFDebugMacInfo; 1533 else if (section_name == g_sect_name_dwarf_debug_pubnames) 1534 sect_type = eSectionTypeDWARFDebugPubNames; 1535 else if (section_name == g_sect_name_dwarf_debug_pubtypes) 1536 sect_type = eSectionTypeDWARFDebugPubTypes; 1537 else if (section_name == g_sect_name_dwarf_debug_ranges) 1538 sect_type = eSectionTypeDWARFDebugRanges; 1539 else if (section_name == g_sect_name_dwarf_debug_str) 1540 sect_type = eSectionTypeDWARFDebugStr; 1541 else if (section_name == g_sect_name_dwarf_apple_names) 1542 sect_type = eSectionTypeDWARFAppleNames; 1543 else if (section_name == g_sect_name_dwarf_apple_types) 1544 sect_type = eSectionTypeDWARFAppleTypes; 1545 else if (section_name == g_sect_name_dwarf_apple_namespaces) 1546 sect_type = eSectionTypeDWARFAppleNamespaces; 1547 else if (section_name == g_sect_name_dwarf_apple_objc) 1548 sect_type = eSectionTypeDWARFAppleObjC; 1549 else if (section_name == g_sect_name_objc_selrefs) 1550 sect_type = eSectionTypeDataCStringPointers; 1551 else if (section_name == g_sect_name_objc_msgrefs) 1552 sect_type = eSectionTypeDataObjCMessageRefs; 1553 else if (section_name == g_sect_name_eh_frame) 1554 sect_type = eSectionTypeEHFrame; 1555 else if (section_name == g_sect_name_cfstring) 1556 sect_type = eSectionTypeDataObjCCFStrings; 1557 else if (section_name == g_sect_name_objc_data || 1558 section_name == g_sect_name_objc_classrefs || 1559 section_name == g_sect_name_objc_superrefs || 1560 section_name == g_sect_name_objc_const || 1561 section_name == g_sect_name_objc_classlist) 1562 { 1563 sect_type = eSectionTypeDataPointers; 1564 } 1565 1566 if (sect_type == eSectionTypeOther) 1567 { 1568 switch (mach_sect_type) 1569 { 1570 // TODO: categorize sections by other flags for regular sections 1571 case S_REGULAR: 1572 if (segment_sp->GetName() == g_sect_name_TEXT) 1573 sect_type = eSectionTypeCode; 1574 else if (segment_sp->GetName() == g_sect_name_DATA) 1575 sect_type = eSectionTypeData; 1576 else 1577 sect_type = eSectionTypeOther; 1578 break; 1579 case S_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1580 case S_CSTRING_LITERALS: sect_type = eSectionTypeDataCString; break; // section with only literal C strings 1581 case S_4BYTE_LITERALS: sect_type = eSectionTypeData4; break; // section with only 4 byte literals 1582 case S_8BYTE_LITERALS: sect_type = eSectionTypeData8; break; // section with only 8 byte literals 1583 case S_LITERAL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only pointers to literals 1584 case S_NON_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only non-lazy symbol pointers 1585 case S_LAZY_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only lazy symbol pointers 1586 case S_SYMBOL_STUBS: sect_type = eSectionTypeCode; break; // section with only symbol stubs, byte size of stub in the reserved2 field 1587 case S_MOD_INIT_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for initialization 1588 case S_MOD_TERM_FUNC_POINTERS: sect_type = eSectionTypeDataPointers; break; // section with only function pointers for termination 1589 case S_COALESCED: sect_type = eSectionTypeOther; break; 1590 case S_GB_ZEROFILL: sect_type = eSectionTypeZeroFill; break; 1591 case S_INTERPOSING: sect_type = eSectionTypeCode; break; // section with only pairs of function pointers for interposing 1592 case S_16BYTE_LITERALS: sect_type = eSectionTypeData16; break; // section with only 16 byte literals 1593 case S_DTRACE_DOF: sect_type = eSectionTypeDebug; break; 1594 case S_LAZY_DYLIB_SYMBOL_POINTERS: sect_type = eSectionTypeDataPointers; break; 1595 default: break; 1596 } 1597 } 1598 1599 SectionSP section_sp(new Section (segment_sp, 1600 module_sp, 1601 this, 1602 ++sectID, 1603 section_name, 1604 sect_type, 1605 sect64.addr - segment_sp->GetFileAddress(), 1606 sect64.size, 1607 sect64.offset, 1608 sect64.offset == 0 ? 0 : sect64.size, 1609 sect64.flags)); 1610 // Set the section to be encrypted to match the segment 1611 1612 bool section_is_encrypted = false; 1613 if (!segment_is_encrypted && load_cmd.filesize != 0) 1614 section_is_encrypted = encrypted_file_ranges.FindEntryThatContains(sect64.offset) != NULL; 1615 1616 section_sp->SetIsEncrypted (segment_is_encrypted || section_is_encrypted); 1617 segment_sp->GetChildren().AddSection(section_sp); 1618 1619 if (segment_sp->IsFake()) 1620 { 1621 segment_sp.reset(); 1622 const_segname.Clear(); 1623 } 1624 } 1625 } 1626 if (segment_sp && is_dsym) 1627 { 1628 if (first_segment_sectID <= sectID) 1629 { 1630 lldb::user_id_t sect_uid; 1631 for (sect_uid = first_segment_sectID; sect_uid <= sectID; ++sect_uid) 1632 { 1633 SectionSP curr_section_sp(segment_sp->GetChildren().FindSectionByID (sect_uid)); 1634 SectionSP next_section_sp; 1635 if (sect_uid + 1 <= sectID) 1636 next_section_sp = segment_sp->GetChildren().FindSectionByID (sect_uid+1); 1637 1638 if (curr_section_sp.get()) 1639 { 1640 if (curr_section_sp->GetByteSize() == 0) 1641 { 1642 if (next_section_sp.get() != NULL) 1643 curr_section_sp->SetByteSize ( next_section_sp->GetFileAddress() - curr_section_sp->GetFileAddress() ); 1644 else 1645 curr_section_sp->SetByteSize ( load_cmd.vmsize ); 1646 } 1647 } 1648 } 1649 } 1650 } 1651 } 1652 } 1653 } 1654 else if (load_cmd.cmd == LC_DYSYMTAB) 1655 { 1656 m_dysymtab.cmd = load_cmd.cmd; 1657 m_dysymtab.cmdsize = load_cmd.cmdsize; 1658 m_data.GetU32 (&offset, &m_dysymtab.ilocalsym, (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1659 } 1660 1661 offset = load_cmd_offset + load_cmd.cmdsize; 1662 } 1663 1664 // StreamFile s(stdout, false); // REMOVE THIS LINE 1665 // s.Printf ("Sections for %s:\n", m_file.GetPath().c_str());// REMOVE THIS LINE 1666 // m_sections_ap->Dump(&s, NULL, true, UINT32_MAX);// REMOVE THIS LINE 1667 } 1668 } 1669 1670 class MachSymtabSectionInfo 1671 { 1672 public: 1673 1674 MachSymtabSectionInfo (SectionList *section_list) : 1675 m_section_list (section_list), 1676 m_section_infos() 1677 { 1678 // Get the number of sections down to a depth of 1 to include 1679 // all segments and their sections, but no other sections that 1680 // may be added for debug map or 1681 m_section_infos.resize(section_list->GetNumSections(1)); 1682 } 1683 1684 1685 SectionSP 1686 GetSection (uint8_t n_sect, addr_t file_addr) 1687 { 1688 if (n_sect == 0) 1689 return SectionSP(); 1690 if (n_sect < m_section_infos.size()) 1691 { 1692 if (!m_section_infos[n_sect].section_sp) 1693 { 1694 SectionSP section_sp (m_section_list->FindSectionByID (n_sect)); 1695 m_section_infos[n_sect].section_sp = section_sp; 1696 if (section_sp) 1697 { 1698 m_section_infos[n_sect].vm_range.SetBaseAddress (section_sp->GetFileAddress()); 1699 m_section_infos[n_sect].vm_range.SetByteSize (section_sp->GetByteSize()); 1700 } 1701 else 1702 { 1703 Host::SystemLog (Host::eSystemLogError, "error: unable to find section for section %u\n", n_sect); 1704 } 1705 } 1706 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) 1707 { 1708 // Symbol is in section. 1709 return m_section_infos[n_sect].section_sp; 1710 } 1711 else if (m_section_infos[n_sect].vm_range.GetByteSize () == 0 && 1712 m_section_infos[n_sect].vm_range.GetBaseAddress() == file_addr) 1713 { 1714 // Symbol is in section with zero size, but has the same start 1715 // address as the section. This can happen with linker symbols 1716 // (symbols that start with the letter 'l' or 'L'. 1717 return m_section_infos[n_sect].section_sp; 1718 } 1719 } 1720 return m_section_list->FindSectionContainingFileAddress(file_addr); 1721 } 1722 1723 protected: 1724 struct SectionInfo 1725 { 1726 SectionInfo () : 1727 vm_range(), 1728 section_sp () 1729 { 1730 } 1731 1732 VMRange vm_range; 1733 SectionSP section_sp; 1734 }; 1735 SectionList *m_section_list; 1736 std::vector<SectionInfo> m_section_infos; 1737 }; 1738 1739 struct TrieEntry 1740 { 1741 TrieEntry () : 1742 name(), 1743 address(LLDB_INVALID_ADDRESS), 1744 flags (0), 1745 other(0), 1746 import_name() 1747 { 1748 } 1749 1750 void 1751 Clear () 1752 { 1753 name.Clear(); 1754 address = LLDB_INVALID_ADDRESS; 1755 flags = 0; 1756 other = 0; 1757 import_name.Clear(); 1758 } 1759 1760 void 1761 Dump () const 1762 { 1763 printf ("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1764 static_cast<unsigned long long>(address), 1765 static_cast<unsigned long long>(flags), 1766 static_cast<unsigned long long>(other), name.GetCString()); 1767 if (import_name) 1768 printf (" -> \"%s\"\n", import_name.GetCString()); 1769 else 1770 printf ("\n"); 1771 } 1772 ConstString name; 1773 uint64_t address; 1774 uint64_t flags; 1775 uint64_t other; 1776 ConstString import_name; 1777 }; 1778 1779 struct TrieEntryWithOffset 1780 { 1781 lldb::offset_t nodeOffset; 1782 TrieEntry entry; 1783 1784 TrieEntryWithOffset (lldb::offset_t offset) : 1785 nodeOffset (offset), 1786 entry() 1787 { 1788 } 1789 1790 void 1791 Dump (uint32_t idx) const 1792 { 1793 printf ("[%3u] 0x%16.16llx: ", idx, 1794 static_cast<unsigned long long>(nodeOffset)); 1795 entry.Dump(); 1796 } 1797 1798 bool 1799 operator<(const TrieEntryWithOffset& other) const 1800 { 1801 return ( nodeOffset < other.nodeOffset ); 1802 } 1803 }; 1804 1805 static void 1806 ParseTrieEntries (DataExtractor &data, 1807 lldb::offset_t offset, 1808 std::vector<llvm::StringRef> &nameSlices, 1809 std::set<lldb::addr_t> &resolver_addresses, 1810 std::vector<TrieEntryWithOffset>& output) 1811 { 1812 if (!data.ValidOffset(offset)) 1813 return; 1814 1815 const uint64_t terminalSize = data.GetULEB128(&offset); 1816 lldb::offset_t children_offset = offset + terminalSize; 1817 if ( terminalSize != 0 ) { 1818 TrieEntryWithOffset e (offset); 1819 e.entry.flags = data.GetULEB128(&offset); 1820 const char *import_name = NULL; 1821 if ( e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT ) { 1822 e.entry.address = 0; 1823 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 1824 import_name = data.GetCStr(&offset); 1825 } 1826 else { 1827 e.entry.address = data.GetULEB128(&offset); 1828 if ( e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER ) 1829 { 1830 //resolver_addresses.insert(e.entry.address); 1831 e.entry.other = data.GetULEB128(&offset); 1832 resolver_addresses.insert(e.entry.other); 1833 } 1834 else 1835 e.entry.other = 0; 1836 } 1837 // Only add symbols that are reexport symbols with a valid import name 1838 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && import_name[0]) 1839 { 1840 std::string name; 1841 if (!nameSlices.empty()) 1842 { 1843 for (auto name_slice: nameSlices) 1844 name.append(name_slice.data(), name_slice.size()); 1845 } 1846 if (name.size() > 1) 1847 { 1848 // Skip the leading '_' 1849 e.entry.name.SetCStringWithLength(name.c_str() + 1,name.size() - 1); 1850 } 1851 if (import_name) 1852 { 1853 // Skip the leading '_' 1854 e.entry.import_name.SetCString(import_name+1); 1855 } 1856 output.push_back(e); 1857 } 1858 } 1859 1860 const uint8_t childrenCount = data.GetU8(&children_offset); 1861 for (uint8_t i=0; i < childrenCount; ++i) { 1862 nameSlices.push_back(data.GetCStr(&children_offset)); 1863 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 1864 if (childNodeOffset) 1865 { 1866 ParseTrieEntries(data, 1867 childNodeOffset, 1868 nameSlices, 1869 resolver_addresses, 1870 output); 1871 } 1872 nameSlices.pop_back(); 1873 } 1874 } 1875 1876 size_t 1877 ObjectFileMachO::ParseSymtab () 1878 { 1879 Timer scoped_timer(__PRETTY_FUNCTION__, 1880 "ObjectFileMachO::ParseSymtab () module = %s", 1881 m_file.GetFilename().AsCString("")); 1882 ModuleSP module_sp (GetModule()); 1883 if (!module_sp) 1884 return 0; 1885 1886 struct symtab_command symtab_load_command = { 0, 0, 0, 0, 0, 0 }; 1887 struct linkedit_data_command function_starts_load_command = { 0, 0, 0, 0 }; 1888 struct dyld_info_command dyld_info = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }; 1889 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 1890 FunctionStarts function_starts; 1891 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1892 uint32_t i; 1893 FileSpecList dylib_files; 1894 Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_SYMBOLS)); 1895 1896 for (i=0; i<m_header.ncmds; ++i) 1897 { 1898 const lldb::offset_t cmd_offset = offset; 1899 // Read in the load command and load command size 1900 struct load_command lc; 1901 if (m_data.GetU32(&offset, &lc, 2) == NULL) 1902 break; 1903 // Watch for the symbol table load command 1904 switch (lc.cmd) 1905 { 1906 case LC_SYMTAB: 1907 symtab_load_command.cmd = lc.cmd; 1908 symtab_load_command.cmdsize = lc.cmdsize; 1909 // Read in the rest of the symtab load command 1910 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 0) // fill in symoff, nsyms, stroff, strsize fields 1911 return 0; 1912 if (symtab_load_command.symoff == 0) 1913 { 1914 if (log) 1915 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 1916 return 0; 1917 } 1918 1919 if (symtab_load_command.stroff == 0) 1920 { 1921 if (log) 1922 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 1923 return 0; 1924 } 1925 1926 if (symtab_load_command.nsyms == 0) 1927 { 1928 if (log) 1929 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 1930 return 0; 1931 } 1932 1933 if (symtab_load_command.strsize == 0) 1934 { 1935 if (log) 1936 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 1937 return 0; 1938 } 1939 break; 1940 1941 case LC_DYLD_INFO: 1942 case LC_DYLD_INFO_ONLY: 1943 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) 1944 { 1945 dyld_info.cmd = lc.cmd; 1946 dyld_info.cmdsize = lc.cmdsize; 1947 } 1948 else 1949 { 1950 memset (&dyld_info, 0, sizeof(dyld_info)); 1951 } 1952 break; 1953 1954 case LC_LOAD_DYLIB: 1955 case LC_LOAD_WEAK_DYLIB: 1956 case LC_REEXPORT_DYLIB: 1957 case LC_LOADFVMLIB: 1958 case LC_LOAD_UPWARD_DYLIB: 1959 { 1960 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 1961 const char *path = m_data.PeekCStr(name_offset); 1962 if (path) 1963 { 1964 FileSpec file_spec(path, false); 1965 // Strip the path if there is @rpath, @executanble, etc so we just use the basename 1966 if (path[0] == '@') 1967 file_spec.GetDirectory().Clear(); 1968 1969 if (lc.cmd == LC_REEXPORT_DYLIB) 1970 { 1971 m_reexported_dylibs.AppendIfUnique(file_spec); 1972 } 1973 1974 dylib_files.Append(file_spec); 1975 } 1976 } 1977 break; 1978 1979 case LC_FUNCTION_STARTS: 1980 function_starts_load_command.cmd = lc.cmd; 1981 function_starts_load_command.cmdsize = lc.cmdsize; 1982 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == NULL) // fill in symoff, nsyms, stroff, strsize fields 1983 memset (&function_starts_load_command, 0, sizeof(function_starts_load_command)); 1984 break; 1985 1986 default: 1987 break; 1988 } 1989 offset = cmd_offset + lc.cmdsize; 1990 } 1991 1992 if (symtab_load_command.cmd) 1993 { 1994 Symtab *symtab = m_symtab_ap.get(); 1995 SectionList *section_list = GetSectionList(); 1996 if (section_list == NULL) 1997 return 0; 1998 1999 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2000 const ByteOrder byte_order = m_data.GetByteOrder(); 2001 bool bit_width_32 = addr_byte_size == 4; 2002 const size_t nlist_byte_size = bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2003 2004 DataExtractor nlist_data (NULL, 0, byte_order, addr_byte_size); 2005 DataExtractor strtab_data (NULL, 0, byte_order, addr_byte_size); 2006 DataExtractor function_starts_data (NULL, 0, byte_order, addr_byte_size); 2007 DataExtractor indirect_symbol_index_data (NULL, 0, byte_order, addr_byte_size); 2008 DataExtractor dyld_trie_data (NULL, 0, byte_order, addr_byte_size); 2009 2010 const addr_t nlist_data_byte_size = symtab_load_command.nsyms * nlist_byte_size; 2011 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2012 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2013 2014 ProcessSP process_sp (m_process_wp.lock()); 2015 Process *process = process_sp.get(); 2016 2017 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2018 2019 if (process) 2020 { 2021 Target &target = process->GetTarget(); 2022 2023 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2024 2025 SectionSP linkedit_section_sp(section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2026 // Reading mach file from memory in a process or core file... 2027 2028 if (linkedit_section_sp) 2029 { 2030 const addr_t linkedit_load_addr = linkedit_section_sp->GetLoadBaseAddress(&target); 2031 const addr_t linkedit_file_offset = linkedit_section_sp->GetFileOffset(); 2032 const addr_t symoff_addr = linkedit_load_addr + symtab_load_command.symoff - linkedit_file_offset; 2033 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - linkedit_file_offset; 2034 2035 bool data_was_read = false; 2036 2037 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__)) 2038 if (m_header.flags & 0x80000000u && process->GetAddressByteSize() == sizeof (void*)) 2039 { 2040 // This mach-o memory file is in the dyld shared cache. If this 2041 // program is not remote and this is iOS, then this process will 2042 // share the same shared cache as the process we are debugging and 2043 // we can read the entire __LINKEDIT from the address space in this 2044 // process. This is a needed optimization that is used for local iOS 2045 // debugging only since all shared libraries in the shared cache do 2046 // not have corresponding files that exist in the file system of the 2047 // device. They have been combined into a single file. This means we 2048 // always have to load these files from memory. All of the symbol and 2049 // string tables from all of the __LINKEDIT sections from the shared 2050 // libraries in the shared cache have been merged into a single large 2051 // symbol and string table. Reading all of this symbol and string table 2052 // data across can slow down debug launch times, so we optimize this by 2053 // reading the memory for the __LINKEDIT section from this process. 2054 2055 UUID lldb_shared_cache(GetLLDBSharedCacheUUID()); 2056 UUID process_shared_cache(GetProcessSharedCacheUUID(process)); 2057 bool use_lldb_cache = true; 2058 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && lldb_shared_cache != process_shared_cache) 2059 { 2060 use_lldb_cache = false; 2061 ModuleSP module_sp (GetModule()); 2062 if (module_sp) 2063 module_sp->ReportWarning ("shared cache in process does not match lldb's own shared cache, startup will be slow."); 2064 2065 } 2066 2067 PlatformSP platform_sp (target.GetPlatform()); 2068 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) 2069 { 2070 data_was_read = true; 2071 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, eByteOrderLittle); 2072 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, eByteOrderLittle); 2073 if (function_starts_load_command.cmd) 2074 { 2075 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 2076 function_starts_data.SetData ((void *)func_start_addr, function_starts_load_command.datasize, eByteOrderLittle); 2077 } 2078 } 2079 } 2080 #endif 2081 2082 if (!data_was_read) 2083 { 2084 if (memory_module_load_level == eMemoryModuleLoadLevelComplete) 2085 { 2086 DataBufferSP nlist_data_sp (ReadMemory (process_sp, symoff_addr, nlist_data_byte_size)); 2087 if (nlist_data_sp) 2088 nlist_data.SetData (nlist_data_sp, 0, nlist_data_sp->GetByteSize()); 2089 // Load strings individually from memory when loading from memory since shared cache 2090 // string tables contain strings for all symbols from all shared cached libraries 2091 //DataBufferSP strtab_data_sp (ReadMemory (process_sp, strtab_addr, strtab_data_byte_size)); 2092 //if (strtab_data_sp) 2093 // strtab_data.SetData (strtab_data_sp, 0, strtab_data_sp->GetByteSize()); 2094 if (m_dysymtab.nindirectsyms != 0) 2095 { 2096 const addr_t indirect_syms_addr = linkedit_load_addr + m_dysymtab.indirectsymoff - linkedit_file_offset; 2097 DataBufferSP indirect_syms_data_sp (ReadMemory (process_sp, indirect_syms_addr, m_dysymtab.nindirectsyms * 4)); 2098 if (indirect_syms_data_sp) 2099 indirect_symbol_index_data.SetData (indirect_syms_data_sp, 0, indirect_syms_data_sp->GetByteSize()); 2100 } 2101 } 2102 2103 if (memory_module_load_level >= eMemoryModuleLoadLevelPartial) 2104 { 2105 if (function_starts_load_command.cmd) 2106 { 2107 const addr_t func_start_addr = linkedit_load_addr + function_starts_load_command.dataoff - linkedit_file_offset; 2108 DataBufferSP func_start_data_sp (ReadMemory (process_sp, func_start_addr, function_starts_load_command.datasize)); 2109 if (func_start_data_sp) 2110 function_starts_data.SetData (func_start_data_sp, 0, func_start_data_sp->GetByteSize()); 2111 } 2112 } 2113 } 2114 } 2115 } 2116 else 2117 { 2118 nlist_data.SetData (m_data, 2119 symtab_load_command.symoff, 2120 nlist_data_byte_size); 2121 strtab_data.SetData (m_data, 2122 symtab_load_command.stroff, 2123 strtab_data_byte_size); 2124 2125 if (dyld_info.export_size > 0) 2126 { 2127 dyld_trie_data.SetData (m_data, 2128 dyld_info.export_off, 2129 dyld_info.export_size); 2130 } 2131 2132 if (m_dysymtab.nindirectsyms != 0) 2133 { 2134 indirect_symbol_index_data.SetData (m_data, 2135 m_dysymtab.indirectsymoff, 2136 m_dysymtab.nindirectsyms * 4); 2137 } 2138 if (function_starts_load_command.cmd) 2139 { 2140 function_starts_data.SetData (m_data, 2141 function_starts_load_command.dataoff, 2142 function_starts_load_command.datasize); 2143 } 2144 } 2145 2146 if (nlist_data.GetByteSize() == 0 && memory_module_load_level == eMemoryModuleLoadLevelComplete) 2147 { 2148 if (log) 2149 module_sp->LogMessage(log, "failed to read nlist data"); 2150 return 0; 2151 } 2152 2153 2154 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2155 if (!have_strtab_data) 2156 { 2157 if (process) 2158 { 2159 if (strtab_addr == LLDB_INVALID_ADDRESS) 2160 { 2161 if (log) 2162 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2163 return 0; 2164 } 2165 } 2166 else 2167 { 2168 if (log) 2169 module_sp->LogMessage(log, "failed to read strtab data"); 2170 return 0; 2171 } 2172 } 2173 2174 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2175 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2176 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2177 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2178 SectionSP text_section_sp(section_list->FindSectionByName(g_segment_name_TEXT)); 2179 SectionSP data_section_sp(section_list->FindSectionByName(g_segment_name_DATA)); 2180 SectionSP objc_section_sp(section_list->FindSectionByName(g_segment_name_OBJC)); 2181 SectionSP eh_frame_section_sp; 2182 if (text_section_sp.get()) 2183 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName (g_section_name_eh_frame); 2184 else 2185 eh_frame_section_sp = section_list->FindSectionByName (g_section_name_eh_frame); 2186 2187 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2188 2189 // lldb works best if it knows the start addresss of all functions in a module. 2190 // Linker symbols or debug info are normally the best source of information for start addr / size but 2191 // they may be stripped in a released binary. 2192 // Two additional sources of information exist in Mach-O binaries: 2193 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each function's start address in the 2194 // binary, relative to the text section. 2195 // eh_frame - the eh_frame FDEs have the start addr & size of each function 2196 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on all modern binaries. 2197 // Binaries built to run on older releases may need to use eh_frame information. 2198 2199 if (text_section_sp && function_starts_data.GetByteSize()) 2200 { 2201 FunctionStarts::Entry function_start_entry; 2202 function_start_entry.data = false; 2203 lldb::offset_t function_start_offset = 0; 2204 function_start_entry.addr = text_section_sp->GetFileAddress(); 2205 uint64_t delta; 2206 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 0) 2207 { 2208 // Now append the current entry 2209 function_start_entry.addr += delta; 2210 function_starts.Append(function_start_entry); 2211 } 2212 } 2213 else 2214 { 2215 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the load command claiming an eh_frame 2216 // but it doesn't actually have the eh_frame content. And if we have a dSYM, we don't need to do any 2217 // of this fill-in-the-missing-symbols works anyway - the debug info should give us all the functions in 2218 // the module. 2219 if (text_section_sp.get() && eh_frame_section_sp.get() && m_type != eTypeDebugInfo) 2220 { 2221 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, eRegisterKindGCC, true); 2222 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2223 eh_frame.GetFunctionAddressAndSizeVector (functions); 2224 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2225 size_t count = functions.GetSize(); 2226 for (size_t i = 0; i < count; ++i) 2227 { 2228 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = functions.GetEntryAtIndex (i); 2229 if (func) 2230 { 2231 FunctionStarts::Entry function_start_entry; 2232 function_start_entry.addr = func->base - text_base_addr; 2233 function_starts.Append(function_start_entry); 2234 } 2235 } 2236 } 2237 } 2238 2239 const size_t function_starts_count = function_starts.GetSize(); 2240 2241 const user_id_t TEXT_eh_frame_sectID = eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() : NO_SECT; 2242 2243 lldb::offset_t nlist_data_offset = 0; 2244 2245 uint32_t N_SO_index = UINT32_MAX; 2246 2247 MachSymtabSectionInfo section_info (section_list); 2248 std::vector<uint32_t> N_FUN_indexes; 2249 std::vector<uint32_t> N_NSYM_indexes; 2250 std::vector<uint32_t> N_INCL_indexes; 2251 std::vector<uint32_t> N_BRAC_indexes; 2252 std::vector<uint32_t> N_COMM_indexes; 2253 typedef std::multimap <uint64_t, uint32_t> ValueToSymbolIndexMap; 2254 typedef std::map <uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2255 typedef std::map <const char *, uint32_t> ConstNameToSymbolIndexMap; 2256 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2257 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2258 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2259 // Any symbols that get merged into another will get an entry 2260 // in this map so we know 2261 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2262 uint32_t nlist_idx = 0; 2263 Symbol *symbol_ptr = NULL; 2264 2265 uint32_t sym_idx = 0; 2266 Symbol *sym = NULL; 2267 size_t num_syms = 0; 2268 std::string memory_symbol_name; 2269 uint32_t unmapped_local_symbols_found = 0; 2270 2271 std::vector<TrieEntryWithOffset> trie_entries; 2272 std::set<lldb::addr_t> resolver_addresses; 2273 2274 if (dyld_trie_data.GetByteSize() > 0) 2275 { 2276 std::vector<llvm::StringRef> nameSlices; 2277 ParseTrieEntries (dyld_trie_data, 2278 0, 2279 nameSlices, 2280 resolver_addresses, 2281 trie_entries); 2282 2283 ConstString text_segment_name ("__TEXT"); 2284 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 2285 if (text_segment_sp) 2286 { 2287 const lldb::addr_t text_segment_file_addr = text_segment_sp->GetFileAddress(); 2288 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) 2289 { 2290 for (auto &e : trie_entries) 2291 e.entry.address += text_segment_file_addr; 2292 } 2293 } 2294 } 2295 2296 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__)) 2297 2298 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been optimized by moving LOCAL 2299 // symbols out of the memory mapped portion of the DSC. The symbol information has all been retained, 2300 // but it isn't available in the normal nlist data. However, there *are* duplicate entries of *some* 2301 // LOCAL symbols in the normal nlist data. To handle this situation correctly, we must first attempt 2302 // to parse any DSC unmapped symbol information. If we find any, we set a flag that tells the normal 2303 // nlist parser to ignore all LOCAL symbols. 2304 2305 if (m_header.flags & 0x80000000u) 2306 { 2307 // Before we can start mapping the DSC, we need to make certain the target process is actually 2308 // using the cache we can find. 2309 2310 // Next we need to determine the correct path for the dyld shared cache. 2311 2312 ArchSpec header_arch; 2313 GetArchitecture(header_arch); 2314 char dsc_path[PATH_MAX]; 2315 2316 snprintf(dsc_path, sizeof(dsc_path), "%s%s%s", 2317 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR */ 2318 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2319 header_arch.GetArchitectureName()); 2320 2321 FileSpec dsc_filespec(dsc_path, false); 2322 2323 // We need definitions of two structures in the on-disk DSC, copy them here manually 2324 struct lldb_copy_dyld_cache_header_v0 2325 { 2326 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 2327 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 2328 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 2329 uint32_t imagesOffset; 2330 uint32_t imagesCount; 2331 uint64_t dyldBaseAddress; 2332 uint64_t codeSignatureOffset; 2333 uint64_t codeSignatureSize; 2334 uint64_t slideInfoOffset; 2335 uint64_t slideInfoSize; 2336 uint64_t localSymbolsOffset; // file offset of where local symbols are stored 2337 uint64_t localSymbolsSize; // size of local symbols information 2338 }; 2339 struct lldb_copy_dyld_cache_header_v1 2340 { 2341 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 2342 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 2343 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 2344 uint32_t imagesOffset; 2345 uint32_t imagesCount; 2346 uint64_t dyldBaseAddress; 2347 uint64_t codeSignatureOffset; 2348 uint64_t codeSignatureSize; 2349 uint64_t slideInfoOffset; 2350 uint64_t slideInfoSize; 2351 uint64_t localSymbolsOffset; 2352 uint64_t localSymbolsSize; 2353 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 and later 2354 }; 2355 2356 struct lldb_copy_dyld_cache_mapping_info 2357 { 2358 uint64_t address; 2359 uint64_t size; 2360 uint64_t fileOffset; 2361 uint32_t maxProt; 2362 uint32_t initProt; 2363 }; 2364 2365 struct lldb_copy_dyld_cache_local_symbols_info 2366 { 2367 uint32_t nlistOffset; 2368 uint32_t nlistCount; 2369 uint32_t stringsOffset; 2370 uint32_t stringsSize; 2371 uint32_t entriesOffset; 2372 uint32_t entriesCount; 2373 }; 2374 struct lldb_copy_dyld_cache_local_symbols_entry 2375 { 2376 uint32_t dylibOffset; 2377 uint32_t nlistStartIndex; 2378 uint32_t nlistCount; 2379 }; 2380 2381 /* The dyld_cache_header has a pointer to the dyld_cache_local_symbols_info structure (localSymbolsOffset). 2382 The dyld_cache_local_symbols_info structure gives us three things: 2383 1. The start and count of the nlist records in the dyld_shared_cache file 2384 2. The start and size of the strings for these nlist records 2385 3. The start and count of dyld_cache_local_symbols_entry entries 2386 2387 There is one dyld_cache_local_symbols_entry per dylib/framework in the dyld shared cache. 2388 The "dylibOffset" field is the Mach-O header of this dylib/framework in the dyld shared cache. 2389 The dyld_cache_local_symbols_entry also lists the start of this dylib/framework's nlist records 2390 and the count of how many nlist records there are for this dylib/framework. 2391 */ 2392 2393 // Process the dsc header to find the unmapped symbols 2394 // 2395 // Save some VM space, do not map the entire cache in one shot. 2396 2397 DataBufferSP dsc_data_sp; 2398 dsc_data_sp = dsc_filespec.MemoryMapFileContents(0, sizeof(struct lldb_copy_dyld_cache_header_v1)); 2399 2400 if (dsc_data_sp) 2401 { 2402 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2403 2404 char version_str[17]; 2405 int version = -1; 2406 lldb::offset_t offset = 0; 2407 memcpy (version_str, dsc_header_data.GetData (&offset, 16), 16); 2408 version_str[16] = '\0'; 2409 if (strncmp (version_str, "dyld_v", 6) == 0 && isdigit (version_str[6])) 2410 { 2411 int v; 2412 if (::sscanf (version_str + 6, "%d", &v) == 1) 2413 { 2414 version = v; 2415 } 2416 } 2417 2418 UUID dsc_uuid; 2419 if (version >= 1) 2420 { 2421 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, uuid); 2422 uint8_t uuid_bytes[sizeof (uuid_t)]; 2423 memcpy (uuid_bytes, dsc_header_data.GetData (&offset, sizeof (uuid_t)), sizeof (uuid_t)); 2424 dsc_uuid.SetBytes (uuid_bytes); 2425 } 2426 2427 bool uuid_match = true; 2428 if (dsc_uuid.IsValid() && process) 2429 { 2430 UUID shared_cache_uuid(GetProcessSharedCacheUUID(process)); 2431 2432 if (shared_cache_uuid.IsValid() && dsc_uuid != shared_cache_uuid) 2433 { 2434 // The on-disk dyld_shared_cache file is not the same as the one in this 2435 // process' memory, don't use it. 2436 uuid_match = false; 2437 ModuleSP module_sp (GetModule()); 2438 if (module_sp) 2439 module_sp->ReportWarning ("process shared cache does not match on-disk dyld_shared_cache file, some symbol names will be missing."); 2440 } 2441 } 2442 2443 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2444 2445 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2446 2447 // If the mappingOffset points to a location inside the header, we've 2448 // opened an old dyld shared cache, and should not proceed further. 2449 if (uuid_match && mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v0)) 2450 { 2451 2452 DataBufferSP dsc_mapping_info_data_sp = dsc_filespec.MemoryMapFileContents(mappingOffset, sizeof (struct lldb_copy_dyld_cache_mapping_info)); 2453 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, byte_order, addr_byte_size); 2454 offset = 0; 2455 2456 // The File addresses (from the in-memory Mach-O load commands) for the shared libraries 2457 // in the shared library cache need to be adjusted by an offset to match up with the 2458 // dylibOffset identifying field in the dyld_cache_local_symbol_entry's. This offset is 2459 // recorded in mapping_offset_value. 2460 const uint64_t mapping_offset_value = dsc_mapping_info_data.GetU64(&offset); 2461 2462 offset = offsetof (struct lldb_copy_dyld_cache_header_v1, localSymbolsOffset); 2463 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2464 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2465 2466 if (localSymbolsOffset && localSymbolsSize) 2467 { 2468 // Map the local symbols 2469 if (DataBufferSP dsc_local_symbols_data_sp = dsc_filespec.MemoryMapFileContents(localSymbolsOffset, localSymbolsSize)) 2470 { 2471 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, byte_order, addr_byte_size); 2472 2473 offset = 0; 2474 2475 // Read the local_symbols_infos struct in one shot 2476 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2477 dsc_local_symbols_data.GetU32(&offset, &local_symbols_info.nlistOffset, 6); 2478 2479 SectionSP text_section_sp(section_list->FindSectionByName(GetSegmentNameTEXT())); 2480 2481 uint32_t header_file_offset = (text_section_sp->GetFileAddress() - mapping_offset_value); 2482 2483 offset = local_symbols_info.entriesOffset; 2484 for (uint32_t entry_index = 0; entry_index < local_symbols_info.entriesCount; entry_index++) 2485 { 2486 struct lldb_copy_dyld_cache_local_symbols_entry local_symbols_entry; 2487 local_symbols_entry.dylibOffset = dsc_local_symbols_data.GetU32(&offset); 2488 local_symbols_entry.nlistStartIndex = dsc_local_symbols_data.GetU32(&offset); 2489 local_symbols_entry.nlistCount = dsc_local_symbols_data.GetU32(&offset); 2490 2491 if (header_file_offset == local_symbols_entry.dylibOffset) 2492 { 2493 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2494 2495 // The normal nlist code cannot correctly size the Symbols array, we need to allocate it here. 2496 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms + unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2497 num_syms = symtab->GetNumSymbols(); 2498 2499 nlist_data_offset = local_symbols_info.nlistOffset + (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2500 uint32_t string_table_offset = local_symbols_info.stringsOffset; 2501 2502 for (uint32_t nlist_index = 0; nlist_index < local_symbols_entry.nlistCount; nlist_index++) 2503 { 2504 ///////////////////////////// 2505 { 2506 struct nlist_64 nlist; 2507 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 2508 break; 2509 2510 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked(&nlist_data_offset); 2511 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2512 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked (&nlist_data_offset); 2513 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked (&nlist_data_offset); 2514 nlist.n_value = dsc_local_symbols_data.GetAddress_unchecked (&nlist_data_offset); 2515 2516 SymbolType type = eSymbolTypeInvalid; 2517 const char *symbol_name = dsc_local_symbols_data.PeekCStr(string_table_offset + nlist.n_strx); 2518 2519 if (symbol_name == NULL) 2520 { 2521 // No symbol should be NULL, even the symbols with no 2522 // string values should have an offset zero which points 2523 // to an empty C-string 2524 Host::SystemLog (Host::eSystemLogError, 2525 "error: DSC unmapped local symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 2526 entry_index, 2527 nlist.n_strx, 2528 module_sp->GetFileSpec().GetPath().c_str()); 2529 continue; 2530 } 2531 if (symbol_name[0] == '\0') 2532 symbol_name = NULL; 2533 2534 const char *symbol_name_non_abi_mangled = NULL; 2535 2536 SectionSP symbol_section; 2537 uint32_t symbol_byte_size = 0; 2538 bool add_nlist = true; 2539 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2540 bool demangled_is_synthesized = false; 2541 bool is_gsym = false; 2542 2543 assert (sym_idx < num_syms); 2544 2545 sym[sym_idx].SetDebug (is_debug); 2546 2547 if (is_debug) 2548 { 2549 switch (nlist.n_type) 2550 { 2551 case N_GSYM: 2552 // global symbol: name,,NO_SECT,type,0 2553 // Sometimes the N_GSYM value contains the address. 2554 2555 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 2556 // have the same address, but we want to ensure that we always find only the real symbol, 2557 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 2558 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 2559 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 2560 // same address. 2561 2562 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 2563 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 2564 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 2565 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 2566 add_nlist = false; 2567 else 2568 { 2569 is_gsym = true; 2570 sym[sym_idx].SetExternal(true); 2571 if (nlist.n_value != 0) 2572 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2573 type = eSymbolTypeData; 2574 } 2575 break; 2576 2577 case N_FNAME: 2578 // procedure name (f77 kludge): name,,NO_SECT,0,0 2579 type = eSymbolTypeCompiler; 2580 break; 2581 2582 case N_FUN: 2583 // procedure: name,,n_sect,linenumber,address 2584 if (symbol_name) 2585 { 2586 type = eSymbolTypeCode; 2587 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2588 2589 N_FUN_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 2590 // We use the current number of symbols in the symbol table in lieu of 2591 // using nlist_idx in case we ever start trimming entries out 2592 N_FUN_indexes.push_back(sym_idx); 2593 } 2594 else 2595 { 2596 type = eSymbolTypeCompiler; 2597 2598 if ( !N_FUN_indexes.empty() ) 2599 { 2600 // Copy the size of the function into the original STAB entry so we don't have 2601 // to hunt for it later 2602 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 2603 N_FUN_indexes.pop_back(); 2604 // We don't really need the end function STAB as it contains the size which 2605 // we already placed with the original symbol, so don't add it if we want a 2606 // minimal symbol table 2607 add_nlist = false; 2608 } 2609 } 2610 break; 2611 2612 case N_STSYM: 2613 // static symbol: name,,n_sect,type,address 2614 N_STSYM_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 2615 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2616 type = eSymbolTypeData; 2617 break; 2618 2619 case N_LCSYM: 2620 // .lcomm symbol: name,,n_sect,type,address 2621 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2622 type = eSymbolTypeCommonBlock; 2623 break; 2624 2625 case N_BNSYM: 2626 // We use the current number of symbols in the symbol table in lieu of 2627 // using nlist_idx in case we ever start trimming entries out 2628 // Skip these if we want minimal symbol tables 2629 add_nlist = false; 2630 break; 2631 2632 case N_ENSYM: 2633 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 2634 // so that we can always skip the entire symbol if we need to navigate 2635 // more quickly at the source level when parsing STABS 2636 // Skip these if we want minimal symbol tables 2637 add_nlist = false; 2638 break; 2639 2640 2641 case N_OPT: 2642 // emitted with gcc2_compiled and in gcc source 2643 type = eSymbolTypeCompiler; 2644 break; 2645 2646 case N_RSYM: 2647 // register sym: name,,NO_SECT,type,register 2648 type = eSymbolTypeVariable; 2649 break; 2650 2651 case N_SLINE: 2652 // src line: 0,,n_sect,linenumber,address 2653 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2654 type = eSymbolTypeLineEntry; 2655 break; 2656 2657 case N_SSYM: 2658 // structure elt: name,,NO_SECT,type,struct_offset 2659 type = eSymbolTypeVariableType; 2660 break; 2661 2662 case N_SO: 2663 // source file name 2664 type = eSymbolTypeSourceFile; 2665 if (symbol_name == NULL) 2666 { 2667 add_nlist = false; 2668 if (N_SO_index != UINT32_MAX) 2669 { 2670 // Set the size of the N_SO to the terminating index of this N_SO 2671 // so that we can always skip the entire N_SO if we need to navigate 2672 // more quickly at the source level when parsing STABS 2673 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2674 symbol_ptr->SetByteSize(sym_idx); 2675 symbol_ptr->SetSizeIsSibling(true); 2676 } 2677 N_NSYM_indexes.clear(); 2678 N_INCL_indexes.clear(); 2679 N_BRAC_indexes.clear(); 2680 N_COMM_indexes.clear(); 2681 N_FUN_indexes.clear(); 2682 N_SO_index = UINT32_MAX; 2683 } 2684 else 2685 { 2686 // We use the current number of symbols in the symbol table in lieu of 2687 // using nlist_idx in case we ever start trimming entries out 2688 const bool N_SO_has_full_path = symbol_name[0] == '/'; 2689 if (N_SO_has_full_path) 2690 { 2691 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2692 { 2693 // We have two consecutive N_SO entries where the first contains a directory 2694 // and the second contains a full path. 2695 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 2696 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2697 add_nlist = false; 2698 } 2699 else 2700 { 2701 // This is the first entry in a N_SO that contains a directory or 2702 // a full path to the source file 2703 N_SO_index = sym_idx; 2704 } 2705 } 2706 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 2707 { 2708 // This is usually the second N_SO entry that contains just the filename, 2709 // so here we combine it with the first one if we are minimizing the symbol table 2710 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 2711 if (so_path && so_path[0]) 2712 { 2713 std::string full_so_path (so_path); 2714 const size_t double_slash_pos = full_so_path.find("//"); 2715 if (double_slash_pos != std::string::npos) 2716 { 2717 // The linker has been generating bad N_SO entries with doubled up paths 2718 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 2719 // and the second is the directory for the source file so you end up with 2720 // a path that looks like "/tmp/src//tmp/src/" 2721 FileSpec so_dir(so_path, false); 2722 if (!so_dir.Exists()) 2723 { 2724 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 2725 if (so_dir.Exists()) 2726 { 2727 // Trim off the incorrect path 2728 full_so_path.erase(0, double_slash_pos + 1); 2729 } 2730 } 2731 } 2732 if (*full_so_path.rbegin() != '/') 2733 full_so_path += '/'; 2734 full_so_path += symbol_name; 2735 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 2736 add_nlist = false; 2737 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 2738 } 2739 } 2740 else 2741 { 2742 // This could be a relative path to a N_SO 2743 N_SO_index = sym_idx; 2744 } 2745 } 2746 break; 2747 2748 case N_OSO: 2749 // object file name: name,,0,0,st_mtime 2750 type = eSymbolTypeObjectFile; 2751 break; 2752 2753 case N_LSYM: 2754 // local sym: name,,NO_SECT,type,offset 2755 type = eSymbolTypeLocal; 2756 break; 2757 2758 //---------------------------------------------------------------------- 2759 // INCL scopes 2760 //---------------------------------------------------------------------- 2761 case N_BINCL: 2762 // include file beginning: name,,NO_SECT,0,sum 2763 // We use the current number of symbols in the symbol table in lieu of 2764 // using nlist_idx in case we ever start trimming entries out 2765 N_INCL_indexes.push_back(sym_idx); 2766 type = eSymbolTypeScopeBegin; 2767 break; 2768 2769 case N_EINCL: 2770 // include file end: name,,NO_SECT,0,0 2771 // Set the size of the N_BINCL to the terminating index of this N_EINCL 2772 // so that we can always skip the entire symbol if we need to navigate 2773 // more quickly at the source level when parsing STABS 2774 if ( !N_INCL_indexes.empty() ) 2775 { 2776 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 2777 symbol_ptr->SetByteSize(sym_idx + 1); 2778 symbol_ptr->SetSizeIsSibling(true); 2779 N_INCL_indexes.pop_back(); 2780 } 2781 type = eSymbolTypeScopeEnd; 2782 break; 2783 2784 case N_SOL: 2785 // #included file name: name,,n_sect,0,address 2786 type = eSymbolTypeHeaderFile; 2787 2788 // We currently don't use the header files on darwin 2789 add_nlist = false; 2790 break; 2791 2792 case N_PARAMS: 2793 // compiler parameters: name,,NO_SECT,0,0 2794 type = eSymbolTypeCompiler; 2795 break; 2796 2797 case N_VERSION: 2798 // compiler version: name,,NO_SECT,0,0 2799 type = eSymbolTypeCompiler; 2800 break; 2801 2802 case N_OLEVEL: 2803 // compiler -O level: name,,NO_SECT,0,0 2804 type = eSymbolTypeCompiler; 2805 break; 2806 2807 case N_PSYM: 2808 // parameter: name,,NO_SECT,type,offset 2809 type = eSymbolTypeVariable; 2810 break; 2811 2812 case N_ENTRY: 2813 // alternate entry: name,,n_sect,linenumber,address 2814 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2815 type = eSymbolTypeLineEntry; 2816 break; 2817 2818 //---------------------------------------------------------------------- 2819 // Left and Right Braces 2820 //---------------------------------------------------------------------- 2821 case N_LBRAC: 2822 // left bracket: 0,,NO_SECT,nesting level,address 2823 // We use the current number of symbols in the symbol table in lieu of 2824 // using nlist_idx in case we ever start trimming entries out 2825 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2826 N_BRAC_indexes.push_back(sym_idx); 2827 type = eSymbolTypeScopeBegin; 2828 break; 2829 2830 case N_RBRAC: 2831 // right bracket: 0,,NO_SECT,nesting level,address 2832 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 2833 // so that we can always skip the entire symbol if we need to navigate 2834 // more quickly at the source level when parsing STABS 2835 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2836 if ( !N_BRAC_indexes.empty() ) 2837 { 2838 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 2839 symbol_ptr->SetByteSize(sym_idx + 1); 2840 symbol_ptr->SetSizeIsSibling(true); 2841 N_BRAC_indexes.pop_back(); 2842 } 2843 type = eSymbolTypeScopeEnd; 2844 break; 2845 2846 case N_EXCL: 2847 // deleted include file: name,,NO_SECT,0,sum 2848 type = eSymbolTypeHeaderFile; 2849 break; 2850 2851 //---------------------------------------------------------------------- 2852 // COMM scopes 2853 //---------------------------------------------------------------------- 2854 case N_BCOMM: 2855 // begin common: name,,NO_SECT,0,0 2856 // We use the current number of symbols in the symbol table in lieu of 2857 // using nlist_idx in case we ever start trimming entries out 2858 type = eSymbolTypeScopeBegin; 2859 N_COMM_indexes.push_back(sym_idx); 2860 break; 2861 2862 case N_ECOML: 2863 // end common (local name): 0,,n_sect,0,address 2864 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2865 // Fall through 2866 2867 case N_ECOMM: 2868 // end common: name,,n_sect,0,0 2869 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 2870 // so that we can always skip the entire symbol if we need to navigate 2871 // more quickly at the source level when parsing STABS 2872 if ( !N_COMM_indexes.empty() ) 2873 { 2874 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 2875 symbol_ptr->SetByteSize(sym_idx + 1); 2876 symbol_ptr->SetSizeIsSibling(true); 2877 N_COMM_indexes.pop_back(); 2878 } 2879 type = eSymbolTypeScopeEnd; 2880 break; 2881 2882 case N_LENG: 2883 // second stab entry with length information 2884 type = eSymbolTypeAdditional; 2885 break; 2886 2887 default: break; 2888 } 2889 } 2890 else 2891 { 2892 //uint8_t n_pext = N_PEXT & nlist.n_type; 2893 uint8_t n_type = N_TYPE & nlist.n_type; 2894 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 2895 2896 switch (n_type) 2897 { 2898 case N_INDR: // Fall through 2899 case N_PBUD: // Fall through 2900 case N_UNDF: 2901 type = eSymbolTypeUndefined; 2902 break; 2903 2904 case N_ABS: 2905 type = eSymbolTypeAbsolute; 2906 break; 2907 2908 case N_SECT: 2909 { 2910 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 2911 2912 if (symbol_section == NULL) 2913 { 2914 // TODO: warn about this? 2915 add_nlist = false; 2916 break; 2917 } 2918 2919 if (TEXT_eh_frame_sectID == nlist.n_sect) 2920 { 2921 type = eSymbolTypeException; 2922 } 2923 else 2924 { 2925 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 2926 2927 switch (section_type) 2928 { 2929 case S_REGULAR: break; // regular section 2930 //case S_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section 2931 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 2932 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 2933 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 2934 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 2935 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 2936 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 2937 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 2938 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 2939 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 2940 //case S_COALESCED: type = eSymbolType; break; // section contains symbols that are to be coalesced 2941 //case S_GB_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 2942 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 2943 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 2944 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 2945 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 2946 default: break; 2947 } 2948 2949 if (type == eSymbolTypeInvalid) 2950 { 2951 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 2952 if (symbol_section->IsDescendant (text_section_sp.get())) 2953 { 2954 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 2955 S_ATTR_SELF_MODIFYING_CODE | 2956 S_ATTR_SOME_INSTRUCTIONS)) 2957 type = eSymbolTypeData; 2958 else 2959 type = eSymbolTypeCode; 2960 } 2961 else if (symbol_section->IsDescendant(data_section_sp.get())) 2962 { 2963 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 2964 { 2965 type = eSymbolTypeRuntime; 2966 2967 if (symbol_name && 2968 symbol_name[0] == '_' && 2969 symbol_name[1] == 'O' && 2970 symbol_name[2] == 'B') 2971 { 2972 llvm::StringRef symbol_name_ref(symbol_name); 2973 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 2974 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 2975 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 2976 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 2977 { 2978 symbol_name_non_abi_mangled = symbol_name + 1; 2979 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 2980 type = eSymbolTypeObjCClass; 2981 demangled_is_synthesized = true; 2982 } 2983 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 2984 { 2985 symbol_name_non_abi_mangled = symbol_name + 1; 2986 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 2987 type = eSymbolTypeObjCMetaClass; 2988 demangled_is_synthesized = true; 2989 } 2990 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 2991 { 2992 symbol_name_non_abi_mangled = symbol_name + 1; 2993 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 2994 type = eSymbolTypeObjCIVar; 2995 demangled_is_synthesized = true; 2996 } 2997 } 2998 } 2999 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 3000 { 3001 type = eSymbolTypeException; 3002 } 3003 else 3004 { 3005 type = eSymbolTypeData; 3006 } 3007 } 3008 else if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 3009 { 3010 type = eSymbolTypeTrampoline; 3011 } 3012 else if (symbol_section->IsDescendant(objc_section_sp.get())) 3013 { 3014 type = eSymbolTypeRuntime; 3015 if (symbol_name && symbol_name[0] == '.') 3016 { 3017 llvm::StringRef symbol_name_ref(symbol_name); 3018 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 3019 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 3020 { 3021 symbol_name_non_abi_mangled = symbol_name; 3022 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 3023 type = eSymbolTypeObjCClass; 3024 demangled_is_synthesized = true; 3025 } 3026 } 3027 } 3028 } 3029 } 3030 } 3031 break; 3032 } 3033 } 3034 3035 if (add_nlist) 3036 { 3037 uint64_t symbol_value = nlist.n_value; 3038 if (symbol_name_non_abi_mangled) 3039 { 3040 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 3041 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 3042 } 3043 else 3044 { 3045 bool symbol_name_is_mangled = false; 3046 3047 if (symbol_name && symbol_name[0] == '_') 3048 { 3049 symbol_name_is_mangled = symbol_name[1] == '_'; 3050 symbol_name++; // Skip the leading underscore 3051 } 3052 3053 if (symbol_name) 3054 { 3055 ConstString const_symbol_name(symbol_name); 3056 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 3057 if (is_gsym && is_debug) 3058 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 3059 } 3060 } 3061 if (symbol_section) 3062 { 3063 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3064 if (symbol_byte_size == 0 && function_starts_count > 0) 3065 { 3066 addr_t symbol_lookup_file_addr = nlist.n_value; 3067 // Do an exact address match for non-ARM addresses, else get the closest since 3068 // the symbol might be a thumb symbol which has an address with bit zero set 3069 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 3070 if (is_arm && func_start_entry) 3071 { 3072 // Verify that the function start address is the symbol address (ARM) 3073 // or the symbol address + 1 (thumb) 3074 if (func_start_entry->addr != symbol_lookup_file_addr && 3075 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 3076 { 3077 // Not the right entry, NULL it out... 3078 func_start_entry = NULL; 3079 } 3080 } 3081 if (func_start_entry) 3082 { 3083 func_start_entry->data = true; 3084 3085 addr_t symbol_file_addr = func_start_entry->addr; 3086 uint32_t symbol_flags = 0; 3087 if (is_arm) 3088 { 3089 if (symbol_file_addr & 1) 3090 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3091 symbol_file_addr &= 0xfffffffffffffffeull; 3092 } 3093 3094 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3095 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3096 if (next_func_start_entry) 3097 { 3098 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3099 // Be sure the clear the Thumb address bit when we calculate the size 3100 // from the current and next address 3101 if (is_arm) 3102 next_symbol_file_addr &= 0xfffffffffffffffeull; 3103 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3104 } 3105 else 3106 { 3107 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3108 } 3109 } 3110 } 3111 symbol_value -= section_file_addr; 3112 } 3113 3114 if (is_debug == false) 3115 { 3116 if (type == eSymbolTypeCode) 3117 { 3118 // See if we can find a N_FUN entry for any code symbols. 3119 // If we do find a match, and the name matches, then we 3120 // can merge the two into just the function symbol to avoid 3121 // duplicate entries in the symbol table 3122 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3123 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3124 if (range.first != range.second) 3125 { 3126 bool found_it = false; 3127 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3128 { 3129 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3130 { 3131 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3132 // We just need the flags from the linker symbol, so put these flags 3133 // into the N_FUN flags to avoid duplicate symbols in the symbol table 3134 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3135 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3136 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3137 sym[pos->second].SetType (eSymbolTypeResolver); 3138 sym[sym_idx].Clear(); 3139 found_it = true; 3140 break; 3141 } 3142 } 3143 if (found_it) 3144 continue; 3145 } 3146 else 3147 { 3148 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3149 type = eSymbolTypeResolver; 3150 } 3151 } 3152 else if (type == eSymbolTypeData) 3153 { 3154 // See if we can find a N_STSYM entry for any data symbols. 3155 // If we do find a match, and the name matches, then we 3156 // can merge the two into just the Static symbol to avoid 3157 // duplicate entries in the symbol table 3158 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3159 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 3160 if (range.first != range.second) 3161 { 3162 bool found_it = false; 3163 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3164 { 3165 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3166 { 3167 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3168 // We just need the flags from the linker symbol, so put these flags 3169 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3170 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3171 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3172 sym[sym_idx].Clear(); 3173 found_it = true; 3174 break; 3175 } 3176 } 3177 if (found_it) 3178 continue; 3179 } 3180 else 3181 { 3182 // Combine N_GSYM stab entries with the non stab symbol 3183 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 3184 if (pos != N_GSYM_name_to_sym_idx.end()) 3185 { 3186 const uint32_t GSYM_sym_idx = pos->second; 3187 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 3188 // Copy the address, because often the N_GSYM address has an invalid address of zero 3189 // when the global is a common symbol 3190 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 3191 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 3192 // We just need the flags from the linker symbol, so put these flags 3193 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3194 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3195 sym[sym_idx].Clear(); 3196 continue; 3197 } 3198 } 3199 } 3200 } 3201 3202 sym[sym_idx].SetID (nlist_idx); 3203 sym[sym_idx].SetType (type); 3204 sym[sym_idx].GetAddress().SetSection (symbol_section); 3205 sym[sym_idx].GetAddress().SetOffset (symbol_value); 3206 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3207 3208 if (symbol_byte_size > 0) 3209 sym[sym_idx].SetByteSize(symbol_byte_size); 3210 3211 if (demangled_is_synthesized) 3212 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3213 ++sym_idx; 3214 } 3215 else 3216 { 3217 sym[sym_idx].Clear(); 3218 } 3219 3220 } 3221 ///////////////////////////// 3222 } 3223 break; // No more entries to consider 3224 } 3225 } 3226 } 3227 } 3228 } 3229 } 3230 } 3231 3232 // Must reset this in case it was mutated above! 3233 nlist_data_offset = 0; 3234 #endif 3235 3236 if (nlist_data.GetByteSize() > 0) 3237 { 3238 3239 // If the sym array was not created while parsing the DSC unmapped 3240 // symbols, create it now. 3241 if (sym == NULL) 3242 { 3243 sym = symtab->Resize (symtab_load_command.nsyms + m_dysymtab.nindirectsyms); 3244 num_syms = symtab->GetNumSymbols(); 3245 } 3246 3247 if (unmapped_local_symbols_found) 3248 { 3249 assert(m_dysymtab.ilocalsym == 0); 3250 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3251 nlist_idx = m_dysymtab.nlocalsym; 3252 } 3253 else 3254 { 3255 nlist_idx = 0; 3256 } 3257 3258 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) 3259 { 3260 struct nlist_64 nlist; 3261 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, nlist_byte_size)) 3262 break; 3263 3264 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3265 nlist.n_type = nlist_data.GetU8_unchecked (&nlist_data_offset); 3266 nlist.n_sect = nlist_data.GetU8_unchecked (&nlist_data_offset); 3267 nlist.n_desc = nlist_data.GetU16_unchecked (&nlist_data_offset); 3268 nlist.n_value = nlist_data.GetAddress_unchecked (&nlist_data_offset); 3269 3270 SymbolType type = eSymbolTypeInvalid; 3271 const char *symbol_name = NULL; 3272 3273 if (have_strtab_data) 3274 { 3275 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3276 3277 if (symbol_name == NULL) 3278 { 3279 // No symbol should be NULL, even the symbols with no 3280 // string values should have an offset zero which points 3281 // to an empty C-string 3282 Host::SystemLog (Host::eSystemLogError, 3283 "error: symbol[%u] has invalid string table offset 0x%x in %s, ignoring symbol\n", 3284 nlist_idx, 3285 nlist.n_strx, 3286 module_sp->GetFileSpec().GetPath().c_str()); 3287 continue; 3288 } 3289 if (symbol_name[0] == '\0') 3290 symbol_name = NULL; 3291 } 3292 else 3293 { 3294 const addr_t str_addr = strtab_addr + nlist.n_strx; 3295 Error str_error; 3296 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, str_error)) 3297 symbol_name = memory_symbol_name.c_str(); 3298 } 3299 const char *symbol_name_non_abi_mangled = NULL; 3300 3301 SectionSP symbol_section; 3302 lldb::addr_t symbol_byte_size = 0; 3303 bool add_nlist = true; 3304 bool is_gsym = false; 3305 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3306 bool demangled_is_synthesized = false; 3307 3308 assert (sym_idx < num_syms); 3309 3310 sym[sym_idx].SetDebug (is_debug); 3311 3312 if (is_debug) 3313 { 3314 switch (nlist.n_type) 3315 { 3316 case N_GSYM: 3317 // global symbol: name,,NO_SECT,type,0 3318 // Sometimes the N_GSYM value contains the address. 3319 3320 // FIXME: In the .o files, we have a GSYM and a debug symbol for all the ObjC data. They 3321 // have the same address, but we want to ensure that we always find only the real symbol, 3322 // 'cause we don't currently correctly attribute the GSYM one to the ObjCClass/Ivar/MetaClass 3323 // symbol type. This is a temporary hack to make sure the ObjectiveC symbols get treated 3324 // correctly. To do this right, we should coalesce all the GSYM & global symbols that have the 3325 // same address. 3326 3327 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O' 3328 && (strncmp (symbol_name, "_OBJC_IVAR_$_", strlen ("_OBJC_IVAR_$_")) == 0 3329 || strncmp (symbol_name, "_OBJC_CLASS_$_", strlen ("_OBJC_CLASS_$_")) == 0 3330 || strncmp (symbol_name, "_OBJC_METACLASS_$_", strlen ("_OBJC_METACLASS_$_")) == 0)) 3331 add_nlist = false; 3332 else 3333 { 3334 is_gsym = true; 3335 sym[sym_idx].SetExternal(true); 3336 if (nlist.n_value != 0) 3337 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3338 type = eSymbolTypeData; 3339 } 3340 break; 3341 3342 case N_FNAME: 3343 // procedure name (f77 kludge): name,,NO_SECT,0,0 3344 type = eSymbolTypeCompiler; 3345 break; 3346 3347 case N_FUN: 3348 // procedure: name,,n_sect,linenumber,address 3349 if (symbol_name) 3350 { 3351 type = eSymbolTypeCode; 3352 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3353 3354 N_FUN_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3355 // We use the current number of symbols in the symbol table in lieu of 3356 // using nlist_idx in case we ever start trimming entries out 3357 N_FUN_indexes.push_back(sym_idx); 3358 } 3359 else 3360 { 3361 type = eSymbolTypeCompiler; 3362 3363 if ( !N_FUN_indexes.empty() ) 3364 { 3365 // Copy the size of the function into the original STAB entry so we don't have 3366 // to hunt for it later 3367 symtab->SymbolAtIndex(N_FUN_indexes.back())->SetByteSize(nlist.n_value); 3368 N_FUN_indexes.pop_back(); 3369 // We don't really need the end function STAB as it contains the size which 3370 // we already placed with the original symbol, so don't add it if we want a 3371 // minimal symbol table 3372 add_nlist = false; 3373 } 3374 } 3375 break; 3376 3377 case N_STSYM: 3378 // static symbol: name,,n_sect,type,address 3379 N_STSYM_addr_to_sym_idx.insert(std::make_pair(nlist.n_value, sym_idx)); 3380 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3381 type = eSymbolTypeData; 3382 break; 3383 3384 case N_LCSYM: 3385 // .lcomm symbol: name,,n_sect,type,address 3386 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3387 type = eSymbolTypeCommonBlock; 3388 break; 3389 3390 case N_BNSYM: 3391 // We use the current number of symbols in the symbol table in lieu of 3392 // using nlist_idx in case we ever start trimming entries out 3393 // Skip these if we want minimal symbol tables 3394 add_nlist = false; 3395 break; 3396 3397 case N_ENSYM: 3398 // Set the size of the N_BNSYM to the terminating index of this N_ENSYM 3399 // so that we can always skip the entire symbol if we need to navigate 3400 // more quickly at the source level when parsing STABS 3401 // Skip these if we want minimal symbol tables 3402 add_nlist = false; 3403 break; 3404 3405 3406 case N_OPT: 3407 // emitted with gcc2_compiled and in gcc source 3408 type = eSymbolTypeCompiler; 3409 break; 3410 3411 case N_RSYM: 3412 // register sym: name,,NO_SECT,type,register 3413 type = eSymbolTypeVariable; 3414 break; 3415 3416 case N_SLINE: 3417 // src line: 0,,n_sect,linenumber,address 3418 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3419 type = eSymbolTypeLineEntry; 3420 break; 3421 3422 case N_SSYM: 3423 // structure elt: name,,NO_SECT,type,struct_offset 3424 type = eSymbolTypeVariableType; 3425 break; 3426 3427 case N_SO: 3428 // source file name 3429 type = eSymbolTypeSourceFile; 3430 if (symbol_name == NULL) 3431 { 3432 add_nlist = false; 3433 if (N_SO_index != UINT32_MAX) 3434 { 3435 // Set the size of the N_SO to the terminating index of this N_SO 3436 // so that we can always skip the entire N_SO if we need to navigate 3437 // more quickly at the source level when parsing STABS 3438 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3439 symbol_ptr->SetByteSize(sym_idx); 3440 symbol_ptr->SetSizeIsSibling(true); 3441 } 3442 N_NSYM_indexes.clear(); 3443 N_INCL_indexes.clear(); 3444 N_BRAC_indexes.clear(); 3445 N_COMM_indexes.clear(); 3446 N_FUN_indexes.clear(); 3447 N_SO_index = UINT32_MAX; 3448 } 3449 else 3450 { 3451 // We use the current number of symbols in the symbol table in lieu of 3452 // using nlist_idx in case we ever start trimming entries out 3453 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3454 if (N_SO_has_full_path) 3455 { 3456 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3457 { 3458 // We have two consecutive N_SO entries where the first contains a directory 3459 // and the second contains a full path. 3460 sym[sym_idx - 1].GetMangled().SetValue(ConstString(symbol_name), false); 3461 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3462 add_nlist = false; 3463 } 3464 else 3465 { 3466 // This is the first entry in a N_SO that contains a directory or 3467 // a full path to the source file 3468 N_SO_index = sym_idx; 3469 } 3470 } 3471 else if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) 3472 { 3473 // This is usually the second N_SO entry that contains just the filename, 3474 // so here we combine it with the first one if we are minimizing the symbol table 3475 const char *so_path = sym[sym_idx - 1].GetMangled().GetDemangledName().AsCString(); 3476 if (so_path && so_path[0]) 3477 { 3478 std::string full_so_path (so_path); 3479 const size_t double_slash_pos = full_so_path.find("//"); 3480 if (double_slash_pos != std::string::npos) 3481 { 3482 // The linker has been generating bad N_SO entries with doubled up paths 3483 // in the format "%s%s" where the first string in the DW_AT_comp_dir, 3484 // and the second is the directory for the source file so you end up with 3485 // a path that looks like "/tmp/src//tmp/src/" 3486 FileSpec so_dir(so_path, false); 3487 if (!so_dir.Exists()) 3488 { 3489 so_dir.SetFile(&full_so_path[double_slash_pos + 1], false); 3490 if (so_dir.Exists()) 3491 { 3492 // Trim off the incorrect path 3493 full_so_path.erase(0, double_slash_pos + 1); 3494 } 3495 } 3496 } 3497 if (*full_so_path.rbegin() != '/') 3498 full_so_path += '/'; 3499 full_so_path += symbol_name; 3500 sym[sym_idx - 1].GetMangled().SetValue(ConstString(full_so_path.c_str()), false); 3501 add_nlist = false; 3502 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3503 } 3504 } 3505 else 3506 { 3507 // This could be a relative path to a N_SO 3508 N_SO_index = sym_idx; 3509 } 3510 } 3511 3512 break; 3513 3514 case N_OSO: 3515 // object file name: name,,0,0,st_mtime 3516 type = eSymbolTypeObjectFile; 3517 break; 3518 3519 case N_LSYM: 3520 // local sym: name,,NO_SECT,type,offset 3521 type = eSymbolTypeLocal; 3522 break; 3523 3524 //---------------------------------------------------------------------- 3525 // INCL scopes 3526 //---------------------------------------------------------------------- 3527 case N_BINCL: 3528 // include file beginning: name,,NO_SECT,0,sum 3529 // We use the current number of symbols in the symbol table in lieu of 3530 // using nlist_idx in case we ever start trimming entries out 3531 N_INCL_indexes.push_back(sym_idx); 3532 type = eSymbolTypeScopeBegin; 3533 break; 3534 3535 case N_EINCL: 3536 // include file end: name,,NO_SECT,0,0 3537 // Set the size of the N_BINCL to the terminating index of this N_EINCL 3538 // so that we can always skip the entire symbol if we need to navigate 3539 // more quickly at the source level when parsing STABS 3540 if ( !N_INCL_indexes.empty() ) 3541 { 3542 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 3543 symbol_ptr->SetByteSize(sym_idx + 1); 3544 symbol_ptr->SetSizeIsSibling(true); 3545 N_INCL_indexes.pop_back(); 3546 } 3547 type = eSymbolTypeScopeEnd; 3548 break; 3549 3550 case N_SOL: 3551 // #included file name: name,,n_sect,0,address 3552 type = eSymbolTypeHeaderFile; 3553 3554 // We currently don't use the header files on darwin 3555 add_nlist = false; 3556 break; 3557 3558 case N_PARAMS: 3559 // compiler parameters: name,,NO_SECT,0,0 3560 type = eSymbolTypeCompiler; 3561 break; 3562 3563 case N_VERSION: 3564 // compiler version: name,,NO_SECT,0,0 3565 type = eSymbolTypeCompiler; 3566 break; 3567 3568 case N_OLEVEL: 3569 // compiler -O level: name,,NO_SECT,0,0 3570 type = eSymbolTypeCompiler; 3571 break; 3572 3573 case N_PSYM: 3574 // parameter: name,,NO_SECT,type,offset 3575 type = eSymbolTypeVariable; 3576 break; 3577 3578 case N_ENTRY: 3579 // alternate entry: name,,n_sect,linenumber,address 3580 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3581 type = eSymbolTypeLineEntry; 3582 break; 3583 3584 //---------------------------------------------------------------------- 3585 // Left and Right Braces 3586 //---------------------------------------------------------------------- 3587 case N_LBRAC: 3588 // left bracket: 0,,NO_SECT,nesting level,address 3589 // We use the current number of symbols in the symbol table in lieu of 3590 // using nlist_idx in case we ever start trimming entries out 3591 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3592 N_BRAC_indexes.push_back(sym_idx); 3593 type = eSymbolTypeScopeBegin; 3594 break; 3595 3596 case N_RBRAC: 3597 // right bracket: 0,,NO_SECT,nesting level,address 3598 // Set the size of the N_LBRAC to the terminating index of this N_RBRAC 3599 // so that we can always skip the entire symbol if we need to navigate 3600 // more quickly at the source level when parsing STABS 3601 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3602 if ( !N_BRAC_indexes.empty() ) 3603 { 3604 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3605 symbol_ptr->SetByteSize(sym_idx + 1); 3606 symbol_ptr->SetSizeIsSibling(true); 3607 N_BRAC_indexes.pop_back(); 3608 } 3609 type = eSymbolTypeScopeEnd; 3610 break; 3611 3612 case N_EXCL: 3613 // deleted include file: name,,NO_SECT,0,sum 3614 type = eSymbolTypeHeaderFile; 3615 break; 3616 3617 //---------------------------------------------------------------------- 3618 // COMM scopes 3619 //---------------------------------------------------------------------- 3620 case N_BCOMM: 3621 // begin common: name,,NO_SECT,0,0 3622 // We use the current number of symbols in the symbol table in lieu of 3623 // using nlist_idx in case we ever start trimming entries out 3624 type = eSymbolTypeScopeBegin; 3625 N_COMM_indexes.push_back(sym_idx); 3626 break; 3627 3628 case N_ECOML: 3629 // end common (local name): 0,,n_sect,0,address 3630 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3631 // Fall through 3632 3633 case N_ECOMM: 3634 // end common: name,,n_sect,0,0 3635 // Set the size of the N_BCOMM to the terminating index of this N_ECOMM/N_ECOML 3636 // so that we can always skip the entire symbol if we need to navigate 3637 // more quickly at the source level when parsing STABS 3638 if ( !N_COMM_indexes.empty() ) 3639 { 3640 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 3641 symbol_ptr->SetByteSize(sym_idx + 1); 3642 symbol_ptr->SetSizeIsSibling(true); 3643 N_COMM_indexes.pop_back(); 3644 } 3645 type = eSymbolTypeScopeEnd; 3646 break; 3647 3648 case N_LENG: 3649 // second stab entry with length information 3650 type = eSymbolTypeAdditional; 3651 break; 3652 3653 default: break; 3654 } 3655 } 3656 else 3657 { 3658 //uint8_t n_pext = N_PEXT & nlist.n_type; 3659 uint8_t n_type = N_TYPE & nlist.n_type; 3660 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3661 3662 switch (n_type) 3663 { 3664 case N_INDR:// Fall through 3665 case N_PBUD:// Fall through 3666 case N_UNDF: 3667 type = eSymbolTypeUndefined; 3668 break; 3669 3670 case N_ABS: 3671 type = eSymbolTypeAbsolute; 3672 break; 3673 3674 case N_SECT: 3675 { 3676 symbol_section = section_info.GetSection (nlist.n_sect, nlist.n_value); 3677 3678 if (!symbol_section) 3679 { 3680 // TODO: warn about this? 3681 add_nlist = false; 3682 break; 3683 } 3684 3685 if (TEXT_eh_frame_sectID == nlist.n_sect) 3686 { 3687 type = eSymbolTypeException; 3688 } 3689 else 3690 { 3691 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 3692 3693 switch (section_type) 3694 { 3695 case S_REGULAR: break; // regular section 3696 //case S_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section 3697 case S_CSTRING_LITERALS: type = eSymbolTypeData; break; // section with only literal C strings 3698 case S_4BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 4 byte literals 3699 case S_8BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 8 byte literals 3700 case S_LITERAL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only pointers to literals 3701 case S_NON_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only non-lazy symbol pointers 3702 case S_LAZY_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; // section with only lazy symbol pointers 3703 case S_SYMBOL_STUBS: type = eSymbolTypeTrampoline; break; // section with only symbol stubs, byte size of stub in the reserved2 field 3704 case S_MOD_INIT_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for initialization 3705 case S_MOD_TERM_FUNC_POINTERS: type = eSymbolTypeCode; break; // section with only function pointers for termination 3706 //case S_COALESCED: type = eSymbolType; break; // section contains symbols that are to be coalesced 3707 //case S_GB_ZEROFILL: type = eSymbolTypeData; break; // zero fill on demand section (that can be larger than 4 gigabytes) 3708 case S_INTERPOSING: type = eSymbolTypeTrampoline; break; // section with only pairs of function pointers for interposing 3709 case S_16BYTE_LITERALS: type = eSymbolTypeData; break; // section with only 16 byte literals 3710 case S_DTRACE_DOF: type = eSymbolTypeInstrumentation; break; 3711 case S_LAZY_DYLIB_SYMBOL_POINTERS: type = eSymbolTypeTrampoline; break; 3712 default: break; 3713 } 3714 3715 if (type == eSymbolTypeInvalid) 3716 { 3717 const char *symbol_sect_name = symbol_section->GetName().AsCString(); 3718 if (symbol_section->IsDescendant (text_section_sp.get())) 3719 { 3720 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 3721 S_ATTR_SELF_MODIFYING_CODE | 3722 S_ATTR_SOME_INSTRUCTIONS)) 3723 type = eSymbolTypeData; 3724 else 3725 type = eSymbolTypeCode; 3726 } 3727 else 3728 if (symbol_section->IsDescendant(data_section_sp.get())) 3729 { 3730 if (symbol_sect_name && ::strstr (symbol_sect_name, "__objc") == symbol_sect_name) 3731 { 3732 type = eSymbolTypeRuntime; 3733 3734 if (symbol_name && 3735 symbol_name[0] == '_' && 3736 symbol_name[1] == 'O' && 3737 symbol_name[2] == 'B') 3738 { 3739 llvm::StringRef symbol_name_ref(symbol_name); 3740 static const llvm::StringRef g_objc_v2_prefix_class ("_OBJC_CLASS_$_"); 3741 static const llvm::StringRef g_objc_v2_prefix_metaclass ("_OBJC_METACLASS_$_"); 3742 static const llvm::StringRef g_objc_v2_prefix_ivar ("_OBJC_IVAR_$_"); 3743 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) 3744 { 3745 symbol_name_non_abi_mangled = symbol_name + 1; 3746 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3747 type = eSymbolTypeObjCClass; 3748 demangled_is_synthesized = true; 3749 } 3750 else if (symbol_name_ref.startswith(g_objc_v2_prefix_metaclass)) 3751 { 3752 symbol_name_non_abi_mangled = symbol_name + 1; 3753 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3754 type = eSymbolTypeObjCMetaClass; 3755 demangled_is_synthesized = true; 3756 } 3757 else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) 3758 { 3759 symbol_name_non_abi_mangled = symbol_name + 1; 3760 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3761 type = eSymbolTypeObjCIVar; 3762 demangled_is_synthesized = true; 3763 } 3764 } 3765 } 3766 else 3767 if (symbol_sect_name && ::strstr (symbol_sect_name, "__gcc_except_tab") == symbol_sect_name) 3768 { 3769 type = eSymbolTypeException; 3770 } 3771 else 3772 { 3773 type = eSymbolTypeData; 3774 } 3775 } 3776 else 3777 if (symbol_sect_name && ::strstr (symbol_sect_name, "__IMPORT") == symbol_sect_name) 3778 { 3779 type = eSymbolTypeTrampoline; 3780 } 3781 else 3782 if (symbol_section->IsDescendant(objc_section_sp.get())) 3783 { 3784 type = eSymbolTypeRuntime; 3785 if (symbol_name && symbol_name[0] == '.') 3786 { 3787 llvm::StringRef symbol_name_ref(symbol_name); 3788 static const llvm::StringRef g_objc_v1_prefix_class (".objc_class_name_"); 3789 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) 3790 { 3791 symbol_name_non_abi_mangled = symbol_name; 3792 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 3793 type = eSymbolTypeObjCClass; 3794 demangled_is_synthesized = true; 3795 } 3796 } 3797 } 3798 } 3799 } 3800 } 3801 break; 3802 } 3803 } 3804 3805 if (add_nlist) 3806 { 3807 uint64_t symbol_value = nlist.n_value; 3808 3809 if (symbol_name_non_abi_mangled) 3810 { 3811 sym[sym_idx].GetMangled().SetMangledName (ConstString(symbol_name_non_abi_mangled)); 3812 sym[sym_idx].GetMangled().SetDemangledName (ConstString(symbol_name)); 3813 } 3814 else 3815 { 3816 bool symbol_name_is_mangled = false; 3817 3818 if (symbol_name && symbol_name[0] == '_') 3819 { 3820 symbol_name_is_mangled = symbol_name[1] == '_'; 3821 symbol_name++; // Skip the leading underscore 3822 } 3823 3824 if (symbol_name) 3825 { 3826 ConstString const_symbol_name(symbol_name); 3827 sym[sym_idx].GetMangled().SetValue(const_symbol_name, symbol_name_is_mangled); 3828 if (is_gsym && is_debug) 3829 { 3830 N_GSYM_name_to_sym_idx[sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()] = sym_idx; 3831 } 3832 } 3833 } 3834 if (symbol_section) 3835 { 3836 const addr_t section_file_addr = symbol_section->GetFileAddress(); 3837 if (symbol_byte_size == 0 && function_starts_count > 0) 3838 { 3839 addr_t symbol_lookup_file_addr = nlist.n_value; 3840 // Do an exact address match for non-ARM addresses, else get the closest since 3841 // the symbol might be a thumb symbol which has an address with bit zero set 3842 FunctionStarts::Entry *func_start_entry = function_starts.FindEntry (symbol_lookup_file_addr, !is_arm); 3843 if (is_arm && func_start_entry) 3844 { 3845 // Verify that the function start address is the symbol address (ARM) 3846 // or the symbol address + 1 (thumb) 3847 if (func_start_entry->addr != symbol_lookup_file_addr && 3848 func_start_entry->addr != (symbol_lookup_file_addr + 1)) 3849 { 3850 // Not the right entry, NULL it out... 3851 func_start_entry = NULL; 3852 } 3853 } 3854 if (func_start_entry) 3855 { 3856 func_start_entry->data = true; 3857 3858 addr_t symbol_file_addr = func_start_entry->addr; 3859 if (is_arm) 3860 symbol_file_addr &= 0xfffffffffffffffeull; 3861 3862 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 3863 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 3864 if (next_func_start_entry) 3865 { 3866 addr_t next_symbol_file_addr = next_func_start_entry->addr; 3867 // Be sure the clear the Thumb address bit when we calculate the size 3868 // from the current and next address 3869 if (is_arm) 3870 next_symbol_file_addr &= 0xfffffffffffffffeull; 3871 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 3872 } 3873 else 3874 { 3875 symbol_byte_size = section_end_file_addr - symbol_file_addr; 3876 } 3877 } 3878 } 3879 symbol_value -= section_file_addr; 3880 } 3881 3882 if (is_debug == false) 3883 { 3884 if (type == eSymbolTypeCode) 3885 { 3886 // See if we can find a N_FUN entry for any code symbols. 3887 // If we do find a match, and the name matches, then we 3888 // can merge the two into just the function symbol to avoid 3889 // duplicate entries in the symbol table 3890 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3891 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 3892 if (range.first != range.second) 3893 { 3894 bool found_it = false; 3895 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3896 { 3897 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3898 { 3899 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3900 // We just need the flags from the linker symbol, so put these flags 3901 // into the N_FUN flags to avoid duplicate symbols in the symbol table 3902 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3903 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3904 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3905 sym[pos->second].SetType (eSymbolTypeResolver); 3906 sym[sym_idx].Clear(); 3907 found_it = true; 3908 break; 3909 } 3910 } 3911 if (found_it) 3912 continue; 3913 } 3914 else 3915 { 3916 if (resolver_addresses.find(nlist.n_value) != resolver_addresses.end()) 3917 type = eSymbolTypeResolver; 3918 } 3919 } 3920 else if (type == eSymbolTypeData) 3921 { 3922 // See if we can find a N_STSYM entry for any data symbols. 3923 // If we do find a match, and the name matches, then we 3924 // can merge the two into just the Static symbol to avoid 3925 // duplicate entries in the symbol table 3926 std::pair<ValueToSymbolIndexMap::const_iterator, ValueToSymbolIndexMap::const_iterator> range; 3927 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 3928 if (range.first != range.second) 3929 { 3930 bool found_it = false; 3931 for (ValueToSymbolIndexMap::const_iterator pos = range.first; pos != range.second; ++pos) 3932 { 3933 if (sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled) == sym[pos->second].GetMangled().GetName(Mangled::ePreferMangled)) 3934 { 3935 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 3936 // We just need the flags from the linker symbol, so put these flags 3937 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3938 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 3939 sym[pos->second].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3940 sym[sym_idx].Clear(); 3941 found_it = true; 3942 break; 3943 } 3944 } 3945 if (found_it) 3946 continue; 3947 } 3948 else 3949 { 3950 // Combine N_GSYM stab entries with the non stab symbol 3951 ConstNameToSymbolIndexMap::const_iterator pos = N_GSYM_name_to_sym_idx.find(sym[sym_idx].GetMangled().GetName(Mangled::ePreferMangled).GetCString()); 3952 if (pos != N_GSYM_name_to_sym_idx.end()) 3953 { 3954 const uint32_t GSYM_sym_idx = pos->second; 3955 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 3956 // Copy the address, because often the N_GSYM address has an invalid address of zero 3957 // when the global is a common symbol 3958 sym[GSYM_sym_idx].GetAddress().SetSection (symbol_section); 3959 sym[GSYM_sym_idx].GetAddress().SetOffset (symbol_value); 3960 // We just need the flags from the linker symbol, so put these flags 3961 // into the N_STSYM flags to avoid duplicate symbols in the symbol table 3962 sym[GSYM_sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3963 sym[sym_idx].Clear(); 3964 continue; 3965 } 3966 } 3967 } 3968 } 3969 3970 sym[sym_idx].SetID (nlist_idx); 3971 sym[sym_idx].SetType (type); 3972 sym[sym_idx].GetAddress().SetSection (symbol_section); 3973 sym[sym_idx].GetAddress().SetOffset (symbol_value); 3974 sym[sym_idx].SetFlags (nlist.n_type << 16 | nlist.n_desc); 3975 3976 if (symbol_byte_size > 0) 3977 sym[sym_idx].SetByteSize(symbol_byte_size); 3978 3979 if (demangled_is_synthesized) 3980 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3981 3982 ++sym_idx; 3983 } 3984 else 3985 { 3986 sym[sym_idx].Clear(); 3987 } 3988 } 3989 } 3990 3991 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 3992 3993 if (function_starts_count > 0) 3994 { 3995 char synthetic_function_symbol[PATH_MAX]; 3996 uint32_t num_synthetic_function_symbols = 0; 3997 for (i=0; i<function_starts_count; ++i) 3998 { 3999 if (function_starts.GetEntryRef (i).data == false) 4000 ++num_synthetic_function_symbols; 4001 } 4002 4003 if (num_synthetic_function_symbols > 0) 4004 { 4005 if (num_syms < sym_idx + num_synthetic_function_symbols) 4006 { 4007 num_syms = sym_idx + num_synthetic_function_symbols; 4008 sym = symtab->Resize (num_syms); 4009 } 4010 uint32_t synthetic_function_symbol_idx = 0; 4011 for (i=0; i<function_starts_count; ++i) 4012 { 4013 const FunctionStarts::Entry *func_start_entry = function_starts.GetEntryAtIndex (i); 4014 if (func_start_entry->data == false) 4015 { 4016 addr_t symbol_file_addr = func_start_entry->addr; 4017 uint32_t symbol_flags = 0; 4018 if (is_arm) 4019 { 4020 if (symbol_file_addr & 1) 4021 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4022 symbol_file_addr &= 0xfffffffffffffffeull; 4023 } 4024 Address symbol_addr; 4025 if (module_sp->ResolveFileAddress (symbol_file_addr, symbol_addr)) 4026 { 4027 SectionSP symbol_section (symbol_addr.GetSection()); 4028 uint32_t symbol_byte_size = 0; 4029 if (symbol_section) 4030 { 4031 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4032 const FunctionStarts::Entry *next_func_start_entry = function_starts.FindNextEntry (func_start_entry); 4033 const addr_t section_end_file_addr = section_file_addr + symbol_section->GetByteSize(); 4034 if (next_func_start_entry) 4035 { 4036 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4037 if (is_arm) 4038 next_symbol_file_addr &= 0xfffffffffffffffeull; 4039 symbol_byte_size = std::min<lldb::addr_t>(next_symbol_file_addr - symbol_file_addr, section_end_file_addr - symbol_file_addr); 4040 } 4041 else 4042 { 4043 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4044 } 4045 snprintf (synthetic_function_symbol, 4046 sizeof(synthetic_function_symbol), 4047 "___lldb_unnamed_function%u$$%s", 4048 ++synthetic_function_symbol_idx, 4049 module_sp->GetFileSpec().GetFilename().GetCString()); 4050 sym[sym_idx].SetID (synthetic_sym_id++); 4051 sym[sym_idx].GetMangled().SetDemangledName(ConstString(synthetic_function_symbol)); 4052 sym[sym_idx].SetType (eSymbolTypeCode); 4053 sym[sym_idx].SetIsSynthetic (true); 4054 sym[sym_idx].GetAddress() = symbol_addr; 4055 if (symbol_flags) 4056 sym[sym_idx].SetFlags (symbol_flags); 4057 if (symbol_byte_size) 4058 sym[sym_idx].SetByteSize (symbol_byte_size); 4059 ++sym_idx; 4060 } 4061 } 4062 } 4063 } 4064 } 4065 } 4066 4067 // Trim our symbols down to just what we ended up with after 4068 // removing any symbols. 4069 if (sym_idx < num_syms) 4070 { 4071 num_syms = sym_idx; 4072 sym = symtab->Resize (num_syms); 4073 } 4074 4075 // Now synthesize indirect symbols 4076 if (m_dysymtab.nindirectsyms != 0) 4077 { 4078 if (indirect_symbol_index_data.GetByteSize()) 4079 { 4080 NListIndexToSymbolIndexMap::const_iterator end_index_pos = m_nlist_idx_to_sym_idx.end(); 4081 4082 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); ++sect_idx) 4083 { 4084 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == S_SYMBOL_STUBS) 4085 { 4086 uint32_t symbol_stub_byte_size = m_mach_sections[sect_idx].reserved2; 4087 if (symbol_stub_byte_size == 0) 4088 continue; 4089 4090 const uint32_t num_symbol_stubs = m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4091 4092 if (num_symbol_stubs == 0) 4093 continue; 4094 4095 const uint32_t symbol_stub_index_offset = m_mach_sections[sect_idx].reserved1; 4096 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; ++stub_idx) 4097 { 4098 const uint32_t symbol_stub_index = symbol_stub_index_offset + stub_idx; 4099 const lldb::addr_t symbol_stub_addr = m_mach_sections[sect_idx].addr + (stub_idx * symbol_stub_byte_size); 4100 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4101 if (indirect_symbol_index_data.ValidOffsetForDataOfSize(symbol_stub_offset, 4)) 4102 { 4103 const uint32_t stub_sym_id = indirect_symbol_index_data.GetU32 (&symbol_stub_offset); 4104 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4105 continue; 4106 4107 NListIndexToSymbolIndexMap::const_iterator index_pos = m_nlist_idx_to_sym_idx.find (stub_sym_id); 4108 Symbol *stub_symbol = NULL; 4109 if (index_pos != end_index_pos) 4110 { 4111 // We have a remapping from the original nlist index to 4112 // a current symbol index, so just look this up by index 4113 stub_symbol = symtab->SymbolAtIndex (index_pos->second); 4114 } 4115 else 4116 { 4117 // We need to lookup a symbol using the original nlist 4118 // symbol index since this index is coming from the 4119 // S_SYMBOL_STUBS 4120 stub_symbol = symtab->FindSymbolByID (stub_sym_id); 4121 } 4122 4123 if (stub_symbol) 4124 { 4125 Address so_addr(symbol_stub_addr, section_list); 4126 4127 if (stub_symbol->GetType() == eSymbolTypeUndefined) 4128 { 4129 // Change the external symbol into a trampoline that makes sense 4130 // These symbols were N_UNDF N_EXT, and are useless to us, so we 4131 // can re-use them so we don't have to make up a synthetic symbol 4132 // for no good reason. 4133 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 4134 stub_symbol->SetType (eSymbolTypeTrampoline); 4135 else 4136 stub_symbol->SetType (eSymbolTypeResolver); 4137 stub_symbol->SetExternal (false); 4138 stub_symbol->GetAddress() = so_addr; 4139 stub_symbol->SetByteSize (symbol_stub_byte_size); 4140 } 4141 else 4142 { 4143 // Make a synthetic symbol to describe the trampoline stub 4144 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4145 if (sym_idx >= num_syms) 4146 { 4147 sym = symtab->Resize (++num_syms); 4148 stub_symbol = NULL; // this pointer no longer valid 4149 } 4150 sym[sym_idx].SetID (synthetic_sym_id++); 4151 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4152 if (resolver_addresses.find(symbol_stub_addr) == resolver_addresses.end()) 4153 sym[sym_idx].SetType (eSymbolTypeTrampoline); 4154 else 4155 sym[sym_idx].SetType (eSymbolTypeResolver); 4156 sym[sym_idx].SetIsSynthetic (true); 4157 sym[sym_idx].GetAddress() = so_addr; 4158 sym[sym_idx].SetByteSize (symbol_stub_byte_size); 4159 ++sym_idx; 4160 } 4161 } 4162 else 4163 { 4164 if (log) 4165 log->Warning ("symbol stub referencing symbol table symbol %u that isn't in our minimal symbol table, fix this!!!", stub_sym_id); 4166 } 4167 } 4168 } 4169 } 4170 } 4171 } 4172 } 4173 4174 4175 if (!trie_entries.empty()) 4176 { 4177 for (const auto &e : trie_entries) 4178 { 4179 if (e.entry.import_name) 4180 { 4181 // Make a synthetic symbol to describe re-exported symbol. 4182 if (sym_idx >= num_syms) 4183 sym = symtab->Resize (++num_syms); 4184 sym[sym_idx].SetID (synthetic_sym_id++); 4185 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4186 sym[sym_idx].SetType (eSymbolTypeReExported); 4187 sym[sym_idx].SetIsSynthetic (true); 4188 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4189 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) 4190 { 4191 sym[sym_idx].SetReExportedSymbolSharedLibrary(dylib_files.GetFileSpecAtIndex(e.entry.other-1)); 4192 } 4193 ++sym_idx; 4194 } 4195 } 4196 } 4197 4198 4199 4200 // StreamFile s(stdout, false); 4201 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4202 // symtab->Dump(&s, NULL, eSortOrderNone); 4203 // Set symbol byte sizes correctly since mach-o nlist entries don't have sizes 4204 symtab->CalculateSymbolSizes(); 4205 4206 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4207 // symtab->Dump(&s, NULL, eSortOrderNone); 4208 4209 return symtab->GetNumSymbols(); 4210 } 4211 return 0; 4212 } 4213 4214 4215 void 4216 ObjectFileMachO::Dump (Stream *s) 4217 { 4218 ModuleSP module_sp(GetModule()); 4219 if (module_sp) 4220 { 4221 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4222 s->Printf("%p: ", static_cast<void*>(this)); 4223 s->Indent(); 4224 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4225 s->PutCString("ObjectFileMachO64"); 4226 else 4227 s->PutCString("ObjectFileMachO32"); 4228 4229 ArchSpec header_arch; 4230 GetArchitecture(header_arch); 4231 4232 *s << ", file = '" << m_file << "', arch = " << header_arch.GetArchitectureName() << "\n"; 4233 4234 SectionList *sections = GetSectionList(); 4235 if (sections) 4236 sections->Dump(s, NULL, true, UINT32_MAX); 4237 4238 if (m_symtab_ap.get()) 4239 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4240 } 4241 } 4242 4243 bool 4244 ObjectFileMachO::GetUUID (const llvm::MachO::mach_header &header, 4245 const lldb_private::DataExtractor &data, 4246 lldb::offset_t lc_offset, 4247 lldb_private::UUID& uuid) 4248 { 4249 uint32_t i; 4250 struct uuid_command load_cmd; 4251 4252 lldb::offset_t offset = lc_offset; 4253 for (i=0; i<header.ncmds; ++i) 4254 { 4255 const lldb::offset_t cmd_offset = offset; 4256 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4257 break; 4258 4259 if (load_cmd.cmd == LC_UUID) 4260 { 4261 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4262 4263 if (uuid_bytes) 4264 { 4265 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4266 // We pretend these object files have no UUID to prevent crashing. 4267 4268 const uint8_t opencl_uuid[] = { 0x8c, 0x8e, 0xb3, 0x9b, 4269 0x3b, 0xa8, 4270 0x4b, 0x16, 4271 0xb6, 0xa4, 4272 0x27, 0x63, 0xbb, 0x14, 0xf0, 0x0d }; 4273 4274 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4275 return false; 4276 4277 uuid.SetBytes (uuid_bytes); 4278 return true; 4279 } 4280 return false; 4281 } 4282 offset = cmd_offset + load_cmd.cmdsize; 4283 } 4284 return false; 4285 } 4286 4287 4288 bool 4289 ObjectFileMachO::GetArchitecture (const llvm::MachO::mach_header &header, 4290 const lldb_private::DataExtractor &data, 4291 lldb::offset_t lc_offset, 4292 ArchSpec &arch) 4293 { 4294 arch.SetArchitecture (eArchTypeMachO, header.cputype, header.cpusubtype); 4295 4296 if (arch.IsValid()) 4297 { 4298 llvm::Triple &triple = arch.GetTriple(); 4299 if (header.filetype == MH_PRELOAD) 4300 { 4301 // Set OS to "unknown" - this is a standalone binary with no dyld et al 4302 triple.setOS(llvm::Triple::UnknownOS); 4303 return true; 4304 } 4305 else 4306 { 4307 struct load_command load_cmd; 4308 4309 lldb::offset_t offset = lc_offset; 4310 for (uint32_t i=0; i<header.ncmds; ++i) 4311 { 4312 const lldb::offset_t cmd_offset = offset; 4313 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4314 break; 4315 4316 switch (load_cmd.cmd) 4317 { 4318 case LC_VERSION_MIN_IPHONEOS: 4319 triple.setOS (llvm::Triple::IOS); 4320 return true; 4321 4322 case LC_VERSION_MIN_MACOSX: 4323 triple.setOS (llvm::Triple::MacOSX); 4324 return true; 4325 4326 default: 4327 break; 4328 } 4329 4330 offset = cmd_offset + load_cmd.cmdsize; 4331 } 4332 4333 if (header.cputype == CPU_TYPE_ARM || header.cputype == CPU_TYPE_ARM64) 4334 triple.setOS (llvm::Triple::IOS); 4335 else 4336 triple.setOS (llvm::Triple::MacOSX); 4337 } 4338 } 4339 return arch.IsValid(); 4340 } 4341 4342 bool 4343 ObjectFileMachO::GetUUID (lldb_private::UUID* uuid) 4344 { 4345 ModuleSP module_sp(GetModule()); 4346 if (module_sp) 4347 { 4348 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4349 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4350 return GetUUID (m_header, m_data, offset, *uuid); 4351 } 4352 return false; 4353 } 4354 4355 4356 uint32_t 4357 ObjectFileMachO::GetDependentModules (FileSpecList& files) 4358 { 4359 uint32_t count = 0; 4360 ModuleSP module_sp(GetModule()); 4361 if (module_sp) 4362 { 4363 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4364 struct load_command load_cmd; 4365 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4366 const bool resolve_path = false; // Don't resolve the dependend file paths since they may not reside on this system 4367 uint32_t i; 4368 for (i=0; i<m_header.ncmds; ++i) 4369 { 4370 const uint32_t cmd_offset = offset; 4371 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4372 break; 4373 4374 switch (load_cmd.cmd) 4375 { 4376 case LC_LOAD_DYLIB: 4377 case LC_LOAD_WEAK_DYLIB: 4378 case LC_REEXPORT_DYLIB: 4379 case LC_LOAD_DYLINKER: 4380 case LC_LOADFVMLIB: 4381 case LC_LOAD_UPWARD_DYLIB: 4382 { 4383 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 4384 const char *path = m_data.PeekCStr(name_offset); 4385 // Skip any path that starts with '@' since these are usually: 4386 // @executable_path/.../file 4387 // @rpath/.../file 4388 if (path && path[0] != '@') 4389 { 4390 FileSpec file_spec(path, resolve_path); 4391 if (files.AppendIfUnique(file_spec)) 4392 count++; 4393 } 4394 } 4395 break; 4396 4397 default: 4398 break; 4399 } 4400 offset = cmd_offset + load_cmd.cmdsize; 4401 } 4402 } 4403 return count; 4404 } 4405 4406 lldb_private::Address 4407 ObjectFileMachO::GetEntryPointAddress () 4408 { 4409 // If the object file is not an executable it can't hold the entry point. m_entry_point_address 4410 // is initialized to an invalid address, so we can just return that. 4411 // If m_entry_point_address is valid it means we've found it already, so return the cached value. 4412 4413 if (!IsExecutable() || m_entry_point_address.IsValid()) 4414 return m_entry_point_address; 4415 4416 // Otherwise, look for the UnixThread or Thread command. The data for the Thread command is given in 4417 // /usr/include/mach-o.h, but it is basically: 4418 // 4419 // uint32_t flavor - this is the flavor argument you would pass to thread_get_state 4420 // uint32_t count - this is the count of longs in the thread state data 4421 // struct XXX_thread_state state - this is the structure from <machine/thread_status.h> corresponding to the flavor. 4422 // <repeat this trio> 4423 // 4424 // So we just keep reading the various register flavors till we find the GPR one, then read the PC out of there. 4425 // FIXME: We will need to have a "RegisterContext data provider" class at some point that can get all the registers 4426 // out of data in this form & attach them to a given thread. That should underlie the MacOS X User process plugin, 4427 // and we'll also need it for the MacOS X Core File process plugin. When we have that we can also use it here. 4428 // 4429 // For now we hard-code the offsets and flavors we need: 4430 // 4431 // 4432 4433 ModuleSP module_sp(GetModule()); 4434 if (module_sp) 4435 { 4436 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4437 struct load_command load_cmd; 4438 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4439 uint32_t i; 4440 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 4441 bool done = false; 4442 4443 for (i=0; i<m_header.ncmds; ++i) 4444 { 4445 const lldb::offset_t cmd_offset = offset; 4446 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4447 break; 4448 4449 switch (load_cmd.cmd) 4450 { 4451 case LC_UNIXTHREAD: 4452 case LC_THREAD: 4453 { 4454 while (offset < cmd_offset + load_cmd.cmdsize) 4455 { 4456 uint32_t flavor = m_data.GetU32(&offset); 4457 uint32_t count = m_data.GetU32(&offset); 4458 if (count == 0) 4459 { 4460 // We've gotten off somehow, log and exit; 4461 return m_entry_point_address; 4462 } 4463 4464 switch (m_header.cputype) 4465 { 4466 case llvm::MachO::CPU_TYPE_ARM: 4467 if (flavor == 1) // ARM_THREAD_STATE from mach/arm/thread_status.h 4468 { 4469 offset += 60; // This is the offset of pc in the GPR thread state data structure. 4470 start_address = m_data.GetU32(&offset); 4471 done = true; 4472 } 4473 break; 4474 case llvm::MachO::CPU_TYPE_ARM64: 4475 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 4476 { 4477 offset += 256; // This is the offset of pc in the GPR thread state data structure. 4478 start_address = m_data.GetU64(&offset); 4479 done = true; 4480 } 4481 break; 4482 case llvm::MachO::CPU_TYPE_I386: 4483 if (flavor == 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 4484 { 4485 offset += 40; // This is the offset of eip in the GPR thread state data structure. 4486 start_address = m_data.GetU32(&offset); 4487 done = true; 4488 } 4489 break; 4490 case llvm::MachO::CPU_TYPE_X86_64: 4491 if (flavor == 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 4492 { 4493 offset += 16 * 8; // This is the offset of rip in the GPR thread state data structure. 4494 start_address = m_data.GetU64(&offset); 4495 done = true; 4496 } 4497 break; 4498 default: 4499 return m_entry_point_address; 4500 } 4501 // Haven't found the GPR flavor yet, skip over the data for this flavor: 4502 if (done) 4503 break; 4504 offset += count * 4; 4505 } 4506 } 4507 break; 4508 case LC_MAIN: 4509 { 4510 ConstString text_segment_name ("__TEXT"); 4511 uint64_t entryoffset = m_data.GetU64(&offset); 4512 SectionSP text_segment_sp = GetSectionList()->FindSectionByName(text_segment_name); 4513 if (text_segment_sp) 4514 { 4515 done = true; 4516 start_address = text_segment_sp->GetFileAddress() + entryoffset; 4517 } 4518 } 4519 4520 default: 4521 break; 4522 } 4523 if (done) 4524 break; 4525 4526 // Go to the next load command: 4527 offset = cmd_offset + load_cmd.cmdsize; 4528 } 4529 4530 if (start_address != LLDB_INVALID_ADDRESS) 4531 { 4532 // We got the start address from the load commands, so now resolve that address in the sections 4533 // of this ObjectFile: 4534 if (!m_entry_point_address.ResolveAddressUsingFileSections (start_address, GetSectionList())) 4535 { 4536 m_entry_point_address.Clear(); 4537 } 4538 } 4539 else 4540 { 4541 // We couldn't read the UnixThread load command - maybe it wasn't there. As a fallback look for the 4542 // "start" symbol in the main executable. 4543 4544 ModuleSP module_sp (GetModule()); 4545 4546 if (module_sp) 4547 { 4548 SymbolContextList contexts; 4549 SymbolContext context; 4550 if (module_sp->FindSymbolsWithNameAndType(ConstString ("start"), eSymbolTypeCode, contexts)) 4551 { 4552 if (contexts.GetContextAtIndex(0, context)) 4553 m_entry_point_address = context.symbol->GetAddress(); 4554 } 4555 } 4556 } 4557 } 4558 4559 return m_entry_point_address; 4560 4561 } 4562 4563 lldb_private::Address 4564 ObjectFileMachO::GetHeaderAddress () 4565 { 4566 lldb_private::Address header_addr; 4567 SectionList *section_list = GetSectionList(); 4568 if (section_list) 4569 { 4570 SectionSP text_segment_sp (section_list->FindSectionByName (GetSegmentNameTEXT())); 4571 if (text_segment_sp) 4572 { 4573 header_addr.SetSection (text_segment_sp); 4574 header_addr.SetOffset (0); 4575 } 4576 } 4577 return header_addr; 4578 } 4579 4580 uint32_t 4581 ObjectFileMachO::GetNumThreadContexts () 4582 { 4583 ModuleSP module_sp(GetModule()); 4584 if (module_sp) 4585 { 4586 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4587 if (!m_thread_context_offsets_valid) 4588 { 4589 m_thread_context_offsets_valid = true; 4590 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4591 FileRangeArray::Entry file_range; 4592 thread_command thread_cmd; 4593 for (uint32_t i=0; i<m_header.ncmds; ++i) 4594 { 4595 const uint32_t cmd_offset = offset; 4596 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 4597 break; 4598 4599 if (thread_cmd.cmd == LC_THREAD) 4600 { 4601 file_range.SetRangeBase (offset); 4602 file_range.SetByteSize (thread_cmd.cmdsize - 8); 4603 m_thread_context_offsets.Append (file_range); 4604 } 4605 offset = cmd_offset + thread_cmd.cmdsize; 4606 } 4607 } 4608 } 4609 return m_thread_context_offsets.GetSize(); 4610 } 4611 4612 lldb::RegisterContextSP 4613 ObjectFileMachO::GetThreadContextAtIndex (uint32_t idx, lldb_private::Thread &thread) 4614 { 4615 lldb::RegisterContextSP reg_ctx_sp; 4616 4617 ModuleSP module_sp(GetModule()); 4618 if (module_sp) 4619 { 4620 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4621 if (!m_thread_context_offsets_valid) 4622 GetNumThreadContexts (); 4623 4624 const FileRangeArray::Entry *thread_context_file_range = m_thread_context_offsets.GetEntryAtIndex (idx); 4625 if (thread_context_file_range) 4626 { 4627 4628 DataExtractor data (m_data, 4629 thread_context_file_range->GetRangeBase(), 4630 thread_context_file_range->GetByteSize()); 4631 4632 switch (m_header.cputype) 4633 { 4634 case llvm::MachO::CPU_TYPE_ARM64: 4635 reg_ctx_sp.reset (new RegisterContextDarwin_arm64_Mach (thread, data)); 4636 break; 4637 4638 case llvm::MachO::CPU_TYPE_ARM: 4639 reg_ctx_sp.reset (new RegisterContextDarwin_arm_Mach (thread, data)); 4640 break; 4641 4642 case llvm::MachO::CPU_TYPE_I386: 4643 reg_ctx_sp.reset (new RegisterContextDarwin_i386_Mach (thread, data)); 4644 break; 4645 4646 case llvm::MachO::CPU_TYPE_X86_64: 4647 reg_ctx_sp.reset (new RegisterContextDarwin_x86_64_Mach (thread, data)); 4648 break; 4649 } 4650 } 4651 } 4652 return reg_ctx_sp; 4653 } 4654 4655 4656 ObjectFile::Type 4657 ObjectFileMachO::CalculateType() 4658 { 4659 switch (m_header.filetype) 4660 { 4661 case MH_OBJECT: // 0x1u 4662 if (GetAddressByteSize () == 4) 4663 { 4664 // 32 bit kexts are just object files, but they do have a valid 4665 // UUID load command. 4666 UUID uuid; 4667 if (GetUUID(&uuid)) 4668 { 4669 // this checking for the UUID load command is not enough 4670 // we could eventually look for the symbol named 4671 // "OSKextGetCurrentIdentifier" as this is required of kexts 4672 if (m_strata == eStrataInvalid) 4673 m_strata = eStrataKernel; 4674 return eTypeSharedLibrary; 4675 } 4676 } 4677 return eTypeObjectFile; 4678 4679 case MH_EXECUTE: return eTypeExecutable; // 0x2u 4680 case MH_FVMLIB: return eTypeSharedLibrary; // 0x3u 4681 case MH_CORE: return eTypeCoreFile; // 0x4u 4682 case MH_PRELOAD: return eTypeSharedLibrary; // 0x5u 4683 case MH_DYLIB: return eTypeSharedLibrary; // 0x6u 4684 case MH_DYLINKER: return eTypeDynamicLinker; // 0x7u 4685 case MH_BUNDLE: return eTypeSharedLibrary; // 0x8u 4686 case MH_DYLIB_STUB: return eTypeStubLibrary; // 0x9u 4687 case MH_DSYM: return eTypeDebugInfo; // 0xAu 4688 case MH_KEXT_BUNDLE: return eTypeSharedLibrary; // 0xBu 4689 default: 4690 break; 4691 } 4692 return eTypeUnknown; 4693 } 4694 4695 ObjectFile::Strata 4696 ObjectFileMachO::CalculateStrata() 4697 { 4698 switch (m_header.filetype) 4699 { 4700 case MH_OBJECT: // 0x1u 4701 { 4702 // 32 bit kexts are just object files, but they do have a valid 4703 // UUID load command. 4704 UUID uuid; 4705 if (GetUUID(&uuid)) 4706 { 4707 // this checking for the UUID load command is not enough 4708 // we could eventually look for the symbol named 4709 // "OSKextGetCurrentIdentifier" as this is required of kexts 4710 if (m_type == eTypeInvalid) 4711 m_type = eTypeSharedLibrary; 4712 4713 return eStrataKernel; 4714 } 4715 } 4716 return eStrataUnknown; 4717 4718 case MH_EXECUTE: // 0x2u 4719 // Check for the MH_DYLDLINK bit in the flags 4720 if (m_header.flags & MH_DYLDLINK) 4721 { 4722 return eStrataUser; 4723 } 4724 else 4725 { 4726 SectionList *section_list = GetSectionList(); 4727 if (section_list) 4728 { 4729 static ConstString g_kld_section_name ("__KLD"); 4730 if (section_list->FindSectionByName(g_kld_section_name)) 4731 return eStrataKernel; 4732 } 4733 } 4734 return eStrataRawImage; 4735 4736 case MH_FVMLIB: return eStrataUser; // 0x3u 4737 case MH_CORE: return eStrataUnknown; // 0x4u 4738 case MH_PRELOAD: return eStrataRawImage; // 0x5u 4739 case MH_DYLIB: return eStrataUser; // 0x6u 4740 case MH_DYLINKER: return eStrataUser; // 0x7u 4741 case MH_BUNDLE: return eStrataUser; // 0x8u 4742 case MH_DYLIB_STUB: return eStrataUser; // 0x9u 4743 case MH_DSYM: return eStrataUnknown; // 0xAu 4744 case MH_KEXT_BUNDLE: return eStrataKernel; // 0xBu 4745 default: 4746 break; 4747 } 4748 return eStrataUnknown; 4749 } 4750 4751 4752 uint32_t 4753 ObjectFileMachO::GetVersion (uint32_t *versions, uint32_t num_versions) 4754 { 4755 ModuleSP module_sp(GetModule()); 4756 if (module_sp) 4757 { 4758 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4759 struct dylib_command load_cmd; 4760 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4761 uint32_t version_cmd = 0; 4762 uint64_t version = 0; 4763 uint32_t i; 4764 for (i=0; i<m_header.ncmds; ++i) 4765 { 4766 const lldb::offset_t cmd_offset = offset; 4767 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4768 break; 4769 4770 if (load_cmd.cmd == LC_ID_DYLIB) 4771 { 4772 if (version_cmd == 0) 4773 { 4774 version_cmd = load_cmd.cmd; 4775 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 4776 break; 4777 version = load_cmd.dylib.current_version; 4778 } 4779 break; // Break for now unless there is another more complete version 4780 // number load command in the future. 4781 } 4782 offset = cmd_offset + load_cmd.cmdsize; 4783 } 4784 4785 if (version_cmd == LC_ID_DYLIB) 4786 { 4787 if (versions != NULL && num_versions > 0) 4788 { 4789 if (num_versions > 0) 4790 versions[0] = (version & 0xFFFF0000ull) >> 16; 4791 if (num_versions > 1) 4792 versions[1] = (version & 0x0000FF00ull) >> 8; 4793 if (num_versions > 2) 4794 versions[2] = (version & 0x000000FFull); 4795 // Fill in an remaining version numbers with invalid values 4796 for (i=3; i<num_versions; ++i) 4797 versions[i] = UINT32_MAX; 4798 } 4799 // The LC_ID_DYLIB load command has a version with 3 version numbers 4800 // in it, so always return 3 4801 return 3; 4802 } 4803 } 4804 return false; 4805 } 4806 4807 bool 4808 ObjectFileMachO::GetArchitecture (ArchSpec &arch) 4809 { 4810 ModuleSP module_sp(GetModule()); 4811 if (module_sp) 4812 { 4813 lldb_private::Mutex::Locker locker(module_sp->GetMutex()); 4814 return GetArchitecture (m_header, m_data, MachHeaderSizeFromMagic(m_header.magic), arch); 4815 } 4816 return false; 4817 } 4818 4819 4820 UUID 4821 ObjectFileMachO::GetProcessSharedCacheUUID (Process *process) 4822 { 4823 UUID uuid; 4824 if (process) 4825 { 4826 addr_t all_image_infos = process->GetImageInfoAddress(); 4827 4828 // The address returned by GetImageInfoAddress may be the address of dyld (don't want) 4829 // or it may be the address of the dyld_all_image_infos structure (want). The first four 4830 // bytes will be either the version field (all_image_infos) or a Mach-O file magic constant. 4831 // Version 13 and higher of dyld_all_image_infos is required to get the sharedCacheUUID field. 4832 4833 Error err; 4834 uint32_t version_or_magic = process->ReadUnsignedIntegerFromMemory (all_image_infos, 4, -1, err); 4835 if (version_or_magic != static_cast<uint32_t>(-1) 4836 && version_or_magic != MH_MAGIC 4837 && version_or_magic != MH_CIGAM 4838 && version_or_magic != MH_MAGIC_64 4839 && version_or_magic != MH_CIGAM_64 4840 && version_or_magic >= 13) 4841 { 4842 addr_t sharedCacheUUID_address = LLDB_INVALID_ADDRESS; 4843 int wordsize = process->GetAddressByteSize(); 4844 if (wordsize == 8) 4845 { 4846 sharedCacheUUID_address = all_image_infos + 160; // sharedCacheUUID <mach-o/dyld_images.h> 4847 } 4848 if (wordsize == 4) 4849 { 4850 sharedCacheUUID_address = all_image_infos + 84; // sharedCacheUUID <mach-o/dyld_images.h> 4851 } 4852 if (sharedCacheUUID_address != LLDB_INVALID_ADDRESS) 4853 { 4854 uuid_t shared_cache_uuid; 4855 if (process->ReadMemory (sharedCacheUUID_address, shared_cache_uuid, sizeof (uuid_t), err) == sizeof (uuid_t)) 4856 { 4857 uuid.SetBytes (shared_cache_uuid); 4858 } 4859 } 4860 } 4861 } 4862 return uuid; 4863 } 4864 4865 UUID 4866 ObjectFileMachO::GetLLDBSharedCacheUUID () 4867 { 4868 UUID uuid; 4869 #if defined (__APPLE__) && (defined (__arm__) || defined (__arm64__)) 4870 uint8_t *(*dyld_get_all_image_infos)(void); 4871 dyld_get_all_image_infos = (uint8_t*(*)()) dlsym (RTLD_DEFAULT, "_dyld_get_all_image_infos"); 4872 if (dyld_get_all_image_infos) 4873 { 4874 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 4875 if (dyld_all_image_infos_address) 4876 { 4877 uint32_t *version = (uint32_t*) dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 4878 if (*version >= 13) 4879 { 4880 uuid_t *sharedCacheUUID_address = 0; 4881 int wordsize = sizeof (uint8_t *); 4882 if (wordsize == 8) 4883 { 4884 sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 160); // sharedCacheUUID <mach-o/dyld_images.h> 4885 } 4886 else 4887 { 4888 sharedCacheUUID_address = (uuid_t*) ((uint8_t*) dyld_all_image_infos_address + 84); // sharedCacheUUID <mach-o/dyld_images.h> 4889 } 4890 uuid.SetBytes (sharedCacheUUID_address); 4891 } 4892 } 4893 } 4894 #endif 4895 return uuid; 4896 } 4897 4898 uint32_t 4899 ObjectFileMachO::GetMinimumOSVersion (uint32_t *versions, uint32_t num_versions) 4900 { 4901 if (m_min_os_versions.empty()) 4902 { 4903 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4904 bool success = false; 4905 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 4906 { 4907 const lldb::offset_t load_cmd_offset = offset; 4908 4909 version_min_command lc; 4910 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 4911 break; 4912 if (lc.cmd == LC_VERSION_MIN_MACOSX || lc.cmd == LC_VERSION_MIN_IPHONEOS) 4913 { 4914 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 4915 { 4916 const uint32_t xxxx = lc.version >> 16; 4917 const uint32_t yy = (lc.version >> 8) & 0xffu; 4918 const uint32_t zz = lc.version & 0xffu; 4919 if (xxxx) 4920 { 4921 m_min_os_versions.push_back(xxxx); 4922 if (yy) 4923 { 4924 m_min_os_versions.push_back(yy); 4925 if (zz) 4926 m_min_os_versions.push_back(zz); 4927 } 4928 } 4929 success = true; 4930 } 4931 } 4932 offset = load_cmd_offset + lc.cmdsize; 4933 } 4934 4935 if (success == false) 4936 { 4937 // Push an invalid value so we don't keep trying to 4938 m_min_os_versions.push_back(UINT32_MAX); 4939 } 4940 } 4941 4942 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) 4943 { 4944 if (versions != NULL && num_versions > 0) 4945 { 4946 for (size_t i=0; i<num_versions; ++i) 4947 { 4948 if (i < m_min_os_versions.size()) 4949 versions[i] = m_min_os_versions[i]; 4950 else 4951 versions[i] = 0; 4952 } 4953 } 4954 return m_min_os_versions.size(); 4955 } 4956 // Call the superclasses version that will empty out the data 4957 return ObjectFile::GetMinimumOSVersion (versions, num_versions); 4958 } 4959 4960 uint32_t 4961 ObjectFileMachO::GetSDKVersion(uint32_t *versions, uint32_t num_versions) 4962 { 4963 if (m_sdk_versions.empty()) 4964 { 4965 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4966 bool success = false; 4967 for (uint32_t i=0; success == false && i < m_header.ncmds; ++i) 4968 { 4969 const lldb::offset_t load_cmd_offset = offset; 4970 4971 version_min_command lc; 4972 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 4973 break; 4974 if (lc.cmd == LC_VERSION_MIN_MACOSX || lc.cmd == LC_VERSION_MIN_IPHONEOS) 4975 { 4976 if (m_data.GetU32 (&offset, &lc.version, (sizeof(lc) / sizeof(uint32_t)) - 2)) 4977 { 4978 const uint32_t xxxx = lc.reserved >> 16; 4979 const uint32_t yy = (lc.reserved >> 8) & 0xffu; 4980 const uint32_t zz = lc.reserved & 0xffu; 4981 if (xxxx) 4982 { 4983 m_sdk_versions.push_back(xxxx); 4984 if (yy) 4985 { 4986 m_sdk_versions.push_back(yy); 4987 if (zz) 4988 m_sdk_versions.push_back(zz); 4989 } 4990 } 4991 success = true; 4992 } 4993 } 4994 offset = load_cmd_offset + lc.cmdsize; 4995 } 4996 4997 if (success == false) 4998 { 4999 // Push an invalid value so we don't keep trying to 5000 m_sdk_versions.push_back(UINT32_MAX); 5001 } 5002 } 5003 5004 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) 5005 { 5006 if (versions != NULL && num_versions > 0) 5007 { 5008 for (size_t i=0; i<num_versions; ++i) 5009 { 5010 if (i < m_sdk_versions.size()) 5011 versions[i] = m_sdk_versions[i]; 5012 else 5013 versions[i] = 0; 5014 } 5015 } 5016 return m_sdk_versions.size(); 5017 } 5018 // Call the superclasses version that will empty out the data 5019 return ObjectFile::GetSDKVersion (versions, num_versions); 5020 } 5021 5022 5023 //------------------------------------------------------------------ 5024 // PluginInterface protocol 5025 //------------------------------------------------------------------ 5026 lldb_private::ConstString 5027 ObjectFileMachO::GetPluginName() 5028 { 5029 return GetPluginNameStatic(); 5030 } 5031 5032 uint32_t 5033 ObjectFileMachO::GetPluginVersion() 5034 { 5035 return 1; 5036 } 5037 5038 5039 bool 5040 ObjectFileMachO::SetLoadAddress (Target &target, 5041 lldb::addr_t value, 5042 bool value_is_offset) 5043 { 5044 bool changed = false; 5045 ModuleSP module_sp = GetModule(); 5046 if (module_sp) 5047 { 5048 size_t num_loaded_sections = 0; 5049 SectionList *section_list = GetSectionList (); 5050 if (section_list) 5051 { 5052 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5053 const size_t num_sections = section_list->GetSize(); 5054 5055 const bool is_memory_image = (bool)m_process_wp.lock(); 5056 const Strata strata = GetStrata(); 5057 static ConstString g_linkedit_segname ("__LINKEDIT"); 5058 if (value_is_offset) 5059 { 5060 // "value" is an offset to apply to each top level segment 5061 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) 5062 { 5063 // Iterate through the object file sections to find all 5064 // of the sections that size on disk (to avoid __PAGEZERO) 5065 // and load them 5066 SectionSP section_sp (section_list->GetSectionAtIndex (sect_idx)); 5067 if (section_sp && 5068 section_sp->GetFileSize() > 0 && 5069 section_sp->IsThreadSpecific() == false && 5070 module_sp.get() == section_sp->GetModule().get()) 5071 { 5072 // Ignore __LINKEDIT and __DWARF segments 5073 if (section_sp->GetName() == g_linkedit_segname) 5074 { 5075 // Only map __LINKEDIT if we have an in memory image and this isn't 5076 // a kernel binary like a kext or mach_kernel. 5077 if (is_memory_image == false || strata == eStrataKernel) 5078 continue; 5079 } 5080 if (target.GetSectionLoadList().SetSectionLoadAddress (section_sp, section_sp->GetFileAddress() + value)) 5081 ++num_loaded_sections; 5082 } 5083 } 5084 } 5085 else 5086 { 5087 // "value" is the new base address of the mach_header, adjust each 5088 // section accordingly 5089 5090 // First find the address of the mach header which is the first non-zero 5091 // file sized section whose file offset is zero as this will be subtracted 5092 // from each other valid section's vmaddr and then get "base_addr" added to 5093 // it when loading the module in the target 5094 for (size_t sect_idx = 0; 5095 sect_idx < num_sections && mach_base_file_addr == LLDB_INVALID_ADDRESS; 5096 ++sect_idx) 5097 { 5098 // Iterate through the object file sections to find all 5099 // of the sections that size on disk (to avoid __PAGEZERO) 5100 // and load them 5101 Section *section = section_list->GetSectionAtIndex (sect_idx).get(); 5102 if (section && 5103 section->GetFileSize() > 0 && 5104 section->GetFileOffset() == 0 && 5105 section->IsThreadSpecific() == false && 5106 module_sp.get() == section->GetModule().get()) 5107 { 5108 // Ignore __LINKEDIT and __DWARF segments 5109 if (section->GetName() == g_linkedit_segname) 5110 { 5111 // Only map __LINKEDIT if we have an in memory image and this isn't 5112 // a kernel binary like a kext or mach_kernel. 5113 if (is_memory_image == false || strata == eStrataKernel) 5114 continue; 5115 } 5116 mach_base_file_addr = section->GetFileAddress(); 5117 } 5118 } 5119 5120 if (mach_base_file_addr != LLDB_INVALID_ADDRESS) 5121 { 5122 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) 5123 { 5124 // Iterate through the object file sections to find all 5125 // of the sections that size on disk (to avoid __PAGEZERO) 5126 // and load them 5127 SectionSP section_sp (section_list->GetSectionAtIndex (sect_idx)); 5128 if (section_sp && 5129 section_sp->GetFileSize() > 0 && 5130 section_sp->IsThreadSpecific() == false && 5131 module_sp.get() == section_sp->GetModule().get()) 5132 { 5133 // Ignore __LINKEDIT and __DWARF segments 5134 if (section_sp->GetName() == g_linkedit_segname) 5135 { 5136 // Only map __LINKEDIT if we have an in memory image and this isn't 5137 // a kernel binary like a kext or mach_kernel. 5138 if (is_memory_image == false || strata == eStrataKernel) 5139 continue; 5140 } 5141 if (target.GetSectionLoadList().SetSectionLoadAddress (section_sp, section_sp->GetFileAddress() - mach_base_file_addr + value)) 5142 ++num_loaded_sections; 5143 } 5144 } 5145 } 5146 } 5147 } 5148 changed = num_loaded_sections > 0; 5149 return num_loaded_sections > 0; 5150 } 5151 return changed; 5152 } 5153 5154 bool 5155 ObjectFileMachO::SaveCore (const lldb::ProcessSP &process_sp, 5156 const FileSpec &outfile, 5157 Error &error) 5158 { 5159 if (process_sp) 5160 { 5161 Target &target = process_sp->GetTarget(); 5162 const ArchSpec target_arch = target.GetArchitecture(); 5163 const llvm::Triple &target_triple = target_arch.GetTriple(); 5164 if (target_triple.getVendor() == llvm::Triple::Apple && 5165 (target_triple.getOS() == llvm::Triple::MacOSX || 5166 target_triple.getOS() == llvm::Triple::IOS)) 5167 { 5168 bool make_core = false; 5169 switch (target_arch.GetMachine()) 5170 { 5171 case llvm::Triple::arm: 5172 case llvm::Triple::x86: 5173 case llvm::Triple::x86_64: 5174 make_core = true; 5175 break; 5176 default: 5177 error.SetErrorStringWithFormat ("unsupported core architecture: %s", target_triple.str().c_str()); 5178 break; 5179 } 5180 5181 if (make_core) 5182 { 5183 std::vector<segment_command_64> segment_load_commands; 5184 // uint32_t range_info_idx = 0; 5185 MemoryRegionInfo range_info; 5186 Error range_error = process_sp->GetMemoryRegionInfo(0, range_info); 5187 if (range_error.Success()) 5188 { 5189 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) 5190 { 5191 const addr_t addr = range_info.GetRange().GetRangeBase(); 5192 const addr_t size = range_info.GetRange().GetByteSize(); 5193 5194 if (size == 0) 5195 break; 5196 5197 // Calculate correct protections 5198 uint32_t prot = 0; 5199 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 5200 prot |= VM_PROT_READ; 5201 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 5202 prot |= VM_PROT_WRITE; 5203 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 5204 prot |= VM_PROT_EXECUTE; 5205 5206 // printf ("[%3u] [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 ") %c%c%c\n", 5207 // range_info_idx, 5208 // addr, 5209 // size, 5210 // (prot & VM_PROT_READ ) ? 'r' : '-', 5211 // (prot & VM_PROT_WRITE ) ? 'w' : '-', 5212 // (prot & VM_PROT_EXECUTE) ? 'x' : '-'); 5213 5214 if (prot != 0) 5215 { 5216 segment_command_64 segment = { 5217 LC_SEGMENT_64, // uint32_t cmd; 5218 sizeof(segment), // uint32_t cmdsize; 5219 {0}, // char segname[16]; 5220 addr, // uint64_t vmaddr; 5221 size, // uint64_t vmsize; 5222 0, // uint64_t fileoff; 5223 size, // uint64_t filesize; 5224 prot, // uint32_t maxprot; 5225 prot, // uint32_t initprot; 5226 0, // uint32_t nsects; 5227 0 }; // uint32_t flags; 5228 segment_load_commands.push_back(segment); 5229 } 5230 else 5231 { 5232 // No protections and a size of 1 used to be returned from old 5233 // debugservers when we asked about a region that was past the 5234 // last memory region and it indicates the end... 5235 if (size == 1) 5236 break; 5237 } 5238 5239 range_error = process_sp->GetMemoryRegionInfo(range_info.GetRange().GetRangeEnd(), range_info); 5240 if (range_error.Fail()) 5241 break; 5242 } 5243 5244 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 5245 const ByteOrder byte_order = target_arch.GetByteOrder(); 5246 StreamString buffer (Stream::eBinary, 5247 addr_byte_size, 5248 byte_order); 5249 5250 mach_header_64 mach_header; 5251 mach_header.magic = MH_MAGIC_64; 5252 mach_header.cputype = target_arch.GetMachOCPUType(); 5253 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 5254 mach_header.filetype = MH_CORE; 5255 mach_header.ncmds = segment_load_commands.size(); 5256 mach_header.flags = 0; 5257 mach_header.reserved = 0; 5258 ThreadList &thread_list = process_sp->GetThreadList(); 5259 const uint32_t num_threads = thread_list.GetSize(); 5260 5261 // Make an array of LC_THREAD data items. Each one contains 5262 // the contents of the LC_THREAD load command. The data doesn't 5263 // contain the load command + load command size, we will 5264 // add the load command and load command size as we emit the data. 5265 std::vector<StreamString> LC_THREAD_datas(num_threads); 5266 for (auto &LC_THREAD_data : LC_THREAD_datas) 5267 { 5268 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 5269 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 5270 LC_THREAD_data.SetByteOrder(byte_order); 5271 } 5272 for (uint32_t thread_idx = 0; thread_idx < num_threads; ++thread_idx) 5273 { 5274 ThreadSP thread_sp (thread_list.GetThreadAtIndex(thread_idx)); 5275 if (thread_sp) 5276 { 5277 switch (mach_header.cputype) 5278 { 5279 case llvm::MachO::CPU_TYPE_ARM: 5280 //RegisterContextDarwin_arm_Mach::Create_LC_THREAD (thread_sp.get(), thread_load_commands); 5281 break; 5282 5283 case llvm::MachO::CPU_TYPE_I386: 5284 //RegisterContextDarwin_i386_Mach::Create_LC_THREAD (thread_sp.get(), thread_load_commands); 5285 break; 5286 5287 case llvm::MachO::CPU_TYPE_X86_64: 5288 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD (thread_sp.get(), LC_THREAD_datas[thread_idx]); 5289 break; 5290 } 5291 5292 } 5293 } 5294 5295 // The size of the load command is the size of the segments... 5296 mach_header.sizeofcmds = segment_load_commands.size() * segment_load_commands[0].cmdsize; 5297 5298 // and the size of all LC_THREAD load command 5299 for (const auto &LC_THREAD_data : LC_THREAD_datas) 5300 { 5301 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 5302 } 5303 5304 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", 5305 mach_header.magic, 5306 mach_header.cputype, 5307 mach_header.cpusubtype, 5308 mach_header.filetype, 5309 mach_header.ncmds, 5310 mach_header.sizeofcmds, 5311 mach_header.flags, 5312 mach_header.reserved); 5313 5314 // Write the mach header 5315 buffer.PutHex32(mach_header.magic); 5316 buffer.PutHex32(mach_header.cputype); 5317 buffer.PutHex32(mach_header.cpusubtype); 5318 buffer.PutHex32(mach_header.filetype); 5319 buffer.PutHex32(mach_header.ncmds); 5320 buffer.PutHex32(mach_header.sizeofcmds); 5321 buffer.PutHex32(mach_header.flags); 5322 buffer.PutHex32(mach_header.reserved); 5323 5324 // Skip the mach header and all load commands and align to the next 5325 // 0x1000 byte boundary 5326 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 5327 if (file_offset & 0x00000fff) 5328 { 5329 file_offset += 0x00001000ull; 5330 file_offset &= (~0x00001000ull + 1); 5331 } 5332 5333 for (auto &segment : segment_load_commands) 5334 { 5335 segment.fileoff = file_offset; 5336 file_offset += segment.filesize; 5337 } 5338 5339 // Write out all of the LC_THREAD load commands 5340 for (const auto &LC_THREAD_data : LC_THREAD_datas) 5341 { 5342 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 5343 buffer.PutHex32(LC_THREAD); 5344 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 5345 buffer.Write(LC_THREAD_data.GetData(), LC_THREAD_data_size); 5346 } 5347 5348 // Write out all of the segment load commands 5349 for (const auto &segment : segment_load_commands) 5350 { 5351 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", 5352 segment.cmd, 5353 segment.cmdsize, 5354 segment.vmaddr, 5355 segment.vmaddr + segment.vmsize, 5356 segment.fileoff, 5357 segment.filesize, 5358 segment.maxprot, 5359 segment.initprot, 5360 segment.nsects, 5361 segment.flags); 5362 5363 buffer.PutHex32(segment.cmd); 5364 buffer.PutHex32(segment.cmdsize); 5365 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 5366 buffer.PutHex64(segment.vmaddr); 5367 buffer.PutHex64(segment.vmsize); 5368 buffer.PutHex64(segment.fileoff); 5369 buffer.PutHex64(segment.filesize); 5370 buffer.PutHex32(segment.maxprot); 5371 buffer.PutHex32(segment.initprot); 5372 buffer.PutHex32(segment.nsects); 5373 buffer.PutHex32(segment.flags); 5374 } 5375 5376 File core_file; 5377 std::string core_file_path(outfile.GetPath()); 5378 error = core_file.Open(core_file_path.c_str(), 5379 File::eOpenOptionWrite | 5380 File::eOpenOptionTruncate | 5381 File::eOpenOptionCanCreate); 5382 if (error.Success()) 5383 { 5384 // Read 1 page at a time 5385 uint8_t bytes[0x1000]; 5386 // Write the mach header and load commands out to the core file 5387 size_t bytes_written = buffer.GetString().size(); 5388 error = core_file.Write(buffer.GetString().data(), bytes_written); 5389 if (error.Success()) 5390 { 5391 // Now write the file data for all memory segments in the process 5392 for (const auto &segment : segment_load_commands) 5393 { 5394 if (segment.fileoff != core_file.SeekFromStart(segment.fileoff)) 5395 { 5396 error.SetErrorStringWithFormat("unable to seek to offset 0x%" PRIx64 " in '%s'", segment.fileoff, core_file_path.c_str()); 5397 break; 5398 } 5399 5400 printf ("Saving data for segment at 0x%llx\n", segment.vmaddr); 5401 addr_t bytes_left = segment.vmsize; 5402 addr_t addr = segment.vmaddr; 5403 Error memory_read_error; 5404 while (bytes_left > 0 && error.Success()) 5405 { 5406 const size_t bytes_to_read = bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 5407 const size_t bytes_read = process_sp->ReadMemory(addr, bytes, bytes_to_read, memory_read_error); 5408 if (bytes_read == bytes_to_read) 5409 { 5410 size_t bytes_written = bytes_read; 5411 error = core_file.Write(bytes, bytes_written); 5412 bytes_left -= bytes_read; 5413 addr += bytes_read; 5414 } 5415 else 5416 { 5417 // Some pages within regions are not readable, those 5418 // should be zero filled 5419 memset (bytes, 0, bytes_to_read); 5420 size_t bytes_written = bytes_to_read; 5421 error = core_file.Write(bytes, bytes_written); 5422 bytes_left -= bytes_to_read; 5423 addr += bytes_to_read; 5424 } 5425 } 5426 } 5427 } 5428 } 5429 } 5430 else 5431 { 5432 error.SetErrorString("process doesn't support getting memory region info"); 5433 } 5434 } 5435 return true; // This is the right plug to handle saving core files for this process 5436 } 5437 } 5438 return false; 5439 } 5440 5441